摘要
针对污水处理是个非线性、时变性、滞后性和不确定性的系统,利用软测量技术以进水化学需氧量(COD)、进水氨氮、进水p H值和曝气池溶解氧(DO)为输入建立RBF神经网络预测模型,并用遗传算法(GA)对RBF神经网络中的中心矢量值C和网络权值W进行优化,来预测出水参数值.以预测得到的出水参数为系统的反馈信号,通过BP神经网络动态优化溶解氧的设定值.最后采用自适应PID控制方法,对溶解氧的设定值进行跟踪优化.仿真结果表明,采用遗传算法优化的RBF神经网络软测量模型对出水COD具有很好的预测效果,基于神经网路的预测、优化和自适应PID控制的溶解氧智能控制系统具有动态跟踪性能好、控制误差小、速度快的优点.
The wastewater treatment is a nonlinear,time-varying,hysteretic and uncertain system.Using soft-sensing technology,RBF neural network prediction model was established with influent COD,influent ammonia nitrogen,influent pH value and aeration tank DO as input.Using genetic algorithm(GA),the center vector value C and network weights W was optimized to predict the effluent parameter value in RBF neural network.Then the predicted effluent parameters were the feedback signals of the system.The BP neural network was used to dynamically optimize the set value of dissolved oxygen.Finally,an adaptive PID control method was used to track and optimize the set value of dissolved oxygen.The simulation results show that the RBF neural network soft-sensing model optimized by genetic algorithm has good prediction effect on effluent COD.The dissolved oxygen intelligent control system based on neural network prediction,optimization and adaptive PID control has good dynamic tracking performance with small control error and high speed.
作者
胡赛飞
李明河
陈甫前
HU Saifei;LI Minghe;CHEN Fuqian(School of Electrical and Information Engineering, Anhui University of Technology, Ma..anshan, Anhui 243002, China)
出处
《宜宾学院学报》
2018年第12期56-60,共5页
Journal of Yibin University
