Artificial Neural Network and Fuzzy Logic for Water Demand forecasting (In case of Mekelle city)

No Thumbnail Available

Date

2018-05

Journal Title

Journal ISSN

Volume Title

Publisher

Addis Ababa University

Abstract

The methodology emphasizes the importance of the new predictive model development, and to account for water demand uncertainty, through evaluation of the reliability of model predictive. In this thesis Artificial Neural Networks (ANN) and fuzzy logic models were used for Mekelle city water consumption predictions analysis. Water consumption data collected from Mekelle city water and sewerage authority and independent monthly climate data were obtained from the National Metrological Agency. Sensitivity analysis was applied to selection most relevant minimum explanatory variables in the forecasting process. These data were used in both ANN and the fuzzy model setting up, testing and validation. To build ANN model the available dataset were divided into 3 subsets: 70% of the data for model development; (15%) of the data are used for training; and (15%) of the data are used validation to determine the optimal number of inputs and optimal number of hidden neurons. Both input variables and the output variable of the water consumption were fuzzified and triangular fuzzy membership functions were created. The Mamdani fuzzy rules in If–Then format with the centroid defuzzification method were employed. Seven ANN model were developed with different weather combination as input variable and Model one were found best with a root mean square error (RMSE) of 30.96, mean absolute percentage error (MAPE) 2.54, correlation coefficient (R2) of 0.98 and 97.46% of forecasting accuracy. The average absolute percentage error of the fuzzy model was found as 19.2%. Therefore, in this research ANN model is successfully presented for predicting water consumption in Mekelle city with climate inputs, cost of water and population compared with the fuzzy prediction system.

Description

Keywords

water demand forecasting, ANN, fuzzy logic, climate and sensitivity analysis

Citation