Open access dataset, code library and benchmarking deep learning approaches for state-of-health estimation of lithium-ion batteries

Fujin Wang,Zhi Zhai,Bingchen Liu,Shiyu Zheng,Zhibin Zhao,Xuefeng Chen
DOI: https://doi.org/10.1016/j.est.2023.109884
IF: 9.4
2024-01-01
Journal of Energy Storage
Abstract:Great progress has been made in deep learning (DL) based state-of-health (SOH) estimation of lithium-ion batteries, which helps to provide recommendations for predictive maintenance and replacement of lithium-ion batteries. However, despite the abundance of articles, few open-source codes are publicly available. While there are several public datasets, they tend to be more oriented toward simulating laboratory environments rather than real-world usage scenarios. Moreover, they solely provide raw data without any corresponding preprocessing codes, resulting in inconsistencies in preprocessing methods across different papers. These reasons lead to unfair comparisons and ineffective improvements. In response to these problems, this paper publishes a large-scale lithium-ion battery run-to-failure dataset, consisting of 55 batteries, and provides a unified data preprocessing method. Besides, we comprehensively evaluate 5 well-known DL-based models to provide benchmark research. To be specific, first, the existing DL-based SOH estimation methods are reviewed in detail. Second, we provide a comprehensive evaluation of DL-based models on 2 large-scale datasets, including 100 batteries, with 3 input types and 3 normalization methods. Third, we make the complete evaluation codes and dataset publicly available for better comparison and model improvement. Fourth, we discuss future DL-based SOH estimation, including unsupervised learning, transfer learning, interpretability, and physics-informed machine learning. We emphasize the importance of open-source code, provide baseline estimation errors (error upper bounds), and discuss existing issues in this field. The code library is available at: https://github.com/wang-fujin/SOHbenchmark.
energy & fuels
What problem does this paper attempt to address?
### What problem does this paper attempt to solve? This paper primarily aims to address several key issues in the estimation of lithium-ion battery State-of-Health (SOH): 1. **Lack of open-source code**: Although there is a plethora of research literature on lithium-ion battery SOH estimation, there is a scarcity of publicly available source code, leading to unfair comparisons between different studies. 2. **Dataset issues**: Existing public datasets are often biased towards laboratory simulations rather than real-world applications, and these datasets typically only provide raw data without preprocessing code, resulting in inconsistencies in preprocessing methods across different studies. 3. **Inconsistent model evaluation standards**: Due to the varying input types, hyperparameters, and other factors used by different researchers, it is challenging to make fair comparisons between different models. To address these issues, the paper makes the following contributions: - Released a large-scale dataset of lithium-ion battery run-to-failure data, containing data from 55 batteries, along with a unified data preprocessing method. - Conducted a comprehensive evaluation of 5 well-known deep learning models to provide benchmark studies. - Made the complete evaluation code and dataset publicly available to enable better comparison and improvement of models by other researchers. - Discussed future research directions for SOH estimation, including unsupervised learning, transfer learning, interpretability, and physics-informed machine learning, and emphasized the importance of open-source code. Through these efforts, the paper aims to provide a fair comparison environment and explore the issues present in existing methods.