Premium
Multi‐phase method of estimation and adaptation of parameters of electrical battery models
Author(s) -
Binelo Marcia F. B.,
Sausen Airam T. Z. R.,
Sausen Paulo S.,
Binelo Manuel O.,
Campos Maurício
Publication year - 2021
Publication title -
international journal of energy research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.808
H-Index - 95
eISSN - 1099-114X
pISSN - 0363-907X
DOI - 10.1002/er.6149
Subject(s) - battery (electricity) , computer science , artificial neural network , genetic algorithm , adaptation (eye) , process (computing) , set (abstract data type) , algorithm , power (physics) , artificial intelligence , machine learning , physics , quantum mechanics , optics , programming language , operating system
Summary Mathematical modeling of the battery lifetime is an important tool for the design of more efficient batteries, as well as for the optimization of their use. The electrical models class is among the classes of mathematical models used for this purpose, and a fundamental step to their application is the correct estimation of their parameters. This paper performs the mathematical modeling of Lithium‐Ion Polymer batteries lifetime through the electrical model of Tremblay, in which a multi‐phase method of estimation and adaptation of parameters is proposed, divided into three phases: discovery, learning, and inference. The multi‐phase method is based on two Artificial Intelligence techniques: genetic algorithms and artificial neural networks. The proposed method is validated by the simulation and experimental studies. From the results, it is concluded that the application of the multi‐phase method improves the effective accuracy of the Tremblay model, when it comes to adapt its parameters to the battery during runtime. For constant discharge currents, the average error reduction was 79 % , when compared to the best set of parameters obtained by GA without the adaptation process. For variable current discharge curves, the method was able to reduce the error more than 35 % . This method can be applied to other battery lifetime prediction models.