How the model builds future supply stacks for the WECC region
A techno-economic capacity expansion model builds the WECC generation stack from 2027 to 2050. It starts from today’s fleet and adds capacity year by year.
The starting fleet is the EIA-860 inventory of existing units: those operating, on standby or temporarily out of service. Planned projects enter through the interconnection queue instead. Committed queue projects are built on online dates adjusted for expected delays. The rest of the queue sets near-term build limits.
Economic build decisions start in 2027. For each candidate technology at each node, the model weighs project economics against the cost of running the system without it.
Each year, investment decisions are optimized together with an hourly dispatch of new and existing assets. A generator is built if it lowers the total cost of running the system.
The capacity expansion model and the dispatch model share the same inputs. These include generator technical parameters, fuel and carbon prices, renewable profiles, outages, transmission limits, wheeling charges and the emergency reserve curve. Build decisions therefore respond to the same market the dispatch model sees, not a simplified proxy.
Read more about the core capacity expansion methodology.
At a glance
| Parameter | Description |
|---|---|
| Geography | WECC: CAISO (as the PG&E, SCE and SDG&E zones) and neighboring balancing areas |
| Horizon | Annual decisions from 2027 to 2050 |
| Starting fleet | EIA-860 existing units, plus committed queue projects built on delay-adjusted online dates |
| Temporal resolution | Hourly dispatch on 30 representative days per model year |
| Windowing | Rolling one-year optimization (co-optimized investment and dispatch, with state carried into the next year) |
| Reliability | CAISO: slice-of-day resource adequacy pooled across its three zones. Other balancing areas: seasonal planning reserve margins aligned to the Western Resource Adequacy Program (WRAP) |
| Policy | California and Washington carbon prices, the California emissions cap, state renewable and clean energy standards |
| Outputs | Annual capacity additions and retirements by zone and technology, reliability metrics, and dispatch time series for sampled days |
Building a future generation stack
Technologies governed by economics are included as build candidates in the model:
- Solar PV
- Onshore wind
- Geothermal
- Gas CCGT (combined cycle gas turbine)
- Gas CT (combustion turbine)
- Battery energy storage (BESS): 2-, 4-, 6- and 8-hour durations
Note: Storage is modeled with power and energy limits, round-trip efficiency, and state-of-charge tracking.
Technologies outside this list, such as nuclear, large hydro and offshore wind, are not built beyond the existing fleet and committed queue projects.
CAISO queue projects are committed in line with their expected completion
The CAISO interconnection queue holds far more capacity than will come online. Each CAISO queue project therefore gets an expected capacity: its nameplate capacity, multiplied by its probability of completion and by the share of it that is deliverable.
- Near-term projects: Projects in CAISO’s GIRID report hold a Notification of Resource Interconnection (NRI). Their probability of completion depends on their NRI status.
- Long-term projects: Projects only in CAISO’s Public Queue Report are still in the Interconnection Study process. Their probability of completion is the historical completion rate of projects at the same study stage. Suspended projects get the rate of the first stage.
- Deliverability: Deliverable capacity is the capacity a generator is eligible to contract to a utility for Resource Adequacy (RA). Near-term projects are fully deliverable. Long-term projects are credited with their Transmission Planning Deliverability (TPD) allocation: full, partial or energy-only. A project with no deliverable capacity cannot finance itself with an RA contract, so energy-only projects are assumed not to come online.
CAISO queue online dates are adjusted for delays
Queue projects often come online later than their stated date. Near-term projects keep their stated online date. For long-term projects, the time to the stated date is stretched by the delay measured for projects at the same study stage.
Some projects have already missed their stated online date. A long-term project in this group gets a new date from completed projects that spent at least as long in the queue. A project that has waited longer than the completed projects supports is dropped. A near-term project in this group comes online at the start of the projection.
Whole CAISO projects are selected to match expected capacity
Expected capacity is a fraction of the queue, but the model builds whole projects. Near-term queue projects that appear in CAISO’s Master List with a commercial operation date are removed first, because they already operate. Long-term projects that are also in the near-term list, and queue projects whose name matches an operating unit in the WECC ADS, are removed too. No plant is counted twice.
For each technology at each node, projects are then selected year by year until their nameplate capacity matches that year’s expected capacity. Projects most likely to complete are selected first, then the earliest, then the largest. Any difference carries into the next year. Selected projects are committed and built on their adjusted online dates. The rest of the CAISO queue is not committed and sets build limits only.
Outside CAISO, projects under construction are committed
Outside CAISO, planned projects come from the WECC Anchor Data Set (ADS). Projects under construction (tier 1) are committed. Planned (tier 2) and conceptual (tier 3) projects are not committed and set build limits only.
ADS online dates are spaced so that no node adds more of a technology in one year than its annual cap allows. Tier 1 projects take the earliest years. Tier 1 and tier 2 projects that recently missed their online date move to the start of the projection. Other projects that missed their online date are dropped.
Build limits follow the queue to 2029, then build history and peak load
Build limits cap the total new capacity of each technology at each node in each year. Committed capacity counts toward the limit.
- 2027 to 2029: The limit is the nameplate capacity of the queue projects that are not committed. A technology with no queue projects at a node cannot be built there.
- From 2030: The limit is the larger of two values. The first is the node’s largest historical build in one year, scaled by its peak load growth. The second is CAISO’s historical ratio of builds to peak load, applied to the node’s peak load. This uses CAISO’s deployment as the benchmark for how fast a balancing area can bring a technology online, where the node has little or no build history of its own. Committed capacity in that year is subtracted.
- WECC-wide limit: From 2030, total builds of each technology across WECC are also capped at the largest regional build in one historical year, less that year’s committed capacity.
Peak loads for the PG&E, SCE and SDG&E zones come from the CPUC SERVM 2026 net load forecast, which is anchored on the CEC’s 2025 IEPR. This is the same demand forecast the model dispatches against.
Retirements follow the CPUC 2026-27 transmission planning assumptions
CAISO gas plant retirements follow the CPUC’s assumptions for the 2026-27 Transmission Planning Process (TPP):
- Once-through cooling (OTC) plants: The remaining OTC units retire at the end of 2026, about 2.9 GW.
- Combined heat and power (CHP): The CHP fleet, about 2.0 GW, phases out linearly between 2031 and 2040.
- Economic retirements: A further 1.7 GW or so of gas retires by 2028, following the CPUC’s RESOLVE capacity expansion results.
Diablo Canyon is assumed to receive a third operating life extension from the Nuclear Regulatory Commission (NRC). This follows the US trend toward longer nuclear lifetimes. No nuclear plant retires within the forecast horizon. Modo Energy’s research team sets this and other asset-level assumptions.
Policy and market features
Five policy mechanisms drive investment in the model: renewable and clean energy standards, carbon prices, California’s emissions cap, renewable energy credits and resource adequacy.
Renewable and clean energy standards are enforced separately
Renewable Portfolio Standards (RPS) require load-serving entities (LSEs) to buy a minimum share of electricity from qualifying renewable sources. Some states also set broader clean energy standards (CES), which count other zero-carbon sources too.
- RPS: Only RPS-eligible renewable generation counts. Renewable energy credits (RECs) can be banked and traded between states.
- CES: RPS-eligible generation counts, plus nuclear and large hydro.
- Shortfalls: A shortfall against either target is allowed but priced, in the same way as an alternative compliance payment.
- Glide path: A target may rise by at most 10 percentage points a year. Steep future steps are phased in over the preceding years.
California’s RPS, set by SB 100 (2018), requires 60% of retail sales from RPS-eligible renewables by 2030. By 2045, 100% must come from renewable and zero-carbon resources. SB 1020 (2022) added interim clean energy targets of 90% by 2035 and 95% by 2040.
RPS-eligible technologies
Under California’s RPS Eligibility Guidebook, the following generation technologies qualify as eligible renewable resources:
- Solar: Photovoltaic (PV) and solar thermal electric
- Wind: Onshore and offshore wind generation
- Geothermal: Geothermal power plants
- Biomass and biogas: Biomass, biogas (including pipeline biomethane), biodiesel, digester gas, landfill gas, and municipal solid waste conversion
- Small hydroelectric: Facilities of 30 MW or less, and conduit hydroelectric
- Fuel cells: Fuel cells using an RPS-eligible renewable fuel
- Ocean energy: Ocean wave, ocean thermal, and tidal current technologies
Large hydroelectric facilities and nuclear generation are not RPS-eligible, although they count toward the clean energy standard.
Carbon prices apply in California and Washington
Two carbon markets price emissions in WECC. Both add to the operating cost of fossil-fueled generation, in the capacity expansion model and the dispatch model alike.
- California cap-and-trade: Run by the California Air Resources Board (CARB). Forecast allowance prices follow the mid case of the California Energy Commission’s (CEC) 2025 Integrated Energy Policy Report (IEPR). Settled years use the carbon price built into CAISO’s gas price indices, which publish each fuel region’s price with and without carbon. The price applies to generators in the California balancing areas. Biogenic emissions from biomass are exempt.
- Washington Climate Commitment Act (CCA): Settled years use the average of the Department of Ecology’s four quarterly allowance auctions. A year counts as settled once all four have cleared. Forecast years follow a mid-case forecast that assumes the Washington and California markets link. The price applies to generators in the Washington balancing areas.
Imports into California and Washington pay the carbon price too. The charge starts from the regulators’ default emission rate for unspecified power, 0.428 tonnes per MWh. It is then scaled down by the clean share of the exporting balancing area’s generation. In 2027, an import from an area with no clean generation pays about $14/MWh. Imports from the Bonneville Power Administration (BPA), whose generation is 91% clean, pay about $1/MWh.
California’s emissions cap follows the CPUC 25 MMT by 2035 trajectory
On top of the allowance price, the model caps total emissions from the California balancing areas each year. The cap reflects the state’s target of carbon neutrality by 2045 under AB 1279.
The cap follows the “25 MMT by 2035” greenhouse gas (GHG) trajectory. The CPUC adopted it in its Final 2025 Inputs & Assumptions for the 2024-26 Integrated Resource Plan (IRP) cycle. The CPUC sets the cap for the CAISO footprint, and the model rescales it to the seven California balancing areas it represents. The cap declines linearly between the years below and is held flat after 2045. Import emissions count toward it at the same clean-share-adjusted rate as the import carbon charge.
| Emissions cap (million tonnes CO2) | 2025 | 2030 | 2035 | 2040 | 2045 |
|---|---|---|---|---|---|
| CPUC trajectory, CAISO footprint | 40.9 | 24.3 | 20.3 | 13.8 | 7.4 |
| Model cap, California balancing areas | 46.7 | 27.7 | 23.2 | 15.7 | 8.4 |
Emissions above the cap are allowed, but priced at CARB’s allowance price ceiling: about $103/t in 2026, rising 5% a year in real terms. In a year where the cap cannot be met physically, the model pays this price rather than shedding load or cutting imports.
Renewable Energy Credits
Renewable energy credits (RECs) give qualifying renewable generators a second revenue stream. The model includes REC values when it weighs new renewable investments. RECs can make marginal projects viable, so renewables build out faster than energy revenues alone would support.
Slice-of-Day Resource Adequacy
CAISO’s resource adequacy (RA) framework requires load-serving entities to hold enough capacity to meet reliability standards. In 2025, the CPUC replaced the monthly peak-based RA program with a slice-of-day framework. The model applies it as a constraint in the capacity expansion, pooled across the PG&E, SCE and SDG&E zones.
The old program set one capacity target per month, at the forecast peak. Slice-of-day sets a target for each hour of the day, based on each month’s worst day. This changes how each technology counts toward reliability:
- Hourly capacity targets: Every resource receives a capacity value for each hour. Dispatchable technologies get a flat credit across the day. Solar and wind get hourly profiles that follow their expected output.
- Exceedance-based accreditation: Solar and wind are credited on historical generation profiles. The previous ELCC method derated California’s solar fleet to as little as 11% of nameplate. Under exceedance, solar earns up to 90% credit at midday.
- Storage: Batteries must show enough spare capacity in other hours to cover their charging. They can count their capacity across several daily cycles. This moves their value toward the evening net load ramp.
The model therefore builds a portfolio that is reliable in every hour, not only at the system peak. In WECC this matters most in the evening, when solar output falls and net load peaks.
For more detail on how slice-of-day reforms affect resource revenues, see Modo Energy’s analysis of CAISO resource adequacy under slice-of-day.
Planning reserve margins outside CAISO follow WRAP
Balancing areas outside CAISO must hold a planning reserve margin (PRM) above their peak load. The margins follow the Western Resource Adequacy Program (WRAP), the regional program most of these areas belong to.
- Seasonal requirements: Summer and winter (November to March) margins are enforced separately, as in WRAP. In the hydro-dependent Northwest, the Bonneville Power Administration holds 20% in summer and 30% in winter.
- Declining capacity credits: Wind, solar and storage are credited on effective load carrying capability (ELCC) curves. Credits fall as more of a technology is installed. Thermal units are credited net of their forced outage rates. Each year re-prices these credits against the fleet carried in from the year before.
- Imports: Deliverable net imports can cover up to 25% of the requirement, subject to transmission limits.
- Shortfalls: A shortfall is allowed, but its penalty is set above the cost of any new build. The model accepts one only when no build can close the gap.
Technology cost projections
Capital and fixed operating costs for generation follow the Moderate case of NREL’s Annual Technology Baseline (ATB). BESS costs follow Modo Energy’s survey of system costs, with an uplift for financing. All costs are in real 2026 dollars.
Costs are net of federal tax credits. Under the Section 48E clean electricity investment tax credit (ITC), solar receives 30% of its capital cost through 2027. Standalone storage receives 30% through 2033, 22.5% in 2034, 15% in 2035 and nothing after. Wind and geothermal keep the production tax credit (PTC). The steps in the charts below mark where the credits phase out.
Data sources
Policy and emissions
| Source | Description | Link |
|---|---|---|
| CEC Integrated Energy Policy Report (IEPR) | California GHG allowance price projections (2025 IEPR, mid case) | CEC IEPR |
| CAISO gas price indices | Settled-year California carbon prices, from fuel region prices with and without GHG cost | CAISO |
| CEC RPS Eligibility Guidebook | Eligible renewable technology definitions and certification requirements | RPS Guidebook, 10th Edition |
| CPUC IRP Inputs & Assumptions | California emissions cap trajectory (Final 2025 I&A, 25 MMT by 2035) and gas retirement assumptions | CPUC IRP |
| CPUC Resource Adequacy | Slice-of-day RA framework and filing requirements | CPUC RA Homepage |
| Washington Department of Ecology | Climate Commitment Act auction results | WA Ecology |
| Western Power Pool | WRAP planning reserve margins and capacity accreditation | WRAP |
Technology costs and fleet data
| Source | Description | Link |
|---|---|---|
| CAISO interconnection queue | GIRID report (near-term projects) and Public Queue Report (long-term projects). Sets CAISO committed projects and build limits, and supplies the completion and delay history | CAISO Interconnection |
| EIA-860 | Existing generator fleet: technology, capacity and operating status | EIA-860 |
| NREL ATB | CAPEX/OPEX projections for generation candidate technologies | NREL ATB |
| WECC ADS | Planned projects outside CAISO (tier 1 committed, tiers 2 and 3 set build limits) and generator technical parameters | WECC ADS |
Financial assumptions
| Category | Source |
|---|---|
| CapEx, generation | NREL Annual Technology Baseline (ATB), Moderate case |
| CapEx, BESS | Modo Energy cost survey, with a financing uplift |
| California GHG allowance prices | CEC 2025 IEPR, mid case |
| Washington allowance prices | Washington Department of Ecology auctions, then a mid-case forecast |
| OpEx | NREL Annual Technology Baseline (ATB) |
| Tax credits | Section 48E ITC for solar and storage; PTC for wind and geothermal |