作　　者： ; ; ; ;

机构地区： 华南理工大学

出　　处： 《微特电机》 2013年第3期52-55,共4页

摘　　要： 针对双馈感应风力发电系统结构复杂,受参数变化和外部干扰较显著,具有非线性、时变、强耦合的特点,在传统矢量控制的基础上,提出一种基于强化学习的自校正控制方法。引入Q学习算法作为强化学习核心算法,快速自动地在线优化PI控制器的输出。基于MATLAB/Simulink环境,在风速低于额定风速时对系统进行仿真,结果表明,引入强化学习自校正控制后,保持了原系统最大风能捕获的能力,同时改善了其动态性能,增强了鲁棒性和自适应性。 The doubly-fed induction wind power generator is a complex, strong coupling,time-varying and highly non- linear system, whose performances are easily affected by parameter changes and external disturbances. On the foundation of the traditional vector control, a self-tuning control method based on reinforcement learning （RL） theory was proposed. The Q-learning method was introduced as the core algorithm of reinforcement learning to optimize the control signals of the PI controller through rapidly automatic on-line learning. The simulation was conducted in MATLAB/Simulink when wind speed was under the rated value. The results show that the control system with an additional self-tuning controller maintains the a- bility to capture the maximum wind power, while improving its dynamic performance and enhancing the robustness and a- daptability, compared with the traditional vector control.

关 键 词： 风力发电 双馈感应发电机 强化学习 学习算法 自校正控制

领　　域： [电气工程]