电力系统及其自动化学报2025，Vol.37Issue(2)：68-77,10.DOI:10.19635/j.cnki.csu-epsa.001464
计及尾流的改进深度确定性策略梯度风电场功率优化控制策略
Improved Deep Deterministic Policy Gradient for Wind Farm Power Optimal Control Strategy Accounting for Wake Effect
摘要
Abstract
Compared with the conventional maximum power point tracking for one single machine,it is more practical to consider the wake effect of upstream turbines on downstream turbines,thus optimizing the overall output power from a wind farm.In this paper,an optimal control strategy for the wake effect in wind farms is proposed on the basis of deep reinforcement learning.The improved deep deterministic policy gradient(DDPG)algorithm is used to optimize the yaw angles of turbines in dynamic wind farms.Based on the conventional DDPG,the prioritized experience replay mecha-nism is designed to replace the random sampling method.The importance of experience samples in the sample pool is ranked,so that the agent can learn the experience which is more valuable in priority to improve the optimization effi-ciency and performance.Meanwhile,delayed updating actor network is applied to speed up the action function and find the optimal solution.Finally,the simulations of a dynamic wind farm WFSim show that under the proposed control strat-egy,the wake distribution is effectively optimized,and the overall output power from the wind farm is improved,which can also be adapted to different wind speeds and layouts.关键词
风电场/尾流效应/偏航控制/深度强化学习/功率优化Key words
wind farm/wake effect/yaw control/deep reinforcement learning/power optimization分类
信息技术与安全科学引用本文复制引用
张勇,刘春,褚梦珂,杨兴武..计及尾流的改进深度确定性策略梯度风电场功率优化控制策略[J].电力系统及其自动化学报,2025,37(2):68-77,10.基金项目
上海市科技计划资助项目(23010501200). （23010501200）
