Multi-scenario collaborative optimization scheduling for source-grid-load-storage with carbon flexibility market and CVaR-based risk assessment

Efficient coordination of multi-type energy storage (batteries, pumped hydro, EVs) on source, grid, and load sides enhances power system flexibility and renewable integration. This paper constructs a source-grid-load-storage (SGLS) collaborative scheduling architecture. Using Latin hypercube sampling and k-medoids clustering, multiple emergency scenarios are generated, and flexibility indices are proposed. Three novel contributions are introduced: (1) Carbon Flexibility Providers (CFPs) with quantitative metrics; (2) a bilevel Carbon-Intensity Coupled Flexibility Market (CICFM) where the Carbon Marginal Emission Factor (CMEF) is rigorously derived from KKT conditions; (3) a CVaR-based Dynamic Carbon Risk Index (DCRI-CVaR) normalized by deterministic carbon cost to preserve tail risk amplification. The bilevel problem is solved as an MPEC using a MIBE-data-driven algorithm that approximates lower-level best responses. Validation on real regional grid data shows significant improvements: 11.8% lower costs, 9.0% higher renewable utilization, 22–31% longer storage life, 47% higher carbon reduction, and 39.5% higher market revenue compared to the best alternatives. The optimal risk aversion coefficient β = 0.8 is case-specific, but the risk-return frontier structure is universally applicable.