Research On Wastewater Treatment Plant Carbon Source Dosing System Based On Intelligent Algorithm

In the face of ever-increasing sewage discharge standards,wastewater treatment plants often use the method of adding chemicals to assist the effluent quality to meet the standards.At present,many wastewater treatment plants in our country have the problem of insufficient influent carbon sources,so the addition of carbon sources has become an important guarantee for the removal of total nitrogen in wastewater treatment plants.In the process of carbon source dosing,most wastewater treatment plants are adopting the dosing calculation method based on manual experience and traditional mathematical formulas.This method is not only low in accuracy,poor in adaptability,but also has a strong time lag.To this end,this research has developed a carbon source dosing control model based on intelligent algorithms,which has the characteristics of high accuracy,strong robustness and fast response,in order to reduce the error caused by manual dosing.In this study,the amount of carbon source dosing in the Biostyr biological filter process of a sewage treatment plant in Wuhan is the research object.Aiming at the problem that the amount of carbon source in the denitrification filter is difficult to accurately calculate,a carbon source input based on BP neural network is designed.Increase the amount of prediction model.The model is based on the online monitoring data of the sewage treatment plant,using a three-layer BP neural network,and the network topology is determined through experiments [13,12,1].At the same time,determine that the best training parameters of the network are the hidden layer transfer function tansig,the output layer transfer function purelin,the training function trainlm,the weight learning function learngdm,the learning rate is set to 0.02,the average error of the network training set is 25.29%,and the prediction set The average error is 26.68%.At this time,the network error is high,mainly because the BP neural network randomly selects the corresponding weight of the neuron every time it is trained,which makes the network easy to fall into a local minimum and the network convergence is poor.Aiming at the problem of insufficient training accuracy of the BP neural network,the particle swarm algorithm with fast optimization ability and the genetic algorithm with strong global search ability are used to optimize the weights corresponding to each node in the BP neural network,and adjust them separately.Experiments show that the accuracy of the optimized two models is greatly improved compared to using the BP neural network alone.The BP neural network optimized by the genetic algorithm has a higher prediction accuracy,with a training set error of 8.82% and a test set error of 9.23%.The accuracy of the test set is 65.4% higher than that of the BP neural network,and 71.24% higher than the prediction accuracy of the traditional mathematical formula method.It can change in real time according to the water inlet and outlet data,and the response speed is faster.In order to further verify the accuracy of the optimization model,use the model prediction data as a carbon source addition strategy for the wastewater treatment plant to investigate the changes in the effluent quality of the wastewater treatment plant.This study uses the Bio Win5.2 software based on the activated sludge mechanism model to construct the wastewater plant operation process The model is validated.Through field measurement and the influent expert evaluation system of Bio Win,the composition parameters were determined to be Fbs=0.172,Fus=0.082,Fup=0.303,Fac=0.310,Fna=0.858.The calibration of the model was further completed through steady-state simulation and dynamic simulation.Using this model to dynamically simulate the prediction results of the BP neural network model prediction set optimized by the genetic algorithm,the results show that using the neural network prediction results as the carbon source addition strategy,the effluent quality of the wastewater treatment plant can remain stable and all can reach the wastewater level A discharge The standard shows that the BP neural network model optimized by genetic algorithm has the feasibility of engineering practice.This model can assist the wastewater treatment plant in the process of determining the amount of carbon source to be added,reducing human error,reducing processing costs,and realizing intelligent control.