Optimize Decision Making And Intelligent Control Study Of Wastewater Treatment Process

Because of containing many microbial reactions,complex physical reactions and chemical reactions,the sewage treatment process is a system with strong nonlinearity,coupling,uncertainty and variability.Due to these characteristics,the sewage treatment process is difficult to be controlled.There are many difficulties in establishing the process model and optimizing the key parameters,which affects effluent water quality.At the same time,China’s sewage treatment plants generally have the problems of high power consumption and high operation cost.In order to inhibit the peak of pollutant concentration in effluent and reduce energy consumption,the thesis carries out the research on intelligent optimization control method and decision control method.In order to realize the energy saving of wastewater treatment process,the intelligent optimization control method is used to control the wastewater treatment process.According to the complicated dynamic characteristics of the WWTP,the hierarchical control strategy is used to realize the energy saving optimization control of the WWTP.The control structure includes the modeling layer,optimization layer and tracking control layer.The modeling layer is used to establish the optimization target model;the optimization layer realizes the optimization of the target variables and obtains the optimized set value of the key variables;the tracking control layer realizes the tracking control of the key variables.Based on the in-depth analysis of the characteristics of sewage treatment process,in the modeling layer we adopt the feedforward neural network to establish the optimized target model.The set-values of dissolved oxygen and nitrate concentration are determined as the optimization variables,and the multi-objective optimization scheme is designed mainly considering the performances of energy consumption and effluent quality in optimization layer.Fuzzy control method is adopts to control the dissolved oxygen concentration and nitrate nitrogen concentration.Intelligent optimization control can effectively reduce energy consumption.However,it cannot solve the problem of effluent ammonia nitrogen and total nitrogen peak.To inhibit the peak of ammonia nitrogen and total nitrogen concentration in effluent water,this thesis further studies the combination of intelligent optimization control and decision control to achieve the goal of reducing energy consumption and make sure the effluent water quality reach the standard.The main studies are as follows:(1)Mutual information based weight initialization method for feed forward neural networksDue to the strong coupling and non-linearity of the wastewater treatment process variables,the optimization model is established by feed forward neural network(FNN).To improve the stability of network and avoid local minima,a Mutual Information based weight initialization(MIWI)method is proposed for FNNs.The MIWI is mainly based on three theorems that have been verified:(1)If the initial weights are close to the global optimum,any descent algorithm can train the weights toward the optimum reliably;(2)The network with optimal weights can fully reflect the relationship between the input variables and output variables;(3)Restrain the range of initial weights reasonably can shorten the training time dramatically.The useful information contained in input variables is measured with the mutual information(MI)between input variables and output variables.The initial distribution of weights is consistent with the information distribution in the input variables.The lower and upper bounds of the weights range are calculated to ensure the neurons inputs are within the active region of sigmoid function.The experimental results show that the proposed method can avoid local minima and guarantee the convergence rate is fast.(2)A mutual information and sensitivity analysis algorithm for building self-organizing feed forward neural networksThe structure of neural network has important influence on network performance: FNN with excessively small network size cannot handle problems well,whereas an excessively large network size will lead to over-fitting and poor generalization performance.For the modeling of wastewater treatment process,the optimal network structure is unknown.In order to determine the appropriate network structure,this paper proposes a network structure adjustment mechanism based on sensitivity and mutual information(HCPS).HCPS uses mutual information to measure the correlation between hidden layer neurons.The two hidden neurons with high mutual information have similar information processing power,which can be combined into one neuron.HCPS also uses sensitivity to analyze the contribution of hidden neurons to network output.HCPS makes full use of information within hidden layer and information between layers to adjust network structure to obtain an effective and concise network structure.The experimental results show that HCPS algorithm can obtain a more compact network structure with better network accuracy.(3)Multi-objective dynamic optimization based intelligent optimization control for wastewater treatment plantsThe wastewater treatment process is a time-varying and nonlinear system.Energy consumption and water quality are two mutually contradictory optimization goals,so the optimal control of wastewater treatment process is a dynamic multi-objective optimization control problem.Considering the characteristics of irregular change in wastewater treatment process,we propose a dynamic multi-objective optimization algorithm based on memory(NSGA2-DM).NSGA2-DM set up a memory library to store the central solution and the corresponding environment variable values.The initial population is generated according to the memory when the environment changes.In the genetic stage,NSGA2-DM calculates the sparse degree of the solution according to the density of each solution in the target space,and the lowest sparse solution is defined as the sparse solution.In each genetic process,a local search is performed around the sparse solution.NSGA2-DM adopts limit optimization local search strategy and random local search strategy synchronously to improve the quality and convergence speed of solutions.The experimental results show that this method can effectively deal with the dynamic stochastic system,which has better precision and speed.(4)Advanced decision and optimization control system for wastewater treatment plantsIn order to realize the energy saving and consumption reduction of sewage treatment process and inhibit the peak of pollutant concentration,this thesis proposes intelligent decision control method for wastewater treatment process.Firstly,establish the prediction model of ammonia nitrogen(SNH,e)and total nitrogen(SNtot,e)with neural network.Secondly,optimize the set points of dissolved oxygen concentration and nitrate nitrogen concentration with multiobjective evolutionary algorithm.Lastly,select control strategy(optimal control strategy or inhibitory control strategy)based on the outcome of prediction model.If the predicted results exceed the standard,the fuzzy control method is used to control the external carbon source and internal return flow(Qa),which can suppress the peak value.If the predicted results do not exceed the standard,the fuzzy control method is used to control the Qa and aeration rate to tracking the set value.The experiment results show that: the proposed method restrains the peak of SNH,e and SNtot,e effectively;the energy consumption is less than the compared inhibitory control methods significantly.