Research On The Prediction Model Of Anaerobic Fermentation Gas Production Based On Mixed Kernel Function LS-SVM

The double crisis of energy and environment makes large and medium-sized biogas projects become the focus and key point for the development of green energy.Using this important means can reduce the pollution caused by livestock waste and its economic and environmental benefits are self-evident,while how to achieve high efficiency is very important.However,anaerobic fermentation is a very complex process of microbial metabolism,which is nonlinear.Due to the lack of advanced equipment or equipment for online monitoring,it is difficult to realize online monitoring of key factors.It severely restricts the efficiency and steady-state performance of anaerobic fermentation systems,making it difficult to achieve further optimization and control.It is a long-term goal for the research design and operation management to make the anaerobic fermentation system run under the condition of high efficiency and steady state.And gas production plays an important role in improving the stability and efficiency of anaerobic fermentation.So it is of great significance to establish a prediction model for biogas production in anaerobic fermentation.In this paper,a laboratory mixed anaerobic fermentation system is used as the research object.Data collection and treatment of influencing factors based on the theory of anaerobic fermentation mechanism,and use grey correlation analysis to determine the input variables of the model.Secondly,the BP neural network and LS-SVM in machine learning are used to establish the prediction model of anaerobic fermentation gas production.It is concluded that LS-SVM has higher predictive ability than BP neural network for the finite sample.Then we do a further study on LS-SVM,and use the polynomial kernel function and the Gauss radial basis kernel function to construct the LS-SVM model of the mixed kernel function to predict the gas production.After that,the mechanism of anaerobic fermentation is integrated,and the fermentation data is effectively processed through the mechanism function.Then the prediction model is built through the mixed kernel LS-SVM,and the kernel parameters are optimized by genetic algorithm.The simulation results show that the combined kernel LS-SVM prediction model combined with the mechanism functionhas a more accurate prediction of gas production and a smaller error than the LS-SVM model of mixed kernel function.The prediction accuracy is 89.42%.Finally,based on the advantages of LabVIEW interface design,it combines it with MATLAB to design a anaerobic fermentation gas production prediction system.Input variables can be used to predict gas production,and it can provide theoretical basis for setting up anaerobic fermentation parameters and increase production of gas production for large and medium-sized biogas engineering in the future.