Northern Territory Low Emissions Carbon Capture Storage and Utilisation Hub
Power Generation Options Analysis Ð Task 7 Report
David Green, Paul Graham, Lisa Havas, Giles Browne, James Foster, Jenny Hayward, Luke Reedman, Taj Khandoker, Andrew Ross, Jody Rogers

December 2024


CSIRO Energy
Citation
Green, D., Graham, P., Havas, L., Browne, G., Foster, J., Hayward, J., Reedman, L., Khandoker, T., Ross, A., Rogers, J. (2024) Northern Territory Low Emissions Carbon Capture Storage and Utilisation Hub, Power Generation Options AnalysisÐ Task 7 Report. CSIRO report number EP2024-6162, pp 72. CSIRO, Australia.
Copyright 
© Commonwealth Scientific and Industrial Research Organisation 2024. To the extent permitted by law, all rights are reserved and no part of this publication covered by copyright may be reproduced or copied in any form or by any means except with the written permission of CSIRO.
Important disclaimer
CSIRO advises that the information contained in this publication comprises general statements based on scientific research. The reader is advised and needs to be aware that such information may be incomplete or unable to be used in any specific situation. No reliance or actions must therefore be made on that information without seeking prior expert professional, scientific and technical advice. To the extent permitted by law, CSIRO (including its employees and consultants) excludes all liability to any person for any consequences, including but not limited to all losses, damages, costs, expenses and any other compensation, arising directly or indirectly from using this publication (in part or in whole) and any information or material contained in it.
CSIRO is committed to providing web accessible content wherever possible. If you are having difficulties with accessing this document, please contact csiro.au/contact. 
Foreword
Transitioning the global energy system while rapidly reducing emissions to net zero by 2050 is a vast and complex global challenge. 
Modelling of a range of emissions pathways and decarbonisation scenarios from the Intergovernmental Panel on Climate Change (IPCC, 2024), International Energy Agency (IEA, 2024) and Net Zero Australia (Net Zero Australia, 2024) shows that to meet net zero 2050 greenhouse gas emissions targets, a wide range of emissions reduction technologies will be required to decarbonise existing and future industries globally (IPCC, 2023).
These organisations identify that emissions elimination from hard-to-abate and high-emissions industries will require using carbon capture and storage (CCS) alongside other abatement strategies, such as electrification, underpinned by power generation from renewable energy sources such as photovoltaics and wind. 
Globally, there is considerable effort to identify industrial hubs and clusters where common user infrastructure can enable rapid decarbonisation of existing industries and future low-emissions industrial development. 
Australia has an opportunity to create new low-carbon growth industries and jobs in these areas, but lacks the infrastructure, skills base and business models to realise this. The transition to net zero will have greater impact on regional communities, particularly those reliant on industries in transition, but it may also create economic opportunities through a wide range of new industries and jobs suited to regional areas.
The Commonwealth Scientific and Industrial Research Organisation (CSIRO) is working to identify decarbonisation and transition pathways for existing and potential future industries that may be established in the Northern Territory (NT) by developing a Low Emissions Hub concept in the Darwin region. 
CSIRO has established a portfolio of projects to explore and evaluate a range of emissions reduction and emerging transition technologies and approaches. This includes research into Northern Territory renewable energy potential, hydrogen demand generation and storage, and carbon capture utilisation and storage (CCUS). CSIRO is working collaboratively with industry and government to understand their needs, drivers and strategic directions so that our research is informed and relevant. This includes establishing appropriate pathways and partnerships to understand and incorporate the perspectives of First Nations peoples.
A key activity is the research into a business case project (CSIRO, 2024; Ross et al., 2022) that aims to enhance understanding of the viability of a CCUS hub centred on the Middle Arm of Darwin Harbour. 
The work has three elements comprising 15 tasks: 
(1) analysing macroeconomic drivers, Northern Territory and regional emissions, low-emissions product markets (Ross et al., 2023), identifying key learnings from other low emissions hubs being developed globally, and cross-sector coupling opportunities (Tasks 0?5) 
(2) completing CCUS hub technical definition and technical risk reduction studies, including detailed studies on the infrastructure requirements for a CCUS hub, renewable power requirements for existing and potential future industries, and road-mapping for CO2 utilisation industries that could be established to produce low or net zero products (e.g. zero-emission chemical feedstocks) (CSIRO, 2023) (Tasks 6?9) 
(3) creating a business case to appreciate the scale of investment required to develop a Low Emissions Hub and the economic returns from doing so; this will lead to suggested business models and routes of execution (Tasks 10?14). 
The CCUS business case project will involve research that is based on possible industrial development scenarios, models of future potential emissions, market demand, technologies and costs. The project is intended to provide an understanding of possible future outcomes. Industry development will be determined by individual industry proponent investment decisions, government policies and regulations, and the development trajectories of technologies essential to the energy and emissions transition. 
On completion of this research, outcomes of the CCUS business case project will be made publicly available. 
The Low Emissions Hub vision identification work summarised in this report comprises Task 7 of the Northern Territory CCUS business case project. The purpose of this task is to understand the potential range in the costs to supply electricity (or hydrogen as a vector) to the Middle Arm Sustainable Development Precinct (MASDP), such that it can inform the potential emission avoidance technologies that could be implemented in the MASDP and the relative demand that may be expected for CCUS.


Contents
Acknowledgements	vi
Abbreviations		vii
Summary		x
1	Introduction	1
1.1	Northern Territory electricity networks	1
1.2	Planning for the future	3
2	Scenario, method and model description	7
2.1	Scenario design	7
2.2	Key questions for the power systems optimisation modelling	12
2.3	Model overview, key features and limitations	13
3	Model results	23
3.1	Generation technology mix and ACoE (unconstrained thermal)	24
3.2	Generation technology mix and ACoE (constrained thermal)	29
3.3	Implications for electrolysis-based hydrogen production	33
4	Conclusions	39
References		40
Appendices		43
A.1.	REZ capacity factor estimates	43
A.2.	Wind generation capacity factors	44
A.3.	Optimistic ACoH and ACoE sensitivities	46
A.4.	Model definition	48
A.5.	Sets (model dimensionality)	48
A.6.	Variables (unknown outputs of the model)	48
A.7.	Parameters (known inputs)	49
A.8.	Objective function	51
A.9.	Constraints	53


Figures
Figure 1: Northern Territory regulated electricity networks, and key transport, electricity and pipeline infrastructure, with modelled REZs and large-scale renewable energy projects identified	2
Figure 2: Australian solar irradiance map dataset (Bureau of Meteorology and Australia Government, 2020)	4
Figure 3: Example pictorial layout of regions and their connections used to describe the modelling results.	9
Figure 4 Example model solution showing the generation, storage and transmission mix	15
Figure 5: Example of electrical power demand and supply breakdown over a one-week period	17
Figure 6: ACoE and total cost breakdown for the example above	18
Figure 7: Comparison of the GenCost scenarios for forecast capital cost	19
Figure 8: Historical and base forecast wholesale price of natural gas	20
Figure 9: Constant load solution for unconstrained gas generation	27
Figure 10: 50/50 flat/variable load mix for unconstrained gas generation	28
Figure 11: Solution for constant load, 90% renewable fraction	31
Figure 12: Model solution for constant load, 100% renewable fraction	32
Figure 13: Optimal electrolyser sizing for 90% VRE fraction with wind	36
Figure 14: Tornado chart of ACoH and ACoE sensitivities	38

Tables
Table 1: ACoE versus load type and VRE fraction.	xiii
Table 2: NT installed but not operational renewable power generation	5
Table 3: NT proposed renewable power generation projects (to be connected to the grid, standalone facilities not included)	5
Table 4: REZ locations and available transmission and pipeline options	10
Table 5: Capacity limits applied to model regions	11
Table 6: Capacities selected by the model and resulting ACoE for the example case	16
Table 7: Range of ACoE and % VRE generation for different load types and unconstrained gas	26
Table 8: ACoE versus load type and VRE fraction	30
Table 9: Selected capacities for constant load, 90% renewable fraction	30
Table 10: Selected capacities for constant load, 100% renewable fraction	33
Table 11: ACoH and ACoE overload type and electrolyser capacity factor	35
Table 12: Model selected capacities for 90% renewables with hydrogen	37
Table 13: Sensitivity of ACoH and ACoE to various parameters	38
Table A.14 REZ renewable resource related capacity factors	43


Acknowledgements
CSIRO acknowledges the Traditional Owners of the land, sea and waters, of the area that we live and work on across Australia. We acknowledge their continuing connection to their culture, and we pay our respects to their Elders past and present.
The authors of this report acknowledge the support and funding provided by CSIRO to undertake this work. We thank the internal CSIRO independent peer reviewers for their review of the report and valuable comments and suggestions.
While this report is an output from a CSIRO-funded initiative, we thank our industry and government collaborators for their insights, contributions and suggestions, which have improved the report outcomes. In addition, we would like to thank Territory Generation and SunCable teams for their feedback, which has improved input assumptions in the models.
Although CSIRO has sought feedback from government and industry on the technical content of the report, CSIRO has sole discretion on including such feedback. 

Abbreviations
AC
alternating current 
ACoE
average cost of electricity 
ACoH
average cost of hydrogen
AE
alkaline electrolysis
AEMO
Australian Energy Market Operator
APGA
Australian Pipelines and Gas Association 
ARENA
Australian Renewable Energy Agency
BESS
battery energy storage system
CapEx
capital expenditure
CCGT
combined cycle gas turbine
CCS
carbon capture and storage
CCUS
carbon capture utilisation and storage
cf
compare
CP
Current Policies
DC
direct current 
DK BESS
Darwin-Katherine Battery Energy Storage System
DKIS
Darwin-Katherine Interconnected System
DKTL
Darwin-Katherine Transmission Line
DLNG
Darwin LNG
EDL
Energy Developments Pty Ltd
ElectSto
Electricity Storage
ElectStoTech
Electricity Storage Technology (e.g. batteries)
Gas Ex
gas exported
GenCost 
CSIRO Energy Generation Cost Report
GJ 
gigajoule
GW
gigawatt
GWh
gigawatt hours
ha
hectare 
H2
hydrogen 
Hr
hour
HVAC
high-voltage alternating current
HVDC
high-voltage direct current
IASR
Inputs, Assumptions and Scenarios Report
IEA 
International Energy Agency
ILNG
Inpex facility LNG
IPCC
Intergovernmental Panel on Climate Change
km
kilometres 
kV
kilovolts 
LAT
latitude
LNG
liquified natural gas 
LONG
longitude
LTCCF
least total cost capacity factor
M
Million?
MASDP
Middle Arm Sustainable Development Precinct
Max
maximum
Min
minimum
MW
megawatt
MWh
megawatt hour
NaN
Not a Number
NEM
National Electricity Market
NT
Northen Territory
NT-DIPL
Northern Territory Department of Infrastructure, Planning and Logistics
NTG 
Northern Territory Government
NT-LEH
Northern Territory Low Emissions Hub
NZ
net zero 
O&M
operating and maintenance
OpEx
operational expenditure
PV
Photovoltaic 
PWC
Power and Water Corporation 
QLD
Queensland 
RenNin
Renewables ninja
REZs
Renewable Energy Zones
RFSU
ready for startup
SCADA
supervisory control and data acquisition
SMR
steam methane reforming
SODAR
sonic detection and ranging
SolLargePV
large-scale solar PV
t
tonnes 
TGen 
Territory Generation
tpd
tonnes per day 
T/hr
tonnes per hour
Tx
transmission
TWh
terawatt hour
VRE
variable renewable energy
WindOn
Wind is turned on in the optimisation model
Summary
Transitioning the global energy system while reducing emissions to net zero by 2050 is a complex challenge. In understanding decarbonisation of existing and future industries in the Northern Territory, a range of mechanisms that can be used to achieve decarbonisation objectives are being investigated, including the use of renewable electricity. Integrating sources of renewable electricity is critical to avoid emissions from not only the power sector but also industrial users of energy. This, in turn, reduces demand for abatement technologies such as CCUS and/or offsets.
The Northern Territory Government (NTG) is looking to invest in the regionÕs ability to take advantage of the global clean energy transition in the form of the proposed Middle Arm Sustainable Development Precinct (MASDP). The MASDP is envisaged to include energy-intensive industries such as renewable hydrogen generation, lithium processing, liquified natural gas (LNG), methanol, ethylene, ammonia, urea synthesis and the operation of a precinct CCUS system.
The purpose of this study is to understand the potential range in the costs to supply energy (electricity and hydrogen) to the MASDP. Various cost estimates are provided by examining least-cost electrical power generation options to meet both the electrical and hydrogen production demand required by the potential MASDP industries. The range of costs results from the range in the allowable fraction of gas-based generation, the type of electrical load (flat versus variable) and model inputs such as the wholesale price of gas and the technology cost year. The modelling additionally investigates which aspects of energy infrastructure the cost of supply is most sensitive to, such as the incorporation of wind generation, as well as the interdependence of power generation and electrolysis-based hydrogen production.
As with all tasks within the CCUS business case, the CSIRO team has consulted widely with industry and the NTG for guidance on the inputs into the models used. However, it is important to note that the results presented herein do not consider detailed proponent design considerations, their individual needs or commercial arrangements, but rather seek to understand system-level costs only and therefore should only be used for this purpose.
Assumptions
While the Darwin-Katherine Interconnected System (DKIS) regulated electricity network is being upgraded for modest increases in maximum demand, industrial electrical demand from the MASDP is likely to be much larger than the entirety of the DKIS demand. As such, for the purposes of this modelling, the energy infrastructure for the MASDP is considered as a separate network to the existing DKIS.
For the electrical load, the modelling represents all industries except electrolysis-based hydrogen production as presenting an aggregated constant load. In power terms, that flat component of the load was assumed as roughly 0.78 GW. For electrolysis-based hydrogen production, the model assumes 110,000 t of hydrogen per year, which represents about half the annual electrical load (of just under 13 TWh). Since there is considerable flexibility in electrolyser operation, this was assumed to be a variable load with the model able to determine the optimal electrolyser capacity factor. 
For renewable energy infrastructure siting, four defined Renewable Energy Zones (REZs) were nominated. These REZs have associated distances from Middle Arm that are approximate representations for the purposes of estimating the cost to build transmission and/or a pipeline to link them to Middle Arm in a new electrical and/or hydrogen gas network.
Assumed transmission options available to the model include lower power (500 MW) high-voltage alternating current (HVAC) to connect the three capacity-constrained and closer REZs to Middle Arm, and two high-power (2,500 or 5,500 MW) high-voltage direct current (HVDC) costing options for connecting Powell Creek to Middle Arm. These high-power HVDC transmission options are further constrained in the model to have a minimum 90% capacity factor to reflect an expected requirement on the economic viability of such a project. Renewable electricity generation is constrained to the REZs, while new natural-gas-based electricity generation is constrained to Middle Arm, as this is the location of both existing and new LNG production and it already has an incoming gas supply pipeline. The transmission routes do not consider current or future easement restrictions and take the shortest path.
Key questions
Given an approximate 50/50 mix of flat and variable electrical load in the Middle Arm Balanced Scenario, the probability that industrial customers requiring a variable load may be expecting different pricing for electricity than customers requiring a constant load, the constraints on individual system component utilisation (i.e. the HVDC transmission link to Powell Creek) and the general uncertainty in how to best size electrolysers, the study was focused on answering the following three key questions:
1. How does the relative makeup of load type (flat versus variable) impact the least-cost solution for electrical power generation infrastructure and range for the average cost of electricity (ACoE)? 
2. How does the constraint on gas generation impact the total cost of the system and the resulting ACoE without CCS?
3. What does the above imply for electrolysis-based hydrogen production in terms of the choice of power system infrastructure, electrolyser sizing and the production cost of hydrogen?
The model 
Based on these key questions, a least-cost optimisation model was developed to explore the mix of generation, locations and associated transmission and pipelines that could meet the industrial electricity demand and the hydrogen production target estimates for the MASDP. While such models are useful in providing consistent infrastructure configurations and cost estimates, and for rapidly assessing how different choices in future scenarios impact the entire system configuration, they are only one part of the broader activity of developing an actionable plan for an industrial hub. By necessity, the models employed here contain limited detail on individual proponents, and take a high-level, Ôperfect coordinationÕ view, whereas the reality of the individual aspects may be quite different due to their individual costs, access to land, competitive advantage or other aspects relevant to them.
As there is uncertainty associated with future gas and technology costs, model realisations have been conducted in 2025 and 2028 technology years and gas prices have been modelled at A$10.4/GJ in 2025 and $8/GJ in 2028 for the base price, and $15.5/GJ in 2025 and $12.1/GJ in 2028 for a 1.5x higher priced sensitivity.
Typically, it would be expected that generation mix and ACoE would vary strongly with load type, and as such the model is applied to a purely constant load, a 50/50 flat/variable load mix and a purely variable load (for each of the above model realisations). To remove the complication of electrolyser and hydrogen infrastructure in answering questions 1 and 2, for those the variable load is not associated with any particular process but is directly associated with the production of hydrogen for question 3.
Model results
In addressing the key study questions, the model results are organised into three categories aligned to the three key questions for the study.
Question 1: How does the relative makeup of load type (flat versus variable) impact the least-cost solution for electrical power generation infrastructure and range for the ACoE? 
With no constraint on gas generation, the model does not build battery storage or onshore wind, even at the higher end of gas prices (Table 1). The 90% utilisation constraint on the HVDC transmission from Powell Creek to Middle Arm, which requires battery storage to meet demand, means that without a gas generation constraint, the model does not use the Powell Creek REZ.
For the constant load model realisation, gas price is the driver of whether solar photovoltaic (PV) is built.  Even though the ACoE from solar PV (with transmission) is significantly lower (approximately $91/MWh compared with $106/MWh for gas at $10.4/GJ), the model does not choose solar until the savings offset the capital investment in the panels Ð this happens at the higher $15.5/GJ and $12.1/GJ gas prices. For those higher gas price cases, the model chooses to fully use the solar PV capacity at the three closer REZs. HVAC transmission cost represents between 3% and 7% for those closer REZs and is used at the 29% capacity factor.
For models where only constant loads are allowed, the 50/50 mix and purely variable load model realisations at all gas fuel prices (considered here) and both 2025 and 2028 technology costs, 14% of the generation is supplied by solar PV in the REZs, with the cost for gas generation being $88/MWh for $8/GJ in 2028 and $81/MWh for solar (slightly higher at $84/MWh in Koolpinyah). The capacity limits at the three closer REZs mean that variable renewable energy (VRE) is able to meet only 14% of total demand, so even when the variable fraction of load is higher, there is no capacity to meet that demand with VRE (without using Powell Creek Ð which it does not do due to the transmission utilisation constraint).
Question 2: How does the constraint on gas generation impact the total cost of the system and the resulting ACoE without CCS?
This question was tested using the same three load types, technology years and gas price variances as above, with four constraints on the fraction of annual generation from gas, unconstrained (100%), 50%, 10% and 0% (i.e. renewable electricity only).
The results for ACoE for this modelling are shown in later in the report. For a constant load, the ACoE rises sharply (about 1.8x) when increasing from 90% renewables to 100%, driven by the cost of battery storage. There is little difference between the 50/50 load mix and a purely variable load due to the 90% utilisation constraint on the HVDC transmission link to Powell Creek. The addition of onshore wind reduces ACoE by approximately 15% when a constraint on gas generation is present. For a constant load, the spilled VRE energy is large (50%) for the 100% renewable fraction, and for the 90% renewable fraction the gas generation capacity factor is low (25%).

Table 1: ACoE versus load type and VRE fraction. 
Minimum and maximum are over the range of results for 4 combinations of technology cost year and wholesale gas price considered in Section 3.1, and with the lower HVDC costing. The coloured shading of the cells is only to help guide the eye as to where the significant variations occur. The values in bold ($174/MWh and $316/MWh) correspond to the solutions shown in Figure 11 and Figure 12 respectively
$/MWh
VRE Only
Gas <= 10%
Gas <= 50%
Gas <= 100%
Load Type
Min
Max
Min
max
Min
Max
Min
Max
Constant load, no wind
367
422
203
240
135
181
94
150
50/50 load, no wind
195
225
176
209
128
173
93
150
Variable load, no wind
187
219
176
209
128
173
93
146
Constant load, with wind
316
367
174
207
120
162
94
150
50/50 load, with wind
162
188
150
178
109
150
93
146
Variable load, with wind
162
188
150
178
109
150
93
146

Consistency with GenCost Estimates of LCoE
This project utilises the capital and operating costs of generation technologies published in the CSIRO GenCost report. However, the costs of high variable renewable share electricity presented here are generally higher than those that are published in GenCost for three key reasons. First, the Northern Territory has a cost premium for infrastructure investment relative to the more populous states. Second, industrial load curves are flatter than the load curves that are found in larger mixed-customer grids, and flatter demand creates a higher requirement for capacity to maintain constant supply. And third, the Middle Arm project represents investment in mostly new standalone infrastructure, whereas the GenCost modelling of renewable integration costs is in the context of adding to existing grids.

Question 3: What does the above imply for electrolysis-based hydrogen production in terms of the choice of power system infrastructure, electrolyser sizing and the production cost of hydrogen?
In examining this question, variable loads were assigned to electrolysis-based hydrogen production and the average cost of hydrogen (ACoH) was derived as well as the impact that hydrogen production has on system design and the ACoE.
The least cost of ACoH varies from $7/kg to $9/kg at the 50% and 100% renewable fractions, respectively. At the 90% renewable fraction, the ACoH is near $8.2/kg with a corresponding ACoE of $146/MWh. For the low (50%) renewable fraction, the optimal electrolyser capacity factor is near 90%, while for high (90% and 100%) renewable fractions it is near 60%. The latter result is mostly driven by the 90% capacity factor constraint on the HVDC link to Powell Creek. The ACoH is most sensitive to the alkaline electrolysis (AE) efficiency (followed by the renewable fraction and wind capacity correction factor). The ACoE is most sensitive to the renewable fraction (followed by the wind capacity correction factor and the discount rate). A hydrogen pipeline from Koolpinyah to Middle Arm is chosen by the model as a least-cost option.
Conclusions
This report provides a range of estimates for the average cost of supply for electricity and hydrogen to meet the demand required by the potential MASDP industries. The ACoE ranges from $93/MWh when no constraint is applied to the renewable fraction, to $176?209/MWh for a 50/50 mix of flat and variables loads and a 90% renewable fraction. The cost is most sensitive to the renewable fraction, followed by the capacity factor of onshore wind Ð which, when included, can decrease the cost of electricity by approximately 15%. Increasing the renewable fraction from 90% to 100% nearly doubles the cost of electricity, driven by the cost of battery storage. 
ACoH at the 90% renewable fraction is near $8.2/kg, with a corresponding ACoE of $146/MWh and an optimal 67% electrolyser capacity factor. The cost of hydrogen supply is most sensitive to the assumed electrolyser efficiency, followed also by the renewable fraction and the onshore wind capacity factor. The optimal electrolyser capacity factor is driven mostly by the 90% capacity factor constraint on the HVDC link to the large REZ at Powell Creek. 
Key needs for the future deployment of renewable electricity in the Northern Territory are the development of a greater understanding of the potential wind resources in the REZs and the identification of low-cost energy storage technologies.


1 
1 Introduction
In understanding decarbonisation of existing industries and the development of low-emissions future industries in the Northern Territory, a range of mechanisms that can be used to achieve decarbonisation objectives have been investigated. The integration of sources of renewable electricity is critical to avoid emissions from not only the power sector but also industrial users of energy. This, in turn, helps determine the demand for abatement technologies such as CCUS and/or offsets. As has been discussed in the Task 1 report of this study, there is strong potential demand for renewable electrification from existing and future industries that can contribute to meeting the Northern TerritoryÕs decarbonisation goals. When used in combination with CCS, renewable sources of electricity and hydrogen could lead to significant reductions in Northern Territory emissions.
1.1 Northern Territory electricity networks
When considering the addition of future renewable electricity generation to the Northern Territory electricity networks, it is useful to consider the existing electricity network and generation facilities.
The Northern TerritoryÕs regulated electricity system is one of the smallest, by capacity, in Australia and comprises three discrete networks: the Alice Springs network, the Darwin-Katherine Interconnected system (DKIS) and the Tennent Creek network. This regulated electricity system is managed by the Power and Water Corporation (PWC).
Of the three networks (Figure 1), the DKIS is the largest with a peak demand of 290 MW, which provided 1500 GWh of electricity to residential and business customers in 2023 PWC (2023). In this system, a 132 kV transmission line runs from Channel Island, south of Darwin, and connects Manton, Batchelor, Pine Creek and Katherine. At each of these locations there is a zone substation and distribution infrastructure. A 66 kV transmission line transports power from Channel Island to the Darwin urban area via the Hudson Creek sub-transmission substation. The 66 kV network also supplies the rural areas of Darwin from Strangways to Humpty Doo and Mary River.
In 2021, electricity consumption in the DKIS was primarily from thermal energy generation (88%), large-scale solar (3%) and small-scale solar (9%) (NTG, 2021). Thermal electricity generation is predominantly provided by Territory Generation (TGen) operated gas turbines located at the Channel Island (278.4 MW capacity), Weddell (129.0 MW capacity) and Katherine power stations (37.0 MW capacity) (TGen, 2024). Supplemental electricity contributions are from gas turbines located at Energy DevelopmentÕs (EDLÕs) Pine Creek Power Station (27 MW capacity) (EDL, 2018) and from British Solar RenewablesÕ solar farms at Manton Dam (10 MW capacity), Batchelor (20 MW capacity) and Katherine (25 MW capacity) (British Solar Renewables, 2024).

Figure 1: Northern Territory regulated electricity networks, and key transport, electricity and pipeline infrastructure, with modelled REZs and large-scale renewable energy projects identified

The Alice Springs network (Figure 1) provides electricity to the township of Alice Springs and surrounding areas. The network has a peak demand of 50 MW and provided 250 GWh of electricity to residential and business customers in 2023 (PWC, 2023). Energy is supplied at 11 kV from the Lovegrove zone substation. South of ÔThe GapÕ are predominantly commercial and industrial customers with energy supplied at 22 kV. Power is generated at the TGen-operated Ron Goodin/Sadadeen Valley Power Station (35.3 MW capacity; reciprocating sets and one industrial gas turbine; 5 MW battery energy storage system [BESS]) and from Owen Springs Power Station (80.9 MW capacity; reciprocating sets and one industrial gas turbine) (TGen, 2024). 
The smallest of the regulated networks is Tennant Creek (Figure 1). The Tennant Creek network has a peak demand of 7 MW capacity and provided 30 GWh of electricity to residential and business customers in 2023 (PWC, 2023). The network is relatively simple, with a single power station and an adjacent zone substation. Electricity is generated at 11 kV from reciprocating sets and one industrial gas turbine operated by TGen (21.9 MW capacity) and is stepped up to 22 kV at the zone substation where power is transported to customers through 22 kV distribution feeders (TGen, 2024).
In addition, several unregulated smaller electricity networks service small towns and communities. These areas are not connected to the three regulated electricity networks. However, although the networks are unregulated, they are still managed by PWC. These smaller electricity networks are: Daly River; Jabiru; Borroloola; Timber Creek; Daly Waters; Elliott; Newcastle Waters; Yulara; Ti Tree; Kings Canyon; Nhulunbuy ? surrounding rural areas only; Groote Eylandt ? Angurugu and Umbakumba only; and Indigenous communities under the Indigenous Essential Services program.
Furthermore, several industries generate their own power for consumption on-premises. These include remote mine sites and the LNG facilities found on the Middle Arm of Darwin Harbour. For example, the INPEX facility (ILNG) on Middle Arm has an installed gas turbine generation capacity of 490 MW, and the Darwin LNG (DLNG) facility, also located on Middle Arm, has an installed gas turbine generation capacity of 180 MW (Territory, 2018).  
1.2 Planning for the future 
The NTG in its Roadmap to Renewables report (Langworthy et al., 2017), targeted 50%?of electricity to be sourced from renewable energy by2030. This target is a key element of the Northern Territory Climate Change Response: Towards?2050 (NTG, 2020)
As discussed in the CCUS business case Task 1 report (Rogers et al., 2024), the Northern Territory has some of the highest solar irradiance globally (Figure 2). There is also an (as of yet to be fully characterised) potential onshore wind resource that could contribute to future renewable electricity generation in the Territory.
In the Northern Territory Climate Change Response: Towards 2050 discussion papers (NTG, 2020), the NTG identified the Territory as having one of the worldÕs best solar resources (Figure 2) as well as the presence of existing energy corridors and easements, energy export infrastructure and proven roll-out of large- and small-scale solar installations (e.g. Katherine Solar Farm and solar energy installations in remote communities). The NTG is increasingly looking to solar to displace diesel in remote Indigenous communities with support from the Australian Renewable Energy Agency (ARENA) and has already provided 10 MW of solar PV systems across 25 sites, servicing 27 communities (ARENA Australian Government, 2018).


Figure 2: Australian solar irradiance map dataset (Bureau of Meteorology and Australia Government, 2020)
In addition to satisfying emission reduction targets, it is anticipated that maximum demand on the regulated DKIS system will grow through to 2030. The NTG in its Darwin-Katherine Electricity System Plan (NTG, 2021) has explored three scenarios, with maximum demand of 298 MW in 2020 increasing to a range of 312?486 MW by 2030. For the upper demand band, the NTG anticipates that this enhanced demand will come from new industry, mine sites and vehicle electrification. The assumption in the lower band scenario requires an installed renewable capacity of 323 MW, while the upper band scenario requires an installed renewable capacity of 511 MW to meet both the renewable energy targets and demand requirements (NTG, 2021).
In the 2023 Transition and Distribution Annual Planning Report PWC is planning for the retirement of 185 MW of thermal generation between 2027 and 2030 at the Channel Island Power Station, with 37 MW of new thermal generation, 267 MW of solar capacity and 33 MW of battery capacity additions over the same time period. These changes to the DKIS generation capacity are not, as yet, aligned with the 2020 Northern Territory Darwin-Katherine Electricity System Plan (NTG, 2021). 
To meet both the renewable energy target and enhanced demand, PWC is investing in network improvements through the upgrading of protection relays in the DKIS network and constructing new transmission infrastructure and substations to the south of Darwin on the Darwin-Katherine Transmission Line (DKTL). This is to improve the dispatchability of existing installed renewable electricity generation (e.g. projects identified in Table 2) and to provide further network access to proposed large-scale renewable energy projects (e.g. projects identified in Table 3) (PWC, 2023). In addition, TGen is in the commissioning phase of its 35 MW/35 MWh Darwin-Katherine Battery Energy Storage System (DK BESS) located at Channel Island (TGen, 2023) and has announced the development of a DK BESS 2 project that will satisfy the requirement of the Darwin-Katherine Electricity System Plan of having 105 MW of high-specification batteries (Lawler and Worden, 2024).

Table 2: NT installed but not operational renewable power generation
Facility name
Type
Company
MW
Size (MWh/yr)
Katherine Ð 34 MW (DKIS)1
Solar
Eni
34
74,460
Batchelor/Manton Dam Ð 25 MW2
Solar
Eni
25
54,660
Manton2
Solar
Rimfire
10
21,900
Notes: 1connected not dispatching; 2limitations in DKIS transmissions system being able to adequately manage variable power from solar systems.
Table 3: NT proposed renewable power generation projects (to be connected to the grid, standalone facilities not included)
Facility name
Type
Company
MW
Size (MWh/yr)
Sun Cable (RFSU 2030, DKIS)*
Solar/Wind
Grok Ventures
900
2,299,500
Larrakia*
Solar
Larrakia Energy
300
657,000
Darwin Renewable Energy Hub (Proposed)
Solar
TBA
180Ð230
TBA
Berry Springs
Solar
Livingstone Solar
~40
90,000
Notes: *current announcements on these projects do not include connection to the DKIS. RFSU Ð ready for startup. Does not include full Sun Cable capacity, just that allocated for the Darwin region

The NTG is looking to invest in the regionÕs ability to take advantage of the global clean energy transition in the form of the proposed MASDP. While the DKIS regulated electricity network is being upgraded for modest increases in maximum demand, industrial electrical demand from the MASDP is likely to be larger than the entirety of the DKIS demand. The non-regulated electricity generation capacity on Middle Arm from current LNG plants is already 670 MW and the development of proposed energy-intensive industries such as CCUS system operation, renewable hydrogen generation, lithium processing, methanol, ethylene, ammonia and urea synthesis (NTG, 2023a) are envisaged to substantially increase electricity demand. To satisfy this potential demand, both the Larrakia large-sale solar project and the SunCable renewable energy project are intending to provide electricity directly to Middle Arm industrial proponents through dedicated 300 MW and 900 MW capacity (Table 3) non-regulated industrial electricity networks. 
The purpose of this study is to understand the potential range in the costs to supply energy (electricity or hydrogen) to the MASDP, such that it can inform possible uptake of potential emission avoidance technologies and the relative demand that may be expected for CCUS as a result.
In this report, a range of cost estimates are provided by examining least-cost electrical power generation options to meet both electrical and hydrogen production demand required by potential MASDP industries. The range of costs results from the range in the allowable fraction of gas-based generation, the type of electrical load (flat versus variable) and other model inputs such as the wholesale price of gas and technology cost year. The modelling additionally investigates which aspects of energy infrastructure the cost of supply is most sensitive to, such as the incorporation of wind generation, as well as the interdependence of power generation and electrolysis-based hydrogen production.
As with all tasks within the CCUS business case, the CSIRO team has consulted widely with industry and the NTG for guidance on the inputs into the models used. It is important to note, however, that the results presented herein do not consider detailed proponent design considerations, their individual needs or commercial arrangements, but rather seek to understand system-level costs only and therefore should only be used for this purpose.
2 Scenario, method and model description
This section describes the method used to model power generation and transmission options and the resulting cost of energy supply (electricity and hydrogen) for the MASDP (NTG, 2024). The section 2.1 describes the scenario to which the model is applied, the section 2.2 lists the key questions answered to achieve the project goal, and section 2.3 summarises the key model features including the inputs, assumptions and caveats.
2.1 Scenario design
2.1.1 Stakeholder engagement
The scenario described below is based on consultation with the relevant stakeholders from the Northern Territory Department of Infrastructure, Planning and Logistics (NT-DIPL) and the Northern Territory Low Emissions Hub (NT-LEH) working group, which includes both NT-DIPL and existing and future industry proponents. Beyond this group, CSIRO also consulted with TGen and SunCable. Key parameters defined from this engagement were:
* the scale of electrical demand for the MASDP
* the approximate mix of flat and variable loads 
* an approximate annual electrolysis-based hydrogen production target 
* the nominal location of potential REZs that the MASDP may be connected to 
* the constraints on capacities per region (cf.Table 4) 
* expectations for high-value asset utilisation (i.e. capacity factors for HVDC transmission and electrolysers) 
* the decision to consider this as a greenfield project (with an exception to consider the impact of off-islanding existing LNG plant power generation). 
The final scenario configuration and sensitivities examined below are the result of feedback, discussion and subsequent refinement of the scenario with these stakeholders.
2.1.2 Scenario description
The NT-DIPL defined a possible makeup of future industries that could be situated in the MASDP in its Balanced Scenario (this scenario uses the widest range of industries that are envisaged to be established in the MASDP). The real industrial mix in the MASDP may not match the exact composition used here, but the Balanced Scenario offers a way to align with other modelling and design activities that the NTG is pursuing, and therefore this scenario has been chosen for the whole of the NT-LEH CCUS business case project. The industries included are the production of LNG, hydrogen (both from steam methane reforming [SMR] and electrolysis based), methanol, ammonia (one based on the hydrogen from SMR and another on that from electrolysis), urea and ethylene, CCUS and critical minerals processing (e.g. lithium, vanadium).
For electrolysis-based hydrogen production, the annual production target (approximately 110,000 tonnes) represents about half the annual electrical load (just under 13 TWh) defined for the MASDP. Since there is considerable flexibility in electrolyser operation, for AE type electrolysers which are the most cost effective in the near term, this is represented as a variable load, and the remainder of the industries are represented as a constant load. As such, the electrical load aggregated over industries is described as a mix of variable and flat components, with the modelling investigating how best to distribute the variable component. In power terms, the flat component is roughly 0.78 GW.
For renewable energy infrastructure siting, the model is allowed to choose from four defined REZs (Figure 1 and Figure 3):
* the REZ under consideration by the NGT in the Darwin-Katherine Electricity System Plan (here labelled ÒWeddellÓ) (NTG, 2021)
* the REZ at Powell Creek (at 800 km distance), representative of a SunCable-like project (NTG, 2023b; SunCable, 2024)
* the REZ close to Middle Arm (here labelled ÒBlackmoreÓ), representing an installation at the scale and approximate distance of that proposed by Larrakia Energy (2022)
* the REZ further out at 40 km distance (here labelled ÒKoolpinyahÓ) to allow the model additional capacity with more of a distance trade-off. 
The geographic layout of these nominal zones is shown in Figure 3. Their specific locations with available transmission and pipeline options for connecting to Middle Arm are listed in Table 4 and their capacity limits are listed in Table 5.
These distances are approximate representations for the purposes of estimating the cost to build transmission and/or pipeline to link them to Middle Arm in a new electrical and/or hydrogen gas network (i.e. not connected to the existing DKIS). The routes do not consider current or future easement restrictions as they take the shortest path. The REZs are named after their nearest town ? they may be known by other names, and the final locations selected will likely differ.

Figure 3: Example pictorial layout of regions and their connections used to describe the modelling results.
The representation for the Powell Creek location is closer than the actual location (800 km away from Middle Arm) for ease of visualisation. The various coloured circles represent selected capacities for any model run: CCGT gas generation (red), large-scale solar PV (green), onshore wind (blue), battery storage (aqua), electrolyser capacity (yellow), electricity transmission (blue) and hydrogen pipeline (pink). While the green circles are calibrated to represent the land area needed for that capacity solar farm, the other colours are only relative to that same calibration, not to their own land use estimate.
The quality of renewable resources at each REZ can broadly be thought of as increasing with distance from Middle Arm, particularly to the south. The capacity factors for large-scale solar PV and wind generation at these locations (cf. Figure 2) are approximately: Blackmore (solar PV) 23.3%, (wind) 19.4%; Weddell 23.3%, 19.9%; Koolpinyah 23.4%, 20.3%; and Powell Creek 28.2%, 34.3% (Pfenninger and Staffell, 2016). Typical ranges for solar PV are 10?25% and for onshore wind are 25?45%, so the Powell Creek REZ stands out as having excellent solar irradiance and significantly higher wind capacity factors than the others (although still only in the middle of the typical ranges found elsewhere in Australia). 
Note also that the wind resource estimate is based on ÒreanalysisÓ of satellite observations for a 4.5 MW turbine at 150 m height. The reanalysis method calibrates atmospheric models to satellite observations of globally observable quantities that these models include, and then uses the resulting local wind speed and direction from the calibrated model. This has the advantage that wind resource estimates are available everywhere, but the disadvantage of lacking the accuracy that a more local observation provides. 
This lack of accuracy is typically corrected for by applying a local bias correction (Staffell and Pfenninger, 2016) when more accurate local data are available (e.g. from sonic detection and ranging [SODAR] or wind measurement towers). While SODAR data have recently been collected in Alice Springs and Tennant Creek, there is no reliable reanalysis bias correction calibration database for the Northern Territory as there is for the National Electricity Market (NEM) regions. Here, a crude capacity correction factor of 0.7 is applied, based on the differences between the reanalysis-based estimates of power generated from Renewables.ninja (Pfenninger and Staffell, 2021)) and actual observations (from supervisory control and data acquisition [SCADA] data) for north-western Queensland locations in the NEM. Thus, the wind resource and associated capacity factors have considerable uncertainty. 
Further discussion of the correction and wind resource estimates can be found in Appendix A.2, and the impact of the correction factor is considered a sensitivity in the results section.
Table 4: REZ locations and available transmission and pipeline options
REZ 
Lat
Long
Distance to Middle Arm [km]
Transmission options
H2 pipeline options
Weddell
12.6677
131.0615
20
HVAC 500 MW
M$2.14/km
$3,698/MW/km
N/A
Blackmore
12.6662
130.9291
20
HVAC 500 MW
N/A
Koolpinyah
12.3940
131.1746
40
HVAC 500 MW
17.4 T/hr
$850,000/km
$48,911/T/hr/km
Powell Creek
18.1154
133.5310
800
(high) HVDC 2,500 MW 
M$2.54/km 
$878/MW/km
³ 90% utilisation
Or
(low) HVDC 5,500 MW
M$2.54/km
$439/MW/km
³ 90% utilisation
N/A
All transmission and hydrogen pipeline costing data are sourced from Australian Pipelines and Gas Association (APGA, 2021). The costing for long-distance (>500 km), high-power HVDC in that reference is the same for 2.5 and 5.5 GW options on a per 500 km basis. We use the costing on the 2.5 GW option as a high value, and the 5.5 GW option as a low value. These approximate HVDC costings include the DC to AC Òconvertor stationsÓ required to connect the HVDC lines to an AC network.
For the purposes of this modelling, the energy infrastructure for the MASDP is considered as a separate network to the existing DKIS. This is largely the result of stakeholder discussions around generation assets connected to the DKIS being subject to additional system connection and control requirements, which may not be required to service MASDP industry or export customers. However, further modelling could incorporate the integration of the two systems should that become the preferred option in the future. 
For this report, the transmission options made available to the model in the scenario are listed in Table 4 and include lower power (500 MW) HVAC to connect the three capacity-constrained and closer REZs to Middle Arm, and two high-power (2,500 or 5,500 MW) HVDC costing options for connecting Powell Creek to Middle Arm. The high-power HVDC transmission options are further constrained in the model to have a minimum 90% capacity factor to reflect an expected requirement on the economic viability of such a project (Pers. Com SunCable). 
While each costing option has a nominal power label, the model can build any fraction of that value and should be considered a costing label. For the only REZ (Koolpinyah) where the model can choose to place hydrogen production via electrolysis (in addition to Middle Arm), it can choose to build new hydrogen pipeline (at the cost of 17.4 T/hr capacity). 
Renewable generation is constrained to the REZs, while new natural-gas-based generation is constrained to Middle Arm, as this is the location of both existing and new LNG production and already has an incoming gas supply pipeline. While other configurations of energy infrastructure to service Middle Arm are both possible and plausible, this choice was deemed sensible in conversation with stakeholders.
Table 5: Capacity limits applied to model regions
Region/REZ
Solar PV 
[MW]
Onshore wind [MW]
CCGT gas 
[MW]
H2 electrolyser capacity 
[MW]
Battery storage [MWh] 
(1, 2, 4 & 8 hrs)
H2 storage (tank) [T]
Middle Arm
0
0
Unlimited
Unlimited
Unlimited
Unlimited
Weddell
200
200
0
0
Unlimited
Unlimited
Blackmore
300
200
0
0
Unlimited
Unlimited
Koolpinyah
400
200
0
Unlimited
Unlimited
Unlimited
Powell Creek
Unlimited
1,000
0
0
Unlimited
Unlimited
CCGT Ð combined cycle gas turbine. 
These limits do not represent any regulatory or physical limits but are nominal values used to facilitate the modelling. In the case of electrolyser capacity limits, and since the model does not consider water availability, the restriction on electrolyser siting to Middle Arm and Koolpinyah was the result of stakeholder feedback.
2.1.3 Flat and variable demand load
The energy demand was simplified by assuming that all industries other than electrolysis-based hydrogen production require a constant load and by aggregating over the industry mix to give a flat electrical power demand of 0.78 GW. The electrolyser operation is required to produce an annual quota equivalent to 300 tpd (tonnes per day), but the model can choose when to run the electrolyser(s) to meet that demand and which generation mix to power it (subject to any constraint of allowable gas-based generation, as described below). This means a variable demand of approximately the same scale as the flat electrical load. See Figure 5 for an example electrical demand trace with flat and variable components. Since the fraction of the load is variable and is a significant driver of the resulting ACoE, the model was also run across a range of load mixes to investigate this relationship (see Section 3.2).
2.1.4 Approach to emissions constraints
As the MASDP is a greenfield project, the modelling results aim to help in understanding the expected costs of supply associated with any particular choice or solution for that initial energy infrastructure. As such, the modelling assumes a greenfield build for all cases and is run only for the 2025 and 2028 technology cost years (see Section 2.3.3). This contrasts with a staged approach where an initially unconstrained (by the Balanced Scenario) emissions scenario result is subsequently constrained. This will likely be the case under the commitments by the Northern Territory and Commonwealth governments to reach net zero emissions by 2050 (Australian Government et al., 2022; NTG, 2020) and the associated Safeguard Mechanism. 
Given the breadth of possible solutions under various stakeholder interests, examining options for the staged reduction of emissions is left for future work once the initial energy infrastructure has been installed. Here a range of emissions ambitions in the initial infrastructure is investigated via a simple constraint on the allowable percentage of electrical power generation from gas-based generation. In Section 3.1 gas-based generation is left unconstrained, while in Sections 3.2 and 3.3 solutions where gas is constrained to be no more than 100%, 50%, 10% and 0% of annual energy production (electricity and hydrogen) are investigated. 
2.2 Key questions for the power systems optimisation modelling 
Given an approximate 50/50 mix of flat and variable electrical load in the Middle Arm Balanced Scenario, the probability that industrial customers requiring a variable load may be expecting different pricing for electricity than customers requiring a constant load, the constraints on individual system component utilisation (i.e. the HVDC transmission link to Powell Creek) and the general uncertainty in how to best size electrolysers, the study is organised to answer the following three key questions:
1. How does the relative makeup of load type (flat versus variable) impact the least-cost solution for electrical power generation infrastructure and range for the ACoE? 
2. How does the constraint on gas generation impact the total cost of the system and the resulting ACoE without CCS?
3. What does the above imply for electrolysis-based hydrogen production in terms of the choice of power system infrastructure, electrolyser sizing and the production cost of hydrogen?
These questions are addressed by application of the model in Sections 3.1, 3.2 and 3.3, respectively.
2.3 Model overview, key features and limitations
Based on the three key questions, a least-cost optimisation model was developed to explore the mix of generation, locations and associated transmission and pipelines that could meet the industrial electricity demand and the hydrogen production target estimates for the MASDP. It is important to note that, while such models are useful in providing consistent infrastructure configurations and cost estimates, and for rapidly assessing how different choices in future scenarios impact the entire system configuration, they are only one part of the broader activity of developing an actionable plan for an industrial hub. 
By necessity, the models employed contain limited detail on individual proponents and take a high-level, Ôperfect coordinationÕ view, whereas the reality of individual aspects may be quite different due to their individual costs, access to land, competitive advantage or other aspects relevant to them. The model results are intended to provide context to guide decisions around how various choices or requirements impact the multi-use system as a whole. Such models, when used to provide quantitatively consistent scenarios and variations, can help in understanding the larger set of activities that make up the industrial hub design and implementation phases.
2.3.1 Model features
In general, the least-cost optimisation model chooses between power generation from gas (specifically CCGT generation), large-scale solar PV and (when enabled) onshore wind. The capacity limits for each of these technologies per region are shown in Figure 4Ð where they limit gas generation to Middle Arm and renewable generation to the four REZs. All of the REZs, except Powell Creek, also have solar PV and wind generation capacity limits. There is a single demand centre in Middle Arm, meaning that the imposed demand for electricity and/or hydrogen must be met at that location by building a transmission line to connect a REZ to Middle Arm, and/or a hydrogen pipeline if the model chooses to produce hydrogen in the REZ (as opposed to locating an electrolyser in Middle Arm and relying on transmission). 
Given any combination of flat industrial electrical load in Middle Arm and an annual quota for electrolyser-produced (here alkaline electrolysers1) hydrogen, the model optimises for the least total system cost configuration of generation capacity, transmission, battery storage, electrolyser capacity, hydrogen pipeline and hydrogen storage (tank based). A variable electrical load can also be specified without being associated with hydrogen production. The optimisation accounts for the intraday variation in (variable) demand and renewable resources by solving at the hourly (or two-hourly2) temporal resolution, and the process can allocate variable demand throughout any time period across the time horizon (one calendar year Ð here 2018, see Section 2.3.4). 
The spatial dependence on renewables is incorporated through the Renewables.ninja platform (Pfenninger and Staffell, 2021). Total system cost is the annualised cost of the entire system covering the capital, operating and maintenance (O&M) costs amortised over the lifetime of the assets according to a fixed discount rate of 7%. The mathematical description of the model is detailed in Appendix A.4. ACoE and ACoH are calculated and can be thought of as the whole-of-system optimised equivalent to the levelised cost more commonly calculated for a single facility.
2.3.2 Example model calculation
To further describe the model, in this section an example solution output is examined. For this example, a 10% limit was imposed on the fraction of total electrical energy generation that can be met by gas generation3 over the year, such that gas fills a peaking role only. The demand at the Middle Arm region is approximately 50% flat electrical load and 50% variable load implied by a 322 tpd electrolysis-produced hydrogen production quota (with the quota applied per year). 
For this example, the capacity factor of any electrolysers was fixed at 70%, as was the HVDC link to Powell Creek (90% or higher, the model can also solve for these). Other choices made for this particular example are the use of the ÒCurrent PoliciesÓ GenCost technology cost scenario, 2028 as the technology cost year, $12.1/GJ as the price of gas fuel, the lower HVDC costing from Figure 7, a 5.5% transmission line loss coefficient and a 2018 weather year. While single-number results are presented in this example, the results section presents ranges of results to capture the range of model input choices.
Geographic representation of the optimal (least-total-cost) solution, the built capacities at each location, a one-week extract of the electrical load demand and supply breakdown indicating where variability and storage are used, and the ACoE regional breakdown, are shown in Figure 4, Figure 5 and Figure 6, respectively.

Figure 4 Example model solution showing the generation, storage and transmission mix 
As indicated by the legend, the transmission lines and hydrogen pipeline are indicated by the blue and magenta connecting lines, respectively. In this case the 1200 MW of HVDC transmission line is built (at the lower 5,500 MW costing and at 90% utilisation) to connect Powell Creek with Middle Arm, and the remaining REZs are connected by approximately 200 MW each (at the 500 MW HVAC costing). Koolpinyah is additionally connected by 1.5 T/hr (at the 20 T/hr costing) hydrogen pipeline. The green circles indicate the size (8,560 ha) of the 3.4 GW solar farm at Powell Creek, and smaller (1,000 ha or less) installations at the closer REZs. The electrolysers are indicated by yellow circles in Middle Arm (close to 1 GW) and Koolpinyah (86 MW), with wind being represented by the blue circles, and battery storage of 8.6 GWh by the aqua circle in Powell Creek only. The wind capacity in Powell Creek and Koolpinyah, as well as the solar PV in Blackmore, Koolpinyah and Weddell, are all at their maximum allowable capacities.

For this example, most of the generation is provided by 3.4 GW of solar PV and 1 GW of wind, firmed with 8.6 GWh of 4-hour battery storage in the Powell Creek REZ (provided via 1.2 GW of HVDC transmission at 90% utilisation). Gas generation in Middle Arm of 235 MW represents the maximum of what the model was allowed to build. While wind generation is cheaper than battery storage, the model only builds the maximum allowable wind generation in Koolpinyah and Powell Creek due to the 90% constraint on the HVDC transmission utilisation Ð which results in larger battery storage in Powell Creek (and less wind at the two smaller REZs) than would have been built without that constraint. The model also chooses to place an 86 MW electrolyser in the Koolpinyah REZ since it had to build wind generation at this location to minimise firming costs. Based on this optimisation, it is more cost-effective to meet hydrogen demand by co-locating hydrogen production with generation and piping it to the demand centre. The model therefore builds an additional 400 MW of solar PV at this location and builds a hydrogen pipeline back to Middle Arm.
Since the only regions where electrolysers are permitted are Middle Arm and Koolpinyah, the model does not implement the same solution for the Blackmore and Weddell REZs, where it only builds generation and the associated HVAC transmission line. The model does not use hydrogen tank storage as there is no schedule for the hydrogen demand at Middle Arm outside of the annual quota. The spilled energy fraction (i.e. the amount of excess energy produced during the year not needed to meet the imposed demand from the flat and variable loads) is 18%.

Table 6: Capacities selected by the model and resulting ACoE for the example case
Region
Gas 
[MW]
Solar PV [MW]
Wind 
[MW]
Battery [MWh]
Electrolyser [MW]
H2 storage 
[T]
ACoE 
[$/MWh]
Middle Arm
235
0
0
0
968
0
148
Powell Creek
0
3,424
1,000
8,567
0
0
159
Blackmore
0
300
0
0
0
0
86
Koolpinyah
0
400
200
0
86
0
121
Weddell
0
200
148
0
0
0
125

For this example, the model is run with a 2-hour temporal resolution. Figure 5 shows the electrical power demand and supply mix, every 2 hours, for one week of the calendar year. The flat electrical demand (light blue), plus the variable electrical demand from electrolysis (darker blue), and the charging of battery storage (magenta), are shown with the peaks following the solar resource, as the model places the variable aspect of demand when the electricity is lowest cost. The supply mix is indicated with components from gas, battery discharge and variable renewables (solar PV and wind, VRE). The areas of the supply curves not filled with block colour demand show the spilled energy fraction.


Figure 5: Example of electrical power demand and supply breakdown over a one-week period 
Note that this is aggregated across all regions, and as such the spillage (generation above demand) or the need for gas peaking may be hidden in that aggregation. The variable demand due to electrolysis (darker blue) is placed mostly during the peak solar hours, with some being met by battery discharge or gas peaking. 

The model also calculates the ACoE supplied by each REZ. Figure 6 shows that the region with solar PV generation only (Blackmore) supplies at a lower cost ($86/MWh) than the regions with both solar PV and wind (e.g. Weddell with close to 50% wind at $125/MWh). The lowest cost (unfirmed) is from the Powell Creek REZ, but with the included storage to provide 90% firmed supply (due to the capacity factor constraint on the HVDC transmission). Much of the cost is made up by battery storage, and a smaller proportion by transmission Ð again, due to the high utilisation of transmission model constraint. The summary results show Middle Arm ACoE costs of $148 MWh. The model also indicates the ACoH supplied to Middle Arm at $9.6/kg. Section 3.2 examines which factors impact that cost. 



Figure 6: ACoE and total cost breakdown for the example above 
For the REZs, the ACoE bar chart shows the breakdown of the cost due to VRE (O&M and capital) and that due to transmission and storage. As expected, Powell Creek exhibits a large fraction of storage costs. The Middle Arm bar breakdown is representative of the fraction of energy imported from the REZs (light green) and that generated by gas peaking in Middle Arm (the red and darker red for gas fuel and expenses, respectively). The contribution to each regionÕs ACoE from energy that is spilled (i.e. unused) is also indicated. The annualised system cost shows the contribution to the total annualised system cost (the value that is minimised), which is dominated by capital costs (including 7% interest) for generation and battery storage.
2.3.3 Costs and cost uncertainty
Technology costs and technology cost year uncertainty
Technology capital, O&M and fuel (gas for power generation) costs come from the CSIRO GenCost 2023?24 Report (Graham et al., 2024), and the transmission and hydrogen pipeline costings from the Australian Pipelines and Gas Association (APGA, 2021). The GenCost report includes several scenarios for technology cost evolution (Current Policies, Net Zero by 2050 and Net Zero Post-2050). However, since this report only considers initial greenfield build, the technology cost years are limited to 2025 and 2028. In those years it is mostly the capital cost of alkaline electrolysers for hydrogen production that vary significantly between GenCost scenarios, as shown in Figure 7. As such, only the sensitivity to GenCost alkaline electrolyser costs area is addressed in the section 3.3. 
The costs are further regionalised, in a simplification of the approach of the Australian Energy Market Operator (AEMO, 2023), where all costs except fuel (gas) are regionalised by applying a regional markup factor. A factor of 1.35 is applied, which is consistent with the high end of AEMOÕs locational cost factors listed in its 2023 Inputs, Assumptions and Scenarios Report (IASR) Workbook (Australian Energy Marketing Operator, 2023). The rationale for this regional markup factor is based on the NTÕs remote location, relatively small industrial and workforce capacity, and operator experience.
The model results in Section 3 are mostly presented in terms of the ACoE or ACoH and are shown as a range of costs. One dimension defines the range over the 2025?2028 technology cost year; the other dimension of that range is the wholesale price of gas for generation fuel, discussed below.

Figure 7: Comparison of the GenCost scenarios for forecast capital cost 
For each technology (colour) there are curves for the ÒCurrent PoliciesÓ (CP, solid lines), ÒNet Zero by 2050Ó (dashed lines) and ÒNet Zero Post-2050Ó (dotted lines) GenCost scenarios (Graham et al., 2024). The only difference between the scenarios of relevance to this work (where only 2025 and 2028 values are used) is the 2028 values for alkaline electrolysis (H2 [AE], green), where the cost is significantly lower for Net Zero by 2050.
Fuel cost uncertainty
The model uses a wholesale gas (for generation fuel) price forecast, which is strongly linked to the international gas market and has proven to be highly variable and uncertain. Similar to the technology cost year above, uncertainty in the price of gas is therefore incorporated in all results by running the model over a range of gas prices. Figure 8 shows historical gas prices, as well as several forecasts out to 2050. The base case choice is the value from the IEAÕs 2022 World Energy Outlook (IEA, 2022) and a 1.5x sensitivity case (these are both at 2025 and 2028, so in total four case prices of A$10.4/GJ in 2025 and $8/GJ in 2028 for the base price, and $15.5/GJ in 2025 and $12.1/GJ in 2028 are used for the 1.5x higher priced sensitivity). The values sit within the spread of forecasted prices shown in Figure 8. 
For the results in Section 3 where minimum/maximum ranges of values are presented, those minima and maxima are taken from the results of simulations over the four (two x two) combinations of gas price and technology cost year. Note also that gas generation is not solely for the purpose of LNG production (as is the case for two of the existing facilities in Middle Arm), where the cost of gas for fuel may be considered close to zero as the process chain for that industry includes burning of excess supply that is not able to be used. For any new gas generation the model considers, it would be supplying power to a range of industries beyond those that extract the gas (itself), such that a zero price of gas is not appropriate.

Figure 8: Historical and base forecast wholesale price of natural gas
The base case used here is the IEA 2022 World Energy Outlook (thick, black) (IEA, 2022) and 1.5x sensitivity (thick, blue) forecast traces. Historical data source is (Australian Energy Regulator, 2024), low and high values (thick, orange) are from GenCost 23?24 (Graham et al., 2024), the 2023 IASR Step Change scenario for Darwin and Sydney (Australian Energy Marketing Operator, 2023), and the low and high cases are from the Western Australian 2022 Whole of System Plan (thick, green) (Western Australian Government, 2022).

2.3.4 Model limitations and caveats
Temporal resolution and weather year
For the results presented here, the model was run for the calendar year of 2018. This year was chosen as it represents the closest to an average year in terms of renewable (solar irradiance and wind speed) resources over the span of years from 2010 to 2021 (see Appendix A.1). A 2-hour temporal resolution is used after benchmarking with results at a 1-hour resolution indicated that a 2-hour resolution was sufficient. Over the range of years there are significant variations in the renewable resources (e.g. 2011 exhibited significantly lower resources, and 2019 higher) such that if either of those extreme years is chosen, the resulting energy infrastructure is impacted. The choice of the single, close-to-average weather year of 2018 was to balance accuracy and model runtime.
Generation technologies
For all generation technologies, the fixed and variable O&M costs do not vary with the year (in contrast to the capital costs, due to the GenCost learning approach).
Onshore wind
For onshore wind (offshore wind is not included), a single height (150 m) and turbine size (4.5 MW) is used. As discussed in Section 2.1.2, the location-specific capacity factors for wind generation are based on a crude 0.7 local correction factor to the reanalysis data available on the Renewables.ninja platform. The uncertainty in the wind resource and associated capacity factors are discussed further in Appendix A.2. 
The inclusion of wind is calculated at the whole-of-system level, so while the model typically chooses to build wind generation as a partial firming solution before resorting to the more expensive battery storage, it does not consider the economic viability of any particular wind farm (i.e. it looks solely at the total system cost as if it were being funded by a single entity).
Gas-based generation
For gas-based generation, only CCGT gas generation is available to the model, which does not include CCS cost. To approximate emissions constraints, the model imposes a constraint on the fraction of generation from CCGT without CCS. Future work should include adding unconstrained CCGT+CCS as an option to meet demand in the presence of a carbon price; the results of the wider study will provide a more detailed understanding of the costs of CCS that could be used in such a model. 
Hydrogen production
Since the model does not include water demand, supply or transport, water restrictions are manually imposed by allowing only electrolysers in Middle Arm and Koolpinyah where discussions with stakeholders indicated that an appropriate water supply was available.
The model does not explicitly distinguish a ÔtypeÕ of hydrogen generation; that is, even when produced by electrolysis (except for when the gas generation fraction is constrained to be zero) the power is always supplied from a mixture of gas and renewable sources, and there were no constraints implemented, specifying which electricity generators can supply the electrolysers (although when hydrogen is produced in the Koolpinyah REZ, only renewable electricity is available).
An alkaline electrolyser efficiency of 55.014 MWh/t is applied, and this is not varied (unless otherwise stated) with year or power. Unless the electrolyser capacity factor is fixed, the model solves for it endogenously down to an applied 20% minimum and the model does not include the potential for ancillary grid services as a revenue stream for electrolyser variable demand to help offset the cost of electricity.
The model does not include the cost to store hydrogen after it has been produced in, or piped to, Middle Arm Ð since it is unclear on what schedule the hydrogen would be exported, transported or consumed. However, the model does have hydrogen tank storage as an option to effectively firm hydrogen production energy supply, although it was never implemented by the model due to there being no prescribed rate at which the hydrogen must be delivered. 
ACoE
While ACoE and ACoH are used as some of the primary reporting variables, the model optimises for minimum annualised total system cost, not for the cost of supply of electricity or hydrogen. ACoE and ACoH are average system-level cost, not market price, which would be a marginal price (likely higher than average cost, but it depends on the cost of hydrogen through the capacity factor of the electrolyser chosen by the model [cf Section 3.3 on the trade-off between ACoE and ACoH via electrolyser capacity factor]). The ACoE is simply calculated by dividing the total electrical energy supplied (or demanded to account for spillage) over the model time horizon from a region and dividing by the total annualised cost required to supply it. The ACoE can be thought of as a system-wide levelised cost of supply.
While capacity factor and spill fraction are calculated by the model, there are no minimum constraints to ensure the economic viability of any particular wind or solar farm (i.e. the model might choose to build wind to benefit the overall system cost, even though the capacity factor is lower than what might be considered economically viable by an individual developer). In such cases, the model results are perhaps more usefully interpreted as being a basis from which to identify which aspects of the system might have the largest gap between standalone economic viability and their role in the least-cost system. 
Transmission
Transmission cost is not a function of time (in contrast to the generation capital technologies), the lengths/distances are rough estimates based on road map distance, and the cost is on a MW per km basis. The results use a continuous transmission capacity variable, rather than discrete instances of a fixed-line capacity. 
Connection to existing infrastructure
The model does not include the DKIS, only the new MASDP and possible off-islanding of existing gas-fired generation in Middle Arm.
2.3.5 Possible future model extensions
As the concept for the MASDP energy infrastructure evolves, it would be advantageous to incorporate a more realistic model for the Safeguard Mechanism, such as CCGT+CCS, offsets and a price of carbon to allow for a staged approach to emissions reduction ? for example, allowing unconstrained thermal generation in 2025, then applying tightening constraints over time to mimic the effects of the Safeguard Mechanism to calculate the least-cost modification of the previously built infrastructure to meet the new constraint. 
3 Model results
When examining the business case for decarbonisation of existing, or the design of greenfield, low-emissions industrial precincts like that proposed within the MASDP, it is important to develop an understanding of how the choice of energy infrastructure Ð especially those common user aspects that are costly to upgrade Ð impacts the expected range of the cost of electricity to industrial customers. For precincts that aim to produce hydrogen via electrolysis, it is also important to know how large variable load can be most cost-effectively integrated. Perhaps most important is how that cost range varies with carbon emission intensity, especially as the industries within the precinct will fall under the Safeguard Mechanism or may be specifically producing commodities that require low- or zero-carbon emission electricity generation. 
The following sections investigate how ACoE (see callout below on how to interpret this metric as it is not a price forecast) and ACoH vary with the generation and transmission technology mix (and electrolyser sizing) that the model chooses to build under various costings and constraints. 
This section has three parts that aim to address the key questions identified in Section 2.2:
* Section 3.1 provides estimates of ACoE and the associated mix of generation technologies, locations and connecting transmission, without any constraint on gas generation, to meet a range of load mixes from purely flat to purely variable 
* Section 3.2 examines how ACoE and the generation technology mix vary as gas generation is constrained for various flat/variable load mixes 
* Section 3.3 examines the implications for electrolysis-based hydrogen production (and ACoH), and the impact that hydrogen production has on system design and ACoE.
How to interpret ACoE 
ACoE is not a price forecast. It is only representative of the annualised total costs over the model horizon (1 year, including the time value of money out to the lifetime of the built technologies) to supply the prescribed demand over the same time period, either within a specific region (e.g. renewables powering electrolysis in the Koolpinyah REZ) or supplying one region from another, which includes the cost of transmission (e.g. supplying Middle Arm from the Powell Creek REZ). 
While these costs may be close to those available in long-term power purchase agreements, electricity prices on the spot market are, in contrast, typically set in each time period by the marginal cost Ð that is, the cost to supply the last unit of energy required to meet the demand at that particular time Ð which will be the most expensive. Thus, the average marginal market price is typically higher than the ACoE. 
Additionally, individual suppliers may choose different strategies for pricing their energy supply. Furthermore, the model has perfect foresight, and builds the precise amount of capacity required over the model horizon, something that is not possible in the real world, which translates to an unknown that typically contributes to a higher cost of supply than the model calculates. 
It is also noted that any overbuild/spilled/curtailed generation capacity contributes towards the MWh included in the $/MWh calculation (i.e. there can be significant generated energy that acts to increase the $/MWh values, which could potentially be avoided if other uses for that energy could be found). The ACoE can be thought of as a system-level levelised cost. 
3.1 Generation technology mix and ACoE (unconstrained thermal)
This section investigates which mix of generation technologies and their siting is the most cost-effective for flat, variable and mixed (flat and variable) demand profiles with no constraints on the fraction of electrical power demand that can be met by gas (CCGT) generation. 
In this work the variable load is represented in two ways. The first (and the approach for this section) is a separate load associated with some unknown, variable demand process that does not have associated capital and operating costs for the process itself (i.e. in the same way the model does not account for the capital and operating expenses of the processes that make up the constant load demand). This contrasts with the second approach, which is for electrolysis-based hydrogen production. In that case, there is the potential for great variability for a large fraction of the load, the variability of which is dependent on the scale of the capital investment (the size/capacity factor) in the electrolyser, such that the capital and operating costs of the electrolysis are incorporated into the total system cost. In this way, the model can size the electrolyser to best match the rest of the system. This second approach is employed in Section 3.3 as it reveals how hydrogen demand is linked to electrical demand.
Key results for unconstrained thermal generation
* With no constraint on gas generation, the model does not build battery storage or onshore wind, even at the higher end of gas prices. 
* The 90% utilisation constraint on the HVDC transmission from Powell Creek to Middle Arm, which requires battery storage, means that without a gas generation constraint, the model does not use the Powell Creek REZ.
For a constant load
* Gas price is the driver of whether solar PV is built. Even though the ACoE from solar PV (with transmission) is significantly lower (approximately $91/MWh compared with $106/MWh for gas at $10.4/GJ), the model does not choose solar until the savings offset the capital investment in the panels Ð this happens at the higher $15.5/GJ and $12.1/GJ gas prices.
* For those higher gas price cases, the model chooses to fully use the solar PV capacity at the three closer REZs. 
* HVAC transmission costs represent between 3% and 7% for those closer REZs and are used at 29% capacity factor.
For 50/50 mix and purely variable loads
* At all gas fuel prices considered here and for both 2025 and 2028 technology costs, 14% of the generation is supplied by solar PV in the REZs, with the cost for gas generation being $88/MWh for $8/GJ in 2028 and $81/MWh for solar (slightly higher at $84/MWh in Koolpinyah).
* The capacity limits at the three closer REZs mean that VRE is only able to meet 14% of total demand, so even when the variable fraction of load is higher, there is no capacity to meet that demand with VRE (without using Powell Creek Ð which it does not do due to the transmission utilisation constraint).
3.1.1 Detailed results
The MASDP Balanced Scenario has an approximately equal mix of flat and variable load types due to the presence of electrolysis-based hydrogen production. Typically, it would be expected that generation mix and ACoE vary strongly with load type, and so in this section the model is applied to a purely constant load, a 50/50 flat/variable load mix and a purely variable load. To remove the complication of electrolyser and hydrogen infrastructure, the variable load is not associated with any particular process (in contrast to Section 3.3, where electrolysis is incorporated). The minimum/maximum range of values comes from the results from four separate optimisations, one for each of the four combinations of fuel price and technology year described in Section 2.3.3.





Table 7: Range of ACoE and % VRE generation for different load types and unconstrained gas

ACoE ($/Mwh)
% VRE generation

Min
Max
Min
Max
Constant load
89
142
0
14
50/50 load mix
88
138
14
14
Variable load
88
138
14
14
Load types are purely flat, 50/50 mix of flat and variable, and purely variable electrical. The minimum and maximum values are over the range of results from the four combinations of technology cost year and wholesale gas price.

The results in Table 7 show the ACoE and percentage of generation from VRE (in this case, it is only solar PV) for unconstrained gas generation. These values indicate that at the lower gas prices ($10.4/GJ and 2025 technology costs, and $8/GJ and 2028 technology costs), large-scale solar PV is not cost-effective, while for higher gas prices the model uses all available solar PV in the closer REZs. 
For the solutions where solar is not built, it is still lower cost for that part of the day it can supply, and when the savings realised by building solar are larger than the capital expenses to build the panels (this occurs at the higher gas prices), solar is built to supply some part of the constant load. However, the REZ penetration is limited to 14%, even for purely variable loads. This is a result of the combination of the solar PV capacity limit constraints and the 90% utilisation constraint on the transmission link to the Powell Creek REZ (which does not have a constraint on solar PV capacity). 
The only option to meet more of the variable load with VRE is to build wind generation at those REZs closer to Darwin. However, this is not chosen as it is more expensive than gas generation, even at the higher gas prices. The other alternative is to build the HVDC transmission link to Powell Creek and the corresponding solar PV. However, the 90% utilisation constraint on the HVDC link requires battery storage to meet this utilisation rate, which results in the cost of supply of VRE from Powell Creek exceeding that of gas generation in Middle Arm, so the model does not choose that option. 
It is noted that without the constraint on HVDC utilisation, the model does choose Powell Creek, and chooses it preferentially over the closer REZs even with the cost of transmission, but the HVDC link utilisation is approximately only 30%. Figure 9 shows the model solution for the constant load, 2025 technology costs and $10.4/GJ gas price; and Figure 10 shows the model solution for the 50/50 and purely variable loads (the solutions are the same) for 2028 technology costs and $8/GJ gas price.
The 50/50 load mix solution (and purely variable solution, as they are the same) in Figure 10 shows that the three closest REZs are used to their capacity limits (for solar PV, Blackmore ² 300 MW, Weddell ² 200 MW and Koolpinyah ² 400 MW), with the Powell Creek REZ not being used. None of the solutions in this section uses battery storage or onshore wind (even though they are available to the model) due to gas providing the lowest cost firming solution.



Figure 9: Constant load solution for unconstrained gas generation 
For 2025 technology costs and $10.4/GJ gas price. In the model summary, NaN (Not a Number) means that the technology for that metric was not built, so the metric cannot be calculated. For the regional ACoE breakdown, Gas Ex encompasses capital (CapEx), and operating and maintenance costs (OpEx).



Figure 10: 50/50 flat/variable load mix for unconstrained gas generation 
For 2028 technology costs and $8/GJ gas price. As with the example in Figure 6, for the Middle Arm region where electricity supply can be imported, the ACoE breakdown represents the volume-weighted fraction of the total cost of supply to Middle Arm; that is, for the above figure, 14% of the supply to Middle Arm is imported (at $81/MWh from Blackmore and Weddell, and $84/MWh from Koolpinyah) and the remainder is provided by gas generation in Middle Arm, where almost two-thirds of the cost is on fuel.



3.2 Generation technology mix and ACoE (constrained thermal)
In this section, ACoE is estimated over a range of initial infrastructure emissions targets (for the new greenfield infrastructure only) by further constraining the approach in section 3.1 to include a maximum limit on the fraction of annual electrical energy production that can be met by gas generation. Since the Safeguard Mechanism (DCCEEW, 2024) requirement is a progressive reduction in allowable net emissions over time relative to a baseline, an understanding of how strongly the ACoE might vary with its associated emissions production will be important as each industry considers when and how to decarbonise. As shown in the section 3.1, that cost depends strongly on how variable the demand load is, the cost of gas for power generation and whether wind generation is included. This section shows how the addition of a limit on the fraction of annual electricity generation that can be met by gas impacts the generation technology mix and ACoE. 
Key results
* For a constant load, the ACoE rises sharply (about 1.8x) when increasing from 90% renewables to 100%, driven by the cost of battery storage.
* There is little difference between the 50/50 load mix and a purely variable load due to the 90% utilisation constraint on the HVDC transmission link to Powell Creek. 
* The addition of onshore wind reduces ACoE by approximately 15% when a constraint on gas generation is present.
* For constant loads, the spilled VRE is large (50%) for the 100% renewable fraction, and for the 90% renewable fraction the gas generation capacity factor is low (25%).
3.2.1 Detailed results
The dependence on ACoE and generation mix as a function of the allowable fraction of annual electrical energy production by thermal (CCGT gas) generation is tested using the following approach: the same three load types as section 3.1 (i.e. purely flat, 50/50 [flat/variable] and purely variable loads); four constraints on the fraction of annual generation that can come from gas (i.e. unconstrained [100%], 50%, 10% and 0% [VRE/renewables only]); and the same four combinations of technology year and wholesale gas price as above. Note that these constraints are upper limits such that even in the unconstrained case, renewables can still be built should they be the least-cost option. Table 8 shows the ACoE results for this range of parameters.
The first notable feature of Table 8 is the sharp increase in ACoE from $164/MWh to $292/MWh between the 90% renewable fraction and 100% for constant loads. The increase is almost 1.8x. To examine this further, model solution figures for two of these cases are included: for a constant load, with wind generation, 2028 technology costs and $8/GJ gas fuel price with a 10% limit on gas (Figure 11 Ð 90% renewable fraction) and another for a 0% limit on gas (Figure 12 Ð 100% renewable fraction). Table 9 and Table 10 show the built capacities per region. The increase in ACoE is largely driven by the almost 3x increase in battery storage (from around 10 GWh to almost 30 GWh) and a 2x increase in spilled energy from VRE (from around 25% to 55%). This last point is reflected further in the per-region ACoE values in Table 8 and Table 10. These are $164/MWh for the 90% renewable fraction and $292/MWh for 100%. The large overbuild of VRE is representative of the relative costs of battery storage and VRE generation capacity (i.e. the model selects overbuilding the VRE capacity to minimise battery storage).
Table 8: ACoE versus load type and VRE fraction 

$/MWh

VRE only
Gas ² 10%
Gas ² 50%
Gas ² 100%
Load type
Min
Max
Min
Max
Min
Max
Min
Max
Constant load, no wind
334
385
191
227
127
171
89
142
50/50 load, no wind
184
213
166
198
121
163
88
138
Variable load, no wind
178
207
166
198
121
163
88
138
Constant load, with wind
292
339
164
195
113
153
89
142
50/50 load, with wind
153
177
141
169
103
142
88
138
Variable load, with wind
153
177
141
169
103
142
88
138
Minimum and maximum are over the range of results for the four combinations of technology cost year and wholesale gas price considered in Section 3.1, and with the lower HVDC costing. The coloured shading highlights the significant variations. The values in bold correspond to the solutions shown in Figure 11 and Figure 12, respectively.
Table 9: Selected capacities for constant load, 90% renewable fraction
Region
Gas 
[MW]
Solar PV [MW]
Wind 
[MW]
Battery [MWh]
Electrolyser [MW]
H2 storage [T]
ACoE 
[$/MWh]
Middle Arm
782
0
0
0
0
0
165
Powell Creek
0
4,261
1,000
10,866
0
0
163
Blackmore
0
0
0
0
0
0
-
Koolpinyah
0
153
126
0
0
0
164
Weddell
0
200
200
0
0
0
146
These capacities correspond to the solution shown in Figure 11.



Figure 11: Solution for constant load, 90% renewable fraction 
For 2028 technology costs, $8/GJ gas fuel price and 10% limit on gas-based generation. The HVDC transmission link to Powell Creek is 1.4 GW and is used at 90% capacity factor.

Table 8 also shows that the ACoE does not vary significantly between the 50/50 and purely variable load types for cases with the 90% renewable fraction or less. This is due to the 90% utilisation constraint on the HVDC transmission link to Powell Creek. The results also indicate an approximately 15% reduction in the ACoE when wind generation is included for cases where there is a constraint on gas-based generation. Figure 11 shows that the 782 MW of gas generation built in Middle Arm operates in somewhat of a peaking role at a capacity factor of 25%.



Figure 12: Model solution for constant load, 100% renewable fraction 
The HVDC transmission link to Powell Creek is 1.45 GW and is used at more than 98% capacity factor. Note the scale difference to the other figures.

Table 10: Selected capacities for constant load, 100% renewable fraction
Region
Gas 
[MW]
Solar PV [MW]
Wind 
[MW]
Battery [MWh]
Electrolyser [MW]
H2 storage [T]
ACoE 
[$/MWh]
Middle Arm
0
0
0
420
0
0
292
Powell Creek
0
9,073
1,000
25,284
0
0
289
Blackmore
0
300
200
546
0
0
343
Koolpinyah
0
400
200
901
0
0
284
Weddell
0
200
200
395
0
0
370
3.2.2 Uncertainty: Off-islanding existing Middle Arm gas generation
A possible configuration of the MASDP energy infrastructure may include off-islanding existing gas (LNG facility) generation in Middle Arm. To test the effect of this scenario, the model was configured with two additional regions representing the ILNG and DLNG facilities, each with its own required 2 km of HVAC 500 MW transmission to connect to Middle Arm. The capital expense for gas generation in these two regions was set to zero, but they still had fuel and operating expenses. 
The inclusion of these two regions only impacts the <100% renewables cases and each was initially allotted a capacity of 50 MW. However, even when allowing for 350 MW of generation from ILNG, that only represents about 20% of the load, and since the capital cost represents about 35% of the cost of gas generation, 35% of 20% is about 7%, which would be the maximum decrease in ACoE expected. The model solutions gave a 4.5% decrease, and this is only for cases that allow for more than 10% of the generation to be met by gas. Thus, considering the more plausible 50 MW off-islanding from both LNG facilities, the impact on the ACoE is not significant.
3.3 Implications for electrolysis-based hydrogen production
In section 3.2 it was shown that when the allowable generation fraction from gas is constrained and the load contains a variable component, variable loads result in lower ACoE. This poses the question of how best to size an electrolyser for hydrogen production ? that is, either a higher capital cost for a higher capacity electrolyser which is run at a lower capacity factor to use the lower cost electricity associated with a variable load, or a lower capital cost for a lower capacity electrolyser run at a higher capacity factor but with higher electricity costs associated with a constant load. This section examines this question to infer ACoH under various assumptions. 
An additional question examines the least-cost location of an electrolyser ? that is, co-locate the electrolyser with renewable generation (and perhaps storage) in a REZ and pipe the hydrogen back to Middle Arm for export, or rely on transmission to link the REZ with an electrolyser placed in Middle Arm. In this work, only the Koolpinyah REZ is allowed an electrolyser (in addition to one at the Middle Arm demand centre). The sensitivity of the ACoH to various model input parameters is also examined.
Key results
* At the 90% renewable fraction, the ACoH is almost $9.4/kg with a corresponding ACoE of $144/MWh.
* For a low (50%) renewable fraction, the optimal electrolyser capacity factor is nearly 90%, while for the high (90% and 100%) renewable fractions it is nearly 60%. The latter result is mostly driven by the 90% capacity factor constraint on the HVDC link to Powell Creek. 
* The ACoH is most sensitive to the AE efficiency (followed by the renewable fraction and wind capacity correction factor).
* The ACoE is most sensitive to the renewable fraction (followed by the wind capacity correction factor and the discount rate).
* A hydrogen pipeline from Koolpinyah to Middle Arm is chosen by the model as the least-cost option.
3.3.1 Detailed results
This section replaces the variable load from the previous sections, which was not connected to a particular industrial processÕs capital expenditure or capacity factor, with a variable load associated specifically with an annual quota for hydrogen production from electrolysis, where the model sees the cost to build and operate the electrolysers and transport the hydrogen if produced outside of Middle Arm. It does not include storage at Middle Arm for the yearÕs quota, but rather assumes that hydrogen is exported or consumed as it is delivered, and that (as detailed in Table 11) electrolysers are restricted to Middle Arm and the Koolpinyah REZ. 
The load mix is restricted to 50/50 flat to variable, and the 50%, 90% and 100% renewable fractions (via the constraint on maximum fraction of energy generation from gas) are examined. To address the question of electrolyser sizing, results for 30%, 80% and model-determined electrolyser capacity factors are included. The 30% and 80% values are imposed via a constraint, while the model-determined value is a result of least total system cost optimisation. Table 3.5 shows the resulting values for ACoE and ACoH for this range in inputs for both wind and no-wind options.


Table 11: ACoH and ACoE overload type and electrolyser capacity factor
Electrolyser capacity factor
30%
Least total cost Capacity Factor (LTCCF)
80%
Wind generation
VRE fraction
[%]
Min LTCCF
[%]
ACoE ($/MWh)
ACoH ($/kg)
ACoE ($/MWh)
ACoH ($/kg)
ACoE ($/MWh)
ACoH ($/kg)



Min
Max
Min
Max
Min
Max
Min
Max
Min
Max
Min
Max
No
50
90
123
167
10.2
13.6
125
167
8.0
10.7
123
165
8.2
9.8
No
90
67
171
204
12.5
15.3
171
204
10.7
12.9
173
206
10.9
13.2
No
100
63
191
221
13.8
16.5
191
222
12.0
14.2
195
227
12.2
14.3
Yes
50
89
104
145
9.3
12.5
106
146
7.0
9.6
104
144
7.1
9.7
Yes
90
67
143
171
11.3
13.9
144
172
9.4
11.4
148
177
9.5
11.5
Yes
100
57
156
181
11.9
14.3
156
181
10.2
12.2
167
194
10.6
12.5
Note that there are additional optimisations not included in the model that may further drive down the ACoH, such as dynamic efficiency (Rezaei et al., 2024) and the 1.35 regional markup factor on all costs except gas fuel. For the 30% and 80% capacity factor columns, the model is constrained to those values, whereas for the least total cost capacity factor (LTCCF) the model is allowed to solve for its value and lists that optimal result in the ÒMin LTCCFÓ column. The row highlighted in purple (and bold values) represents the base case to which the sensitivities of Figure 14 are applied. The minimum and maximum values are determined over the 2 x 2 set of model runs over the technology cost year and gas-fuel price multiplier as in previous sections.
The results in Table 11 show that for high renewable fractions (90% and 100%), a lower cost of electricity supply is matched with a higher cost of hydrogen supply for low (30%) capacity factor electrolyser sizing, with the opposite being the case for high (80%) electrolyser sizing. Additionally, the least total cost capacity factor (LTCCF) values show that an optimal sizing closer to 60% results in the lowest combinations of both ACoE and ACoH. This trade-off in ACoH versus ACoE via the electrolyser capacity factor was also observed by (Bruce et al., 2018). The general observation that the inclusion of wind generation reduces costs by almost 15% continues here. The model solution for the LTCCF run with wind and a 90% renewable fraction is shown in Figure 13 and Table 12.


Figure 13: Optimal electrolyser sizing for 90% VRE fraction with wind 
Model solution for 2028 technology costs, $8/GJ gas fuel price and 50/50 variable load mix with the variability representing electrolyser power demand to meet an annual hydrogen production quota. Capacity values are shown in Table 12. The yellow circles indicate electrolyser sizing, and the magenta connecting line from Koolpinyah to Powell Creek indicates a hydrogen pipeline.
To further examine which model parameters ACoH and ACoE are most sensitive to, the change in ACoH and ACoE for lower and higher deviations from the base case was investigated (Figure 13). The parameters included in this sensitivity analysis were the efficiency of the AE-type electrolysers, the renewable fraction, the factor by which the wind generation capacity factor is adjusted, whether another use can be found for the spilled VRE such that it can be cost recovered, the discount rate, the GenCost scenario, the choice of weather year, the HVDC transmission capital cost and gas fuel price. The results of the scan and the values of the parameter variations are shown in Figure 14 and Table 13, respectively.
Table 12: Model selected capacities for 90% renewables with hydrogen 
Region
Gas 
[MW]
Solar PV [MW]
Wind 
[MW]
Battery [MWh]
Electrolyser [MW]
H2 storage [T]
ACoE 
[$/MWh]
Middle Arm
217
0
0
0
893
0
143
Powell Creek
0
3,507
1,000
8,733
0
0
159
Blackmore
0
300
0
0
0
0
82
Koolpinyah
0
400
187
0
255
0
114
Weddell
0
200
7
0
0
0
85
For this case the model also built 1.2 GW of HVDC transmission between Powell Creek and Middle Arm which is used at 90%. The hydrogen pipeline from Koolpinyah to Middle Arm has a capacity of 4.6 T/hr.
Figure 14 shows that the strongest drivers of ACoH are (in order) the electrolyser efficiency, the amount of spilled VRE that can be cost recovered, the renewable fraction, the assumptions on wind generation capacity factor and the discount rate. For ACoE, the drivers are the same except for the electrolyser efficiency, which has little impact on electricity cost. While these results show the sensitivity of a base case to each parameter variation independently, we also include an ÒoptimisticÓ version of this study in Appendix A.3 (Figure ), where for each subsequent sensitivity it is applied to the lower cost solution from the prior sensitivity to examine how low the ACoH and ACoE could get if all factors were chosen optimistically.




Figure 14: Tornado chart of ACoH and ACoE sensitivities
For a base case indicated by the ÒmidÓ column of Table 13 and for the technology costs year of 2028. While the sensitivity of a particular base case is applied to each parameter individually, stacking these sensitivities (e.g. applying the next sensitivity to the lower result of the previous sensitivity) is also possible to explore optimistic choices for all parameters. An example result is shown in Appendix A.3.

Table 13: Sensitivity of ACoH and ACoE to various parameters




ACoH [$/kg]
ACoE [$/MWh]
Variable
Low
Mid
High
Low
Mid
High
Low
Mid
High
AE efficiency
45 MWh/T
55 MWh/T
65 MWh/T
7.5
9.4
11.2
145
144
148
Recover spilled VRE
Yes
No
No
8.2
9.4
9.4
120
144
144
Renewable fraction
0.8
0.9
1
8.6
9.4
10.2
133
144
156
Wind cap. fac. multiplier
1
0.7
0.5
8.3
9.4
10
125
144
158
Discount rate
0.06
0.07
0.08
8.7
9.4
10.1
133
144
155
GenCost scenario
NZ by 2050
Current Policies
NZ post-2050
8.6
9.4
9.4
136
144
145
Weather year
Best (2019)
Average (2018)
Worst (2010)
8.8
9.4
9.7
132
144
152
HVDC capital cost
$220/km/MW
$439/km/MW
$878/km/MW
9.3
9.4
9.6
142
144
148
Gas price
$6.4/GJ
$8/GJ
$9.6/GJ
9.3
9.4
9.5
142
144
145
The base case for the results in Figure 14 and from which the sensitivities are calculated is represented by the ÒmidÓ column parameters of this table and are for a 2028 technology cost year.
4 Conclusions
This report provides a range of estimates for the ACoE and ACoH to meet the demand required by potential MASDP industries. These costs are the output of a least-cost optimisation model that seeks to minimise the annualised cost of the whole system including generation, transmission, hydrogen production and transport, and geographic location constrained by prospective REZs. 
The ACoE ranges from $88/MWh when no constraint is applied to the renewable fraction, to $141?169/MWh for a 50/50 mix of flat and variables loads and a 90% renewable fraction. The cost is most sensitive to (in order) how much of the spilled VRE can be cost recovered, the capacity factor of onshore wind (which when included can decrease the cost of electricity by approximately 15%) and the renewable fraction. Increasing the renewable fraction from 90% to 100% nearly doubles the cost of electricity, driven by the cost of battery storage. 
The ACoH at the 90% renewable fraction is almost $9.4/kg with a corresponding ACoE of $144/MWh and an optimal 67% electrolyser capacity factor. The cost of hydrogen supply is most sensitive to (in order) the assumed electrolyser efficiency, the ability to recover cost for spilled VRE, onshore wind capacity factor and the renewable fraction. The optimal electrolyser capacity factor is driven mostly by the 90% capacity factor constraint on the HVDC transmission link to the large REZ at Powell Creek. 
As outlined above, key needs for the future deployment of renewable electricity in the Northern Territory are the development of a greater understanding of the potential wind resources in REZs and the identification of low-cost energy storage technologies. This would lower the cost of implementing the 90% HVDC transmission capacity factor requirements and enable enhanced electrolyser capacity utilisation. 
As the concept for the MASDP energy infrastructure evolves, it would be advantageous to incorporate a more realistic model for the Safeguard Mechanism, such that CCGT+CCS, offsets and a price of carbon to allow for a staged approach to emissions reduction ? for example, allowing unconstrained thermal generation in 2025, then applying tightening constraints over time to mimic the Safeguard Mechanism to calculate the least-cost modification of the previously built infrastructure to meet the new constraint. 
References
APGA (2021) Final Report: Pipelines vs Powerlines Ð A Technoeconomic Analysis in the Australian Context. Australia Pipelines and Gas Association. <https://39713956.fs1.hubspotusercontent-na1.net/hubfs/39713956/pipelines_vs_powerlines_-_a_summary.pdf>.
ARENA Australian Government (2018) NT SETuP: A first look at the integration of PV and Diesel Power Stations in remote communities.
Australian Energy Marketing Operator (2023) 2023 Inputs, Assumptions and Scenarios Report (IASR) Workbook. <https://aemo.com.au/-/media/files/major-publications/isp/2023/2023-inputs-assumptions-and-scenarios-report.pdf?la=en>.
Australian Energy Regulator (2024) Gas Market Prices. <https://www.aer.gov.au/industry/registers/charts/gas-market-prices>.
Australian Government, Department of Industry, Science, Energy and Resources (2022) Australia's Nationally Determined Contribution: Communication 2022.
British Solar Renewables (2024) BSR Energy Developing AustraliaÕs Clean Energy Powerhouse. <https://britishrenewables.com/australia/bsr-energy>.
Bruce S, Temminghoff M, Hayward J, Schmidt E, Munnings C, Palfreyman D and Hartley P (2018) National hydrogen roadmap. Australia: CSIRO 92. 
Bureau of Meteorology and Australia Government (2020) Average daily solar exposure. <http://www.bom.gov.au/jsp/ncc/climate_averages/solar-exposure/index.jsp>.
CSIRO (2023) Opportunities for CO2 Utilisation in the Northern Territory. Australia.
CSIRO (2024) A business case for a low-emissions CCUS hub in the Northern Territory. <https://www.csiro.au/en/research/technology-space/energy/NT-low-emissions-hub>.
DCCEEW (2024) Safeguard Mechanism. Department of Climate Change, Energy, the Environment and Water.
EDL (2018) Pine Creek Power Station. <https://edlenergy.com/wp-content/uploads/2018/10/EDL-Fact-Sheet-Pine-Creek_Final.pdf>.
Graham P, Hayward J and Foster J (2024) GenCost 2023-24: Final Report.
IEA (2022) World Energy Outlook 2022. International Energy Agency. <https://www.iea.org/reports/world-energy-outlook-2022>.
IEA (2024) IEA Website. <https://www.iea.org/>.
IPCC (2023) Climate Change 2023: Synthesis Report. Intergovernmental Panel on Climate Change.
IPCC (2024) AR6 Synthesis Report: Climate Change 2023. Intergovernmental Panel on Climate Change. <https://www.ipcc.ch/report/ar6/syr/>.
Langworthy A, Bourne G, Frearson L, Howard K, McKenzie A and Peake O (2017) Northern Territory; Roadmap to Renewables; Fifty per cent by 2030.
Larrakia Energy (2022) Media Release: Larrakia and partners forging a renewable future. <https://www.larrakia.com.au/wp-content/uploads/2022/11/Larrakia-Energy-Media-Release-20221116.pdf>.
Lawler E and Worden K (2024) Media release: Keeping power prices low and securing our energy future. Northern Territory Government. <https://territorystories.nt.gov.au/10070/935595/0>.
Net Zero Australia (2024) About Net Zero Australia. <https://www.netzeroaustralia.net.au/>.
NTG (2020) Northern Territory Climate Change Response: Towards 2050. In: Envrionment and Natural Resources (ed.). Northern Territory Government.
NTG (2021) Darwin-Katherine Electricity System Plan: Cleaner, more affordable and secure energy system by 2030. Northern Territory Government. <https://territoryrenewableenergy.nt.gov.au/__data/assets/pdf_file/0011/1056782/darwin-katherine-electricity-system-plan-web.pdf?v=0.1.1>.
NTG (2023a) Middle Arm Precinct. Northern Territory Government. <https://middlearmprecinct.nt.gov.au/__data/assets/pdf_file/0007/1204747/middle-arm-precinct-industries.pdf >.
NTG (2023b) NT Infrastructure Plan and Pipeline 2023. Northern Territory Government. <https://infrastructure.nt.gov.au/__data/assets/pdf_file/0010/1341289/nt-infrastructure-plan-pipeline-2023.PDF>.
NTG (2024) The Middle Arm Sustainable Development Precinct: The Precinct. Northern Territory Government. <https://middlearmprecinct.nt.gov.au/about-the-precinct>.
Pfenninger S and Staffell I (2016) Long-term patterns of European PV output using 30 years of validated hourly reanalysis and satellite data. Energy 114, 1251Ð1265. 
Pfenninger S and Staffell I (2021) Renewables.ninja. <https://www.renewables.ninja/about>.
PWC (2023) Transmission and Distribution Annual Planning Report 2023. <https://www.powerwater.com.au/__data/assets/pdf_file/0020/216425/2023-Transmission-and-Distribution-Annual-Planning-Report.pdf>.
Rezaei M, Akimov A and Gray EM (2024) Economics of renewable hydrogen production using wind and solar energy: A case study for Queensland, Australia. Journal of Cleaner Production 435, 140476. 
Rogers JL, Gee R, Ross A, Ironside M and Squiers I (2024) Northern Territory Low Emissions Carbon Capture Storage and Utilisation Hub. Northern Territory Economy, Industries and Emissions Ð Task 1 Report. CSIRO, Australia.
Ross A, Ironside M and Gee R (2023) The Northern Territory low-emissions carbon capture, utilisation and storage hub development Ð the collaborative business case development. The APPEA Journal 63. DOI: https://doi.org/10.1071/AJ22210.
Ross A, Stewart M, Richardson C and Clifford A (2022) Collaborative development of the Northern Territory low-emissions carbon capture, utilisation and storage hub Ð a blueprint for the rapid decarbonisation of Northern Australia. The APPEA Journal 62. DOI: https://doi.org/10.1071/AJ21185.
Staffell I and Pfenninger S (2016) Using bias-corrected reanalysis to simulate current and future wind power output. Energy 114, 1224Ð1239. 
SunCable (2024) SunCable Projects. <https://www.suncable.energy/our-projects>.
Territory UCN (2018) Power System Review 2016?17. Darwin. <https://territorystories.nt.gov.au/10070/363579>.
TGen (2024) Our Power Stations. Territory Generation. <https://territorygeneration.com.au/about-us/our-power-stations/>.
Western Australian Government (2022) Whole of System Plan. <https://www.wa.gov.au/government/document-collections/whole-of-system-plan>.

Appendices
A.1. REZ capacity factor estimatesÊ
Table A.14 REZ renewable resource related capacity factors
TechnologyÊ
RegionÊ
2010Ê
2011Ê
2012Ê
2013Ê
2014Ê
2015Ê
2016Ê
2017Ê
2018Ê
2019Ê
2020Ê
2021Ê
AVGÊ
Solar PVÊ
Middle ArmÊ
22.5Ê
22.9Ê
23.6Ê
22.7Ê
24.0Ê
24.1Ê
23.6Ê
22.7Ê
23.4Ê
24.2Ê
23.6Ê
22.1Ê
23.3Ê
Solar PVÊ
Powell CreekÊ
26.4Ê
26.6Ê
28.5Ê
28.0Ê
27.8Ê
28.2Ê
27.7Ê
28.6Ê
29.0Ê
30.5Ê
28.4Ê
28.5Ê
28.2Ê
Solar PVÊ
BlackmoreÊ
22.6Ê
22.9Ê
23.6Ê
22.7Ê
24.0Ê
24.1Ê
23.5Ê
22.7Ê
23.4Ê
24.1Ê
23.6Ê
22.1Ê
23.3Ê
Solar PVÊ
KoolpinyahÊ
22.5Ê
23.0Ê
23.8Ê
22.8Ê
24.1Ê
24.2Ê
23.7Ê
22.8Ê
23.5Ê
24.2Ê
23.7Ê
22.2Ê
23.4Ê
Solar PVÊ
WeddellÊ
22.6Ê
23.0Ê
23.7Ê
22.7Ê
24.1Ê
24.2Ê
23.6Ê
22.8Ê
23.5Ê
24.2Ê
23.6Ê
22.2Ê
23.3Ê
Wind (x0.7)Ê
Middle ArmÊ
22.3Ê
23.6Ê
17.5Ê
18.7Ê
18.3Ê
21.4Ê
14.6Ê
19.6Ê
18.6Ê
19.9Ê
20.4Ê
18.0Ê
19.4Ê
Wind (x0.7)Ê
Powell CreekÊ
35.1Ê
34.2Ê
33.9Ê
32.6Ê
32.6Ê
36.6Ê
31.2Ê
34.2Ê
35.4Ê
38.8Ê
34.2Ê
33.2Ê
34.3Ê
Wind (x0.7)Ê
BlackmoreÊ
22.3Ê
23.9Ê
17.8Ê
18.8Ê
18.5Ê
21.5Ê
14.6Ê
19.5Ê
18.6Ê
20.0Ê
20.5Ê
18.1Ê
19.5Ê
Wind (x0.7)Ê
KoolpinyahÊ
23.1Ê
23.8Ê
17.9Ê
19.6Ê
19.1Ê
22.5Ê
15.7Ê
21.1Ê
19.7Ê
21.0Ê
21.5Ê
18.7Ê
20.3Ê
Wind (x0.7)Ê
WeddellÊ
22.5Ê
24.0Ê
18.1Ê
19.1Ê
18.9Ê
22.0Ê
15.2Ê
20.2Ê
19.1Ê
20.7Ê
21.0Ê
18.4Ê
19.9Ê
Ê
Sourced from ?Staffell and Pfenninger (2016)?. For wind, the turbine size is 4.5 MW at 150 m.Ê

A.2. Wind generation capacity factorsÊ
This work employs a wind capacity factor multiplier to add additional local bias correction to the reanalysis-based estimates from Renewables.ninja. The value is chosen as 0.7 for all aspects of this work except the sensitivity analysis presented in the tornado charts (Figure 14 and A.2). The need to apply local correction factors to account for discrepancies between reanalysis (global, satellite-based) estimates of local ground wind speed and actuals is common, and several studies have been published (Pfenninger and Staffell, 2016; Staffell and Pfenninger, 2016) indicating that the magnitude of the discrepancies is significant in the context of determining optimal energy generation mix. 

Figure  shows the uncorrected reanalysis-based capacity factors across the NT from Renewables.ninja. As those values lie at the upper end of (or beyond) the typical ranges for the capacity factor for onshore wind (25?45%), a comparison with on-the-ground actuals was completed to derive the 0.7 local bias correction factor for this work. The crude nature of this correction factor is representative of the lack of on-the-ground wind speed studies throughout the NT.


Figure A.1 Reanalysis estimates of onshore wind capacity factor
Estimates from http://www.renewables.ninja website for a Vestas V126/3000 at 3.5 MW at 100 m height. 
Two approaches were taken to estimate the 0.7 broad correction factor. The first was to compare correction factors where actual wind SCADA data are available in the NEM at sites as close to the NT as possible Ð these were AEMO REZs Q2 and Q5 in north-west QLD.4 Since the AEMO capacity factor values are corrected against actual SCADA data for existing wind generation installations, comparison with the reanalysis estimates allows calculation of a local bias correction for those sites. Table A.2 Comparison of AEMO and Renewables.ninja onshore wind capacity factors for QLDshows this comparison, with the value of the correction factor for the two REZs closest to the NT being 0.79 and 0.61 such that an average value near 0.7 is reasonable. 
Table A.2 Comparison of AEMO and Renewables.ninja onshore wind capacity factors for QLD
REZ ID
Average yearly capacity factor (2021)


AEMO high
AEMO med
RenNin
AEMO med/RenNin
Q1
0.44
0.40
0.33
1.21
Q2
0.41
0.32
0.40
0.79
Q3
0.00
0.00
0.36
0.00
Q4
0.36
0.31
0.43
0.74
Q5
0.32
0.29
0.48
0.61
Q6
0.38
0.33
0.44
0.74
Q7
0.29
0.28
0.33
0.84
Q8
0.38
0.33
0.43
0.78
Q9
0.30
0.25
0.38
0.67
The locations used to download wind data from Renewables.ninja only approximate those used for the wind generation sites in the QLD REZs. While those locations are within the REZ itself, they may not match the exact locations of specific wind generation installations within that REZ. Rows in bold are those closest to the NT.

The second approach was direct calibration of the reanalysis estimates using SODAR-based wind speed estimates for two locations within the NT where ongoing (Table A.3) or recently completed (Table A.4) studies were available. Using a variety of turbine power curves and turbine heights, the SODAR data were converted to an average capacity factor over the period for which the data were available. 
This showed that for Owen Springs the reanalysis data required a local bias correction factor of 0.57, while in Tennant Creek it was 0.87. Combining these values with those above from NEM sites in north-west QLD suggested the value of 0.7 as being reasonable. It is noted that reanalysis-based estimates of local wind speed alone are insufficient for use in a business case. 
Table A.3 Tennant Creek reanalysis correction factor, December 2022 to June 2023
Height (m)
100
150
200
Turbine



RN VS66 (2 MW)
0.82
0.98
0.95
RN V117 (4 MW)
0.83
0.91
0.86
RN V164 (9.5 MW)
0.83
0.93
0.89
Siemens Gamesa SG 4.5 145
0.84
0.89
0.83
FG V126 (3 MW)
0.82
0.87
0.80


Average:
0.87

Table A.4 Owen Springs reanalysis correction factor, November 2021 to November 2022
Height (m)
100
150
200
Turbine



RN VS66 (2 MW)
0.46
0.55
0.62
RN V117 (4 MW)
0.48
0.58
0.65
RN V164 (9.5 MW)
0.47
0.56
0.63
Siemens Gamesa SG 4.5 145
0.51
0.62
0.68
FG V126 (3 MW)
0.47
0.57
0.64


Average:
0.57


A.3. Optimistic ACoH and ACoE sensitivities 
Here we ÒstackÓ the sensitivities from Figure 14 to look at the most optimistic case for the ACoH. This is done by using the low setting from each row for the row below, such that each subsequent sensitivity is to the low setting of the variable before rather than always relative to a base case. These results still have the 1.35 regional markup factor applied to all costs except gas fuel.Ê

Ê
Figure A.2 Optimistic ACoH and ACoE sensitivities 
These are the same range of parameter sensitives as in Figure 14. However, here they apply to the low setting from the row above.Ê




Table A.5 Optimistic ACoH and ACoE sensitivities
Ê
Ê
Ê
Ê
ACOH [$/KG]Ê
ACOE [$/MWH]Ê
Variable
LowÊ
MidÊ
HighÊ
LowÊ
MidÊ
HighÊ
LowÊ
MidÊ
HighÊ
AE efficiencyÊ
45 MWh/tÊ
55 MWh/tÊ
65 MWh/tÊ
7.5Ê
9.4Ê
11.2Ê
143Ê
144Ê
147Ê
Recover spilled VRE
Yes
No
No
6.6
7.5
7.5
120
143
143
Renewable fractionÊ
0.8Ê
0.9Ê
1Ê
6.1Ê
6.6Ê
7.5Ê
110Ê
120Ê
135Ê
Wind cap. fac. multiplierÊ
1Ê
0.7Ê
0.5Ê
5.4Ê
6.1Ê
6.6Ê
96Ê
110Ê
121Ê
Discount rateÊ
0.06Ê
0.07Ê
0.08Ê
5.0Ê
5.4Ê
5.8Ê
89Ê
96Ê
103Ê
GenCost scenarioÊ
NZ by 2050Ê
Current PoliciesÊ
NZ post-2050Ê
4.9Ê
5.0Ê
5.1Ê
89Ê
89Ê
90Ê
Weather yearÊ
Best (2019)Ê
Average (2018)Ê
Worst (2010)Ê
4.6Ê
4.9Ê
5.0Ê
83Ê
89Ê
92Ê
HVDC capital costÊ
$220/km/MWÊ
$439/km/MWÊ
$878/km/MWÊ
4.6Ê
4.6Ê
4.7Ê
82Ê
83Ê
85Ê
Gas priceÊ
$6.4/GJÊ
$8/GJÊ
$9.6/GJÊ
4.5Ê
4.6Ê
4.7Ê
78Ê
82Ê
84Ê

These are for a technology cost year of 2028 and a 1.35 regional markup factor.Ê

A.4. Model definition
The model applied for this work is a linear optimisation. The following details the objective function for total annualised system cost Ð which is minimised, as well as the variables (the unknowns the model solves for), the parameters (known input data), and sets (the dimensions of the problem).
Notes on notation:
* Sets over which variables and parameters are defined are indicated with brackets rather, e.g., foo(T) for a variable foo defined over set T. 
* Summation notation is written as sum(T,foo(T)) indicating a sum of foo over T, or for multiple sets sums as sum((T,R),bar(T,R)) indicating the sum over bar over both T and R for which it is defined.
A.5. Sets (model dimensionality)
h hourly time interval in a single year
Month Month of the year
GenTech generation technology options
       GasGenTech subset of GenTech which includes gas technologies
       VREGenTech subset of GenTech which includes variable renewables
H2Tech hydrogen electrolyser technology options
ElecStoTech storage technology options where each technology has a defined duration
H2StoTech hydrogen storage technology options
TxTech transport technology
ElecTxTech transmission technology
PipeTxTech pipeline technology
Region energy supply regions available
       RegionD subset of regions that are demand regions
       NotDemandRegion subset of regions that are not demand regions
A.6. Variables (unknown outputs of the model)
TotalCost_AUD annualised total costs of meeting demand for energy in the precinct in Australian dollars
ElecConvCap(GenTech,Region) electricity conversion technology capacity (MW)
H2ConvCap(H2Tech,Region) hydrogen conversion (electrolyser) technology capacity (MW)
ElecStoCap(ElecStoTech,Region) electricity storage capacity (MWh)
H2StoCap(H2StoTech,Region) hydrogen storage capacity (T)
TxCap(TxTech,Region,RegionD) transport capacity (ElecTxTech in MW, PipeTxTech in tonne/hr)
ElecEnergyProd(GenTech,Region,h) hourly energy production (MW)
H2EnergyProd(H2Tech,Region,h) hourly energy production (tonnes)
ElecRegionExport(Region,h) total exported energy including storage discharge (GenTech in MW)
H2RegionExport(Region,h) total exported energy including storage discharge (H2Tech in tonnes)
ElecStateofCharge(ElecStoTech,Region,h) electricity storage state of charge (MWh)
ElecDischar(ElecStoTech,Region,h) electricity storage discharge (MW)
ElecChar(ElecStoTech,Region,h) electricity storage charge (MW)
H2StateofCharge(H2StoTech,Region,h) hydrogen storage state of charge (tonnes)
H2Dischar(H2StoTech,Region,h) hydrogen storage discharge (tonnes)
H2Char(H2StoTech,Region,h) hydrogen storage charge (tonnes)
A.7. Parameters (known inputs)
dt fraction of one hour in which demand load and renewable profile data are stepped
ElecConvCapCost(GenTech) conversion technology capital cost ($/MW)
H2ConvCapCost(H2Tech) conversion technology capital cost ($/MW)
ElecStoCapCost(ElecStoTech) electricity storage capacity cost in $/MWh
H2StoCapCost(H2StoTech) hydrogen storage capacity cost in $/tonne
TxCapCost(TxTech) transmission capacity cost in $/MW/km or $/tonne/hr/km
Distance(region,regionD) distance from region to region
Duration_Hrs(ElecStoTech) duration of electricity storage technology (hours)
H2Duration_H2s(H2StoTech) duration of hydrogen storage technology (hours)
DiscountRate discount rate or weighted cost of capital
ElecConvEconomicLife_Yrs(GenTech) electricity conversion capacity cost amortisation period
ElecStoEconomicLife_Yrs(ElecStoTech) electricity storage capacity cost amortisation period
H2ConvEconomicLife_Yrs(H2Tech) hydrogen conversion capacity cost amortisation period
H2StoEconomicLife_Yrs(H2StoTech) hydrogen storage capacity cost amortisation period
TxEconomicLife_Yrs(TxTech) transport capacity cost amortisation period
ElecConvConstructPeriod_Yrs(GenTech) capacity construction period
ElecStoConstructPeriod_Yrs(ElecStoTech) storage capacity cost construction period
H2ConvConstructPeriod_Yrs(H2Tech) capacity construction period
H2StoConstructPeriod_Yrs(H2StoTech) storage capacity cost construction period
TxConstructPeriod_Yrs(TxTech) transport capacity construction period
ElecConvCapCostConversion(GenTech) converts electricity conversion capacity upfront costs to annual cost and accounts for interest lost during construction
ElecStoCapCostConversion(ElecStoTech) converts electricity storage capacity upfront costs to annual cost and accounts for interest lost during construction
H2ConvCapCostConversion(H2Tech) converts hydrogen conversion capacity upfront costs to annual cost and accounts for interest lost during construction
H2StoCapCostConversion(H2StoTech) converts hydrogen storage capacity upfront costs to annual cost and accounts for interest lost during construction
TxCapCostConversion(TxTech) converts transport capacity upfront costs to annual costs and accounts for interest lost during construction
GJperMWh gigjoules per MWh which is 3.6
GasPrice_AUDperGJ natural gas price in $/GJ
FuelEfficiency(GasGenTech) fuel efficiency expressed as the ratio of energy out per energy in
ElecConvVariableOM(GenTech) electricity variable operating and maintenance costs ($/MWh)
ElecConvFixedOM(GenTech) electricity fixed operating and maintenance costs ($/MW)
ElecStoFixedOM(ElecStoTech) electricity storage fixed operating and maintenance costs ($/MWh)
H2ConvVariableOM(H2Tech) hydrogen electrolyser variable operating and maintenance costs ($/Tonne)
H2ConvFixedOM(H2Tech) hydrogen electrolyser fixed operating and maintenance costs ($/MW)
H2StoFixedOM(H2StoTech) hydrogen storage fixed operating and maintenance costs ($/tonne)
VREprofile_MWperMW(VREGenTech,Region,h) variable renewable generation profile in MWs per MW
NonH2ElecDemand_MW(h) the main industrial electricity demand other than for hydrogen production
FixedLoad_MW(h) a fixed constant in MW
IndustrialProductionCapacity_Tonnes(h) tonnes of final product produced per dt
MWperTonne electricity load per tonne of production
ElecStoEfficiency(ElecStoTech) round trip electricity storage efficiency
H2StoEfficiency(H2StoTech) round trip hydrogen storage efficiency
ElecToH2Efficiency(H2Tech) tonnes of hydrogen produced per MWh of electricity input to electrolyser technology
AnnualHydrogen_Tonnes targeted annual production of hydrogen
H2Demand_Tonnes(h) required half-hourly demand for hydrogen to customers
MinHourlyUseFactor(H2Tech) minimum amount of instantaneous utilisation of electrolyser capacity
ElecTxLossFactor transmission loss factor accounting for average electricity lost owing to transmission waste heat
PipeTxLossFactor pipeline loss factor accounting for average energy lost owing to leakage
RenCapMax(VREGenTech,Region) maximum capacity that can be deployed in this region by renewable technology
MaxGasShare maximum share of gas in total electricity supply (including for hydrogen)
ElectrolyserCapFac capacity factor of electrolysers
A.8. Objective function 
Table A.6 describes the terms in the objective function, which minimises the cost of supply over the model time horizon (here 1 year) at a temporal resolution of dt (here dt = 2 hours). 
Table A.6
ObjElecConvCapCost
sum((GenTech,Region),ElecConvCap(GenTech,Region) ? ElecConvCapCost(GenTech) ? ElecConvCapCostConversion(GenTech) ) +
Capital cost of electricity conversion technology
ObjH2ConvCapCost
sum((H2Tech,Region),H2ConvCap(H2Tech,Region) ? H2ConvCapCost(H2Tech) ? H2ConvCapCostConversion(H2Tech) ) +
Capital cost of hydrogen conversion (electrolyser) technology
ObjElecStoCapCost
sum((ElecStoTech,Region), ElecStoCap(ElecStoTech,Region) ? ElecStoCapCost(ElecStoTech) ? ElecStoCapCostConversion(ElecStoTech) ) +
Capital cost of electricity storage
ObjH2StoCapCost
sum((H2StoTech,Region), H2StoCap(H2StoTech,Region) ? H2StoCapCost(H2StoTech) ? H2StoCapCostConversion(H2StoTech) ) +
Capital cost of hydrogen storage
ObjGasGenFuelCost
sum((GasGenTech,Region,h), ElecEnergyProd(GasGenTech,Region,h) ? dt ? GasPrice_AUDperGJ /FuelEfficiency(GasGenTech) ? GJperMWh ) +
Fuel cost of gas
ObjElecConvOMCost
sum((GenTech,Region,h),ElecConvVariableOM(GenTech) ? ElecEnergyProd(GenTech,Region,h) ? dt ) + sum((GenTech,Region),ElecConvFixedOM(GenTech) ? ElecConvCap(GenTech,Region) ) +
Operating and maintenance costs of electricity conversion
ObjH2ConvOMCost
sum((H2Tech,Region,h),H2ConvVariableOM(H2Tech) ? H2EnergyProd(H2Tech,Region,h) ? dt ) + sum((H2Tech,Region),H2ConvFixedOM(H2Tech) ? H2ConvCap(H2Tech,Region) ) +
Operating and maintenance costs of hydrogen conversion (electrolysers)
ObjElecStoOMCost
sum((ElecStoTech,Region), ElecStoFixedOM(ElecStoTech) ? ElecStoCap(ElecStoTech,Region) ) +
Operating and maintenance costs of electricity storage
ObjH2StoOMCost
sum((H2StoTech,Region), H2StoFixedOM(H2StoTech) ? H2StoCap(H2StoTech,Region) ) +
Operating and maintenance costs of hydrogen storage
ObjTxCapCost
Sum((TxTech,Region,RegionD),TxCap(TxTech,Region,RegionD) ? TxCapCost(TxTech) ? Distance(Region,RegionD) ? TxCapCostConversion(TxTech)
Capital cost of transmission
Where:
*Conversion(T) = (1+DiscountRate)^ *ConstructPeriod_Yrs(T) ? (DiscountRate ? ((1+DiscountRate)^ *EconomicLife_Yrs(T) ))/(((1+DiscountRate)^ *EconomicLife_Yrs(T) )-1)

A.9. Constraints
Maximum regional exports and regional energy supply?demand balance constraint: electricity or hydrogen from production and storage discharge net of charging and any intermediate use of energy can exceed exports (i.e. spilled energy can exist). However, exports cannot exceed production and storage discharge net of charging and intermediate use of energy in any half hour.
ElecRegionExport(Region,h) ² sum(GenTech,ElecEnergyProd(GenTech,Region,h)) + sum(ElecStoTech,ElecDischar(ElecStoTech,Region,h)) - sum(ElecStoTech, ElecChar(ElecStoTech,Region,h) ) - sum((H2Tech,Region),H2EnergyProd(H2Tech,Region,h) ? ElecToH2Efficiency(H2Tech) )
ElecRegionImport(RegionD,h) <= sum(NotDemandRegion, ElecRegionExport(NotDemandRegion,h))
ElecRegionImport(NotDemandRegion,h) == 0
H2RegionExport(Region,h) ² sum(H2Tech,H2EnergyProd(H2Tech,Region,h))+ sum(H2StoTech,H2Dischar(H2StoTech,Region,h)) - sum(H2StoTech,H2Char(H2StoTech,Region,h) )
Total demand?supply balance: final industrial demand must be met by exports from all regions. Regional exports are subject to transport losses.
sum(Region,ElecRegionExport(Region,h) ? (1-ElecTxLossFactor)) ³ NonH2ElecDemand_MW(h)
ElecRegionExport(DemandRegion,h) x É >= NonH2ElecDemand_MWh(DemandRegion,h)
sum(Region,H2RegionExport(Region,h) ? (1-PipeTxLossFactor)) ³ H2Demand_Tonnes(h)
Variable electricity generation defined: capacity multiplied by the variable production profile relevant to the selected generation site.
ElecEnergyProd(VREGenTech,Region,h) = VREprofile_MWperMW(VREGenTech,Region,h) ? ElecConvCap(VREGenTech,Region)
Maximum energy production constraint: energy production cannot exceed capacity. Variable renewables not included here because they are fixed by the previous equation.
ElecEnergyProd(GasGenTech,Region,h) ² ElecConvCap(GasGenTech,Region)
(H2EnergyProd(H2Tech,Region,h)/ElecToH2Efficiency(H2Tech)) ² H2ConvCap(H2Tech,Region)
State of charge balance: state of charge is the sum of charging and state of charge minus discharging in the previous period.
ElecStateofCharge(ElecStoTech,Region,h+1) = ElecStateofCharge(ElecStoTech,Region,h)+ ElecStoEfficiency(ElecStoTech)^1/2 ? ElecChar(ElecStoTech,Region,h) ? dt - ElecStoEfficiency(ElecStoTech)^-1/2 ? ElecDischar(ElecStoTech,Region,h) ? dt 
H2StateofCharge(H2StoTech,Region,h+1) = H2StateofCharge(H2StoTech,Region,h)+ H2StoEfficiency(H2StoTech)^1/2 ? H2Char(H2StoTech,Region,h) ? dt - H2StoEfficiency(H2StoTech)^-1/2 ? H2Dischar(H2StoTech,Region,h) ? dt
Maximum state of charge constraint: cannot exceed energy capacity of storage.
ElecStateofCharge(ElecStoTech,Region,h) ² ElecStoCap(ElecStoTech,Region)
H2StateofCharge(H2StoTech,Region,h) ² H2StoCap(H2StoTech,Region)
Maximum discharge constraint 1: cannot exceed state of charge.
ElecDischar(ElecStoTech,Region,h) ² ElecStateofCharge(ElecStoTech,Region,h)/dt 
H2Dischar(H2StoTech,Region,h) ² H2StateofCharge(H2StoTech,Region,h)/dt
Maximum discharge constraint 2: cannot exceed power capacity of storage.
ElecDischar(ElecStoTech,Region,h) ² ElecStoCap(ElecStoTech,Region)/Duration_Hrs(ElecStoTech)
H2Dischar(H2StoTech,Region,h) ² H2StoCap(H2StoTech,Region)/H2Duration_Hrs(H2StoTech)
Maximum charge constraint: charge cannot exceed power capacity of storage.
ElecChar(ElecStoTech,Region,h) ² ElecStoCap(ElecStoTech,Region)/Duration_Hrs(ElecStoTech)
H2Char(H2StoTech,Region,h) ² H2StoCap(H2StoTech,Region)/H2Duration_Hrs(H2StoTech)
Minimum annual hydrogen production constraint: a minimum annual hydrogen production target must be met.
Sum((H2Tech,Region,h), H2EnergyProd(H2Tech,Region,h) ? dt ) ³ AnnualHydrogen_Tonnes
Minimum electrolyser production constraint: the minimum amount of instantaneous electrolyser capacity utilisation.
H2EnergyProd(H2Tech,Region,h) ³ MinHourlyUseFactor(H2Tech) ? H2ConvCap(H2Tech,Region) ? ElecToH2Efficiency(H2Tech) ? dt 
Maximum transport constraint: exports from any region in a time period cannot exceed transmission capacity from that region.
ElecRegionExport(Region,h) ² sum((ElecTxTech,RegionD),TxCap(ElecTxTech,Region,RegionD))
H2RegionExport(Region,h) ² sum((PipeTxTech,RegionD),TxCap(PipeTxTech,Region,RegionD)) ? dt
Maximum renewable deployment constraint: the maximum amount of renewable generation capacity that can be deployed by a region.
ElecConvCap(VREGenTech,Region) ² RenCapMax(VREGenTech,Region)
Maximum gas supply constraint: the maximum share of annual electricity generation that can be supplied by gas technologies.
sum((GasGenTech,Region,h),ElecEnergyProd(GasGenTech,Region,h)) 
 ² MaxGasShare ? sum(h,ElecEnergProd(h) )
Electrolyser capacity factor constraint: specifies the capacity factor of electrolysers.
sum(h,H2EnergyProd(H2Tech,Region,h))/length(h) == ElectrolyserCapFac * H2ConvCap(H2Tech,Region) * ElecToH2Efficiency(H2Tech)


As AustraliaÕs national science agency and innovation catalyst, CSIRO is solving the greatest challenges through innovative science and technology.
CSIRO. Unlocking a better future for everyone.
Contact us
1300 363 400
+61 3 9545 2176
csiro.au/contact
csiro.au
For further information

CSIRO Energy 
Andrew Ross
+61 8 6436 8790
Andrew.Ross@csiro.au
csiro.au/Energy





















1 Only alkaline electrolysers are included because the cost-effectiveness of other electrolyser types is assumed not to have dropped sufficiently by 2025 or 2028 (our chosen technology cost years) for the model to take up. 
2 The temporal resolution of the model is typically chosen at two hours when many model runs need to be completed and the model run time is significant Ð as was the case here. A sensitivity analysis to the choice of one- or two-hourly temporal resolution indicated that two-hourly was sufficient for the purposes of this study. 
3 Throughout, gas generation refers to CCGT generation as that is the most cost-effective new-build gas technology. 
4 See the 2023 AEMO IASR Excel workbook for REZ capacity factors and locations. 
---------------


------------------------------------------------------------




---------------


------------------------------------------------------------








ii | CSIRO AustraliaÕs National Science Agency

Northern Territory Low Emissions Carbon Capture Storage and Utilisation Hub | i