The Capacity Expansion Model (CEM) determines the optimal mix of new generation and storage capacity to meet demand reliably and cost-effectively over the forecast horizon. Near-term buildout is anchored to interconnection queue analysis; longer-term decisions are optimised.
At a glance
| Parameter | Value |
|---|---|
| Geography | NYISO, PJM, ISO-NE, MISO (all modeled zones) |
| Horizon | Annual investment decisions to 2050 |
| Temporal resolution | Hourly dispatch on representative days |
| Windowing | Rolling 1-year optimisation (co-optimised investment + dispatch) |
| Reliability | Planning Reserve Margin (PRM) with technology-specific derating |
| Solver | XPRESS (same as PCM) |
For further details on the core CEM methodology, see the core model documentation.
Build candidates
Technologies available for new capacity investment:
| Category | Technologies |
|---|---|
| Renewables | Solar PV, onshore wind, offshore wind |
| Thermal | Gas CCGT, gas CT |
| Storage | Battery energy storage (multiple durations: 2hr, 4hr, 6hr, 8hr) |
Technologies not modeled as build candidates use predetermined buildout from interconnection queues and official plans: nuclear, large hydro, pumped storage.
Retirements
Technology-specific economic lifetimes from NREL ATB. Coal plants retire at 25 years; gas plants at 60. Nuclear and hydro have 200-year lifetimes (effectively no modeled retirement).
Interconnection queue analysis
Near-term capacity additions are anchored to active interconnection queue data from each ISO. Queue projects are filtered by likelihood, technology, and geographic constraints before entering the model.
NYISO queue
NYISO queue processing applies several filters:
- Moratorium town exclusion — BESS projects within 10 km of moratorium towns are excluded
- Phase 1 interconnection cost screening — projects with prohibitively high interconnection costs are filtered out
- Cluster queue filtering — projects filtered by commercial operation date horizon (4 years) and likelihood threshold (50%)
- Annual BESS cap — maximum ~1,000 MW storage per year in the near-term, based on 3 ISC tenders (2027–2029)
Other ISOs
PJM, ISO-NE, and MISO queues are processed from CSV data with technology mapping and likelihood-based attrition rates.
Policy and market features
Renewable Portfolio Standards (RPS)
State-level RPS mandates are enforced as CEM constraints:
| State | Target | Deadline | Legislation | Link |
|---|---|---|---|---|
| New York | 70% renewable, 100% zero-emission | 2030 / 2040 | CLCPA (2019) | NY CLCPA |
| New Jersey | 50% renewable, 100% clean energy | 2030 / 2050 | EMP (2020) | NJ EMP |
| Massachusetts | 80% clean electricity | 2050 | Clean Energy Standard | MA CES |
| Connecticut | 100% zero-carbon | 2040 | PA 22-5 (2022) | CT PA 22-5 |
| Maine | 80% renewable | 2030 | LD 1494 (2019) | ME LD 1494 |
Large hydro and nuclear count toward “zero-emission”/”clean energy” targets but not “renewable” targets — New York’s 70% renewable target requires genuinely new wind, solar, and storage.
NYISO Capacity Market
NYISO uses an ICAP demand curve to set capacity prices across four nested localities. Technology-specific Capacity Accreditation Factors (CAFs) determine how much capacity credit each resource receives. See Capacity Prices for full methodology, CAF tables, and data sources.
RGGI carbon budget
The RGGI regional CO2 budget constraint applies to the CEM as well, influencing build decisions toward lower-emission technologies. The same CCR tier mechanism used in the PCM applies here.
Renewable energy credits
PTC and ITC values by technology are embedded in the investment economics, reducing the effective CAPEX of eligible renewables and storage.
Technology cost projections
CAPEX and OPEX for all candidate technologies are sourced from NREL Annual Technology Baseline (ATB), converted to the model’s currency base year (2025 real USD).
Build constraints
- IC queue anchoring — near-term builds (first 5 years) are constrained to match filtered IC queue projections; CEM does not build in those years
- NYISO BESS cap — 1,000 MW/year near-term from ISC tenders
- Per-node and global limits — loaded from CSVs and validated for consistency
- Minimum project sizes — technology-specific minimums from NREL ATB
Assumptions and caveats
- CEM uses representative days with hourly dispatch, not full-year chronological simulation.
- Rolling 1-year windows mean the model does not co-optimise across multiple investment years simultaneously.
- NYISO BESS CAFs decline over the forecast horizon (e.g., 2hr BESS NYCA CAF drops from 0.74 in 2025 to 0.14 by 2044), reflecting increasing storage saturation.
- Offshore wind capacity is limited to zones with modeled interconnection points.
Data sources
| Source | Description | Link |
|---|---|---|
| NREL ATB | CAPEX/OPEX projections for all candidate technologies | NREL ATB |
| EIA-860 | Existing fleet characterisation and retirement dates | EIA-860 |
| NYISO IC Queue | Active and cluster queue data with project costs | NYISO |
| PJM IC Queue | Queue data with technology mapping | PJM |
| ISO-NE IC Queue | Queue data | ISO-NE |
| MISO IC Queue | Queue data | MISO |
| RGGI | CO2 budget and allowance auction data | RGGI |
| NYISO ICAPWG | CAF methodologies and TSL floor updates | NYISO ICAPWG |