Water Quality Evaluation and Prediction Based on a Combined Model

Abstract

Along with increasingly serious water pollution, water environmental problems have become major factors that hinder the sustainable development of our economy and society. Reliable evaluation of water quality and accurate prediction of water pollution indicators are the key links in water resource management and water pollution control. In this paper, the water quality data of Lanzhou Xincheng Bridge section in the Yellow River Basin and Sichuan Panzhihua Longdong section in the Yangtze River Basin were used to establish a water quality evaluation model and a prediction model. For the water quality evaluation model, we constructed the research samples by means of equal intervals and uniform distribution of interpolated water quality index data according to Environmental Quality Standards for Surface Water. The training samples were determined by a stratified sampling method, and the water quality evaluation model was established using a T-S fuzzy neural network. The experimental results show that the highest accuracy achieved by the evaluation model in water quality classification was 94.12%. With respect to the water quality prediction model, we propose ARIMA-WNN, which combines the autoregressive integrated moving average model (ARIMA) and a wavelet neural network (WNN) with the bat algorithm (BA) to determine the optimal weight of each individual model. The experimental results show that the highest prediction accuracy of ARIMA-WNN is 68.06% higher than that of the original model.