Utilização dos Modelos Lineares e Não Lineares para Estimativa da Curva de Descarga da Bateria de um Quadrirotor

Batista D. Catumba; José A. Monteiro Sobrinho; Adriel de O. Freitas; Vandilberto P. Pinto; Leonardo R. Rodrigues

doi:10.20906/CBA2022/3343

Batista D. Catumba Universidade da Integração Internacional da Lusofonia Afro-Brasileira Campus das Auroras, R. José Franco de Oliveira, s/n - Zona Rural, Redenção - CE, 62790-970
José A. Monteiro Sobrinho Universidade da Integração Internacional da Lusofonia Afro-Brasileira Campus das Auroras, R. José Franco de Oliveira, s/n - Zona Rural, Redenção - CE, 62790-970
Adriel de O. Freitas Universidade da Integração Internacional da Lusofonia Afro-Brasileira Campus das Auroras, R. José Franco de Oliveira, s/n - Zona Rural, Redenção - CE, 62790-970
Vandilberto P. Pinto Universidade da Integração Internacional da Lusofonia Afro-Brasileira Campus das Auroras, R. José Franco de Oliveira, s/n - Zona Rural, Redenção - CE, 62790-970
Leonardo R. Rodrigues Divisão de Eletrônica, Instituto de Aeronáutica e Espaço - IAE, Praça Marechal Eduardo Gomes, 50, Vila das Acácias, São José dos Campos – SP, 12.228-904

DOI: https://doi.org/10.20906/CBA2022/3343

Keywords: Mathematical models, Systems identification, MATLAB, Validation, estimation

Abstract

Process modeling is a way of identifying and predicting possible failures in various systems, thus seeking to eliminate production waste. The use of mathematical models makes it possible to represent a given system in several different ways, depending on the perspective to be considered when choosing the system. This paper aims to model the data obtained from the battery discharge of a quad-rotor drone, Parrot Rolling Spider, in which two new Shoot Li-Po batteries, model XT-412, were used. For each battery, 5 tests were performed, and for each one, the battery discharge profile was recorded, ranging from 3.30V to 4.10V, at intervals of 50(mV). Such data were analyzed with the MATLAB program platform, system identification toolbox, which is used to build mathematical models of dynamic systems from measured input and output data. Within the tool, it was possible to model four mathematical models: ARX, OE, NARX and Hammerstein-Wiener, where the correctness values of these models can be seen in this work.As results it can be highlighted that the non-linear Hammerstein-Wiener model presented the highest accuracy, but presented the highest Computational Execution Time.