Optimal Dispatch of VPPs with Shared Energy Storage and User Response Uncertainty in Coordinated Transmission, Distribution and Microgird Networks

Virtual Power Plants (VPPs) integrate distributed generation (DG), shared energy storage stations (SESS), gas turbines, and flexible loads to optimize resource allocation under the "dual carbon" goals. This paper addresses user response uncertainty by formulating load aggregation as a Combinatorial Multi-Armed Bandit (CMAB) problem and proposes the CUCBAvg (Combinatorial Upper Confidence Bound-Average) algorithm for real-time user selection. A day-ahead dispatch model maximizes operational profit while coordinating SESS, DG, and flexible loads in coordinated transmission, distribution and microgrid Network. Simulations results demonstrate significant improvements in economic performance and grid stability, validating the effectiveness of the proposed approach.