Development of DNN-based LIB State Diagnosis System Using Statistical Feature Extraction

Abstract

LIBs (Lithium-Ion Battery) have been actively applied as a power source for the mobile robots due to their high energy density and high power discharge. The internal resistance of LIBs increases with repeated charging and discharging, which causes the maximum charge capacity to decrease. Since the performance of LIBs, such as usage time and maximum discharge current, is determined by their maximum charge capacity. Therefore, choosing the LIB with maximum charge capacity is a critical task that determines the success of a mobile robot&quots mission. In this paper, we propose a DNN-based LIB state diagnosis system to effectively identify the maximum charge capacity of LIBs. The system is constructed using a Deep Neural Network (DNN)-based classification model, utilizing diagnostic data to analyze features and diagnose the state of the LIB. Additionally, to enhance the practical use of the diagnosed LIB, three states are defined based on the mission requirements of mobile robots and the maximum charging capacity of the LIB. Diagnostic data is generated from time series discharge data, providing an effective reflection of the deterioration characteristics of the LIB. Finally, to evaluate and validate the performance of the proposed system, data not involved in the training process is applied, and the results are analyzed using a confusion matrix.