Mixed Multi-modeling Soft Measurement Research On Biological Aerated Filter Sewage Treatment

Sewage treatment process has the characteristics of strongly nonlinear, time delay and large inertia. Sustained and stable operation of sewage treatment control system is the key to guarantee the quality of treated water, but some water quality parameters testing equipment are so expensive or poor real time detection on line that it is of practical significance to do the research of soft measurement based on sewage disposal.All of the research in this article which joint commitment with a well-known environmental science and technology company are based on science and technology significant project of Anhui Province. In this essay, a study has been done on the key parameters of sewage treatment process based on Least square support vector machine(LS-SVM) method, soft-sensing prediction model of COD(chemical oxygen demand) concentration for exiting water from biological aerated filter exposing pool has been established. COD concentration of exiting water provides the judgment basis of water quality ahead of schedule.The main research contents including:Firstly, the article select Polynomial kernel function, RBF kernel function and Sigmoid kernel function to compose a combined kernel function, and then put forward a new method to settle the combined kernel function coefficient through the use of mean square error of single kernel function LS-SVM, so the COD concentration soft sensor prediction model of exiting water has been established based on improved mixed kernel function LS-SVM.Secondly, on the basis of improved kernel fuzzy C clustering algorithm, this essay classifies the data acquired from the scene, establish multiple kernel LS-SVM model of sub data class, and obtain the output of multiple predictive model. A COD concentration multiple model LS-SVM soft sensor prediction model of exiting water has been established based on improved clustering algorithm eventually.Afterwards, by using multiple modeling method to remedy the prediction error of COD mechanism model which derived on the principle of biochemical reaction of organic compounds and materials conservation relationship, also the influence of uncertain factors in the system, a hybrid prediction model that integrated with BAF mechanism model was established.Finally, the research contents of this paper are summarized, the deficiencies arepointed out and the prospects are put forward. Simulation results show that the improved models improve the prediction accuracy certainly.The contributions of this paper are put forward a new method to determine the coefficient of mixed kernel function, improve clustering validity function, build a prediction model of effluent COD of combined kernel function LS-SVM base on the improved clustering algorithm, and propose a mixed multi-model effluent COD concentration prediction method associated with the BAF mechanism model, so as to make more accurate prediction.