Research On Key Technologies Of Wastewater Biochemical Treatment Process And Development Of Control System

The wastewater treatment process is a typical non-linear system,which is strongly disturbed by the flow and load in real time,and accompanied by the changes in the composition of the wastewater,but the wastewater treatment plant must keep running continuously to meet the increasingly strict discharge standards.Advanced control research has important theoretical and practical significance for reducing operating costs,improving quality,optimizing energy use,and reducing environmental pollution.However,in the actual wastewater treatment process,as the wastewater components become more and more complex,the traditional PID control becomes more and more difficult to meet the process requirements and discharge standards,and because the water quality monitoring sensor is expensive or missing,etc.it is unable to effectively monitor wastewater water quality in real time.Based on this,this thesis takes the actual engineering project control system development of a wastewater treatment center in Changsha as the research background.This thesis is based on the Anaerobic-Anoxic-Oxic(AAO)process used in this wastewater treatment center.First of all,in view of the problem that effluent COD in the wastewater treatment process is difficult to meet the real-time online monitoring due to limited measurement methods when using traditional measurement methods,a soft measurement modeling method based on improved particle swarm optimization Gaussian process regression(IPSO-GPR)is proposed.This model introduces a small sample machine learning-Gaussian Process Regression(GPR)to the COD prediction of effluent in the wastewater treatment process.Because the single kernel function of GPR is difficult to meet the prediction accuracy of effluent COD,this thesis proposes a GPR prediction model based on the combined kernel function of SE kernel function and PER kernel function.The experimental results show that compared with the single kernel function GPR prediction model,the combination proposed in this thesis The kernel function GPR prediction model has higher prediction accuracy for effluent COD.At the same time,an improved particle swarm algorithm is used to optimize the hyperparameters of the combined kernel function,which solves the problem of relying on initial values and low generalization ability based on Conjugate Gradient Method.The simulation results show the problem of initial value and low generalization ability that compared with the traditional LSSVM and BP-ANN prediction models,the IPSO-GPR prediction model proposed in this paper has higher prediction accuracy for the COD of effluent during wastewater treatment.Secondly,for the aeration tank control system of wastewater biochemical treatment process,due to the problems of large lag,non-linearity and large fluctuations,difficult to determine mathematical models in the control of dissolved oxygen(DO)concentration,This thesis proposes a PID control strategy using BP neural network Method for adjusting DO concentration.Using MATLAB to verify the algorithm,the results show that the controller has good robustness,small overshoot and fast response speed.Finally,according to the wastewater treatment process and control requirements,a supervisory control and data acquisition system(SCADA)is designed and implemented as the host computer monitoring system,and Schneider M580 series PLC is used as the lower computer's wastewater biochemical automatic control system.