Prediction of Water Consumption Using NARX Neural Network Based on Grey Relational Analysis

Abstract

Scientific prediction of water consumption is beneficial to the water resources management. The selection of external factors directly affects the applicability and accuracy of model prediction. Aiming at the difficulty of determining the multivariable input of the Nonlinear Auto Regressive Models with Exogenous Inputs (NARX) neural network, a NARX neural network model based on grey relational analysis (GRA) is proposed. First, the GRA was used to analyze the relationship between influencing factors and water consumption, and the main influencing factor was selected as the NARX input based on the correlation coefficient. Then the NARX is applied to predict water consumption. To prove the superiority of the GRA-NARX neural network model, a single NARX neural network (without GRA) and back propagation neural network (BPNN) model are chosen as references. The experimental results show that the proposed model has higher prediction accuracy.