Research On Soft Measurement In Sewage Treatment Effluent COD Based On HS-RBF

Sewage treatment system is a complex nonlinear system involving several areas including physics,chemistry,environmental science etc.COD(Chemical Oxygen Demand)is an important index of showing the degree of organic contamination in water directly.Because of the complexity of various factors and conditions in sewage treatment and the barriers of instrument measurement technique development level,some domestic and overseas manufacturers has developed COD rapid measuring and online instruments by far,but they are quite high priced,expensive to maintain,easy to be influenced by water quality,with great error and hard to be widely promoted.Aiming at this problem,this paper comes up with a COD soft-measuring model based on HS-RBF(Harmony Search,HS)neural network,and achieves prediction and estimation of effluent COD.The main work is summarized as follows:1.Based on introducing the technique development of sewage treatment measurement and soft measuring,this paper gained the auxiliary variables by indepth research and analysis on sewage treatment system craft and theory.The paper processed the collected data by normalization and principle component analysis,then reduced the dimension of auxiliary variables from 7 to 5,and built soft-measuring models based on RBF and BP neural networks individually.The research finding shows that the soft-measuring model has a better prediction ability after dimension reduction and RBF neural network has higher accuracy of prediction and generalization ability than BP neural network.2.For the problem of RBF neural network hidden layer relevant parameters making a great influence on the generalization ability of the function,this paper bring harmony search algorithm into optimize it.To shorten the convergence time,enhance the global search ability of the algorithm in optimization process,the paper comes up with GADHS(Global-best Additional Disturbed Harmony Research)algorithm based on the summary of basic harmony search algorithm and former researches.Compared with several improved harmony search algorithm in benchmark function tests,it performed a better search ability.3.Using the improved algorithm to optimize the data centers and extensive constant in hidden layer and the weights in output layer.The results of research and simulation showed that the accuracy,stability and generalization of the model had been improved for bringing the global optimal dynamic harmony search algorithm in RBF neural network to optimize while using RBF neural network to achieve soft measuring modeling and prediction of COD in wastewater treatment process,and GADHS had the best prediction ability.By comparing with the data measured by wastewater treatment works.This model has high prediction accuracy,and meets the relevant specifications and requirements.The research in this paper is of significant reference value referred in research of online measuring in sewage treatment effluent quality.