Searching for an affordable Neural Network Architecture for Electrochemical Parameter Identification of a Li-ion Battery on an actual Battery Management Board

Abstract

The diagnosis of a lithium-ion battery is essential to operate the battery for safety and life extension. The electrochemical parameter identification is one of the diagnosis methodologies of the battery, which can describe various electrochemical side reactions occurring in multiple situations such as fast charging. The identification methods based on the neural network are conducted to overcome the high complexity of the method based on an electrochemical model. However, the previous methods adopt the networks optimized on the other fields such as image processing and do not consider the operating environments, which have low hardware specifications. In this paper, an affordable neural network architecture for parameter identification is explored through the experiment, and the applicability of the network is verified in the electric vehicle environment using a vehicle control unit board.