Please use this identifier to cite or link to this item:
Title: Predicting the behaviour of water distribution networks with machine learning models
Author: Matos, Pedro
Matos, Sérgio
Andrade-Campos, A.
Keywords: Water supply systems
Machine learning
Water demand forecasting
Richmond’s network
Issue Date: 2019
Abstract: Water supply systems are indispensable infrastructures in any modern society, considering that a modern house is expected to have running water all the time. Water supply systems must pump water to meet their clients demands and face large cost-efficiency problems related to pumping operations. This work presents and analyses a possible solution to this problem using machine learning to both forecast water demands and simulate the consequent behaviour of the network which enables the optimisation of the energy cost. The study was conducted using data from real water demands from the central region of Portugal and previously modelled networks such as Richmond’s network. The results indicate that Artificial Neural Networks (ANNs) and Recurrent Neural Networks (RNNs) with Gated Recurrent Units (GRUs) are capable of achieving good performance in forecasting water demands, and that it is possible to create a model that mimics the behaviour of a water supply network of reasonable size using ANNs.
Peer review: yes
Publisher Version:
Appears in Collections:DEM - Comunicações
DETI - Comunicações
IEETA - Comunicações

Files in This Item:
File Description SizeFormat 
INFORUM2019 - Predicting_behaviour_water_distribution_networks.pdf606.22 kBAdobe PDFView/Open

Formato BibTex MendeleyEndnote Degois 

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.