Urban Residential Water Demand Prediction Based on Artificial Neural Networks and Time Series Models

• Published in: Water resources management 2015-08, Vol.29 (10), p.3651-3662
• Main Authors: Al-Zahrani, Muhammad A., Abo-Monasar, Amin
• Format: Article
• Language: English
• Subjects: Aluminum; Arabia; Artificial neural networks; Atmospheric Sciences; Civil Engineering; Climatic data; Demand; Earth and Environmental Science; Earth Sciences; Economic forecasting; Environment; Family income; Forecasting; Geotechnical Engineering & Applied Earth Sciences; Households; Hydrogeology; Hydrology/Water Resources; Mathematical models; Neural networks; Noise; Predictions; Rain; Regression analysis; Studies; Time series; Water consumption; Water demand; Water resources; Water resources management; Water shortages; Wavelet transforms
• ISSN: 0920-4741; 1573-1650
• DOI: 10.1007/s11269-015-1021-z

Water demand prediction is essential in any short or long-term management plans. For short-term prediction of water demand, climatic factors play an important role since they have direct influence on water consumption. In this paper, prediction of future daily water demand for Al-Khobar city in the Kingdom of Saudi Arabia is investigated. For this purpose, the combined technique of Artificial Neural Networks (ANNs) and time series models was constructed based on the available daily water consumption and climatic data. The paper covers the following: forecast daily water demand for Al-Khobar city, compare the performance of the ANNs [General Regression Neural Network (GRNN) model] technique to time series models in predicting water consumption, and study the ability of the combined technique (GRNN and time series) to forecast water consumption compared to the time series technique alone. Results indicate that combining time series models with ANNs model will give better prediction compared to the use of ANNs or time series models alone.