A Data-Driven Model of Virtual Power Plants in Day-Ahead Unit Commitment

Due to the increasing penetration of distributed energy resources (DERs), power system operators face significant challenges of ensuring the effective integration of DERs. The virtual power plant (VPP) enables DERs to provide their valuable services by aggregating them and participating in the wholesale market as a single entity. However, the available capacity of VPP depends on its DER outputs, which is time varying and not exactly known when the independent system operator runs the day-ahead unit commitment engine. In this study, we develop a model to evaluate the physical characteristics of the VPP, i.e., its maximum capacity and ramping capabilities, given the uncertainty in wind power output and load consumption. The proposed model is based on a distributionally robust optimization approach that utilizes moment information (e.g., mean and covariance) of the unknown parameter. We reformulate the model as a binary second-order conic program and develop a separation framework to address it. We first solve a two-stage problem and then benchmark it with a multi-stage case. Case studies are conducted to show the performance of the proposed approach.