Rethinking Capacity Accreditation in High-Renewable Power Systems: Discrepancies, Drivers, and Practical Insights from a CAISO Case Study

Independent System Operator (CAISO) system as a high-renewable testbed, the study employs a modified version of the CAISO 2024 Stochastic Production Cost Model to evaluate accreditation methods under realistic system stresses.

The comparative analysis reveals that the transition from peak load to peak net load metrics significantly devalues solar resources as risk hours migrate to post-sunset periods, while wind remains relatively stable due to its evening generation patterns. Within probabilistic frameworks, the choice of benchmark has a profound impact on the results. The Equivalent

Firm Capacity (EFC) method, anchored to a theoretical “perfect” generator, tends to yield higher battery storage capacity credits compared to the Effective Load Carrying Capability

(ELCC) metric, which uses incremental load as its benchmark. Furthermore, probabilistic methods “unlock” solar capacity credits by capturing synergies with battery storage that simple performance averaging fails to identify, highlighting the necessity of modeling resource interactions rather than treating technologies in isolation.

The study further observes that the Marginal Reliability Improvement (MRI) consistently delivers lower capacity credits than average counterparts, reflecting the diminishing marginal utility of resources as the system risk window becomes increasingly saturated. Notably, the

MRI for solar can converge to zero under the Loss of Load Hours (LOLH) benchmark, while maintaining a positive value under the Expected Unserved Energy (EUE) metric, demonstrating that metric selection can prematurely signal full saturation. These findings suggest the adoption of integrated, probabilistic frameworks that accurately quantify the synergistic reliability contributions to provide robust investment signals. Future work will focus on high-fidelity storage dispatch forecasting, quantifying diversity premiums, and investigating metric-driven saturation thresholds.