Data-driven Forecast And Analysis Of Effluent Quality Of Sewage Treatment Unit

In order to solve the problems that the effluent quality of most sewage treatment plants is easy to exceed the standard,the failure of sewage treatment process is difficult to diagnose,the operation cost is too high,and the automation level is low,based on a sewage treatment plant in Shenyang,this paper establishes a sewage treatment plant effluent quality forecast model and an analysis model of main influencing factors,which provides theoretical and technical support for realizing the efficient,closed-loop and stable operation of urban sewage treatment plants,and has important theoretical and practical significance.The research contents and achievements of this paper are as follows:（1）Aiming at the problems of numerous variables,strong coupling and difficult analysis in the sewage treatment process,considering the interaction of water quality characteristics,a filtering feature selection algorithm based on random forest was adopted to select water quality characteristics,and finally 4,20,18,16,18,1 and 26 water quality characteristics were selected as the forecast effluent COD concentration,NH₃-N concentration,TP concentration,TN concentration,p H,water temperature and sewage flow rate.（2）Aiming at the problems that the effluent quality of wastewater treatment plant is easy to exceed the standard,the machine learning algorithm has many parameters,and it is difficult to find the global optimal solution,an integrated learning water quality forecast algorithm（SSA-RFR,SSA-XGBR）optimized by sparrow search algorithm is proposed,and the effluent quality forecast model is established,and the effluent quality exceeding the standard is forewarned.This method improves the search method of super parameters of water quality forecast model,and realizes the dynamic optimization of integrated learning super parameters.On the test set,the forecast accuracy of the improved algorithm XGBR effluent COD concentration,NH₃-N concentration,TP concentration,TN concentration,p H,water temperature and sewage flow rate is 75.73%,88.14%,90.64%,96.71%,94.65%,99.77%and 88.15%respectively,compared with the basic ensemble learning algorithm.When the forecast water quality is not up to standard,it can be forecasted and warned in advance,which improves the monitoring efficiency of effluent quality of sewage plant.（3）Aiming at the problems of process fault diagnosis,delayed regulation and control,high operating cost,difficulty in explaining water quality prediction process by black box model,inapplicability of traditional correlation analysis method to sewage characteristics analysis in sewage treatment plants,SHAP algorithm with artificial intelligence interpretability is adopted to analyze the main factors that affect the concentration of COD,NH₃-N,TP and TN in effluent from both local and global perspectives,which will help to control the relevant parameters in the front-end treatment process of sewage treatment plants in time,ensure the effluent quality of sewage treatment plants to meet the discharge standards and avoid pollution.