Water quality modeling and prediction of water supply plants in low-temperature and low-turbidity periods by using black box artificial intelligence models

Abstract

The treatment of low-temperature and low-turbidity water, together with the control of operating parameters, is a big problem in water treatment. In this study, the daily monitoring data of one water supply plant from 2021 to 2022 was used to predict the effluent chemical oxygen demand (COD) during low temperature and turbid periods by using black box artificial intelligence models (AI), such as Support Vector Regression (SVR), Decision Tree (DT), Random Forest (RF) and Backpropagation Neural Network (BP). The results of a single model show that the DT model has better results than the other single models. In ensemble modeling, the performance of single artificial intelligence models can be improved by using neural network integration. In the validation phase, the ensemble model can improve the prediction accuracy by about 15%. At the same time, the model also obtained a reliable prediction effect in the same region, water source, and the process of the water supply plant.