© W. Grace, Published by EDP Sciences, 2023

1 Introduction

1.1 The rise of private solar PV

Worldwide, the amount of solar PV installed each year has grown from around 20 GWp in 2010 to over 200 GWp in 2022 with cumulative installations approaching 1000 GW [1]. The International Energy Agency's ‘Net Zero Emissions by 2050 Scenario’ projects that global solar capacity will reach 11,000 GW by 2040 [2]. In Australia there are over 3 million small scale (rooftop) solar PV installations (a third of houses) with a capacity of 13.5 GWp¹. It has been estimated that the total potential for rooftop in Australia “is 179 gigawatts with an annual energy output of 245 terawatt-hours” which is more than the current annual consumption in the National Electricity Market [3].

This level of existing and potential penetration reflects the solar insolation characteristics of the populated southern regions of Australia in combination with government subsidies, reducing capital costs and relatively high network electricity tariffs [4].

These circumstances are not reflected in all countries with solar insolation levels suitable for small scale solar PV. For example the level of penetration in U.S. residential buildings is currently less than 1% [5]. However, 100% renewable energy scenario modelling of U.S. electricity demand predicts that solar power will dominate the generation mix [6], and undoubtedly this will mean that small-scale systems will play a significant role in the future [7]. In India by 2019 rooftop PV system capacity was about 2 GW but government policy is to reach 40 GW of rooftop PV capacity (of a total of 100 GW total of grid connected solar power by 2022) [8].

The benefits of BTM solar PV in terms of cost savings is limited by daytime-only supply. Coupling such systems with battery energy storage systems (BESS) offers significant additional savings as evening demand is met from stored excess daytime generation [9–12]. Penetration of private BESS is however not presently financially attractive due to the high capital system costs. Presently in Australia the installed cost of BESS is around AUD 1500/kWh². The capital cost of BESS generally, including for small scale systems, is projected to reduce significantly as demand increases, and will become financially attractive when costs fall to Euro 400/kWh (approx. AUD 655) [13].

Although progress varies significantly across the world, it is clear that BTM rooftop solar PV, coupled with battery storage will supply a very significant fraction of electricity demand in the future. This study seeks to identify and quantify the impact of that emerging trend on electricity networks.

1.2 Climate policy context

The goal of the Paris Agreement is “to limit global warming to well below 2, preferably to 1.5°C, compared to pre-industrial levels”. Since the Agreement was signed in 2016 there has been renewed worldwide focus on transitioning energy systems to achieve zero emissions by 2050 [14], with California often cited as leading on policy [15]. So-called “Scope 2” emissions are those related to the purchase of electricity from network generation and are a significant component of the national greenhouse gas (GHG) emission profile in most countries.

In Australia in 2019 the Scope 2 GHG emissions were 179,446 ktCO₂-e which was 34% of total emissions. In the state of Western Australia, the focus of this study, Scope 2 emissions from purchased electricity were 25,048 ktCO₂-e or 27% of total emissions³. This represents an increase of 185% since the year 2000. Most of the state's population is serviced by the South West Interconnected System (SWIS), where the GHG intensity has declined by around 30% since 1990 to 0.69 kg CO₂-e/kWh as gas and more recently, solar PV and wind power, has led to a reduction in coal-fired generation. The Government of Western Australia has declared an intention for the state to achieve “Net Zero” emissions by 2050, and is developing strategies for each sector of the economy to reach that goal. Clearly Scope 2 emissions are a major component of each sector's emissions and will play a central role in achieving Net Zero.

1.3 The south west interconnected system

The SWIS services the south west of the state of Western Australia (WA) including the capital of Perth, with over 1.1 million household and business customers. Western Australia has amongst the highest penetration of behind-the-meter (BTM) private solar PV in the world with over 340,000 systems installed on the SWIS with a nameplate capacity of over 2000 MW producing some 3200 MWh of energy or 12% of electricity consumed annually by SWIS customers [16]. The Australian Energy Market Operator (AEMO) is responsible for operating the electricity system including system planning, dispatching energy and monitoring voltage and frequency. AEMO operates the Wholesale Electricity Market (WEM) in which licensed retailers purchase electricity from generators to on-sell to customers. In 2021, some 17,600 MWh of energy was delivered to customers from the sources set out in Figure 1.

Fig. 1 SWIS generation 2021.

1.4 Private solar PV in western australia

Each year AEMO produces a Statement of Opportunities for the WEM in which it sets out projections for future electricity demand over the coming decade, including the impact of private BTM solar PV [16]. These projections have continually under-estimated the take-up, as illustrated in Figure 2.

The latest projections assume that the annual rate of take-up will decline from 11% pa to 5% by 2031, despite the fact that annual increases have averaged 26% pa since 2015. This is important because of the impact of BTM solar on the SWIS in terms of annual, seasonal and diurnal network demand. Some 36% of homes and businesses have solar PV systems connected to the SWIS, leaving plenty of roof-space available for capacity to grow. The impact of the existing capacity is already being felt with customers supplying 70% of demand and network loads falling to 707 MW in October 2022 requiring network generation to be curtailed⁴. Half hourly network data from 2021 clearly illustrates the emerging impact as illustrated in Figure 3 which shows midday network loads reduced by up to 1,500 MW at some times of year.

Negative loads have already occurred in the South Australia network due to high private solar generation⁵.

Fig. 2 AEMO projections of private solar PV.

Fig. 3 (a) Half hourly electricity demand and distributed PV (DPV) generation. (b) Resulting half hourly SWIS network loads.

1.5 Study objectives

Most published research on the energy transition has a focus on network renewable energy and storage [17,18] but neglects the impact of private generation and storage on electricity networks. The objective of the study reported here is to explore through systems modelling, the likely amount of future private BTM generation and energy storage, and the optimum generation and energy storage profile for a 100% renewables SWIS based on: the underlying demand for electricity as the population grows; homes and businesses electrifying their appliances and vehicles; and private BTM generation and energy storage growth. Although the study is focused on the SWIS, it has relevance for any region with high levels of solar insolation where rooftop solar PV will be favoured, including Africa, Central and Southern America and the United States [19].

2 Methodology

2.1 Model configuration

The model described here is an adaptation of that described in previously published work [20], and utilizes the stock and flow and optimisation capability of the Vensim⁶ system dynamics simulation software. The model (see Fig. 4) incorporates underlying customer demand for electricity, BTM generation in homes, businesses and precinct scale microgrids, and SWIS network generation (solar PV, on and offshore wind, wave) and storage (in the form of pumped hydro [21], batteries [22] and hydrogen). Accordingly, it is a model of the electricity system comprising these interacting elements rather than a model per se of the SWIS.

The model simulates the period from 2021 to 2051 by which time the state of Western Australia aims to have ‘Net Zero’ emissions. The model uses a 1 hour time step for a typical day in each month.

Fig. 4 Model configuration.

2.2 Demand assumptions

Assumptions for the underlying customer demand for such an extended timeframe will always be uncertain. For this study, the overall philosophy of AEMO for Australia's National Energy Market (NEM) set out in their 2022 Integrated System Plan (ISP) [23] has been adopted. This approach was taken as AEMO has not prepared the same ISP for the SWIS. The ISP includes several scenarios in response to the energy transition, including “Progressive Change − pursuing an economy-wide net zero emissions 2050 target progressively, ratcheting up emissions reduction goals over time”. This scenario includes projections associated with energy efficiency and the following transition timeline:

• 2020 s − consumer Distributed Energy Resources (DER) investment, corporate emission abatement, and technology cost reductions;

• 2030s − commercially viable alternatives to emissions-intensive heavy industry;

• 2040s − industrial electrification, Electric Vehicles (EVs) becoming more prevalent and consumer electrification;

• Post 2045–Some domestic hydrogen production supporting the transport sector and as a blended pipeline gas, with some industrial applications after 2045.

The NEM ISP demand projections were normalised for energy intensity of the economy (i.e. as a fraction of projected Gross State Product (GSP) taken from the same report) and factored to apply to Western Australia from 2021. Electric vehicle (EV) projections were taken from a report by the Commonwealth Scientific and Industrial Research Organisation (CSIRO) [24], see Figure 5.

These assumptions were combined with AEMO's 2022 Electricity Statement of Opportunities (ESOO) for the SWIS [16], which contains a ten year outlook, to produce a disaggregated demand projection for the SWIS to 2051 (Fig. 6) which was used in the model.

The diurnal and seasonal demand pattern for homes and businesses was derived as follows. AEMO provides data from the SWIS that sets out the total ‘operational load’ for each half hour of the year⁷. The 2021 data was combined with the calculated output of the private solar (also reported in the ESOO) to derive an hourly demand pattern across the year for the whole of the SWIS. This pattern was then applied to the annual demand (Fig. 6), and disaggregated into user types using a combination of information from the ESOO (existing demand) and the NEM ISP (future demand). Although no projection of demand over this timeframe can be accurate, this approach is deemed suitable for the purposes of this study.

Fig. 5 Electric vehicle projections.

Fig. 6 SWIS Demand projection.

2.3 Private solar PV projections

The diurnal and seasonal generation of household scale solar PV for systems in the SWIS was derived from the National Renewable Energy Laboratory's ‘PVWatts’ calculator⁸ in combination with assumptions of Australia's Clean Energy Regulator, which is an annual capacity factor of 0.183. The battery component of the model assumes 10% losses in charging and a depth of discharge (DOD) of 90% (reflecting the range of different battery types). Battery use is simulated simply, i.e. the battery is charged when generation exceeds demand, and discharged when demand exceeds generation. Excess generation is exported to the network, and energy is imported from the network when the battery is discharged to its DOD.

Projections for private BTM solar PV and battery storage are produced by the model on the basis of payback period as follows. At each time step, the payback period for investment in a solar PV installation is calculated (Annual savings from avoided network purchases/Capital cost of system) and the take-up (percentage of premises without solar PV) determines the number of systems added to the stock of private solar PV in each time step. The take-up rate is based on Figure 7, and is the same as that used in previous models which have proven to match reasonably well with actual take-up over recent years (see Fig. 2).

This approach was applied separately to household and business premises with and without EVs, and the payback period of private batteries is calculated on the basis of the ‘additional’ savings over and above those arising from the solar PV investment.

Although the savings from private systems are dependent on the SWIS tariffs, previous work has indicated that this is secondary to the reducing cost of solar and batteries. Accordingly, the model assumes static SWIS tariffs throughout the simulation period for the purposes of estimating take-up.

The price of residential and business solar PV and battery systems was taken from combining the NEM ISP cost projections with historical market prices⁹. Prices are currently reduced by a discount derived from the Australian Government's Renewable Energy Target (RET) scheme¹⁰, which will diminish until the scheme comes to an end in 2030. The resulting price profile is depicted in Figure 8, which is inclusive of inverter costs.

A similar approach was taken to project the price of residential and business scale battery systems resulting in the price profile set out in Figure 9.

The model also calculates the escalating capacity of private solar systems based on declining prices over time using the same payback logic described above. The model ‘selects’ a suitable battery capacity based on a separate model (not described here) that optimises battery capacity for a given solar PV capacity and the demand profile of average residential and business premises. Take-up of battery systems is then governed by payback in accordance with Figure 7. As costs fall, the penetration of private battery storage systems will become more financially viable [25].

In addition to residential and business solar and battery systems, the system will also be influenced by BTM microgrid solutions, such as those being implemented by the Western Australian Government's land development agency Development WA at the Peel Business Park¹² south of the capital city of Perth, in which the author was involved. These schemes operate as embedded networks with generation and storage, and offer customers high levels of renewable energy as well as discounts on regulated tariffs¹³. It is likely that these schemes will increase in future, particularly in industrial settings, and the model assumes that by 2050 some 20% of large industrial loads will be serviced by network connected microgrids.

Fig. 7 Take-up of private solar PV and battery systems.

Fig. 8 Projected capital cost of residential and business solar PV. 11

Fig. 9 Projected capital cost of residential and business battery systems.

2.4 Network scale generation and storage

The seasonal and diurnal generation patterns and capacity factors of network scale solar and onshore wind generation were derived from AEMO data on existing facilities. The resulting annual capacity factors assumed for these were 0.26 for fixed PV and 0.4 for onshore wind, based on published AEMO historical data. The generation pattern for single axis PV was taken from PVWatts with a capacity factor of 0.26. Generation patterns for offshore wind were obtained from the Global Wind Atlas¹⁴ (capacity factor 0.47), and from Hughes and Heap [26] for wave technology (capacity factor 0.3).

The cost of large scale generation, pumped hydro, battery storage and electrolysers was obtained from CSIRO's GenCost 2021–22 report [27]. The model uses Levelised Cost of Energy (LCOE) values for generation facilities but Equivalent Annual Cost (EAC) for storage facilities (i.e. the annual cost of owning, operating, and maintaining the asset over its entire life). Each set of costs is set out in the Tables 1-3 for generation, battery storage and pumped hydro respectively.

The modelling also considers the possibility of using hydrogen to store excess energy, and use it to fuel Open Cycle Gas Turbines (OCGTs). The model assumes the electrolyser load to be 57 MWh/tH₂ and 10.9 MWh/ tH₂ of energy produced in generation, i.e. a round trip efficiency of around 19% [23].

2.5 Network costs

There are no published generation costs for the WEM but Julius Susanto (Technical Director at Australian Energy Market Commission) has produced some estimates¹⁵ which are summarised in Table 4, excluding the cost of ancillary services and market fees. This has been used to benchmark the cost of future network configurations.

2.6 Simulations

The model was used firstly to project the take-up of private residential, business and microgrid solar PV and battery systems to 2051. The model assumes saturation at 72% of households (80% of separate houses, and 50% of medium and high density dwellings) and 50% of business premises.¹⁶ The balance of demand is assumed to be met by private microgrids and the SWIS.

The next phase of modelling explored the optimum combination of SWIS network generation and energy storage to supply that demand. For these simulations the model produces the hourly network load profile arising from gross system demand less the self-supply from private solar PV and batteries. The capacity of network generation and storage to supply that net demand was then optimised to achieve the lowest annual network cost using Vensim's optimizing engine using Powell hill climbing algorithm searches [28].

3 Results

3.1 Private solar PV and battery storage

The payback profiles generated by the model for residential and business systems are shown in Figures 10 and 11.

It is clear from the declining payback periods in Figures 10 and 11 that both the capacity and penetration of private solar will increase over coming decades. The penetration of private battery systems will somewhat lag, but likely to grow strongly after 2025.

Figures 12a and 12b depict the simulated penetration of solar and batteries in houses and businesses with internal combustion engines (ICE). Penetration is projected be saturated by around 2035 in households but earlier in businesses whose demand pattern is better suited to solar PV generation. Penetration in houses and businesses with electric vehicles (not shown) is slightly higher in houses because of the additional benefit in overnight EV charging.

Figures 13a and b depict the simulated solar PV and battery capacity in households, businesses and microgrids associated with the projected level of penetration. The model projects that by 2030 there could be around 7000 MW of private solar capacity and 4,200 MWh of battery capacity, with solar capacity doubling by 2050 and battery capacity rapidly escalating to 24,000 MWh by 2050.

Fig. 10 Household payback periods.7

Fig. 11 Business payback periods.

Fig. 12 (a) Penetration of solar PV and batteries in ICE households. (b) Penetration of solar PV and batteries in ICE business premises.

Fig. 13 (a) Private solar PV capacity. (b) Private battery capacity.

3.2 Network impacts

The impact of this large increase in private generation and storage capacity becomes clear when viewing the performance of individual private systems.

Figures 14a and b depict an individual house with an ICE vehicle and solar PV on an average day in 2021 and 2050, while Figures 15a and b depict an EV house with a battery in 2021 and 2050. These figures illustrate the growing quantity of energy that will be exported to the network from an individual house, if not curtailed. When this is compounded by the projected large take-up it can be seen that this will have a massive impact on the SWIS network. While the emergence of private storage somewhat ameliorates the impact of solar exports [29], this will not be sufficient to totally absorb private energy exports.

The projected network load arising from these projections is depicted in Figures 16a–16h for a typical day in Summer and Autumn (when solar generation is high and demand is normally lowest) in selected years to 2050. These indicate that network loads will likely become negative before 2030 unless private exports are curtailed.

Recently introduced rules require all new and upgraded inverters installed with a capacity of 5 kVA or greater to be capable of being remotely switched off “during an extreme minimum demand event”. However, as noted here this will become an every-day event in most months in future years. The concern of network managers in respect of large exports, whether leading to low or negative loads, relates to network stability in respect of voltage and frequency regulation [30]. However, curtailment would be a perverse outcome leading to the loss of (almost) free renewable energy to allow expensive and polluting fossil fuel generators to continue operating (for some time at least). Renewable technologies and services including generation, embedded networks and energy storage will ultimately be able to deal with the stability challenges [31], and so the following assumes that the export of private generation to the SWIS will occur eventually, even if not in the near future.

Clearly baseload generation facilities will become untenable to operate well before loads become negative, as these technologies become economically unviable at low capacity factors, and in any case older plants cannot be ramped at a rate necessary to accommodate the rapid changes in network load [32]. Accordingly, due to private exports, network generation will necessarily have to rely on large penetration of renewables and storage, irrespective of emissions policies.

Fig. 14 (a) Solar performance − ICE house in 2021. (b) Solar performance − ICE house in 2050.

Fig. 15 (a) Solar performance − EV house with battery in 2021. (b) Solar performance − EV house with battery in 2050.

Fig. 16 (a) Projected network load in Jan 2021. (b) Projected network load in Oct 2021. (c) Projected network load in Jan 2030. (d) Projected network load in Oct 2030. (e) Projected network load in Jan 2040. (f) Projected network load in Oct 2040. (g) Projected network load in Jan 2050. (h) Projected network load in Oct 2050.

3.3 Optimal network configuration

The second phase of modelling explored a range of potential renewable network generation and energy storage solutions that are compatible with large private energy exports. The model projects that the annual system supply − demand will be as set out in Figure 17. The model calibrates well with the actual SWIS network loads for 2021 identified in Figure 3.

Irrespective of the energy demand pattern and generation technology, it is most efficient to match generation to load as closely as possible. The compatibility of various renewable generation technologies with the projected network load can be illustrated by comparing average hourly generation with network load. Figures 18 to 21a–21d illustrate this for solar PV, onshore wind, offshore wind and wave technologies, with generation capacity matched to the projected 2050 load for average January and July days (i.e. total generation equals load for that day).

Given the diurnal shape of the future network load, large scale solar is not ideal, merely adding to excess generation during the middle of the day. In WA, onshore wind in fact has the best diurnal profile, as generation output is typically lower in the middle of the day. The model was used to explore annual generation and storage options and costs (excluding the costs of transmission and distribution) for these combinations.

Fig. 17 System supply.

Fig. 18 (a) Single axis solar PV cf Network load Jan day 2050. (b) Single axis solar PV cf Network load Jul day 2050.

Fig. 19 (a) Onshore wind cf Network load Jan day 2050. (b) Onshore wind cf Network load Jul day 2050.

Fig. 20 (a) Offshore wind cf Network load Jan day 2050. (b) Offshore wind cf Network load Jul day 2050.

Fig. 21 (a) Wave cf Network load Jan day 20508. (b) Wave cf Network load Jul day 2050.

3.4 2030 Network configuration

Based on the costs set out in Tables 1–3 the model produces a configuration of generation and storage that optimises for cost, i.e. produces the lowest total annual cost. For 2030, the optimal generation is around 4500 MW of onshore wind, supplemented with pumped hydro storage (PHES) and OCGTs (Case 2030-1). The diurnal supply-demand curve for a typical day in each month is depicted in Figure 22a. It is only the OCGTs that create greenhouse gas emissions in this scenario.

Additional cases include the assumption that only 2400 MW of onshore wind (half of the first case) is installed. Optimising for cost results in complementing the onshore wind with wave (Case 2030-2) − see Figure 22b and single axis PV (SAPV) (Case 2030-3) − see Figure 22c. The results are summarised in Table 4.

Fig. 22 (a) System supply and demand for an average day in each month in 2030 optimised for cost − onshore wind generation, energy storage and OCGTs. (b) System supply and demand for an average day in each month in 2030 optimised for cost − onshore wind, wave, energy storage and OCGTs. (c) System supply and demand for an average day in each month in 2050 optimised for cost − onshore wind, SAPV, energy storage and OCGTs.

3.5 2040 network configuration

To match the projected network load for 2040, the model again produces an optimal configuration (Case 2040-1) utilising onshore wind, PHES and OCGTs. Case 2040-2 assumes only half of that onshore wind is installed, giving rise to the next most cost optimal configuration in which onshore wind is supplemented by wave. A similar cost outcome is achieved with onshore wind being supplemented by offshore wind and a small SAPV capacity (Case 2040-3).

The diurnal supply-demand curve for a typical day in each month is depicted in Figures 23a–23c and the results are summarised in Table 6.

3.6 2050 network configuration

By 2050 the optimal configuration remains predominantly onshore wind, supplemented by a small amount of SAPV together with network storage (Case 2050-1). Should only half of the onshore wind be installed the next most cost optimal configuration involves wave, similar SAPV and more storage (Case 2050-2). A similar cost outcome (not shown) is achieved with a small amount of offshore wind, much larger SAPV capacity, and more storage.

By 2050, the cost of hydrogen electrolysers is projected to have fallen significantly (see Table 1) and so an additional Table 1 configuration has been considered. Case 2050-3 assumes that the excess generation is not curtailed, rather is deployed to produce hydrogen at the site of the OCGTs, which is then used as fuel when those facilities are deployed. This ‘free’ energy reduces the cost of operating the OCGTs by around 80% and turns the OCGTs into renewable energy generators.

The diurnal supply-demand curve for a typical day in each month is depicted in Figures 24a–24c and the results are summarised in Table 7.

Fig. 23 (a) System supply and demand for an average day in each month in 2040 optimised for cost– onshore wind generation, energy storage and OCGTs. (b) System supply and demand for an average day in each month in 2040 optimised for cost − onshore wind, wave, energy storage and OCGTs, (c) System supply and demand for an average day in each month in 2040 optimised for cost − onshore wind, SAPV, energy storage and OCGTs.

Fig. 24 (a) System supply and demand for an average day in each month in 2050 optimised for cost– onshore wind generation, energy storage and OCGTs. (b) System supply and demand for an average day in each month in 2050 optimised for cost − onshore wind, wave, SAPV, energy storage and OCGTs. (c) System supply and demand for an average day in each month in 2050 optimised for cost − onshore wind, H2 storage and OCGTs.

3.7 Summary of results

The results for the optimal configurations plus the 2050 hydrogen storage option are summarised in Table 7 and Figure 25.

The other configurations involving offshore wind, SAPV and storage variations are significantly more costly than the optimum solution, but in each case network costs are projected to decrease in the period to 2050.

Fig. 25 Summary of simulation results.

4 Discussion

4.1 Caveat

Like all modelling studies, this one relies on a multitude of assumptions, most importantly the continually reducing costs of renewable energy and energy storage resulting in continuing increases in the number and capacity of BTM systems, and the associated impact on network loads. Also important are the projections of decreasing costs at the network scale. However, if these assumptions prove to be optimistic, the general findings will still be largely applicable, albeit that the magnitudes of demands and capacities will be different.

4.2 Optimal renewable generation

Given the diurnal and seasonal shape of the future network load (see Figure 15), and a projected cost reduction of around 20% by 2050 [27], onshore wind is the optimal generation solution for the SWIS. Despite solar PV being projected to halve in cost by 2050, the mismatch between generation and load (see Figure 18) means significant costly storage would be required. One of the assumed benefits of offshore wind is its higher capacity factor and diurnal/seasonal consistency in output [33]. However, this consistency is not as beneficial when the network load drops so dramatically in the middle of the day in Summer and Spring (see Figure 20). Offshore wind is also projected to have the highest LCOE by 2050. There is a similar problem for wave technology in terms of generation pattern (see Figure 20) despite its projected cost decrease to 2050 (see Table 1).

Network costs for the optimal onshore wind configurations are lower (in $2022) than currently is the case (∼$85/MWh) by 15–25% through 2030–50, but the alternatives involving a lower proportion of onshore wind are marginally higher (13%) until 2050.

Given the benefits of diversifying weather-related generation, it is likely that the optimal network will involve 50-75% onshore wind, supplemented by offshore wind and/or wave, and smaller quantities of SAPV.

4.3 Role of OCGTs

The amount of extra generation and/or storage required without OCGTs means that they will remain important generators into the future, irrespective of their high LCOE. In these simulations the necessary capacity factor for optimal network performance is only around 7%, rising to a maximum of 18% for some of the alternatives. It is likely that they will remain fuelled by natural gas until the cost of hydrogen production makes this alternative fuel cost competitive around 2050. Even so, their utilisation, needed mainly in the winter, means that the overall network emissions produced are minimal, ranging from 0.04-0.13 tCO₂-e/MWh compared with the present SWIS intensity of 0.69 tCO₂-e/MWh.

4.4 Role of storage

Energy storage will be an important component of all future networks with batteries sure to play an important role [34–37], albeit that it is not yet clear which battery technology will dominate [38]. Benefits include ancillary services such as “peak shaving, improvement in voltage profile and reduction in power losses” [39]. This study indicates that storage capacity will need to reach around 1000 MWh/1000 MW by 2030 and increase to 1600 MWh/1600 MW by 2050. The simulations here lead to a preference for PHES as the cheaper alternative to battery storage. However, the difference in cost is rapidly narrowing [27], and for stability reasons it will be likely to be beneficial to have some distributed storage in the form of batteries within the network [40], potentially interfacing with private storage.

4.5 Role of hydrogen

Hydrogen production is being primarily considered as an alternative fuel for heavy transport and industrial processes. However, as set out here, the large amount of excess energy arising from BTM exports and otherwise curtailed network generation provides an opportunity to beneficially use this excess energy within the electricity network [41,42]. This would be a reasonably simple mechanism to incorporate into the market, making electrolysers a load ‘called in’ by the market operator during periods of low/negative demand in order to utilise excess energy while maintaining network stability. The electrolysers would be co-located with and operated by OCGT market participants who in any case require high-capacity transmission connections to the network. Hydrogen would be stored onsite at these facilities, removing the complications of transport, with sufficient hydrogen produced to balance supply and demand over the year [44].

However, it is unlikely that this outcome will be commercially viable until the cost of electrolysers falls significantly, projected to be around 2050 [27].

4.6 Market operations

Since electricity markets have been deregulated around the world in recent decades, complex market mechanisms have been established, trading electricity between generators and retailers at each trading interval [43]. In Western Australia, the SWIS operates differently to many markets (including the NEM) in two respects, firstly that most trading is mostly done via bilateral contracts rather than a spot market, and secondly in respect of the existence of a Reserve Capacity Mechanism (RCM) (see Table 4). Interval trading creates competition between generators, and this makes sense when variable costs (mainly fuel) are a significant component of the bid. However, the future will be dominated by renewable generators and energy storage facilities whose variable costs are very small compared to their capital costs. With variable loads and intermittent generation leading to high levels of curtailment, it will be necessary to pay facility owners for the capacity they make available, whether or not they are contributing to network supply during any particular interval, otherwise investments will be very risky. In this study the network costs have been based on $/MW capacity for generation and $/MWh capacity for energy storage. The RCM provides the opportunity to move the WEM more towards a (predominantly) capacity market over time, allocating capacity credits for both generation and storage. Such an arrangement will still ensure competition, while reducing investment risk.

5 Conclusions

The reducing cost of small-scale solar PV systems, supported by the Australian Government's RET scheme has led to a rapid take-up of private BTM generation, which will continue as capital costs continue to fall. In Western Australia, despite the already high penetration, there remain many more rooftops to accommodate such systems. The reducing cost of small-scale batteries will also lead to a take-up of private energy storage, somewhat easing, but not eliminating the impact on the SWIS network.

Although private schemes in aggregate reduce the necessary network capacity, low and eventually negative loads in the middle of the day are a challenge for system planners and managers. AEMO have consistently underestimated the rate of BTM penetration, which means that preparation for conceptualisation and implementation of the future network has been too slow. Although a phase out of most of the SWIS coal-fired power stations is planned by 2030, it is likely they will be technically and commercially unviable before then, as will any form of baseload network generation.

Given the diurnal and seasonal shape of the future network load, together with projected renewable generation costs, onshore wind energy will be the most cost optimal generation source for the SWIS, supplemented by smaller capacity offshore wind, wave and SAPV facilities. OCGTs will continue to be required to meet short term supply-demand deficits.

Network storage will be necessary to absorb excess energy from BTM imports and renewable generation, but significant curtailment will still be necessary to match supply with demand. Although PHES is the lower cost option, battery storage will likely be required within the transmission/distribution system to balance local supply-demand and maintain network stability. In the longer term, hydrogen production offers an opportunity to provide a zero emission fuel source for OCGTs.

Network generation and storage costs per MWh of network load into the future are likely to be similar to existing costs with the range of technologies considered in this study.

The main focus for future work arising from this study relates to the design of the wholesale electricity market, which is unsuited to the high capital/low operating cost generation and storage plant that will dominate future electricity networks as they transition to net zero emissions. The impact of the future mix of private and network generation/storage on transmission and distribution of electricity is another critical element requiring urgent attention.

References

All Tables

All Figures

	Fig. 1 SWIS generation 2021.
In the text

	Fig. 2 AEMO projections of private solar PV.
In the text

	Fig. 3 (a) Half hourly electricity demand and distributed PV (DPV) generation. (b) Resulting half hourly SWIS network loads.
In the text

	Fig. 4 Model configuration.
In the text

	Fig. 5 Electric vehicle projections.
In the text

	Fig. 6 SWIS Demand projection.
In the text

	Fig. 7 Take-up of private solar PV and battery systems.
In the text

	Fig. 8 Projected capital cost of residential and business solar PV. 11
In the text

	Fig. 9 Projected capital cost of residential and business battery systems.
In the text

	Fig. 10 Household payback periods.7
In the text

	Fig. 11 Business payback periods.
In the text

	Fig. 12 (a) Penetration of solar PV and batteries in ICE households. (b) Penetration of solar PV and batteries in ICE business premises.
In the text

	Fig. 13 (a) Private solar PV capacity. (b) Private battery capacity.
In the text

	Fig. 14 (a) Solar performance − ICE house in 2021. (b) Solar performance − ICE house in 2050.
In the text

	Fig. 15 (a) Solar performance − EV house with battery in 2021. (b) Solar performance − EV house with battery in 2050.
In the text

	Fig. 16 (a) Projected network load in Jan 2021. (b) Projected network load in Oct 2021. (c) Projected network load in Jan 2030. (d) Projected network load in Oct 2030. (e) Projected network load in Jan 2040. (f) Projected network load in Oct 2040. (g) Projected network load in Jan 2050. (h) Projected network load in Oct 2050.
In the text

	Fig. 17 System supply.
In the text

	Fig. 18 (a) Single axis solar PV cf Network load Jan day 2050. (b) Single axis solar PV cf Network load Jul day 2050.
In the text

	Fig. 19 (a) Onshore wind cf Network load Jan day 2050. (b) Onshore wind cf Network load Jul day 2050.
In the text

	Fig. 20 (a) Offshore wind cf Network load Jan day 2050. (b) Offshore wind cf Network load Jul day 2050.
In the text

	Fig. 21 (a) Wave cf Network load Jan day 20508. (b) Wave cf Network load Jul day 2050.
In the text

Fig. 22 (a) System supply and demand for an average day in each month in 2030 optimised for cost − onshore wind generation, energy storage and OCGTs. (b) System supply and demand for an average day in each month in 2030 optimised for cost − onshore wind, wave, energy storage and OCGTs. (c) System supply and demand for an average day in each month in 2050 optimised for cost − onshore wind, SAPV, energy storage and OCGTs.

In the text

Fig. 23 (a) System supply and demand for an average day in each month in 2040 optimised for cost– onshore wind generation, energy storage and OCGTs. (b) System supply and demand for an average day in each month in 2040 optimised for cost − onshore wind, wave, energy storage and OCGTs, (c) System supply and demand for an average day in each month in 2040 optimised for cost − onshore wind, SAPV, energy storage and OCGTs.

In the text

Fig. 24 (a) System supply and demand for an average day in each month in 2050 optimised for cost– onshore wind generation, energy storage and OCGTs. (b) System supply and demand for an average day in each month in 2050 optimised for cost − onshore wind, wave, SAPV, energy storage and OCGTs. (c) System supply and demand for an average day in each month in 2050 optimised for cost − onshore wind, H2 storage and OCGTs.

In the text

	Fig. 25 Summary of simulation results.
In the text