The Research On Mathmatic Model For Forecasting Sewage Load Based On The Time Series

The forecasting technology of sewage treatment has been developed rapidly in many scholars' research and practical application,however,domestic research has just started due to the short rise time.Based on the analysis of the actual sewage data of the cooperative enterprise sewage treatment plant,this paper combines the objectives of the research project and established the forecasting model of the relevant variables,and the effect of the model prediction is analyzed.First,data analysis of influent variables(influent,influent COD,influent ammonia nitrogen)and aeration tank related variables(aeration,liquid level,sludge concentration,etc.)were carried out for the data used.The ARMA model and the VAR model were used to model the relevant variables of the influent load by analyzing the advantages and disadvantages of different prediction models;the neural network was used to model the dissolved oxygen DO in the aeration tank.Then,the ARMA model of the relevant variables(influent,influent COD and influent COD load)were established by analyzing and preprocessing the timing data(including data integration,abnormal data identification and cleaning,data conversion and protocol and data filtering).Based on the influence of linkage variables,the VAR model of influent COD was explored and established.And the two models(ARMA model and VAR model)were compared and analyzed,and the prediction residuals of the VAR model were found in the two predictive models of the influent COD(ARMA model and the VAR model).The results were compared with those of the two predictive models.The prediction residuals of the ARMA model are less accurate than those of the ARMA model,which indicates that the VAR model is more accurate and explains the reasons why the VAR model is not concerned with the requirements and applicability of the VAR model.Based on the principle of BP neural network and thought evolutionary algorithm,the BP neural network model of aeration tank DO is initially established and the prediction results are analyzed.Then,the optimal DO prediction model based on the sample data is established by optimizing the initial weight and threshold of the BP network by using the evolutionary algorithm of BP neural network,and then optimizing the structure,configuration and training samples of the BP network.By comparing with the initial network,the DO model predicted by DO is only 0.0014 and the recognition rate DR is up to 99.4898,which fully shows that the optimized BP network improves the prediction performance of the model to a great extent.Finally,the applicability of the three models is proved to prove the universality of ARMA to the time series modeling.The VAR model is introduced into the sewage treatment process,and the established model is beneficial to the linkage effect between the influent COD correlation variables analysis,which helps us to explore the interaction between the relevant components of sewage.The prediction based on the aeration tank DO will help us to estimate the aeration in advance,so as to optimize the aeration control.In summary,the prediction of the relevant variables for wastewater treatment will help us to fully predict the relevant steps of the wastewater treatment,thereby reducing the processing costs.