摘要
针对污水处理过程(wastewater treatment process,WWTP)中溶解氧质量浓度和硝态氮质量浓度的最优控制问题,提出了一种基于迭代自适应动态规划(adaptive dynamic programming,ADP)算法的最优控制策略.该策略无须知道污水处理过程的非线性动力学模型,只需污水处理系统的输入输出观测信息,设计基于ADP强化学习原理的控制体系结构,并利用神经网络辨识特性,通过在线迭代来逼近性能评价指标和最优控制策略.实验结果表明:该控制器相对于传统的PID控制策略,提高了污水处理过程的控制精度,系统鲁棒性也明显增强,控制性能更优.
In order to improve the control performance of the dissolved oxygen （DO） concentration and nitrate nitrogen （Sno） concentration in a wastewater treatment plant （ WWTP） ,an optimal control strategy based on iterative adaptive dynamic programming （ADP） algorithm was proposed. It only need the input and output of the non-linear dynamics model of the control system, and approximate the performance evaluation index and optimal control strategy by networks identifying the control system. The learning algorithm of controler was implemented on line in the benchmark simulation model （BSM1）. Finally,the experiments validate the effectiveness of the proposed control strategy in improving the control precision and system robustness compared to the traditional PID control strategy.
出处
《北京工业大学学报》
CAS
CSCD
北大核心
2018年第2期200-206,共7页
Journal of Beijing University of Technology
基金
国家自然科学基金重点资助项目(61533002)
北京市科学技术委员会科技计划资助项目(Z141101004414058)
关键词
自适应动态规划
性能评价指标
最优控制策略
迭代自适应动态规划(ADP)算法
adaptive dynamic programming
performance evaluation index
optimal control strategy
iterative adaptive dynamic programming （ ADP） algorithm
