Automation of Water Distribution System by Prediction of Water Consumption and Leakage Detection Using Machine Learning and IoT

Abstract

The distribution of water can be a very challenging task and even a small mistake in consumption estimation can cause huge problem which includes shortage of water in some areas. Also, the leakage in water supply line is a huge problem as these supply lines are underground and it is very difficult to identify and repair if a leakage occurs in those pipelines. To solve these problems a lot of research have been done by different authors which includes the use of regression-based Machine learning algorithms for predicting the water consumption this includes Random Forest, Decision Tree and Support Vector Regression algorithms and for leakage detection they have used classification algorithms like deep autoencoder, hydroacoustic spectrograms and many more. Also, there are some models in which the researchers have used sensors like pressure sensor to use that data for prediction of leakage detection. This previous work will be further discussed in the second section of this paper. In this paper we have combined Machine Learning with IoT to create an automated model which can continuously monitor the date for leakage in water distribution line and predict the water consumption and this whole task is done in real-time. In this paper we have used classification variant of Artificial Neural Network algorithm for leakage detection and regression variant for water consumption prediction and for the performance evaluation of this model we have used R-squared and Adjusted R-Squared for water consumption prediction and confusion matrix, accuracy for leakage detection.