ARTIFICIAL NEURAL NETWORKS AND MATHEMATICAL MODELING APPLIED TO THE CLASSIFICATION OF BIOMASS SPECIES MOST SUITABLE FOR ENERGY PRODUCTION

• Grant number: 24/16751-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: December 01, 2024
• End date: November 30, 2025
• Field of knowledge: Interdisciplinary Subjects
• Principal Investigator: Alfredo Bonini Neto
• Grantee: Júlio César Rocha Montagnani
• Host Institution: Faculdade de Ciências e Engenharia. Universidade Estadual Paulista (UNESP). Campus de Tupã. Tupã , SP, Brazil

Abstract

Energy production is essential for the development of a country. In Brazil, most of the energy demand is met by hydroelectric power. However, the deregulation of the electricity sector and restrictive policies on the construction of new transmission lines have forced the electrical system to operate near its maximum capacity, resulting in failures and energy rationing during drought periods. Given these challenges, it is crucial to diversify the energy matrix and invest in alternative and renewable sources. Thermogravimetric Analysis (TGA) curves of wood often present very similar layouts, making it difficult to obtain specific information about the components or the type of biomass using traditional techniques, which usually rely on weight loss at specific temperature values. In this project, Artificial Neural Networks (ANN) will be applied as an innovative technique to more effectively identify wood species or differences in their components. The research will focus on evaluating the use of TGA curve markers, which function similarly to genetic markers, to obtain detailed information about the biomass. These markers are multiple values of residual weight percentage in relation to the initial weight at specific temperatures on the TGA curve. The ANNs automatically adjust weights based on training patterns from the input marker data, thus enabling a more accurate classification of the samples. Due to its ability to replicate curves for the same sample and its unique characteristics, thermogravimetric analysis becomes a valuable tool for identifying wood species and characterizing biomass composition. In addition to ANNs, regression and correlation methods will be used to analyze the relationship between the obtained data and the estimated values. Regression will create mathematical models that describe the relationship between TGA markers and the type of biomass, while correlation will assess the strength and direction of these relationships. The use of these advanced techniques will provide detailed data on biomass, improving the classification of samples and contributing to the development of a more diversified and sustainable energy matrix by identifying the wood species most suitable for energy generation.