废水中和过程的RBF神经网络预测控制  

RBFNN Predictive Control of Wastewater Neutralization Process

作　　者：周洪煜[1] 梁东义 周松杰 

机构地区：[1]重庆大学动力工程学院,重庆400030 [2]商丘裕东发电厂,河南永城476600

出　　处：《控制工程》2014年第1期79-83,共5页Control Engineering of China

基　　金：重庆市科委重大科技攻关项目(CSTC2009AB108)

摘　　要：废水中和过程有较强的非线性、时变性和滞后特性,对于过程模型的辨识与控制较为困难,采用常规的线性化模型或传统PID控制方法存在模型过于复杂,算法难以在线实施,控制精度不能保证等问题,很难取得满意的控制效果、针对该问题,应用酸碱中和的强酸当量模型,提出了一种基于敏感度(Sensitivity Analysis,SA)和动态粒子群优化算法(Dynamic Particle Swarm Optimization,DPSO)相结合的RBF神经网络模型辨识方法,通过调整网络结构和辨识出系统的滞后时间来提高模型辨识效率和预测精度,将RBF神经网络辨识器与神经网络控制器相结合构成电厂废水处理pH中和过程的预测控制系统。经过仿真研究和试验验证,与电厂实际应用的PID控制方法相比较,该方法能有效地对pH值进行控制,并实现较小的控制误差和节约药剂的效果。Wastewater neutralization process has strong nonlinear, delay, time- variation property. The identification and control of this process is very difficult. The application of common linear model or PID control will encounter problems, like serious complexity of model, hard to be online conduct and low accuracy in control. T0 solve this problem, an RBFNN identification method based on SA - DPSO algorithm is proposed applied on model of strong acid equivalent of pH neutralization process. It can raise the identification effi- ciency and model prediction accuracy through identification of delay time. The combination of RBFNN identifier and predictive controller realized predictive control to pH neutralization process of wastewater. Compared with the PID control method applied in power plant, the simulation results show that the method is efficient for pH value control, and it can obtain the effect of less error and save medicament.

关 键 词：迟延系统 pH中和 SA—DPSO算法 模型辨识 预测控制 RBF神经网络 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]