Water quality prediction using ARIMA-SSA-LSTM combination model

Abstract

The water quality index model is a popular tool for evaluating drinking water quality. To overcome low precision and significant errors in the traditional single prediction model, a novel autoregressive integrated moving average (ARIMA)-sparrow search algorithm (SSA)-long short-term memory (LSTM) combination model is proposed to accurately predict residual chlorine, turbidity, and pH in drinking water. First, the ARIMA model is used to extract the linear part of water quality data and output the nonlinear residual. Then, the LSTM model is used to predict the residual, and the SSA is used to find the optimal hyperparameters of the LSTM model, which plays an essential role in reducing the error of the model. To prove the superiority of the model developed, the ARIMA-SSA-LSTM model is compared with SSA-LSTM, whale optimization algorithm-LSTM, PSO-LSTM, ARIMA-LSTM, ARIMA, and LSTM. The results show that the coefficient of determination (R2) of the combination model for residual chlorine, turbidity, and pH are 0.950, 0.990, and 0.998, respectively, which are greater than all comparison models. Therefore, the model is more suitable for the prediction and analysis of water quality data. HIGHLIGHTS This article presents a new combined forecasting method to predict the trend of water quality in water distribution networks.; The autoregressive integrated moving average (ARIMA)-sparrow search algorithm (SSA)-long short-term memory (LSTM) combined model overcomes the limitations of the traditional single model.; SSA can find more suitable hyperparameters for LSTM model, so as to improve the prediction accuracy of LSTM model.;

Keywords

• autoregressive integrated moving average
• combined model
• long short-term memory
• sparrow search algorithm
• water quality prediction