Urban Water Consumption Estimation Using Artificial Intelligence Techniques

Abstract

The general objective of an urban water supply authority is to match supply and demand at a level of service acceptable to the consumers. In this context, predicting water consumption in urban area is of key importance for water supply management. Unfortunately necessary large database are not properly maintained in a country like India. In this situation, this research work propose at using ANFIS (Adaptive Neuro Fuzzy Inference System) to assess the present condition of daily water consumption and predict the future trend in the urban environment. In the field of water resources and hydrology, Artificial Intelligence system approaches such as fuzzy logic (FL), Artificial Neural Network (ANN), etc. have been used successfully for nonlinear and non-stationary time series modeling. For complex system, both FL and ANN techniques are combined together so as the individual strength can be exploited in a better manner for the construction of powerful hybrid system. The purpose of this study was to develop hybrid (ANFIS) model for the estimation of daily water consumption. The performance and potential of the various developed ANFIS models are examined using a four-year’s time series water consumption data collected from New Mangalore Port Trust in Mangalore city, India. The influence of lagged variables with respect to daily water consumption and their combination are analyzed. The performances of the ANFIS models are evaluated by using three performance indices namely CC, MSE and MRE and best model was selected. The best ANFIS model reveals higher correlation with recorded data when compared to single fuzzy logic model and statistical Multi-Linear Regression model (MLR). Based on this analysis, the concern authority can take an effective decision regarding the short term as well as long term efficient water management strategy in urban environment of this coastal city.