基于深度强化学习的同步发电机最优励磁控制方法

Optimal Excitation Control Method for Synchronous Generators Based on Deep Reinforcement Learning

常规的同步发电机最优励磁控制方法主要使用ARX(AutoRegressive with eXternal input)自回归各态经历模型生成控制传递函数,易受采样周期离散化作用影响,导致控制扰动过高,因此提出基于深度强化学习设计一种全新的同步发电机最优励磁控制方法。构建了同步发电机最优励磁控制模型,利用深度强化学习设计了同步发电机励磁模糊控制器,优化了同步发电机最优励磁控制参数,从而实现了同步发电机最优励磁控制。实验结果表明,同步发电机深度强化学习最优励磁控制方法在不同状态下的控制扰动均较低,说明控制效果较好,具有可靠性,有一定的应用价值。

The conventional optimal excitation control method for synchronous generators mainly uses ARX(AutoRegression with eXternal input)to generate control transfer functions through various state experience models,which is susceptible to the influence of sampling period discretization,leading to excessive control disturbances.Therefore a novel optimal excitation control method for synchronous generators based on deep reinforcement learning is proposed.A model for optimal excitation control of synchronous generators is constructed,and a fuzzy controller for excitation of synchronous generators is designed using deep reinforcement learning.The optimal excitation control of synchronous generators is achieved by optimizing the control parameters.The deep reinforcement learning-based optimal excitation control method for synchronous generators is indicated by experiments to achieve lower control disturbances in different states,and thus have good control utility and reliability,and certain application value.