Battery_SOC_Estimat​ion

Battery_SOC_Estimation

SOC (State of Charge) estimation for a battery using an ensemble approach with Coulomb counting and pre-trained LSTM prediction

An accurate estimation of battery’s State of Charge (SoC) is a prerequisite prior to devising battery management and control systems. Traditional techniques including Coulomb counting and open circuit voltage (OCV) methods still need improvements due to battery’s innate issues such as non-linearity, temperature dependence, and aging effects. We introduce a machine learning-based approach for estimating the SoC of a battery using voltage, current, and temperature data. We utilized battery data from four Tesla Model 3 battery packs with varying temperature and discharge cycle environments. This dataset encompasses the battery's SoC, voltage, current, and temperature measurements over time.

Our findings demonstrated that our model could estimate the battery's SoC with an RMSE of less than 2%. The proposed methodology overcomes challenges inherent in conventional estimation techniques and offers the potential for application across diverse battery technologies while ensuring the explainability of the model's predictions

Files

• Battery_Data.mat: Battery data for validation purposes.
• trained_lstm.mat: Pre-trained LSTM network's weights.
• Model.m: SOC estimator, ensemble approach with Coulomb counting and pre-trained LSTM prediction.

reference

The data and parts of the example code were based on the following reference material.

@inproceedings{kollmeyer2022blind,
  title={A blind modeling tool for standardized evaluation of battery state of charge estimation algorithms},
  author={Kollmeyer, Phillip J and Naguib, Mina and Khanum, Fauzia and Emadi, Ali},
  booktitle={2022 IEEE Transportation Electrification Conference \& Expo (ITEC)},
  pages={243--248},
  year={2022},
  organization={IEEE}
}