基于深度强化学习的电机S-曲线转速控制方法  

S-curve Speed Control Method of Motor Based on Deep Reinforcement Learning

作　　者：毛德辉 李鹏[1] 吴建德 MAO Dehui;LI Peng;WU Jiande(Department of Information,Yunnan University,Kunming Yunnan 650000,China;Yunnan Key Laboratory of Intelligent Systems and Computing,Yunnan University,Kunming Yunnan 650500,China)

机构地区：[1]云南大学信息学院,云南昆明650000 [2]云南大学云南省智能系统与计算重点实验室,云南昆明650500

出　　处：《机床与液压》2025年第5期88-94,共7页Machine Tool & Hydraulics

基　　金：云南省重大科技专项项目(202202AD080005);云南省应用基础研究企业联合专项项目(202101BC070001-012);云南省中青年学术和技术带头人后备人才培养计划(202105AC160094);云南大学第十五届研究生科研创新项目(KC-23236408)。

摘　　要：针对电机暂态控制产生冲击和抖动导致的设备易损坏问题,提出一种基于深度强化学习的电机S-曲线转速控制方法。采用Sigmoid函数设计电机转速控制速度规划曲线,降低加速度不连续变化的冲击影响;基于Actor-Critic框架,结合深度确定性策略算法(DDPG)和混合形式的奖励函数搭建深度强化学习控制器,建立动态奖惩机制;通过MATLAB/Simulink仿真分析,将DDPG控制器与模型预测控制器、PID控制器、BP-PID控制器及GA-PID控制器分别在电机空载与负载时的控制效果进行对比。实验结果表明:DDPG控制器可以实现电机平滑加减速控制,并且可以消除超调量,具有更快的响应速度;当存在外部负载扰动时,电机转速跌落减少了50%~75%,极大提高了系统的抗扰动能力。In view of the equipment vulnerability caused by impact and vibration during motor transient control,a motor S-curve speed control method based on deep reinforcement learning was proposed.The core idea of the proposed method was to mitigate the impact of discontinuous changes in acceleration by utilizing the Sigmoid function to design the motor′s acceleration and deceleration control signals.A deep reinforcement learning controller was designed based on the Actor-Critic framework,in combination with the deep deterministic policy gradient(DDPG)algorithm.A hybrid reward function was employed to enable a dynamic reward-penalty mechanism.Finally,MATLAB/Simulink simulation was made to compare the performance of the DDPG controller with model predictive control,PID control,BP-PID control,and GA-PID control under both no-load and loaded conditions for the motor across various operational scenarios.The experimental results show that the DDPG controller enables smooth acceleration and deceleration control of the motor,while effectively eliminating overshoot and achieving a faster response speed.Additionally,in the presence of external load disturbances,the motor speed drop is reduced by from 50%to 75%,which greatly enhancing the system′s disturbance rejection capability.

关 键 词：深度强化学习 S-曲线 暂态控制 转速控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] TM383.6[自动化与计算机技术—控制科学与工程]