引用本文:吴敏,王晓璐,姜玉东,钟磊,莫飞杨.深度确定性策略梯度与模糊PID的协同温度控制[J].控制理论与应用,2022,39(12):2358~2365.[点击复制]
WU Min,WANG Xiao-lu,JIANG Yu-dong,ZHONG Lei,MO Fei-yang.Collaborative temperature control of deep deterministic policy gradient and fuzzy PID[J].Control Theory and Technology,2022,39(12):2358~2365.[点击复制]
深度确定性策略梯度与模糊PID的协同温度控制
Collaborative temperature control of deep deterministic policy gradient and fuzzy PID
摘要点击 1678  全文点击 451  投稿时间：2021-09-16  修订日期：2023-02-25
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2022.10874
  2022,39(12):2358-2365
中文关键词  模糊PID  温度控制  协同控制  DDPG  遗传算法
英文关键词  fuzzy PID  temperature control  collaborative control  DDPG  genetic algorithm
基金项目  江苏省自然科学基金项目(BK20151099)资助.
作者单位E-mail
吴敏* 金陵科技学院机电工程学院 wumin1@jit.edu.cn 
王晓璐 金陵科技学院机电工程学院  
姜玉东 金陵科技学院机电工程学院  
钟磊 江苏时代新能源科技有限公司  
莫飞杨 澄瑞电力科技(上海)有限公司  
中文摘要
      针对现有温度控制系统控温时间长、误差大的问题, 本文提出了一种基于深度确定性策略梯度(DDPG)和模糊自整定PID的协同温度控制. 首先, 模糊PID在控制大滞后系统时, 控制器不能立刻对产生的干扰起抑制作用, 且无法保证大滞后系统的稳定性等问题, 本文建立了模糊PID和DDPG算法相结合的温度控制模型, 该模型将模糊PID作为主控制器, DDPG算法作为辅助控制, 利用双控制器模型实现温度协同控制. 接着, 利用遗传算法对模糊PID的隶属函数和模糊规则进行寻优, 获得模型参数最优解. 最后, 在仿真实验中验证所提方法的有效性. 仿真实验结果表明, 本文提出的算法可有效减少噪声干扰, 减小控制系统的响应时间、误差和超调量.
英文摘要
      The existing temperature control system takes a long time to control the temperature and has large errors, aiming at which this paper proposes an algorithm of temperature collaborative control. The algorithm is based on deep deterministic policy gradient (DDPG) and fuzzy self-tuning PID. First of all, when fuzzy PID controls a large-lag system, the controller cannot suppress the disturbance immediately, and it cannot guarantee the stability of the large-lag system. In this paper, we establish a temperature control model combining fuzzy PID and DDPG. In this model, fuzzy PID is used as the main controller while DDPG is used as the auxiliary one, so that the dual-controller model can realize the goal of temperature cooperative control. Then, we use genetic algorithm to find the optimal solution of the membership function and fuzzy rules of fuzzy PID, by which we can obtain the optimal solution of model parameters. Finally, the effectiveness of our method is verified in simulation experiments. The results show that the algorithm proposed in this paper can effectively reduce the noise interference and the response time, errors and overshoot of the control system.