Water demand forecasting using time series

• Autor(a) principal: Lysenko, Alina
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/33797
• topic: Water supply system (WSS); Water demand forecasting; Time series; Forecasting methods; ARIMA; Heuristic forecasting model

Process optimization has been an area of growing interest for companies. For a water supply management company, water distribution optimization means coordinating efficiently the water harvesting, treatment and distribution stages in a single process. However, to plan efficiently the operation of water supply systems it is mandatory to forecast the water demands. The forecasting task is possible if some variables are previously known, such as the rainy days that directly influence the harvest stage. As well as weather conditions, there are other variables that affect the success of a good water demand forecast, such as holidays. The large number of variables requires increasingly sophisticated forecasting models that require greater information processing capacity. These models do not always find acceptable solution, and, in practice, the variables needed to train a robust model are not always available. This difficulty is overcome by the implementation of forecast models based solely on the time-series. These are computationally simpler and do not require data from entities outside the Water Supply companies. The focus of this work is to analyze, implement and test forecasting algorithms based on time series applied to water demand, having as input variables only the history of water consumption. The first stage of this work consisted of studying the history of water consumption in the Penacova area to trace temporal patterns in the data. The second stage deals with the implementation of a classical ARIMA prediction model and implementation of a Bakker heuristic model. Both models were compared and different advantages and disadvantages were analyzed.