Previous studies have shown that soft computing models are excellent predictive models for demand management problems. However, their applications in solving water demand forecasting problems have been scantily reported. In this study, feedforward artificial neural networks (ANNs) and a support vector machine (SVM) were used to forecast water consumption. Two ANN models were trained using different algorithms: differential evolution (DE) and conjugate gradient (CG). The performance of these soft computing models was investigated with real-world data sets from the City of Ekurhuleni, South Africa, and compared with conventionally used exponential smoothing (ES) and multiple linear regression (MLR). The results obtained showed that the ANN model that was trained with DE performed better than the CG-trained ANN and other predictive models (SVM, ES and MLR). This observation further demonstrates the robustness of evolutionary computation techniques amongst soft computing techniques.

The purpose of this review is to establish and classify the diverse ways in which evolutionary computation (EC) techniques have been employed in water demand modelling and to identify important research challenges and future directions. This review also investigates the potentials of conventional EC techniques in influencing water demand management policies beyond an advisory role while recommending strategies for their use by policy-makers with the sustainable development goals (SDGs) in perspective. This review ultimately proposes a novel integrated water demand and management modelling framework (IWDMMF) that enables water policy-makers to assess the wider impact of water demand management decisions through the principles of egalitarianism, utilitarianism, libertarianism and sufficientarianism. This is necessary to ensure that water policy decisions incorporate equity and justice.

This paper presents the development of an artificial intelligent water demand forecasting model. The model comprises a single-hidden layer feed-forward neural network trained in using a differential evolution algorithm. Multiple feature selection techniques were employed to identify the minimal subset of features for optimal learning, namely Pearson correlation, information gain, symmetrical uncertainty, Relief-F attribute and principal component analysis. The performance of the feature selection techniques was compared to a baseline scenario comprising a full set of data covering potential casual variables including weather, socioeconomic and historical water consumption data. The performance of the models was evaluated based on accuracy. Results show that the five feature selection techniques outperformed the baseline scenario. More importantly, the subset of features obtained from the Pearson correlation technique produced the most superior model in terms of model accuracy. Findings from the study suggest that the inclusion of weather and socioeconomic variables in water demand modelling could enhance the accuracy of forecasts and cater for the impacts of climate and socioeconomic variations in water demand planning and management.