Research On Modeling And Model Optimization Strategy Of Activated Sludge Process

The activated sludge based wastewater treatment process is a complex nonlinear dynamic system with large time delay,multivariable and strong coupling.In addition,the key process parameters of the activated sludge process（ASP）often need to be measured manually,which consumes a lot of manpower,time and economic cost.At the same time,the detection is not timely.The existence of this problem will inevitably lead to a large number of excessive wastewater discharge,causing secondary pollution to the ecological environment.Therefore,the study of new methods for modeling and model optimization of the wastewater treatment process,to realize the accurate description of wastewater treatment process,providing a model basis for process control and ensuring the control effect and the quality of effluent has very important theoretical value and practical significance.This thesis takes the activated sludge wastewater treatment process as research object,based on the meta-heuristic search algorithm,neural network,feature selection and other intelligent technologies,researches on the modeling and model optimization strategies of activated sludge process.The main innovative research achievements of this thesis are as follows:（1）Based on the in-depth study of the search mechanism of the invasive weeds optimization（IWO）algorithm,a niche-based adaptive invasive weeds optimization（NAIWO）is proposed for the shortcomings of insufficient local search ability in the early stage and lack of population diversity in the later stage of IWO.Firstly,a dynamic adaptive mechanism is introduced in the spatial diffusion step of IWO to balance the global and local search capabilities of the algorithm.Secondly,it combines IWO with the niche idea,and the classification competition mechanism of the niche idea is introduced to enhance the stability and global search ability of IWO.Finally,the effectiveness and stability of the proposed NAIWO algorithm is verified through optimization tests on multiple benchmark functions.（2）Aiming at the problem that the ASM1 model will be affected by external conditions and internal environment in practical application,which leads to the low prediction accuracy of the model,an optimal parameter estimation method of ASM1model based on NAIWO algorithm is proposed.Firstly,the mechanism of pollutant removal in the activated sludge process isanalyzed,and the mechanism model（ASM1）of the activated sludge process isestablished on MATLAB.Then,the NAIWO algorithm is applied to the optimal parameter estimation of the ASM1 model,and the NAIWO-ASM1 model of the activated sludge process is established to realize the estimation of the sensitive parameters in the ASM1 model.Finally,the model accuracy verification experiment iscarried out with the operating data of two wastewater treatment plants of different scales,which verified the validity and superiority of the proposed parameter estimation method.（3）In order to solve the problem that the prediction model of key water quality parameter BOD₅ concentration in ASP is too complex and the accuracy is reduced due to the high dimension of input variables,a gradually filtering-encapsulation hybrid feature selection algorithm based on mutual information is proposed.Firstly,the mutual information evaluation method of m RMR（max-relevance and min-redundancy）is improved,and a new hybrid mutual information evaluation strategy is proposed to select the input features for the first time.Then,combined with the evaluation results of classifier or approximator,the backward strategy is used to eliminate the irrelevant or weakly related variables which are difficult to identify based on mutual information,so as to select the optimal feature subset.Finally,BPNN is used as the approximator to establish the BOD₅ prediction model of the actual wastewater treatment plant to verify the performance of the proposed feature selection algorithm.Experimental results demonstrate that the proposed algorithm realizes the selection of input features of the BOD₅ prediction model,which ensuring the accuracy of the model while reducing its complexity.（4）Aiming at the problem that BPNN can not effectively extract the dynamic characteristics of ASP,resulting in the poor accuracy of the effluent BOD₅ prediction model,a BOD₅ prediction model based on NAIWO-ESN_CRJisproposed.Firstly,cycle reservoir with regular jumping（CRJ）is used to replace the random reservoir of ESN to eliminate the impact of network randomness on prediction performance.Secondly,NAIWO algorithm’s powerful numerical optimization ability is used to calculate the output connection weight of ESN to further improve the performance of the network.Finally,a prediction model of effluent BOD₅ concentration based on NAIWO-ESN_CRJisestablished,and the performance of the model istested with the measured data from a real wastewater treatment plant.The experimental results illustrate that NAIWO-ESN_CRJ is superior to BPNN and ESN model,which can effectively improve the tracking accuracy of BOD₅ concentration in effluent of wastewater treatment plant.