| Title |
Long Short Term Memory-Based State-of-Health Prediction Algorithm of a Rechargeable Lithium-Ion Battery for Electric Vehicle |
| Authors |
권상욱(Sanguk Kwon) ; 한동호(Dongho Han) ; 박성윤(Seongyun Park) ; 김종훈(Jonghoon Kim) |
| DOI |
https://doi.org/10.5370/KIEE.2019.68.10.1214 |
| Keywords |
tric vehicle(EV); Urban dynamometer driving schedule(UDDS); State-of-health(SOH); Machine learning; Artificial neural network(ANN); Long short term memory(LSTM) |
| Abstract |
The state-of-health(SOH) information of a rechargeable lithium-ion battery is dependent on variable electric vehicle(EV)’s output features caused by frequent discharging/charging current; temperature; and state-of-charge(SOC) operating range. Above all; the most important thing to be checked is this SOH information should be correctly predicted for providing guarantee lithium-ion battery statuses to EV users. For this goal; critical aging factors that results in battery management system(BMS) performance should be obtained by various experiments and be reflected in a sophisticated study. Therefore; this paper introduces two steps for accomplishing an improved SOH prediction of a rechargeable lithium-ion battery. The first step is to perform aging factors extraction and their correlation analyses based on experiments using urban dynamometer driving schedule(UDDS) current profile. From this step; the long short term memory(LSTM) used to predict nonlinear and time-series datasets is newly proposed in the second step. According to the comparison with the recurrent neural network(RNN)-based SOH and clear verification; this paper provides the effectiveness of the SOH prediction. |