Virtual Power Plant Scheduling Strategy Based on DQN Deep Learning Algorithm

One of the notable features of the new power system is the extensive integration of renewable energy. However, renewable energy generation has randomness and uncontrollability, and the access locations are scattered, which increases the difficulty of safe and stable operation of the power system. The proposal of virtual power plant (VPP) provides a feasible path for the above-mentioned problems. Solving the VPP optimization scheduling problem is one of the research focuses in the context of new power systems. This article analyzes the output characteristics of the constituent units of a virtual power plant and constructs a multi agent internal operation model for the virtual power plant. Furthermore, this article proposes a virtual power plant scheduling model considering local network congestion and market power prevention by introducing market power penalty. The scheduling model is solved by DQN deep learning algorithm. The validity of this virtual power plant scheduling model is verified through case analysis.