热力发电2025，Vol.54Issue(9)：86-94,9.DOI:10.19666/j.rlfd.202501006
基于强化学习的风电机组偏航系统模型预测控制
Predictive control of wind turbine yaw system model based on reinforcement learning
摘要
Abstract
It is crucial to improve the dynamic performance of the yaw system of wind turbines in multiple operating scenarios.Therefore,a predictive control strategy for wind turbine yaw system model based on reinforcement learning is proposed,which achieves multi-objective parameter dynamic optimization through the dual-delay depth deterministic policy gradient(TD3)algorithm.Firstly,a multi-step model predictive controller for the yaw system(YMPC)is established to address the conflicting control objectives of power loss rate and yaw actuator utilization rate.Secondly,based on the optimization objectives and wind conditions of the yaw system,a dual-delay depth deterministic strategy gradient(TD3)intelligent agent is designed to determine the input state,action,and reward mechanism of the YMPC.The TD3 intelligent agent is then used to tune the weight coefficients and control step size of the YMPC.Finally,the effectiveness of this method was validated using typical daily data from wind farms in northern China.The results indicate that the proposed strategy significantly improves the overall performance of the yaw system compared with the YMPC with fixed control parameters.关键词
风电机组/偏航系统/强化学习/控制参数Key words
wind turbines/yaw system/reinforcement learning/control parameter引用本文复制引用
桑申刚,李桂朋,王向伟,刘毅,王森,申向荣..基于强化学习的风电机组偏航系统模型预测控制[J].热力发电,2025,54(9):86-94,9.基金项目
中国华能集团有限公司总部科技项目(HNKJ22-HF69)Science and Technology Project of China Huanneng Group Co.,Ltd.(HNKJ22-HF69) （HNKJ22-HF69）
