Research And Application Of Hybride Soft-sensing Model For Anaerobic Digestion Process

With the continuous improvement of the degree of industrialization in china,water pollution problem is particularly prominent,and treating is still a heavy task.Anaerobic digestion(AD)technology process is becoming more and more important in wastewater treatment,because it has many advantages,such as high efficiency,low sludge yield and biogas production.However,anaerobic digestion is a complex,nonlinear,biochemical process.In particular,methanogenic bacteria are very sensitive to changing surroundings.Therefore,AD process control and analysis is believed to be one of the main limitations for effective organic matter degradation.Recently,computer technology,automation technology and artificial intelligence(AI)technology are gradually used in wastewater treatment process,and people have achieved good results.And with the in-depth study of these technologies,soft-sensing technology to solve the above problems provided a new way of thinking.Soft-sensing model design,parameter optimization and so on are getting more and more attention,and is an important research topic in the current wastewater treatment monitoring and control field.Base on the all-around review and analysis of the progress of wastewater treatment study,the theories of the soft-sensing technology,including the theories of the support vector machine(SVM),the Particle Swarm Optimization(PSO)and the microbial dynamics,this study systematically study the modeling idea and method of soft sensor model based on SVM.And in this study,a soft-sensing model of COD removal rate and effluent VFA concentration in anaerobic treatment of waste water based on the PSO-SVM model was established,and a mixed soft-sensing model of gas production based on microbial kinetics and the PSO-SVM model was established.Some pioneering and exploratory work on the soft-sensing model used in wastewater treatment was studied in this study.Main contents and results are as follows:1.An anaerobic treatment system based on the IC reactor and the monitoring system with related indexes was built in the laboratory.Activated sludge and granular sludge were used as inoculated sludge to activate two IC anaerobic reactors,respectively.The reactor,which used the granular sludge as inoculum sludge,runs smoothly in the whole start-up process.While,in the early stage of the whole start-up,the other reactor fluctuated greatly.The gas production was less than 0.50L/h,and the pH fluctuated greatly,the minimum pH was up to 5.50.With the granulated process finished,the reactor was gradually stabilized.2.The reactor,which used the granular sludge as inoculum sludge,was operated for 60 d under different operating conditions.And we collected 159 sets of data.When the alkalinity of the anaerobic system is low,the pH is correspondingly low,and the VFA concentration rises accordingly.In this situation,the system is easy to rancidity phenomenon.At the same time,in order to reduce the acidity,the alkalinity of the wastewater should be increased appropriately.3.Based on the analysis of particle swarm optimization(PSO)and the principle of support vector machine,the classification strategy of original data set was introduced in the soft-sensing model.And a soft-sensing model of COD removal rate and effluent VFA concentration in anaerobic treatment of waste water based on the PSO-SVM model was established.Before the original data set being classified,the results of the PSO-SVM model shows well.The correlation coefficient(R)is 65.86% and 85.25 for the COD removal rate and total VFA concentration prediction,respectively.After the data-classification strategy implying,the metadata set divided into two sets,and the performance of the hybrid model was significantly improved.The R of the COD removal rate modeling testing set is 92.34% and 83.41%,respectively.And The R of the total VFA concentration modeling testing set is 99.14% and 99.59%,respectively.4.Based on the analysis of microbial kinetics model,some microbial kinetics equations were intruded in the hybrid model.Thus,a mixed soft-sensing model of gas production based on microbial kinetics and the PSO-SVM model was established.The original data set is roughly evenly distributed after the kinetic model modified,and the messiness and noise of the original data set were controlled.The R value of the traditional model is 86.71%.In particularly,the kinetic model quantified the effects of pH,temperature and VFAs on AD process.And compared with traditional model,the performance of the hybrid model promote a lot,the R value of the hybrid model is 95.73%.