Cut-Selection Enhanced Risk-Averse Stochastic Dual Dynamic Programming for Economic Dispatch with Carbon Capture Systems

This paper investigates a multi-stage stochastic economic dispatch problem for power systems with carbon capture systems (MCCSED). The focus is on effectively managing operational risks through a risk-averse approach. To efficiently solve this challenging problem, a cut-selection-enhanced risk-averse stochastic dual dynamic programming (SDDP) algorithm is developed, incorporating an enhanced test of usefulness (ETOU) strategy to remedy the premature discarding of valid cuts inherent in the conventional level 1 (L1) method. Case studies on IEEE 39-bus, 118-bus, and 300-bus systems yield these findings. First, higher carbon capture installation rates enhance risk management through increased storage flexibility. Second, the risk-averse approach reduces tail-risk costs and enables more intuitive parameter tuning than distributionally robust optimization, revealing a consistent two-phase pattern across risk management approaches and underscoring the importance of coordinating risk management and flexibility. Third, the proposed ETOU-L1 method offers competitive convergence by recovering valid cuts usually discarded by the conventional L1 method, while demonstrating scalability for large-scale problems.