Dynamic Optimization Allocation of Active Power in Wind Farm Clusters Based on Deep Reinforcement Learning

Due to factors such as uneven distribution and significant volatility of wind energy resources, the power allocation of wind farm clusters is imbalanced, and the phenomenon of wind abandonment is becoming increasingly prominent, leading to a further decrease in wind energy utilization efficiency. Therefore, a dynamic optimization allocation of active power in wind farm clusters based on deep reinforcement learning is proposed. Predict the potential power of a single wind farm and consider the influence of wake effects. Use a wake model to correct the predicted power value and obtain the dynamic power demand of the wind farm cluster. Maximizing the active power output, minimizing the wind power curtailment rate, and minimizing the active power loss of the wind power cluster are taken as the objective functions, while system safety operation and power tracking are taken as constraints. Modeling the optimization allocation problem of active power in wind farm clusters as a fully collaborative multi-agent task, by designing joint actions of intelligent agents and calculating the target Q value, learning the optimal power optimization allocation strategy, and obtaining the intelligent agent power adjustment amount, the optimal allocation of active power in wind farm clusters is achieved. The experimental results show that applying the method proposed in this paper for optimizing the allocation of active power in wind farm clusters results in a cluster wind abandonment rate of less than 5%, significantly improving the utilization of wind energy resources.