Robust Procurement Planning for Community Choice Aggregators: Managing Market Price Uncertainty

1 INTRODUCTION

• CCAs submit IRPs that must meet reliability and policy requirements, including California’s clean-electricity requirements, while deciding how much renewable generation and storage to procure and when.CCAs procure power for local governments and customers, while investor-owned utilities retain transmission and distribution infrastructure.Source: §1 · pp. 1–2
• Uncertain export revenues can make additional renewable capacity appear economical, while adverse prices can leave a CCA with a compliant portfolio that costs more than anticipated.Deterministic price trajectories may favor export-oriented investment, while costly imports can also be understated; CCAs must maintain competitive retail rates.Source: §1 · p. 2
• Wholesale price uncertainty has received comparatively limited attention in long-term CCA procurement, especially regarding the value of surplus clean generation.The study isolates market price uncertainty from other commonly studied uncertain inputs.Source: §1 · p. 2
• Budgeted robust optimization lets planners control protection against adverse price realizations, addressing a gap in widely used open-source capacity-expansion tools for CCA procurement.The paper identifies deterministic planning in RESOLVE and a lack of widely used open-source models with budgeted RO for uncertain wholesale import costs and export revenues.Source: §1 · pp. 2–3
• The study extends GenX with budgeted static robust optimization and evaluates procurement, imports and exports, out-of-sample prices, cost exposure, portfolio stability, and market dependence.It examines how uncertainty budgets affect the amount, timing, and composition of renewable and storage procurement.Source: §1 · p. 3
• Robust portfolios can change how a CCA meets fixed clean-energy requirements by reducing the incentive to build renewables whose value depends on uncertain export revenues.The framework helps evaluate the trade-off between price-risk protection and affordability and durability of the clean-energy transition.Source: §1 · p. 3

2 Model Formulation

• The GenX extension minimizes multistage net present cost subject to investment, dispatch, power-balance, policy, and adequacy constraints, adding a reserve margin, market transactions, and imported-electricity emissions accounting.The reserve margin uses resource-specific capacity accreditation.Source: §2 · p. 3

2.1 Wholesale Market Representation

• Imports and exports enter zonal power balance subject to transaction limits and constraints against unsupported exports; emissions from unspecified imports are also counted.The market module uses weighted operating costs for time steps.Source: §2.1 · pp. 3–4

2.2 Robust Optimization Model

• The study isolates wholesale price uncertainty because it has received comparatively limited attention despite its importance to CCAs relying on non-firm market transactions.Other uncertainties, including demand, technology costs, fuel prices, and renewable availability, have been widely studied.Source: §2.2 · p. 4
• The robust formulation represents adverse import-price deviations and protects against worst-case price movements without requiring probability assumptions.The budget-of-uncertainty framework limits how many deviations can jointly be active.Source: §2.2 · p. 4
• The planner selects investment, operating, and market-transaction decisions, while an adversarial decision-maker selects adverse price deviations to maximize additional system cost within the budget.The deviations are represented by activation variables indicating fractions of maximum adverse deviations.Source: §2.2 · p. 4
• The uncertainty budget Γ controls conservatism by limiting simultaneous worst-case deviations; Γ = 0 reproduces deterministic planning, and 50% is the most conservative effective setting in evaluated solutions.Because imports and exports do not occur simultaneously within a time step, budgets above 50% cannot activate further adverse deviations on realized transactions.Source: §2.2 · p. 4
• Strong duality replaces the nested worst-case maximization with its dual minimization, yielding an equivalent single-level linear program that standard solvers can solve efficiently.The resulting portfolio is static: decisions do not adapt after prices are realized, which may make solutions more conservative than adaptive approaches.Source: §2.2 · pp. 4–5

3 Case Study

• The case study uses collaborating-CCA inputs over 2025–2040, represented by five planning periods with finer resolution early and coarser periods later.The five periods begin in 2025, 2026, 2028, 2030, and 2035; model assumptions align with CPUC IRP inputs and requirements.Source: §3 · p. 5

3.1 Capacity Expansion Model Inputs

• The model represents 65 individual resources, including thermal, storage, renewable, must-run, and hybrid resources, with results reported in aggregated technology categories.Demand-response and RA-only categories contribute firm capacity but are structured not to dispatch in the hourly energy balance.Source: §3.1 · p. 5
• Demand rises across planning stages, with late-afternoon and early-evening peaks, lower midday loads, and summer and early-autumn seasonality.The case imposes a declining CO2 cap, increasing clean-energy shares, and stage-specific resource build limits.Source: §3.1 · p. 6, Figure 1 · p. 6

3.2 Market Price Inputs and Uncertainty Set Construction

• The uncertainty set uses six vendor-provided price trajectories and an asymmetric interval around a nominal profile, preserving hourly structure and adverse import-cost and export-revenue deviations.Pbase_base is selected as the nominal import trajectory; the nominal export price is set to 0.95 times the nominal import price.Source: §3.2 · pp. 6–8
• Price spreads widen in later planning years, with the largest deviations generally during evening ramp hours around 18:00–22:00.All scenarios were solved in full-foresight multistage mode across uncertainty budgets from deterministic Γ = 0 to the effective 50% maximum.Source: §3.2 · pp. 7–8, §3.3 · p. 8
• Average price deviations widen over the horizon: import deviations grow from approximately $8/MWh in 2025 to more than $34/MWh by 2035, while export deviations rise from approximately $20/MWh to over $29/MWh.The constructed uncertainty envelope is shown for representative high-price days across planning years.Source: §3.2 · p. 8, Figure 3 · p. 8

3.3 Simulation and Computational Setup

• The experiments use the extended GenX model and leave optional binary constraints against simultaneous imports and exports unenforced; resulting solutions still avoid simultaneous transactions.Consequently, the scenarios are solved as linear programs.Source: §3.3 · p. 8

4.1 Baseline Procurement under Deterministic Prices

• The deterministic baseline satisfies modeled CPUC IRP requirements but relies entirely on the nominal market-price trajectory.Solar capacity grows from approximately 900 MW in 2024 to more than 3,200 MW by 2035, alongside storage expansion.Source: §4.1 · p. 9, Figure 4 · p. 9
• Under the nominal price forecast, market export revenues increasingly offset investment and operating expenditures as export volumes grow, partially financing the deterministic portfolio’s continued expansion and exposing it to future market-price uncertainty.Export volumes reach their highest level in 2030 before declining slightly in 2035, while remaining well above import volumes.Source: §4.1 · p. 10

4.2.1 Portfolio Composition

• As Γ rises, total procurement declines, especially solar and 6-hour batteries, while imports vary little and exports fall sharply.ESR and emissions constraints remain satisfied; four- and 8-hour battery capacities remain comparatively stable.Source: §4.2.1 · pp. 11–12, Figure 7 · p. 11, Figure 6 · p. 11

4.2.2 Cost Trade-offs and Price of Robustness

• PoR reaches approximately 14% by Γ = 28% and about 15% at the highest budgets, with the curve flattening as active market exposure becomes covered.The premium reflects lost export revenues and explicit price protection, partly offset by lower investment expenditures.Source: §4.2.2 · pp. 12–13, Figure 9 · p. 13
• Γcap = 8% provides a modest PoR reduction with little nominal-cost change, while Γcap = 20% and 40% reduce sensitivity at higher upfront cost.Fixed-capacity portfolios’ total costs become increasingly similar as the operational uncertainty budget rises.Source: §4.2.2 · p. 13, Figure 9 · p. 13

4.2.3 Operational Implications of Robust Capacity Planning

• Higher uncertainty shifts operational reliance toward wind and away from market exports in both 2026 and 2035, despite little change in total wind capacity.Internally available firm and renewable resources are used more heavily to meet demand and manage variability.Source: §4.2.3 · p. 14, Figure 10 · p. 14
• Solar curtailment rises initially as exports decline at Γ = 8%, then falls at Γ = 40% as the robust portfolio reduces solar capacity.The pattern reflects both renewable-capacity choices and the operational management of surplus solar generation.Source: §4.2.3 · pp. 14–15, Figure 11 · p. 15

4.3 Sensitivity of Procurement Decisions to Uncertainty Set Specification

• The sensitivity tests vary the nominal price trajectory and scale base_flat deviation bounds by κ = 1.5 to test larger price deviations.The stressed-envelope test is not intended as a specific forecast scenario; analysis focuses on end-of-horizon installed capacity.Source: §4.3 · p. 15
• The stressed_flat envelope produces more persistent capacity differences than base_flat, including lower 6-hour battery, offshore wind, and solar capacity and higher onshore wind capacity.The wider deviation bounds affect both storage selection and renewable procurement composition.Source: §4.3 · p. 15
• Relative to base_base, total capacity is approximately 4% higher at low Γ, then approximately 3% lower under base_flat and 8% lower under stressed_flat at higher budgets.Figure 12 compares end-of-horizon installed capacity by technology relative to base_base at each Γ.Source: §4.3 · pp. 15–16, Figure 12 · p. 16
• Alternative uncertainty-set specifications change technology allocation more than overall procurement scale, while producing comparatively small total-system-cost differences.The stressed_flat specification has slightly higher costs than base_flat; both cost curves stabilize at moderate and high Γ.Source: §4.3 · p. 16, Figure 13 · p. 16

4.4 Out-of-Sample Robustness Validation

• The validation evaluates 157 price profiles, including 150 synthetic trajectories, six vendor profiles, and one scarcity-jump profile.Three Γ = 0% runs and one Γ = 8% run did not solve within the time limit; all Γ = 40% runs solved.Source: §4.4 · pp. 16–17
• Higher robustness shifts the total-cost distribution upward while reducing its dispersion; low_low produces the highest system costs because of reduced export revenues.The low_low and high_high profiles used to construct the uncertainty set bound the total-cost distributions.Source: §4.4 · p. 17, Figure 14 · p. 17
• Import costs do not follow a uniform ordering across extreme profiles, and residual import costs become small relative to total system cost as Γ rises.At high robustness, high_high may produce higher import costs than low_low even as overall cost exposure becomes more stable.Source: §4.4 · p. 17, Figure 14 · p. 17
• As Γ increases, mean, median, 95th-percentile, and CVaR95 costs rise, but the gap between CVaR95 and nominal portfolio cost narrows.Robust procurement reduces upper-tail exposure relative to each portfolio’s planned cost, not absolute tail costs.Source: §4.4 · p. 17, Figure 14 · p. 17

4.5.1 Downside Financial Protection

• DCP falls from 1.3% at Γ∗ = 0% to 0.6% at Γ∗ = 8% and −0.8% at Γ∗ = 40%; hedge effectiveness is approximately 52% at Γ = 8%.The negative DCP at Γ∗ = 40% reflects conservative over-hedging, not an equivalent increase in meaningful risk reduction.Source: §4.5.1 · p. 18

4.5.2 Portfolio Stability

• Portfolio stability stays close to one at Γ = 8%, indicating modest baseline-plan adjustments, then declines and largely plateaus by Γ = 40%.Lower stability scores indicate greater structural deviation from the deterministic portfolio.Source: §4.5.2 · pp. 18–19

4.5.3 Market Dependence

• Mean market dependence falls from 3.27% at Γ = 0% to 2.85% at Γ = 40%, while CVaR95 dependence drops from 4.52% to 3.39%.The narrowing gap between mean and tail dependence indicates reduced exposure to unusually high import requirements as robustness increases.Source: §4.5.3 · p. 19, Table 3 · p. 20

4.5 Policy and Decision Implications

• A moderate budget such as Γ = 8% removes about half the baseline downside premium while preserving high portfolio stability; larger budgets further reduce market and tail exposure with diminishing benefits.The preferred Γ depends on the CCA’s priorities for affordability, procurement stability, market exposure, and low-carbon procurement.Source: §4.5.3 · pp. 19–20, Figure 15 · p. 19

4.5.3 Market Dependence

• All evaluated portfolios meet the modeled clean-energy requirements and emissions limits; Γ changes how those obligations are met, not whether they are met.The affected choices include renewable procurement composition and timing, reliance on uncertain export revenues, and exposure to future market conditions.Source: §4.5.3 · p. 19

5 Conclusions and Future Work

• The framework extends open-source GenX with budgeted robust optimization to hedge against adverse market-price realizations while remaining computationally tractable for large-scale IRP studies.Its implementation in GenX can be adapted to other CCAs, utilities, and long-term planning applications.Source: §5 · p. 20
• Robust portfolios shift away from export-oriented renewable expansion toward more conservative investments that depend less on uncertain wholesale market revenues.The findings suggest deterministic planning may overestimate the long-term value of market exports, leaving procurement strategies more exposed to adverse prices than intended.Source: §5 · p. 20
• Moderate uncertainty budgets deliver a substantial share of robust-planning benefits, while higher budgets bring diminishing economic value as solutions become more conservative.Moderate budgets reduce downside cost exposure with modest changes to the deterministic portfolio; higher budgets further reduce market dependence and relative tail exposure but raise nominal planning cost.Source: §5 · pp. 20–21
• All evaluated portfolios meet clean-energy requirements and emissions limits, while robustness changes the amount, timing, and composition of renewable and storage procurement.The resulting portfolios rely less on uncertain revenues from surplus electricity exports.Source: §5 · p. 20
• Future work could jointly model demand, fuel, renewable-availability, and technology-cost uncertainty, and explore adaptive robust optimization and integration with modeling to generate alternatives.These extensions could examine portfolios with comparable cost and risk but differing technology mixes, market dependence, and implementation requirements.Source: §5 · p. 21

• The passage is demonstration placeholder text and provides no substantive funding information.Source: Funding · p. 21

Data availability

• Input data, model inputs, modified GenX source code, simulation results, and analysis scripts are available on Zenodo to support reproduction of the study’s analyses and figures and tables.The materials support robust optimization, sensitivity, price-of-robustness, and out-of-sample analyses.Source: Data availability · p. 21

A.1 Nomenclature

• The appendix defines notation for resources, planning stages, market transactions, prices, emissions, and the budget of uncertainty used in the CCA formulation.The uncertainty budget is denoted Γ; the nomenclature also distinguishes maximum upward import-price and downward export-price deviations.Source: A Extended GenX Formulation for CCA… · p. 24, §A.1 · p. 24
• The notation includes nominal import and export prices, their allowed deviations, installed capacity, dispatch, and worst-case price activations.The uncertainty budget and these variables specify the price-uncertainty representation described in the appendix.Source: §A.1 · p. 24

A.2 Capacity Expansion Model

• The baseline GenX objective minimizes net present value of system costs, including investment, fixed and variable O&M, fuel, and emissions-related expenses.Costs across planning periods are weighted by discount factors and an operational-expenditure multiplier; the study builds its extensions on this standard objective.Source: §A.2 · pp. 24–25

A.3 Planning Reserve Margin

• The CCA formulation adds an annual PRM constraint requiring each planning zone to procure sufficient ELCC-adjusted capacity for its reliability target.PRM shortfalls are represented by a slack variable and discouraged through an objective-function penalty.Source: §A.3 · p. 25

A.4 Market Module

• The market module represents out-of-area dynamic-scheduled resources as unspecified imports and includes import and export transactions in the objective and zonal power balance.Transaction quantities are bounded, and exports cannot exceed electricity generated within the zone, preventing market-purchase-to-export arbitrage.Source: §A.4 · pp. 25–26
• Optional binary constraints can prevent simultaneous imports and exports, but the study did not use them because simultaneous transactions did not occur in its solutions.The reported scenarios were therefore solved as linear programs under the assumed price structure and export restrictions.Source: §A.4 · p. 26
• Unspecified-import emissions are assigned to the importing zone and included in emissions accounting, aligning imported electricity with the framework applied to in-state generation.The treatment is intended to avoid giving imports an artificial advantage in the planning model.Source: §A.4 · p. 26

B Supplementary Material

• Figure B.1 plots installed capacity against uncertainty budget Γ, with stacked colors marking additions by planning year and 2024 denoting existing capacity.Panel (a) shows total capacity; panels (b)–(i) show selected resource-specific trajectories.Source: Figure B.1 · p. 27