The Study And Application Of Intelligent Identification Method For Membrane Fouling Of MBR

Membrane bioreactor(MBR)has the advantages of convenient management and flexible operation.However,membrane fouling is an inevitable problem in wastewater treatment process(WWTP)with MBR.The membrane fouling will cause the separation performance reduction of MBR,effluent quality,and even the collapse of the entire WWTP,which affects the stability of WWTP.Therefore,how to accurately identify the membrane fouling status and provide the corresponding suggestions,has become an open problem.However,this WWTP with MBR is multi-flow,time-varying and contains some uncertainties.Then,the identification of pollution status of MBR is still a difficult problem in today’s research.To achieve the stable and efficient operation of WWTP and guarantee the performance of MBR,an intelligent identification method for membrane fouling is proposed in this paper.First,a soft-computing method,based on the stacked denoising autoencoder(SDA),is developed to online predict the key indexes of membrane fouling with the good performance by analyzing the the mechanism of membrane fouling.Second,a robust deep neural network(RDNN),based on the fuzzy denoising autoencoder(FDA)is designed to overcome the poor robustness for uncertainties to improves the network performance by introducing the fuzzy set theory.Third,an expert knowledge base is constructed by analyzing the relationship of key index variables and membrane fouling.Then,an identification method based on RDNN for membrane fouling is developed to accurately identify the types of membrane fouling and provid the operational suggestions for operators at the same time to effectively alleviate the membrane fouling.Finally,an intelligent identification system for membrane fouling is developed to realize the monitoring of key indicators of membrane fouling and the online identification of pollution types of MBR to reduce the incidence of membrane pollution and improve the stability of WWTP.The innovations of this paper are showed as below:1.The design of soft-computing method for membrane fouling.In the actual operation of WWTP with MBR,membrane fouling is usually evaluated by the key indexes.However,the key indicators are difficult to obtain online in real WWTP.Therefore,a soft-computing method,based on SDA,is developed to online predict the key indexes of membrane fouling.First,the input and output varibles of method are determined,based on membrane fouling mechanism.Second,an adaptive BP algorithm is designed to update the parameters of SDA to improves the learning speed of the method.Final,an optimization algorithm based on the unsupervised and supervised learning methods was used to update the learning parameters of the proposed a softcomputing method to improv the learning performance.Experimental results show that this proposed soft-computing method based on SDA has better prediction performance than other algorithms.2.The design of RDNN based on FDA.To solve the uncertainties in the process of membrane fouling,a RDNN based on FDA is proposed in this paper.First,a FDA,is developed to replace the general base-building unit in DNN.Then,the proposed RDNN is able to extract the robust representations to weaken the uncertainties.Second,a compact parameter strategy(CPS)is designed to reconstruct the parameters of FDA.Then,the computational burden of FDA can be alleviated to speed up the learning process.Third,an adaptive BP algorithm,with an adaptive learning rate strategy,is proposed to update the parameters of RDNN.Then,the performance of RDNN can be improved.Finally,the results on the benchmark problems and real applications demonstrate the effectiveness of the proposed RDNN.3.The design of identification method based on RDNN for membrane fouling.To solve the problem that membrane fouling type is difficult to be determined accurately,an identification method based on RDNN is proposed to idenfy the membrane fouling.First,a knowledge expert database was established,based on the membrane fouling mechanism and expert experience.Second,an identification method based on RDNN is designed to accurately identify membrane fouling online.Finally,in order to verify the effectiveness of the proposed identification method,it is applied to the actual MBR sewage treatment plant.The experimental results show that the membrane fouling identification method can accurately identify different types of membrane fouling and provid the operational suggestions for operators at the same time to effectively alleviate the membrane fouling.4.The design of intelligent identification system for MBR membrane fouling.To realize the application of the above intelligent identification methods in the actual WWTP with MBR,an intelligent identification system for membrane pollution is developed.First,a data collection and transmission platform of WWTP with MBR is constructed to realize the data collection and transmission.Second,an intelligent identification system of MBR membrane pollution is developed to realize the monitoring of key index variables of WWTP with MBR and the identification of membrane fouling types.Final,this intelligent identification system for membrane pollution was designed and applied to real sewage treatment plants in Beijing.The experimental results show that the proposed intelligent identification system for MBR membrane fouling can effectively reduce the incidence of membrane fouling and is beneficial to the stable operation of WWTP.