Treffer: Prediction of drinking water needs : the case of Bobo-Dioulasso

The objective of this study is to develop a solution to predict daily water consumption in order to optimize the water management system in the city of Bobo-Dioulasso. To achieve this, neural networks are used to predict consumption using historical consumption data and daily temperature in the city as the parameters. Four neural network algorithmswere implemented for this study: Multi-Layer Perceptron (MLP), simple recurrent neural network, Long Short-Term Memory (LSTM) recurrent neural network, and Gated Recurrent Unit (GRU) recurrent neural network. The study focused on the eight distribution zones of the National Water and Sanitation Authority in the city. In view of the results, certain algorithms stood out from others in terms of prediction. The GRU network algorithm performed better on nearly half of the training data, followed by the MLP algorithm. The resulting models allow for the prediction of daily consumption on day J, given the consumption and temperature of day J-1. This work was carried out using tools such as the Jupyter NoteBookenvironment and the Python language.