Self-organizing Modeling And Multiobjective Optimal Control For Municipal Wastewater Treatment Process

Activated sludge process is a widely used aerobic biological treatment method in municipal wastewater treatment plants.The activated sludge is formed by continuously filling air into the wastewater after a certain reaction time due to the multiplication of aerobic microorganisms.Biological coagulation,adsorption and oxidation of activated sludge are used to decompose and remove organic pollutants from wastewater.In order to meet effluent discharge standards and reduce fines,wastewater treatment plants are often in full load operating state,i.e.maintaining the concentration of dissolved oxygen?DO?,S_O,5,in the aerobic zone?the fifth partition?,and the nitrate nitrogen?NO?,S_NO,2,in the anaerobic zone?the second partition?in the high level through blowers and reflux pumps.However,the operation of blower and reflux pump requires a large amount of energy supply,which inevitably increases the operation cost.At the same time,from the point of view of biochemical reaction mechanism,only suitable S_O,5 and S_NO,2 can ensure the smooth progress of nitrification and denitrification.Therefore,the set points of S_O,5 and S_NO,2 should be dynamically optimized according to the actual operation,so as to reduce energy consumption?EC?and effluent quality?EQ?as much as possible.Obviously,in order to improve the wastewater treatment effect and reduce the operation cost,it is imperative to construct a multiobjective optimal control strategy for wastewater treatment process?WWTP?.This paper uses hierarchical control architecture to design the multiobjective optimal control scheme in WWTP.The architecture consists of three layers:modeling layer,optimization layer and control layer.The purpose of the modeling layer is to establish the objective function of multiobjective optimal control using process data;the objective of the optimization layer is to optimize the objective function online and obtain the optimal set points of key control variables dynamically;and the purpose of the control layer is to track the optimal set points in real time to ensure the smooth progress of wastewater biochemical reaction.According to the hierarchical control scheme,the following four issues need to be solved primarily:first,the establishment of evaluation model in WWTP;second,the solving of the complex multiobjective optimization problem in WWTP;third,the tracking control of DO and NO in WWTP;fourth,the hierarchical optimal control of WWTP which is integrated with self-organizing modeling,self-organizing multi-variable control and adaptive multiobjective optimization with constraints.Based on this,the main research contents of this subject are as follows:?1?Evaluation model of wastewater treatment process based on self-organizing fuzzy neural networkAiming at the establishment of evaluation model in multiobjective optimal control of WWTP,a self-organizing fuzzy neural network with efficient online self-organizing scheme?SOFNN-EOS?is proposed.First,asymmetric Gauss function is introduced as membership function,which can construct generalized ellipsoidal basis function?GEBF?,so that input space can be divided more flexibly and effectively,and the ability of describing nonlinear systems by fuzzy rules can be enhanced.Second,while using geometric growth criterion to realize automatic generation of fuzzy rules,a hierarchical pruning strategy is proposed.This pruning strategy is based on rule density and significance and can not only effectively overcome the difficulty of pruning thresholds setting but also prevent the deletion of important rules by mistake.Third,an adaptive allocation strategy of the antecedent parameters of fuzzy rules is designed,which can not only adjust the generalized ellipsoid base region adaptively to obtain better local approximation,but also balance the system precision,rule learning speed and rule base transparency in learning process.Furthermore,the improved recursive least squares algorithm is used to estimate the consequent parameter of the fuzzy rules online to ensure the convergence of the network while the fuzzy rules are self-organized growing and pruning.Finally,the convergence of the SOFNN-EOS is analyzed to ensure that it can be successfully used in practical engineering.Experimental results of benchmark test problems,EC,EQ,and effluent ammonia nitrogen modeling in WWTP show that SOFNN-EOS not only has a compact network structure and good generalization performance,but also has a high transparency and strong interpretability of the resulting fuzzy rule base.?2?Multiobjective optimization method based on adaptive MOEA/DAiming at the complex multiobjective optimization problem?MOP?in WWTP,a decomposition-based multiobjective evolutionary algorithm with angle-based adaptive penalty?MOEA/D-AAP?and a decomposition-based multiobjective evolutionary algorithm with adaptive two-phase?MOEA/D-ATP?are proposed respectively.In MOEA/D-AAP,first,the solution concentration of the weight vector is defined by the angle between the solution in the target space and the weight vector;then,the adaptive adjustment strategy of penalty factor is designed by using the solution concentration information in the evolutionary process;finally,the performance of MOEA/D-AAP algorithm is verified by six complex MOPs.In MOEA/D-ATP,first,by exploiting the advantages of multi-reference points and two-phase optimization strategy,an automatic switching algorithm based on the lifting degree of aggregation function is designed for automatic switching between two stages;then,an adaptive operator selection algorithm based on the update rate of parent solution is designed for the adaptive combination of differential evolution operators in order to balance the exploration and development ability of the algorithm;finally,the performance of MOEA/D-ATP algorithm is verified by 12 complex MOPs.Experimental results show that the proposed MOEA/D-AAP and MOEA/D-ATP can find more boundary solutions in Pareto optimal front,which can enhance the diversity of approximated Pareto solution set.Among them,MOEA/D-ATP performs better.?3?Predictive control of wastewater treatment process based on self-organizing fuzzy neural networkA model predictive control based on multi-input and multi-output?MIMO?SOFNN?SOFNN-MPC?is proposed for the multi-variable control of S_O,5 and S_NO,2 in WWTP.First,the generalized multilevel noise is used as the excitation signal to generate the input and output data for system identification.Second,the generalized ellipsoid basis function,geometric growth criterion,hierarchical pruning strategy and modified recursive least squares algorithm are used to establish the MIMO SOFNN.Third,according to the input and output data,the SOFNN identification model of WWTP is established offline.In the real time control process,the network weight parameters are adjusted online to provide an accurate prediction model for MPC.Furthermore,the convergence and stability analysis of SOFNN-MPC are given to ensure its successful application in practical engineering.Finally,the BSM1 is used to test the proposed SOFNN-MPC.Experimental results show that SOFNN-MPC has more advantages in tracking accuracy and control stability compared with conventional PID control.?4?Hierarchical multiobjective optimal control of wastewater treatment processA hierarchical multiobjective optimal control scheme for WWTP is proposed,which integrates SOFNN prediction model,adaptive MOEA/D optimization algorithm and SOFNN controller.First,through the analysis of wastewater biochemical treatment process,EC and EQ are selected as optimization objectives,the limited values of effluent quality parameters are selected as constraints.Then,by combineing MOEA/D-ATP,SOFNN prediction models and SOFNN controller,the dynamic optimization,intelligent decision-making and bottom tracking control of optimal set points of S_O,5 and S_NO,2 in WWTP are realized.Finally,BSM1 was used to verify the proposed optimal control scheme.Experimental results show that the hierarchical optimal control based on MOEA/D-ATP can effectively reduce the energy consumption of wastewater treatment process on the premise of ensuring the average effluent quality parameters meet the standards.