ARTIFICIAL NEURAL NETWORK APPLIED TO THE PREDICTION OF HYDROELECTRIC PRODUCTION

• Grant number: 23/12125-4
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date until: April 01, 2024
• End date until: March 31, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Guilherme Pina Cardim
• Grantee: Gabriel Mateus Bastida
• Host Institution: Faculdade de Engenharia e Ciências (FEC). Universidade Estadual Paulista (UNESP). Campus de Rosana. Rosana , SP, Brazil

Abstract

In Brazil, electricity production comes largely from hydroelectric plants, which are vulnerable to several external factors, such as rainfall and temperature. The uncertain estimate of water resources and external factors can lead to inefficient production and, consequently, a waste of natural resources. In this way, the prediction of the generation and consumption of electric energy makes it possible to generate greater savings in energy production costs, a more efficient use of available natural resources and a reduction in the energy wasted due to the current inability of the system to store it. Monitoring the electricity production using available resources and estimating production becomes of great importance, a fact proven by several recent scientific publications that seek to predict the consumption and/or production of electricity in different contexts. In this sense, the project's main objective is to predict the electricity production by hydroelectric plants to be carried out in a timely manner from the monitoring of external factors, allowing a prior assessment of the scenario, enabling better decision-making in relation to the available resources. Therefore, it is intended to obtain past data that will serve as a basis for a machine learning application, which is expected to predict the energy to be produced. To achieve the proposed objective, some steps must be fulfilled, such as: evaluation and analysis of different factors that may be related to hydroelectric production; evaluate different machine learning techniques in order to identify the most appropriate one for the proposal; development of a tool for temporal analysis based on constantly acquired data; evaluation of the obtained results and readjustment of the developed algorithm; making the developed tool available to the community.