Application Of Support Vector Machine Based On Simulated Annealing Algorithm In MBR Membrane Pollution

Membrane bioreactor(MBR)is an important way in the sewage treatment process.Which has the advantages of high efficiency and water quality,is easy to implement automatic control.The scope of application and scale is increasing year by year.Which has aroused the attention of many countries.However,membrane fouling is becoming a major obstacle to the rapid development of MBR.Membrane fouling will lead to a decrease of membrane flux.Therefore,It has become a research focus of how to reduce MBR membrane fouling effectively.In this paper,based on the study of various models in the field of MBR before,we can find that the traditional neural network model is easy to fall into the local minimum and the parameter is difficult to be determined in MBR membrane fouling research.This paper proposes a model based on SA-SVM support vector machine for MBR membrane pollution prediction using simulated annealing algorithm and support vector machine.Firstly,the simulated annealing algorithm is applied to global optimization of the three important parameters of support vector machine,including penalty factor,insensitivity coefficient and kernel parameter.Then,using the optimal parameters as initial parameters of support vector machine to build the MBR membrane pollution prediction model.Finally,principal component analysis(PCA)was used to analyze the factors affecting membrane fouling,and the main influencing factors were selected as the input of the model and the flux of membrane as output.Experimental results show that the MBR membrane pollution prediction model based on SA-SVM support vector machine has better fitting effect and higher prediction precision when predicting membrane flux,at the same time,the stability and generalization ability are also improved Compared with the neural network model.In the process of initial parameter optimization of MBR membrane pollution prediction model using simulated annealing algorithm,we also find some problems,SA algorithm has some disadvantages such as slow convergence speed and relatively sensitive initial parameter setting.So we introduce a hybrid optimization algorithm which combines SA and GA to optimize the parameters of the model.This algorithm not only retains the global searching ability of GA,but also has the advantages of local optimization of SA algorithm.Experimental results show that,compared with the SA-SVM model,the ASAGA-SVM model has a high fitting effect,with an average relative error of 0.0263.We can conclude that,in the face of small sample flux data,The MBR membrane pollution prediction model based on ASAGA-SVM has better prediction accuracy than the SA-SVM model.