Research On Soft Sensing Modeling Of Wastewater Based On Improved Least Squares Support Vector Machine

Nowadays,it is necessary and urgent for us that which the scarcity of water resources per capita and the serious pollution of water.Sewage treatment is a complicated industrial process integrating nonlinearity,hysteresis,inertia,etc,which's complex ecological environment and diversified technological processes directly lead to be particularly difficult for the detection of water quality.However,existing detection technologies which cannot be used to reliably detect all types of water quality parameters in real time are subject to cost,the constraints of accuracy and other factors.However,with the development of traditional technology,soft measurement technology has emerged and successfully applied to the actual process of sewage treatment,and also has achieved very good results.In this paper,the biological aerated filter is selected to be the research object,and a soft measurement model of sewage treatment based on the prediction of effluent COD concentration is established,to realize the prediction of the effluent COD concentration,and it will be the effluent guarantee for the accurately determined and advanced prediction.The main contents are as follows:Firstly,this paper selects the LSSVM based on SVM improvement as the basic algorithm,and establishes a soft-sensing model based on LSSVM,which greatly increases the computational efficiency,but the accuracy of the model has decreases slightly.Therefore,in view of this drawback,this paper chooses to use the improved particle swarm optimization algorithm to optimize the parameters from the perspective of hyper parameters,and to improve the accuracy of the model to some extent.Compared with simulation analysis,the improved model will have a great improvement in the accuracy of the model.However,based on the drawback that LSSVM treats each kind of sample equally and is susceptible to noise sensitive points,this paper introduces the method of fuzzy membership degree?_i and adds a membership degree to each sample selection,which has established a soft sensor model based on fuzzy LSSVM,and introduces two different function models that select membership degrees are used to optimize and optimize the model.The simulation results show that the fuzzy LSSVM model not only has a certain degree of accuracy in comparison with the common model,but also has the ability to face outliers of thegreat improvement.The main contributions and innovations of this topic are:This paper chooses to optimize the LSSVM model from two aspects of parameter optimization and fuzzy regression.The improved particle swarm algorithm reasonably has selected the hyper parameters combination needed by the model,greatly improves the accuracy of the model.The addition of the fuzzy algorithm treats various sample data differently,which not only improves the accuracy of the model,but also greatly stabilizes the model.The combination of the two aspects together have realized real-time prediction of the effluent COD concentration.