Research On Continuous Feeding Model Of Anaerobic Digestion Based On SVM

The technological progress of anaerobic digestion is of great significance to the resource utilization of organic waste and the realization of carbon neutrality.This paper studies the control strategy of feeding and discharging process of anaerobic digestion system.The research work includes two progressive parts.First,starting from the chemical kinetic mechanism of anaerobic digestion,using MATLAB software to fit the experimental results of the ADM1 model,and discuss the feasibility of using the ADM1 mathematical model to design an intelligent control scheme.By calibrating the k _La value to simulate the change process of CH₄ and CO₂ gas production,the obtained cumulative gas production model has a good fit,which is higher than 0.98.It can be seen that the calibrated ADM1 model can reflect the gas production law of the anaerobic digestion process with batch feeding and discharging.However,if the actual engineering situation is complex,the composition of biomass waste is complex,and the degree of pretreatment is different,it is very difficult to design the mechanism-based calibration of the ADM1 model.On the basis of biochemical dynamics research,this article attempts to use machine learning methods to take into account the influencing factors of the digestion process on site.Take a variety of biomass anaerobic digestion co-fermentation experimental system as the research object,this paper designs a continuous feeding model of anaerobic digestion system based on support vector machine（SVM）by designing a digestion experiment with sequential batch feeding and analyzing data samples.The biogas production,CH₄concentration and CO₂ concentration of the self-made anaerobic digestion reactor were classified and predicted,the timing of feeding was judged and selected,and the gas production of anaerobic digestion was regression predicted.The accuracy rate of SVM algorithm classification verification reached 0.98,and the average error percentage of the regression prediction of single-day gas production was 0.13.Through the fitting of the above-mentioned calibrated ADM1 model and the research and verification of the SVM model of the small sample anaerobic digestion process,this paper proposes the continuous feeding control strategy of anaerobic digestion.Combining with the real-time biogas production volume and gas composition data monitored online,the status and stage of the anaerobic digestion process can be judged,so that more flexible and accurate feeding and discharging countermeasures can be adopted.On the other hand,by regularly returning to expand the number of training samples,and regularly verifying the algorithm model,screen out the interference factors in the new complex digestion system.