基于参考输入优化的变速风电机组最大化风能捕获方法

Maximizing wind energy extraction for variable-speed wind turbines based on the optimization of reference input

中文关键词  风电机组  最大化风能捕获  最大功率点跟踪  慢动态特性  参考输入优化  强化学习

英文关键词  wind turbines  maximizing wind energy extraction  maximum power point tracking  slow dynamic behavior  optimization of reference input  reinforcement learning

中文摘要

      变速风电机组在额定风速以下应用最大功率点跟踪实现最大化风能捕获. 然而, 大惯量风电机组在面对快 速波动的湍流风速时, 因转速调节慢而难以保持运行于最大功率点. 本文研究进一步发现, 平均转速跟踪误差与整 体的风能捕获效率并非单调关系, 这使得当前以减小转速跟踪误差为目标的控制器设计难以有效提升风电机组的 发电效率. 为此, 本文以提升风能捕获效率(而非减小转速跟踪误差)为目标, 提出一种基于参考输入优化的风电机 组最大化风能捕获方法. 考虑到参考转速对风能捕获效率的复杂影响难以准确建模, 本文借助深度确定性策略梯度 (DDPG)强化学习算法实现参考输入优化. 仿真结果表明该方法能够有效提升湍流风下变速风电机组的风能捕获效 率.

英文摘要

      Variable-speed wind turbines (VSWTs) are expected to maximize their power extraction via maximum power point tracking (MPPT). However, turbines with large inertia are unable to track the optimal rotor speed which continuously fluctuates depending on instantaneous wind speed, leading to the decline in wind energy extraction efficiency. It is found that the average speed tracking error is not monotonically related to the overall wind energy extraction efficiency. This makes it difficult for the MPPT controllers which are designed aiming to reduce the speed tracking error to effectively improve the wind energy extraction efficiency of the turbines with slow dynamic characteristics. Therefore, in order to improve the efficiency of wind energy capture (rather than reduce the speed tracking error) as the goal, this paper proposes a wind turbine maximum wind energy capture method based on reference input optimization. The optimization of reference input is realized with a reinforcement learning algorithm called deep deterministic policy gradient (DDPG), meeting the challenge of the complex effect of reference on performance. The simulation results show that the proposed method can effectively improve the efficiencies of VSWTs under turbulence.