Methodology for Capacity Credit Evaluation of Physical and Virtual Energy Storage in Decarbonized Power System

Technical Summary: Methodology for Capacity Credit Evaluation of Physical and Virtual Energy Storage in Decarbonized Power Systems

Problem Statement
As power systems transition toward decarbonization, the integration of renewable energy sources (RES) introduces significant stochasticity, necessitating flexible resources like Energy Storage (ES) and Virtual Energy Storage (VES, derived from Demand Response) to ensure system adequacy. While the reliability benefits of these resources are acknowledged, existing methodologies for evaluating their Capacity Credit (CC)—the credible capacity value they contribute to system reliability—suffer from critical limitations. Current approaches typically rely on either "fixed dispatch" (deterministic scheduling that ignores real-time failures) or "greedy management" (assuming resources remain fully charged and available at all times). These methods fail to account for:

1. Real-time failures: The inability of existing dispatch models to trade-off day-ahead self-energy management (e.g., arbitrage, self-consumption) with emergency responses to generation or network outages.
2. Decision-Independent Uncertainties (DIUs): Factors such as forced outages, capacity degradation, and baseline consumption that exist independently of system control.
3. Decision-Dependent Uncertainties (DDUs): Factors where the availability of resources (specifically VES) is influenced by human behavior, incentive signals, and discomfort costs, which alter the available capacity dynamically.

Overlooking these factors leads to significant overestimation of CC (reported as 10%-70% in previous works), posing risks to system security and market fairness.

Methodology
The paper proposes a novel evaluation framework centered on a two-stage coordinated dispatch method integrated with a Sequential Monte Carlo Simulation (SMCS).

1. Uncertainty Modeling:

   • DIUs: Modeled for Conventional Generation (CG), Renewable Generation (RG), and ES/VES operation states (Bernoulli distributions) and baseline consumption (Lognormal/Normal distributions).
   • DDUs: Specifically modeled for VES available capacity. The State of Charge (SoC) bounds are treated as decision-dependent variables, expanding or contracting based on the trade-off between incentive payments and user discomfort. This is handled via chance-constrained optimization.
   • Generic Energy Storage (GES): A unified modeling approach is introduced to treat both physical ES and VES under a common framework, distinguishing their specific uncertainty profiles.

2. Two-Stage Coordinated Dispatch:

   • Stage 1 (Day-Ahead/Normal State): Resources follow self-energy management strategies (e.g., energy arbitrage for ES, self-consumption for VES) to maximize economic or operational efficiency under normal conditions.
   • Stage 2 (Real-Time/Emergency State): Upon detection of system inadequacy (loss of load or capacity deficit), the dispatch adjusts to minimize load curtailment. This stage incorporates the baseline consumption and DDUs to determine the actual available capacity, rather than the rated capacity.
   • Recovery State: Post-emergency, resources recover to their baseline profiles, accounting for load rebound effects.

3. Evaluation Framework:

   • The framework uses SMCS to generate chronological failure and repair scenarios.
   • It calculates reliability indices (Expected Energy Not Served - EENS) for both a reference system (without storage) and a test system (with storage).
   • Capacity Credit Indices: The paper introduces and utilizes several indices, including Equivalent Firm Capability (EFC), Equivalent Conventional Capacity (ECC), and a novel Equivalent Physical Storage Capacity (EPSC) to quantify VES's ability to replace physical storage. It also employs Equivalent Generation Capacity Substituted (EGCS) and Effective Load Carrying Capability (ELCC).

Key Contributions

1. Coordinated Dispatch Framework: The proposal of a two-stage dispatch method that balances day-ahead self-management with real-time failure adjustment. Unlike fixed dispatch, it reacts to outages; unlike greedy management, it respects baseline consumption and uncertainty constraints.
2. Incorporation of DIUs and DDUs: The explicit modeling of decision-dependent uncertainties (human behavior/incentives) and decision-independent uncertainties (degradation/outages) within the CC evaluation. This addresses the gap where previous works treated resource availability as static or exogenous.
3. Novel Indices and Practical Assessment: The introduction of EPSC and the distinction between "theoretical" (idealized) and "practical" (real-world with uncertainties) reliability indices. This allows for a realistic assessment of how much generation or storage can be displaced while maintaining equivalent system adequacy.
4. Comprehensive Case Studies: Extensive validation using the IEEE RTS-79 system and real-world data (Belgium's Elia grid for ES, Austin, TX for VES) to compare the proposed method against fixed and greedy dispatch strategies.

Results

• Overestimation in Existing Methods: The study confirms that ignoring DIUs and DDUs leads to a 10%-70% overestimation of CC. For instance, greedy management yields high theoretical CC but fails in practical reliability due to unmodeled degradation and self-discharge.
• Performance of Coordinated Dispatch: The proposed method generates credible, realistic CC values. While its theoretical EENS is slightly higher than greedy management (due to realistic capacity constraints), its practical performance is superior because it accounts for real-world limitations.
• Impact of Uncertainties:
  • For ES, capacity degradation and forced outages significantly reduce practical reliability compared to theoretical models.
  • For VES, the impact of DDUs (incentive/discomfort) is profound. Modeling these uncertainties reduces the available SoC bounds by 40%-50% during peak times, drastically lowering the practical CC compared to models that assume constant availability.
• Factor Sensitivity:
  • ES: Capacity value is highly sensitive to energy capacity ratings and charge/discharge efficiency. Long-duration storage (6h) shows significantly higher CC than short-duration storage.
  • VES: CC is highly sensitive to load factors, correlation factors, and DDU levels. VES is effective primarily in low-to-medium RES penetration scenarios; its ability to replace generation drops to near zero in high-penetration scenarios (>50% RES).
  • Market Implications: Current capacity market regulations (e.g., in CAISO, NYISO) often use fixed derating factors that align with "greedy" theoretical values, leading to over-remuneration and potential security risks.

Significance and Claims
The paper claims that its primary significance lies in providing a credible and realistic evaluation framework for the capacity credit of ES and VES in decarbonized power systems. By moving beyond simplified dispatch assumptions and incorporating the dynamic nature of human and market behaviors (DDUs), the proposed method prevents the overestimation of resource adequacy.

The authors argue that this framework is essential for:

• Capacity Market Design: Providing a fair basis for remuneration that reflects actual performance rather than theoretical potential.
• System Planning: Guiding the allocation of resources, suggesting that long-duration ES is critical for high-RES systems, while VES is best suited for low-to-medium RES penetration and specific load characteristics.
• Policy and Regulation: Offering evidence that current regulatory approaches (fixed derating) are insufficient and recommending the integration of behavioral modeling and chance-constrained optimization into future market rules.

The paper concludes that while the proposed method is computationally more intensive, it is necessary to avoid the substantial risks associated with over-relying on storage resources whose availability is stochastically limited by both physical degradation and human decision-making.