An Effluent Total Nitrogen Prediction Study Of Wastewater Treatment Plant Based On Deep Learning

With the development of the water quality improvement work plan,many cities have clearly stated that the existing wastewater treatment plant effluent standards need to reach Grade A and above,and total nitrogen is one of the main indicators.Since Total Nitrogen does not list national emission reduction targets,most sewage treatment plants are overexposed,resulting in Total Nitrogen failure to meet the standard.Therefore,the establishment of a prediction method for Total Nitrogen in wastewater treatment can help the operators of sewage treatment plants to effectively improve the operation level.According to domestic and foreign research,the early prediction methods are mostly white box models,that is,the ASM model and some commercial software are used to predict the water quality of effluent.With the development of computer technology,data mining methods have gradually matured.Among them,neural network algorithms are favored by scholars.BP neural network,RBF neural network and deep learning network have been applied to the field of effluent water quality prediction.Investigating the shortcomings of the existing research,this thesis proposes the following research ideas,that is,using accurate inflow and outflow data as a training set and test set,and establishing a BP neural network model to predict the total nitrogen in the effluent,the results show that the influent indicators plus part of effluent indicators is used as input data,more accurate prediction results can be obtained.In order to reduce the amount of calculation,the principal component analysis method is used to reduce the dimensionality of the data.The results show that the contribution rate of the influent p H and water flow to the principal component is very low,which can be ignored when the BP neural network predicts the effluent total nitrogen.In order to solve the phenomenon that the traditional BP neural network is easy to fall into the local minimum value,the genetic algorithm and the particle swarm optimization algorithm are combined to optimize the BP neural network to predict the weight of water total nitrogen.The results show that the prediction accuracy is optimized by genetic algorithm and particle swarm optimization,there is a clear improvement.A deep learning network-depth belief network DBN and long-short memory network LSTM were established to predict the total nitrogen in the wastewater treatment plant.The CDBN network prediction results are better than the BP neural network prediction results.The LSTM network predicts the total nitrogen in the wastewater treatment plant.The results are worse than those predicted by the CDBN network and the BP network.