| Title |
Visualization and Estimation of Battery Data Status using the LSTM-AutoEncoder Algorithm |
| Authors |
Tae-Geol Woo ; Eun-Ji Lee ; In-Ho Cho ; Kang-Moon Park |
| DOI |
https://doi.org/10.6113/TKPE.2024.29.3.216 |
| Keywords |
Autoencoder; Deep learning; LSTM; Battery state |
| Abstract |
The growing concern for the environment and changing consumer preferences are driving the increasing demand for electric vehicles in the market. Compared with traditional internal combustion vehicles, electric vehicles are considered the future of transportation due to their low fossil fuel consumption and reduced environmental impact. Among the major components of electric vehicles, their batteries have a finite lifespan and need frequent replacements and thus necessitate further research on their state and performance. Previous studies have used linear regression models and time series prediction models to predict battery state and aging. In this study, we aim to visualize the current state of a battery using a time-series specific LSTM-AutoEncoder. The dataset is derived from repeated charge/discharge experiments, and visualization data can be extracted through algorithmic configuration using an autoencoder. The prediction accuracy for the current state is more than 90%, demonstrating a high level of accuracy. |