ANNs approach to identify water demand drivers for Saf-Saf river basin

• Published in: Journal of applied water engineering and research Vol. 8; no. 1; pp. 44 - 54
• Main Authors: Sakaa, Bachir, Chaffai, Hicham, Hani, Azzedine
• Format: Journal Article
• Language: English
• Published: Taylor & Francis 02.01.2020
• Subjects: GRNN model; MLPNN model; Quarterly water demand; RBFNN model; Saf-Saf river basin municipalities
• ISSN: 2324-9676; 2324-9676
• DOI: 10.1080/23249676.2020.1719220

This paper presents artificial neural network (ANN) techniques such as generalized regression neural networks (GRNNs), radial basis neural networks (RBFNNs) and multilayer perceptron neural networks (MLPNNs) for predicting quarterly water demand (QWD). The data set including total of 720 data records is divided into two subsets, training and testing. Various ANN models depending on the combination of antecedent values of water demand, temperature, rainfall and population are constructed and the best-fit input structure is examined. The performance of ANN models in training and testing phases are compared with the observed water demand values to select the best-fit forecasting model. For this purpose, some performance criteria such as root mean square error, coefficient of determination (R 2 ) and accuracy factor (A f ) are evaluated for different models (GRNN, RBFNN and MLPNN). The results indicated that MLPNN outperforms all other ANN techniques (GRNN and RBFNN) in the forecasting of QWD. Abbreviations: A f , accuracy factor; ANN, artificial neural networks; GRNN, general regression neural network; MLPNN, multilayer perceptron neural network; P, seasonal mean rainfall (in mm); Pop, population (in number); RBFNN, radial basis function neural network; RMSE, Root mean square error; R 2 , Coefficient of determination; T, seasonal mean temperature (in °C); QWD: quarterly water demand (measured by 10 6 m 3 )