Application of the fully Data-Driven combination model for water demand forecasting in the mountainous tourist area

Abstract

For water demand forecasting in the mountainous tourist area this paper proposes a novel approach, namely the combination of autoregressive integrated moving average (ARIMA) model and radial basis function (RBF) neural network. And at the same time, the combination model focuses on the characteristics of the mountainous tourist area, which is relatively closed with a smaller scale and the water supply curve is relatively smooth. This model depends on fully data-driven approach, which means that it only relies on the historical data of water demand and ignores external factors to eliminate various unstable factors including the weather, season, tourists, and others, simplifying the model parameters. Combination model can overcome the limitations of the single model in the nonlinear sequence processing, improve the forecasting accuracy. At last, the historical data, which origins from the Mount Huangshan Scenic Area water supply scheduling in 2012, have been tested and achieved significant forecasting effects.