Evaluation and Prediction of Factory Sewage Discharge Based on ARIMA Model — Incorporating Biological Water Purification Factors for Model Optimization

Abstract

This study focuses on the evaluation and short-term prediction of factory sewage discharge. Based on the detected water quality data, we quantified the temporal characteristics of sewage discharge indicators, defined a comprehensive pollution index, and established a prediction model by integrating methods such as Pearson correlation analysis, standardization, entropy weight method, and ARIMA model. Finally, the model was further optimized by introducing biological factors affecting water purification. For the temporal characteristics analysis (Problem 1), we verified the strong correlation between pollutant concentrations (COD and ammonia nitrogen) and their emissions using Pearson correlation analysis, and revealed intraday (7:00-11:00 and 17:00-20:00 as peak discharge periods) and daily variation trends through line charts, heatmaps, and statistical analysis. For the pollution index definition (Problem 2), pH, COD concentration, and ammonia nitrogen concentration were selected as evaluation indicators; data were standardized with reference to national emission standards, and indicator weights were determined by the entropy weight method to calculate the comprehensive pollution index (values >1 indicate excessive pollution). For the prediction model (Problem 3), the ARIMA model was established after stationarity testing, difference processing, and ACF/PACF analysis, which achieved reliable prediction results for future pollutant indicators and comprehensive pollution index (verified by RMSE and MAPE). The optimized model incorporating biological purification factors (microorganisms and phytoplankton) is more in line with the actual dynamic changes of aquatic ecosystems.