Deployment Potential of Concentrating Solar Power Technologies in California

Abstract

As states within the United States respond to future grid development goals, there is a growing demand for reliable and resilient nighttime generation that can be addressed by low-cost, long-duration energy storage solutions. This report studies the potential of including concentrating solar power (CSP) in the technology mix to support California’s goals as defined in Senate Bill 100. A joint agency report study that determined potential pathways to achieve the renewable portfolio standard set by the bill did not include CSP, and our work provides information that could be used as a follow-up. This study uses a capacity expansion model configured to have nodal spatial fidelity in California and balancing-area fidelity in the Western Interconnection outside of California. The authors discovered that by applying current technology cost projections CSP fulfills nearly 15% of the annual load while representing just 6% of total installed capacity in 2045, replacing approximately 30 GWe of wind, solar PV, and standalone batteries compared to a scenario without CSP included. The deployment of CSP in the results is sensitive to the technology’s cost, which highlights the importance of meeting cost targets in 2030 and beyond to enable the technology’s potential contribution to California’s carbon reduction goals.

1. Introduction

The renewable portfolio standard (RPS) is a regulation implemented by a subset of states since the 1990s that requires the utilities providing electricity to end users to procure a minimum proportion of energy from renewable sources for one or more target years [1]. California’s RPS program was established by Senate Bill 1078 in 2002 [2], and in response to its ability to meet 2020 RPS milestones ahead of schedule, the state enacted Senate Bill 100 (SB100) in September 2018 to increase the statewide RPS for electric utilities from 50% to 60% by 2030 and establish a goal of 100% clean energy by 2045 [2]. To assess potential pathways to achieve the standards imposed by the legislation, the California Energy Commission released a joint agency report that analyzed the potential large-scale deployment of clean energy technologies to replace retiring fossil-fueled plants throughout the state to achieve the RPS goals for both 2030 and 2045, including adoption of solar and storage behind the meter to supplement additional capacity from utilities [3].

The analysis in the joint agency report [3] discovered that there were multiple pathways to expand electricity capacity to achieve the RPS established by SB100. Using the RESOLVE California model to develop a least-cost resource portfolio to meet the RPS, the study showed that a significant resource build-out was necessary to meet 2045 goals, with average capacity installations nearly equal to the 1-year historical high installation in California for utility-scale wind and solar, as well as battery capacity that exceeds the historical maximum. The variability of renewable resources leads to a significant energy storage requirement, and the analysis suggests that natural gas be available to ensure an uninterrupted power supply during the transition to 100% clean energy.

The SB100 study did not consider CSP in its capacity expansion estimates. CSP plants use sun-tracking mirrors to redirect sunlight to a central location, in which the absorbed heat can be deployed to generate electricity or sent to a thermal energy storage (TES) system for later use [3,4]. Although the significant cost reduction in photovoltaics (PV) over the past 20 years makes it the preferred option for new solar installations, the dispatchability of CSP with low-cost TES compared to batteries can make it a cost-competitive option for complementing variable generation [5] and providing ancillary services to the grid [6,7,8] in addition to providing firm capacity to the grid [9]. While deployment of CSP in the United States has been limited in the past decade, many plants are planned and under construction internationally, particularly in China, which has more than 30 projects either planned, deployed, or under construction [10]. The hypothesis behind the present study is that retirements of fossil-fueled plants and increasing renewable energy deployments in the later years of the SB100 timeline will require technologies with longer-duration storage, which will make CSP with TES systems economical in certain regions compared to battery technologies paired with PV and wind.

The primary goal of this study is to model SB100 while including CSP as an option for deployment in the capacity expansion model to determine what role CSP could play in California specifically and the Western Interconnection (WI) in general when it is included in model scenarios. The study also assesses the best locations for CSP in California, the role it plays in the SB100 energy landscape, and the sensitivity of deployments to cost assumptions. The hypothesis for this study, established in prior work [11], is that the combination of aggressive clean energy goals and excellent solar resource in California makes it a promising location to include CSP with TES in capacity expansion planning.

To perform this study, the authors used the National Renewable Energy Laboratory’s (NREL’s) Resource Planning Model (RPM), an integrated resource planning and dispatch tool for modeling regional electric systems, to determine the build-out of CSP and other generation technologies over time and in compliance with SB100 goals. NREL uses RPM to model the expansion of power generation technologies in California and the entire WI [12,13,14,15]. The RPM model can be used to evaluate what technologies are built in which regions to meet different policy objectives, where transmission will be expanded, and the cost and design assumptions required for CSP to be deployed. This level of regional-specific detail differentiates RPM relative to other capacity expansion models such as RESOLVE, the model used to produce the results in the SB100 study.

Prior work has assessed the potential of deploying CSP technologies using various methods. Ammari et al. studied the deployment potential of a parabolic trough CSP plant in Morocco by comparing the levelized cost of energy to energy market pricing [16]. Brumana et al. developed a study of a potential high-renewable (i.e., ≥90%) hybrid energy system for a remote community in the Middle East and North Africa region and found that including CSP with long-duration TES reduced system costs compared to an all-electric storage scenario [17]. Li et al. presented a case study of CSP paired with coal and combined heat and power plants within a broader energy system and show that CSP with TES can reduce renewable energy curtailment as well as total system cost in instances where the renewable penetration is high. These studies are useful in determining economic viability as a standalone project but lack the context of electricity markets. Qualitative studies have assessed potential barriers to CSP deployment and potential policies to mitigate them in the United States [18], Europe [19,20], Iran [21], and globally [22].

Production cost models have been employed to assess the economic potential of CSP in energy markets as well, which include lower-fidelity models than CSP price-taker models such as those in [23,24], but provide systems context and enable energy market participation [25]. Norambuena-Guzman et al. presented a review of power system capacity planning models that incorporate CSP with TES [26]. Narimani et al. explored the value of CSP systems in the Australian national electricity market using production cost simulations in PLEXOS under both a generator short run marginal cost-based stack model and a generator bid stack model [7]. Madaeni and Sioshansi estimated the added capacity value of CSP when TES is included for a system deployed in the Southwestern United States [27]. Borge-Diez et al. showed the potential of CSP to stabilize grid prices due to the flexibility of scheduling output when TES is incorporated into the system.

The novelty of our contribution is the inclusion of CSP into a large-scale capacity planning model represented at high spatial fidelity over a long time horizon for the Western Interconnection, combined with our focus on developing system plans with and without CSP in the technology mix and comparing the system-level planning over time. Our proposed extension of the SB100 study including CSP has a similar representation of SB100 to what was included in NREL’s LA100 study [28]. Specifically, our analysis includes (1) annual targets for the state of California, (2) all eligible technologies represented within RPM, (3) representation of a sales-based RPS, and (4) renewable energy certificate (REC) tracking and enforcement of limitations by REC category.

The rest of this report includes the following: Section 2 details the methodology of our study and the key assumptions associated with the SB100 model. Section 3 details the results of the study in which we compare capacity expansion plans with and without CSP available as a technology option. Section 4 provides our conclusions.

2. Materials and Methods

This section describes our methodology by introducing RPM (v. 2018.01), discussing model inputs and key assumptions behind our analysis, and describing the model scenarios that we adopt to showcase the potential for CSP in capacity expansion planning to meet the California RPS determined in SB100.

2.1. Resource Planning Model

RPM is one of NREL’s utility-scale capacity expansion models [29]. The main advantage that RPM has over other capacity expansion models is that it can model the region of interest at a nodal level rather than zonal. The model obtains the least-cost system expansion (e.g., capacity, transmission, and reduced-order dispatch) decisions up to the year 2050 in 5-year increments at annual fidelity.

RPM is used to model a single regional grid (i.e., Western or Eastern Interconnection). In this study, we are modeling the Western Interconnection. Many studies have modeled this interconnection [12,13,15,28,30]. Within the focus region, RPM models everything at the transmission nodal level, whereas the rest of the Western Interconnection (outside of the focus region) is modeled zonally. Each zone is analogous to a Western Electricity Coordinating Council (WECC) balancing area. The nodal-zonal structure used in this study is shown in Figure 1. For more details on the overall structure and typical data sources of the model, see [31]. Recent improvements in modeling methods for storage, CSP with thermal storage, and PV plus battery storage systems are described in [32,33].

The overall structure of the RPM algorithm is shown in Figure 2. The right of the diagram (Investment and Dispatch Co-Optimization) shows the least-cost optimization problem at a high level. The optimization problem is a linear or mixed-integer program that minimizes the overall system costs, such as capital costs, fixed operations and maintenance costs, variable operations and maintenance costs, and fuel costs subject to the constraints listed in the lower half of the box on the right in Figure 2. RPM reduces the temporal dimension of its optimal dispatch by modeling five single-day dispatch periods classified according to load or variable-generation resources (e.g., low, mid, high, low variable generation, and peak) at an hourly resolution within each model year. The constraints on the system at each hour include load balancing, maintenance of reserve capacity (i.e., planning, spinning, flexibility, and regulation reserves), and operational bounds for individual generators and transmission paths. There are also constraints that link consecutive hours within each dispatch period, as well as consecutive dispatch periods, to capture limits on ramp rates and reconcile the state of charge for each energy storage system. The latter applies to stand-alone batteries, PV plus batteries, pumped storage hydropower, and concentrating solar power with TES. The ramp rates inform both operational constraints in RPM decisions and the ability of a technology to provide spinning reserves; operating reserve margins are set to 23% to be consistent with the LA100 study [28]. CSP ramp rates are modeled in the production cost modeling in PLEXOS that validates the viability of the RPM results.

The methods described on the left-hand side of Figure 2 estimate capacity credits and curtailment impacts for variable-generation and energy-limited resources such as storage using hourly data. Capacity credits for variable renewable energy resources are determined by calculating load duration curves and net load duration curves with the existing grid build-out, then assessing the impact to the net load duration curve after adding incremental capacity of a given technology. The method used to calculate capacity credits is summarized in Figure 3 and detailed calculations are available in Zhou et al. [34]. The authors acknowledge the frequent use of similar capacity measures that are time-dynamic, such as effective load carrying capacity [35], and use the method described above to remain consistent with the recent LA100 study [28].

A separate production cost model, PLEXOS, is used to validate the viability of the procurement decisions over the course of the given year, and a separate probabilisitic model, the Probabilistic Resource Adequacy Suite (PRAS) [36], is then run to simulate simplified system operations under randomly drawn generator outages. PRAS calculates probabilistic metrics to quantify the risk of failing to serve demand due to insufficient resource availability. These metrics include loss of load expectation (LOLE), and RPM is reoptimized with new constraints to represent previously unmet stress periods if the required LOLE target is not satisfied.

These estimates are key inputs into the least-cost investment and dispatch problem, because they improve the valuation of these resources as potential investments when only 5 days’ worth of data are used to represent annual operations. These key parameters, along with the supporting maximum-capacity credit dispatch profiles, are shown in the blue “parameters” box in Figure 2 as being fed into the optimization problem to ensure that all technologies’ flexibility is appropriately valued.

2.2. Model Input Assumptions

Capacity expansion model results depend on input data and assumptions. This section discusses the data inputs to the RPM model and the assumptions used for each scenario.

2.2.1. Baseline System Description

The model runs in 5-year increments from 2010 to 2050. From 2010 through 2020, RPM builds the existing grid, transmission, and load for the Western Interconnection through “prescribed builds.” New generation builds by the model—along with known projects under construction—start in 2025 for energy and energy storage technologies except for fossil-based, nuclear, and hydropower technologies. Fossil-based technologies are excluded in California by our SB100 assumptions but permissible outside of the state. Similarly, we do not allow the model to build new nuclear or hydropower to be consistent with the SB100 study. Rather, for this project, we are running the model until 2045. The fuel prices are based on the U.S. Energy Information Administration’s Annual Energy Outlook 2022 Reference case [37]. The technology cost assumptions are based on NREL’s 2022 Annual Technology Baseline (ATB) [38]. RPM also includes a production tax credit and investment tax credit that align with the Inflation Reduction Act and the 2022 Standard Scenarios [39].

NREL worked to develop the RPM focus region (nodal model) for this study, which is shown on the left of Figure 1, whereas the zonal representation of WECC used for import and export of electricity between California and other parts of the interconnection are shown on the right. Based on study objectives, the focus region includes the area covered by Pacific Gas and Electric, Southern California Edison, San Diego Gas and Electric, Imperial Irrigation District, and Los Angeles Department of Water and Power (LADWP). We included only transmission lines at 220 kilovolts and above to make model runtime reasonable. Additionally, portions of northern California were left out of the nodal representation to reduce computational time because we did not expect that CSP would play a role in that region due to the relatively low solar resource compared to other parts of the state. We enforced a realistic set of transmission parameters, including DC optimal power flow and planning reserve margins. We collected RPS data used in RPM from the Database of State Incentives for Renewables and Efficiency (https://www.dsireusa.org/, accessed on 18 September 2025) [29,31], which are summarized in Figure 4 for California.

2.2.2. Electricity Generation Technologies and Costs

The 2022 NREL ATB provides current and future costs for all technologies, including CSP [38]. CSP plants are assumed to be 100 megawatt electric (MWe) in size using molten-salt power tower CSP technology, in which TES is always included. The moderate cost scenario (from the ATB) is assumed for all technologies for our study as shown in Figure 5. Note that the costs for CSP from ATB are for the default configuration that includes 10 h of TES, whereas PV plus battery systems have either 4 or 8 h of storage available. CSP capital expenditure changes depend on the configuration assumed (see Section 2.3). CSP exhibits the largest cost reduction between 2020 and 2030 of the candidate technologies for expansion planning in this study. Recent worldwide project installations indicate that industrywide costs are trending downward, but cost reductions may need to accelerate to meet projections for 2030 [40]. The assumption of future cost reductions is consistent with other studies of CSP’s deployment potential abroad, such as the work of Viebahn et al. [41].

2.2.3. SB100 Policy and Regulatory Assumptions

The RPS target assumptions in the model are 60% by 2030, 90% by 2035, 95% by 2040, and 100% by 2045, as shown in Figure 4. The RPS targets ensure that a specific percentage of the electricity provided by utilities comes from clean energy resources. The resources in the model that are qualified to contribute to the RPS are renewable combustion turbine, geothermal, hydropower, CSP, utility PV plus battery, utility PV, offshore wind, land-based wind, and nuclear; however, to match the SB100 study, new generation is restricted to the technologies summarized in Figure 5. RECs are allowed to contribute to meeting RPS targets, subject to a few limitations. RECs are issued when a megawatt hour is generated and delivered to the power grid from renewable energy sources. These RECs can be purchased either as stand-alone (unbundled) or bundled with energy delivery. Consistent with SB100, our model only allows bundled RECs for electricity consumed within California, though these bundled RECs may be purchased from outside the state. Unbundled RECs, which allow fossil fuel consumption to be balanced by a purchased REC from elsewhere and at a different time, are not allowable within California per SB100, but are modeled in other states in the WECC region as permitted by state-level regulations.

California’s SB100 legislation states that the RPS applies to end-use consumption. A consequence of the language in the legislation is that non-RPS-compliant generation sources can be used in California to offset transmission and distribution losses. RPM uses end-use consumption in California to validate that RPS quotas are met and therefore allows non-RPS generators to contribute to the California grid if their contributions do not exceed transmission and distribution losses. This provision allows the limited generation by fossil-fueled plants in 2045 shown in the results in Section 3.

2.3. Model Scenarios

The goal of this study is to demonstrate the deployment potential of CSP technologies in California when it is included in capacity expansion models under SB100. The main runs for this study include two core scenarios: one in which CSP technologies are included in the model, and one in which CSP technologies are not allowed to deploy. The difference in capacity deployment of and generation from CSP plants between the two scenarios demonstrates the impact of including CSP technologies in the grid analysis. The generation profiles also illustrate how CSP is being used by the grid in the results that follow, and the nodal analysis shows potential deployment locations in California for CSP plants.

CSP deployment potential depends heavily on the configuration assumed for the plant. CSP plants can be configured to meet a range of grid needs, from directly generating electricity for the grid (similar to a solar PV plant) to using TES to provide peaking or load-following capabilities. The CSP plant configuration is determined by two factors: the solar multiple (SM) and the TES duration. The SM is the thermal-delivery capacity of the heliostat field relative to the thermal demand of the plant’s power turbine. An SM greater than one means that the heliostat field should at times produce more thermal energy than the power plant can use. Excess thermal energy can then be put into available storage.

In existing CSP plants that include TES, the thermal energy is stored in large tanks of molten nitrate salt. Stored thermal energy from the salt in the tanks can run the power plant regardless of solar conditions. The TES duration is the maximum amount of time (in hours) that the plant can operate using only storage (assuming the tanks start full). The SM and TES of a system influences the system’s total cost and levelized cost of electricity, with a higher SM (larger heliostat field) and higher TES duration (larger molten salt solar tanks) incurring higher project capital costs and yielding a higher capacity factor for the plant. The SM and solar resource determine how much energy is available for storage on a given day. If the SM is too low, then the TES tanks will not fill to capacity frequently and will not be fully utilized. If the SM is too high, then the field will generate more thermal energy than the tanks can store and excess thermal energy that could have been generated will be lost (curtailed). Grid operators dispatch the CSP plants along with other resources to satisfy demand at the lowest cost, with the timing and duration of CSP dispatch varying with the TES capacity as well as grid needs. Describing CSP plant configurations that best serve the California grid under SB100 is a key outcome of this study.

Costs for CSP configurations were calculated using a correlation developed as part of the ATB to also allow NREL capacity expansion models to examine various CSP configurations [38]. The correlation is based on the breakout of the turbine, storage, and heliostat field capital costs for the base case (SM = 2.4, TES = 10 h) used in the ATB, which are calculated using the System Advisor Model (SAM) [42]. We then validated the correlation using the SAM model for a range of SM and TES values by letting SAM optimize the heliostat field layout, tower height, and receiver dimensions, simulating and recording the results, and comparing the resulting overnight capital costs (OCC) to the correlation values. The correlation assuming a 100 MWe CSP plant is given in Equation (1):

$$\begin{array}{ccc}\hfill & \mathrm{OCC}=\mathrm{turbine}\mathrm{cost}+\mathrm{TES}·\mathrm{storage}\mathrm{cost}+\mathrm{solar}\mathrm{multiple}·\mathrm{field}\mathrm{cost}\hfill & \hfill \\ & \mathrm{OCC}(\$/\mathrm{kW})=\$1910/\mathrm{kWe}+TES\left(h\right)·\$77/\mathrm{kWhe}+\mathrm{SM}·\$1486/\mathrm{kWe}.\hfill \end{array}$$

(1)

For this study, preliminary model runs showed that the CSP without TES did not deploy (as expected) so that configuration was not included in the full model runs.

Table 1. Summary of eight candidate CSP configurations included in core scenarios.

Solar Multiple	TES Duration (Hours)
1.0	6/9
2.0	9/12/15
3.0	12/15/18

summarizes the eight candidate CSP configurations considered in the core scenario. The configurations range from SM 1.0/TES 6 h, which act as a flexible generator with low-capacity factor similar to a peaking power plant, to SM 3.0/TES 18 h, which behave similar to a load-following plant. The RPM model selects the configuration that is most competitive depending on grid demands. Specifically, RPM solves a linear program to obtain a minimum-cost collection of investment and operations decisions that meet a prespecified demand, in which the costs include transmission expansion, new capacity investments, fuel, and operations and maintenance.

2.4. Resource Potential

The goal of the RPM analysis is to evaluate specific zones of interest for CSP development and evaluate under what assumptions CSP projects could be deployed to meet the SB100 goals. To accomplish this, NREL performed a spatial analysis of the region using geographic information systems to determine the siting and resource potential for CSP. Specifically, we used the National Solar Resource Radiation Database (NSRDB) to obtain historical solar resource time-series inputs to the different nodes and zones in RPM; the NSRDB has solar resource and weather data at the spatial fidelity of a 2 km-by-2 km grid [43]. The variable generator profiles (e.g., wind, solar PV, and CSP) are generated using the Renewable Energy Potential Model with weather data from the NSRDB to generate capacity factors profiles (by percentage), aggregated to a nodal or zonal level [44]. We chose 2012 for sourcing weather data to be consistent with the LA100 study [28]. Similar to prior RPM case studies, resource potential (in squared kilometers of land area), system performance (i.e., annual and hourly capacity factors), and grid interconnection distances were used to characterize the solar or wind resources available for capacity expansion in each region [31].

3. Results

This section summarizes the results of the RPM instances described in Section 2 to obtain capacity expansion plans with and without CSP as an option, comparing generation and capacity installation in California. In addition to the baseline or “midline” case, which employs the technology cost and performance assumptions summarized in Section 2, we perform a sensitivity analysis on the cost of CSP to determine the elasticity of its adoption in capacity expansion planning under higher technology cost scenarios.

3.1. Summary of RPM Results in California

Figure 6 shows the installed capacity from 2020 to 2045 for the No CSP and the CSP core scenarios. Given that the model is not allowed to build new fossil-based technologies, in the No CSP scenario, we see that the model relies mostly on variable-generation technologies, battery storage, and geothermal to meet the increasing demand as we approach 2045. We see with the CSP scenario that when it is allowed, the model selects CSP to displace variable generation and batteries, likely because of the high-capacity factor and relatively low cost of TES compared to batteries.

Similarly, Figure 7 shows the annual generation from 2020 to 2045 for the No CSP and CSP scenarios. The role of CSP is more evident in Figure 7, where it makes up about 15% of all generation in California by 2045, compared to Figure 6 where CSP is only about 6% of total installed capacity in California by 2045. The 9.1 gigawatt-electric (GWe) of CSP displaces about 30 GWe of other technologies in 2045 because of its higher capacity factor compared to other renewable technologies.

The RPM results for technologies other than CSP shown in Figure 5 and Figure 6 show some insights into how SB100 is implemented by RPM. First, although the goal of SB100 is to minimize emissions from fossil-fuel generators, the RPM results show that these technologies stay on the grid. Figure 6 shows that the installed capacity of fossil-based technologies is relatively constant after the coal fleet is retired in 2025, i.e., the results show limited retirement of natural gas assets. However, Figure 7 shows that the annual generation of these technologies drops from 2020 to 2045, diminishing progressively to comply with SB100 RPS goals. We interpret this as the model utilizing natural gas assets to provide critical reserves at a relatively low fixed cost, while the limited energy production offsets system losses as is allowed by SB100.

Figure 8 shows the capacity difference between the CSP and the No CSP runs. A positive value in the plot denotes a relative increase in a technology’s deployment in the CSP run, and a negative value in the plot denotes a relative decrease in a technology’s deployment in the CSP run compared to the No CSP run. When CSP technologies are included, RPM first builds about 2 GWe of installed CSP capacity in 2030. The amount of installed CSP grows steadily after that, reaching a value of 9.1 GWe by 2045. Figure 8 shows that CSP mostly replaces wind, stand-alone solar PV, and stand-alone battery storage when it is included, with reductions in capacity of approximately 22%, 29% and 38% respectively; in total, the 9.1 GWe of CSP displaces about 30 GWe of other technologies.

Figure 9 shows the annual generation differences by technology. The results show that CSP generates approximately 60 terawatt hours electric (TWhe) of energy in 2045. The technology primarily displaces wind and standalone PV systems which produce approximately 26% and 33% less energy respectively; additionally, storage losses are reduced by approximately 50% relative to the No CSP scenario in 2045. While total systemwide costs are similar, California’s electricity imports from other states in the WI are reduced by approximately 2 TWh annually (or approximately 24%) in 2045 the CSP scenario.

Figure 10 shows annual generation by region in California under SB100 for both core scenarios, whereas Figure 11 shows CSP deployments by region for the CSP-included scenario. CSP deploys in four of the six regions modeled in California, with most deployments occurring in Southern California Edison and Pacific Gas and Electric Valley and a small amount in San Diego Gas and Electric and Imperial Irrigation District. The LADWP region includes new generation from a combined-cycle natural gas plant built in 2025 that offsets transmission and distribution losses in out years, which differs from the projections outlined in the LA100 study [28] in part because we only enforce clean generation goals at the state level in our study. However, the lack of CSP deployment in LADWP shown in Figure 10 is consistent with LA100.

Another notable outcome is that of the eight configurations considered. In California, RPM only one was deployed—the SM3.0/TES 12 h configuration. However, other configurations were deployed in other parts of WECC. A CSP plant with 12 h of storage capacity acts like a load-following generator on the grid. This is confirmed by Figure 12, which shows the aggregated charging as dispatch for all CSP installed capacity in California under SB100 in RPM for 2045. The CSP plants charge during the day at rates higher than the plant’s installed capacity due to the high SM in the deployed configuration. The CSP plants operate at close to maximum capacity during the afternoon and evening hours for all representative days. This is the time of day when generation from solar PV is falling while consumer demand is increasing, leading to a severe ramp in nonvariable generation requirements. The RPM results show that CSP plays a substantial role in supporting the grid during this critical time of day. Except for the peak day in Figure 12, CSP generation is suppressed during each representative day’s midday because of the abundance of solar PV generation at that time.

3.2. CSP and Combustion Generator Results Outside of California

The results of Section 3.1 show that achieving the clean electricity goal established by the RPS via SB100 is achievable with generation within California. Because of the lower RPS in other states, the 2045 RPM results include installations of 120 GWe of natural gas combustion turbines and 28 GWe of combined-cycle plants in WECC outside of California. The electricity imports to California from outside the state combined with the natural gas resource capacity inside the state provide the buffer of carbon-emitting generators to offset system losses. This result is consistent with the findings in the joint agency report, which estimates about 11 million metric tons of annual greenhouse gas emissions attributable to the grid in a “combustion-free” scenario (see Figure 6 in [3]).

Figure 12. Aggregated dispatch of all CSP plants deployed in California under SB100 for the five representative load days modeled in RPM for 2045.

Figure 12. Aggregated dispatch of all CSP plants deployed in California under SB100 for the five representative load days modeled in RPM for 2045.

3.3. Sensitivity to CSP Technology Cost

To understand the impact of CSP technology cost on its adoption in RPM under the RPS guidelines established by SB100, we performed a sensitivity analysis to obtain capacity expansion decisions when CSP pricing is 10% and 20% higher than the baseline cost over time.

Table 2. Summary of CSP deployments in 2045 from RPM results to meet the California RPS established by SB100 as a function of CSP technology costs at three different levels.

CSP Installations	Baseline (GWe)	+10% (GWe)	+20% (GWe)
California	9.1	3.9	2.2
Western Interconnection	4.5	3.2	1.4
Total	13.6	7.1	3.7
% of Baseline	100%	52%	27%

summarizes RPM’s prescribed installed capacity and annual generation in California as a function of CSP technology cost. The results show that CSP deployment is highly elastic to the technology cost, as 2045 generation capacity decreases to less than one-third of that of the baseline case when the cost increases by 20%.

3.4. CSP Project Siting and Transmission Discussion

The structure of RPM, including nodal representation of California’s transmission system and five hourly representative days, allows load and generation to be represented with high spatial and temporal resolution. This, overlaid with CSP resource potential, allows RPM to select optimal locations for new generating resources. Figure 13 shows the locations in California where new CSP generation facilities interconnect to the grid in the RPM instance with CSP included at the baseline cost assumption. Note that locations outside of California are simply located at the geographic center of the balancing area. The deployment of CSP systems in California is generally located within the existing transmission network as shown in Figure 13, and the dispatch of the systems primarily in the evening, nighttime and early morning shown in Figure 12 largely complements the PV production in the daytime along existing transmission infrastructure. As a result, the investment in the CSP scenario is approximately 21% lower than in the No CSP scenario under baseline assumptions.

While some deployment outside of California is present in Idaho and Nevada, there is a notable lack of deployment in Arizona and New Mexico despite high solar resource potential and relatively low construction costs in the area. Results suggest the lack of deployment in Arizona and New Mexico may be related to low RPS targets and low-cost wind resources, respectively.

The results in Figure 13 suggest that southern California and southern Nevada exhibit potential for CSP deployment to support California RPS goals. However, while the model considers the cost and resource potential of each region, more detailed analysis at a local level is required to determine the feasibility of siting, additional costs associated with environmental impact mitigation, and other deployment factors not considered in RPM.

4. Discussion

4.1. Comparison to Previous Work

In addition to the results of the RESOLVE model in the joint agency report, similar capacity expansion plans have been developed as part of the LA100 study, which focuses exclusively on the LADWP region [28]. Both studies highlight that wind and PV comprise most new generation assets when planning to meet the California RPS and that load flexibility strategies can offset a significant part of expensive energy storage investments. We observe similar results using RPM. There is no deployment of CSP in LADWP in our results, which is consistent with those of the SB100 scenario of the LA100 study. Additionally, the SB100 scenario of the LA100 study allows limited natural gas-fired electricity production and as a result includes some gas generation in 2045, like our study. The RESOLVE model runs excluded CSP as an option and had a similar build-out to that of our study in the No CSP scenario, except for the presence of natural gas in our analysis to offset system losses. No offshore wind is built in any of our RPM scenarios, which may be attributable to high technology cost compared to land-based wind. The LA100 study also lacks offshore wind deployment, except for the sensitivity in which offshore wind costs are reduced; see Table 19 and Figure 48 of [28]. More broadly, the notion that CSP with TES exhibits value in its ability to provide nighttime dispatch in high-renewable scenarios is consistent with other prior work [6,9,45].

4.2. Discussion of Cost Projections

The results in this study suggest that CSP has potential to contribute to California’s RPS goals established in SB100 at the technology cost projections outlined in the 2022 Annual Technology Baseline. The availability of CSP at assumed costs in the ATB assumes significant cost reductions compared to current costs; scaling up the technology’s deployment to achieve these cost reductions may require significant support and technology development as noted in a study by Lillestam et al. [46]. The cost reduction assumptions in the ATB are under the “Moderate Scenario” which is in line with the forecast scenario employed for all other technologies in this study. For CSP, the moderate scenario includes several improvements to the technology by 2030, including the use of advanced receiver coatings, improvements to the performance and reliability of the thermal energy storage system, heliostat cost reductions, and the use of a supercritical CO₂ cycle at current molten-salt operating temperatures at plants like Crescent Dunes. While some of these improvements, such as that of the power cycle, are unlikely to be deployed at scale in 2030 given the current state of the technology, the ongoing deployment of CSP globally has led to a downward trend in capital costs; a recent IRENA cost analysis calculated a 46% year-over-year reduction in LCOE of CSP projects in 2024 [47], suggesting that continued deployment of CSP may lead to the forecasted cost reduction in the ATB if the United States is able to realize some of the technology improvements experienced internationally in recent years. The technologies displaced by new CSP capacity in the CSP scenario of this study compared to the No CSP scenario, primarily utility-scale wind and PV with battery storage, are likely to be deployed instead of CSP if either (a) the cost reductions are not realized for the CSP technology or (b) the cost reductions in the displaced technologies exceed their forecasts such that they provide the services of nighttime and peak generation that CSP offers, but at relatively lower cost.

There are many other considerations related to the deployment of CSP that are not included in this analysis, such as environmental impact, fire danger, and other siting and permitting requirements. The results in this study are not intended to prescribe future policy decisions or predict future capacity expansion, but rather to provide insights into CSP’s potential for deployment when the technology is available.

5. Conclusions

California’s joint agency study to achieve planning goals for the electricity grid in the state by 2045 excluded CSP as a generation technology option because of their assessment of deployment experience and elevated costs. We present a study of CSP’s potential to support California’s SB100 goals using RPM, a capacity expansion model that provides high spatial fidelity in California and balancing-area spatial fidelity in the rest of the Western Interconnection. We first demonstrate similar results to the joint agency study when we exclude CSP from the RPM model. Next, we use projections of current and future technology costs from NREL’s 2022 ATB, including CSP, to show that CSP could provide a significant proportion of the technology mix starting in 2035 and approximately 9.1 GWe of CSP deployed in California by 2045. CSP provides dispatchable solar energy, capable of generating in off-peak hours, as evidenced by the predominant build-out of CSP plants with 12 h of TES. These plants replace some 30 GWe of wind and solar when compared to scenarios that exclude CSP.

The model selection of CSP in out-year capacity expansion in California is sensitive to the technology’s cost, highlighting a need for the CSP industry to attain the necessary deployment experience and technology development to meet cost targets in 2030 and beyond. Achieving the cost reductions in the projections assumed in this study (i.e., in the 2022 ATB) requires the development of a stable supply chain of CSP projects to realize the projected cost reductions. Additional performance improvements currently under research and development, such as high-temperature particle storage technology [48] and the U.S. Department of Energy’s Heliostat Consortium [49], may also contribute to CSP’s cost competitiveness. Results suggest that CSP remains part of a potential solution for dispatchable solar energy to meet California’s clean energy targets in SB100, even under cost increases of up to 20% from the baseline case. This is consistent with previous findings that if CSP costs decline according to projections, deployment may expand across the southern United States [50].