Univariate Long-Term Municipal Water Demand Forecasting

TL;DR

This paper evaluates various time series models for long-term municipal water demand forecasting in London, Canada, finding Prophet to be the most accurate and interpretable, with open-source implementation for practical use.

Contribution

It compares multiple forecasting techniques and identifies Prophet as the best model for municipal water demand prediction, emphasizing interpretability and robustness.

Findings

Prophet achieved 2.51% MAPE in testing.

Prophet handles missing data gracefully.

Open-source implementation available for municipalities.

Abstract

This study describes an investigation into the modelling of citywide water consumption in London, Canada. Multiple modelling techniques were evaluated for the task of univariate time series forecasting with water consumption, including linear regression, Facebook's Prophet method, recurrent neural networks, and convolutional neural networks. Prophet was identified as the model of choice, having achieved a mean absolute percentage error of 2.51%, averaged across a 5-fold cross validation. Prophet was also found to have other advantages deemed valuable to water demand management stakeholders, including inherent interpretability and graceful handling of missing data. The implementation for the methods described in this paper has been open sourced, as they may be adaptable by other municipalities.