ANFIS在混凝投药前馈控制器中应用的仿真研究

Simulative research of ANFIS applied in coagulant dosing feedforward controller

在水处理中混凝投药前馈控制器的应用效果好坏关键在于控制器是否对混凝投药过程具有良好的模型辨识能力,传统的控制器效果都不太理想,而且存在沉淀池出水浊度波动大,药剂浪费严重等问题。为了解决该问题,介绍了一种用多层前馈神经网络优化模糊逻辑系统的自适应模糊推理系统——ANFIS。它具有良好的非线性函数逼近能力,在ANFIS投药前馈控制器的设计中,运用减法聚类对样本数据进行空间划分,获取初始模糊隶属函数和模糊规则,得到ANFIS模型的初始结构。用烧杯试验历史数据进行了仿真验证,并与传统的回归模型前馈投药控制仿真比较,结果表明ANFIS投药前馈控制模型明显优于回归模型,它能够根据原水水质适时有效预测混凝投药量。

In the water treatment process,the key of applied quality of coagulant dosing feed forward controller lies in whether the controller has a good ability to model identification for coagulant dosing process.But the results of all traditional coagulant control schemes are not satisfying.There are some serious questions of chemicals wastage,and the turbidity of output water of sediment pool fluctuates quickly.In order to solve the problem, neural network control（NNC） and fuzzy control of IC,adaptive-network-based fuzzy inference system（ANFIS）were introduced.It had a good ability of non-linear function approximation.In the design of ANFIS scheme,the data of sample was classified by subtractive cluster method,and some fuzzy membership functions and rules were obtained,finally a initial ANFIS structure was established.ANFIS control model were simulated and tested by using jar-test historical data,and compared with the simulation of the traditional regression model.The results of comparison suggested that ANFIS feed-forward control model was distinct superior to regression model.According to the quality of raw water,it could predict coagulant dosage effectively in time.