Two-stage MPC-based energy optimization scheduling for a virtual power plant with multiple adjustable resources in electricity spot markets

To address the reliable and economical operation of virtual power plants containing a large number of adjustable resources in the day-ahead and intraday markets, a two-stage model predictive control (MPC) based optimization scheduling strategies for virtual power plants (VPPs) in the electricity spot market is proposed and discussed in this paper. Firstly, a TCN-GRU-attention hybrid prediction algorithm is developed for forecasting output of inelastic loads, wind power and photovoltaic (PV) in VPPs. Secondly, a hierarchical bi-level mixed-integer linear programming model integrating multiple resources with EV participation is established to enable that EV charging load, GT generation and ESS can dynamically adjust their behaviour under different operation cost. Furthermore, an improved MPC-based intraday dispatch strategy embedded with a bidirectional dynamic penalty mechanism is proposed to rapidly respond to real-time fluctuations of uncontrollable resources. The proposed two-stage MPC-based scheduling strategies can thereby enhance the flexibility and operational efficiency of the VPP system. Simulation results demonstrate that the two-stage collaborative optimization improves the total revenue of VPPs by 5.06%, fully verifying the robust economic advantages of the proposed solution in complex market environments.