Use of time series models with machine learning tools: study about forecasting demand and water reservation in Peruibe County (Sao Paulo, Brazil).

• Grant number: 23/03883-2
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: May 01, 2023
• End date: April 30, 2024
• Field of knowledge: Physical Sciences and Mathematics - Probability and Statistics - Statistics
• Principal Investigator: Chang Chiann
• Grantee: Mateus de Faria Baptista
• Host Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated research grant: 17/50343-2 - Institutional development plan in the area of digital transformation: advanced manufacturing and smart and sustainable cities (PDIp), AP.PDIP

Abstract

Nowadays, one of the disruptive aspects of industrial automation is the machine learning technology, a segment based on programming languages and IoT (Internet of Things) platforms for system integration and digital interconnection. In industrial monitoring, it is possible to extract datasets from sensors and other equipments at fixed interval points in time in order to analyze different variables. This research project consists in a time series analysis using data of Peruibe (São Paulo, Brazil) reservoir, including measurements of water leak, level and pressure. Using programming languages such as Python and R, the data analysis is divided into 3 stages: outlier detection, reconstruction of missing values and forecasting data. For the first two, the techniques include adjustment of curves by regression, classical decomposition combined with regression and STL (Seasonal-Trend Decomposition Procedure Based on Loess). For the last one, the procedures include BATS and TBATS models (Trigonometric, Box-Cox transformation, ARMA erros, Trend and Seasonal components), regression models with autocorrelated errors and rolling analysis techniques. This is a continuation of the following paper: LARRUBIA, L.F. Detecção de anomalias, interpolação e previsão em tempo real de séries temporais para operação de reservatórios e distribuição de água. The efficiency of the proposed methods will be measured and compared by a new database from 2021 to 2022. Additionally, this research also contemplates the hydraulic project of the water reservoir system from Peruibe, examining parameters such as useful and available volume and operational limit in order to evaluate the quality of the water supply domain.