Parameter Estimation of Battery Modules in a Modular Reconfigurable Battery Using Deep Neural Network
This paper introduces a groundbreaking method employing feedforward neural networks (FNN) for the precise estimation of internal parameters in modular reconfigurable batteries (MRBs). MRBs, integral in numerous applications, particularly Electric Vehicles (EVs), often face challenges in accurately measuring individual battery parameters due to extensive sensor requirements. This study addresses this issue through neural networks. Highlighting the widespread use of Lithium-ion batteries in energy storage, especially in EVs, due to their high energy density and durability, this research centers on two critical MRB parameters: internal resistance and open circuit voltage. The FNN, meticulously trained on a well-curated dataset comprising MMC net output voltage, current, duty cycle, and submodule switching states adeptly predicts the mentioned parameters. Results show that the model could reach an accuracy of 97.91% in internal resistance estimation and 99.85% in open circuit voltage prediction. This methodology holds potential for cost reduction and simplification of battery monitoring systems, enhancing MRB efficiency across diverse applications.
