Journal
INTERNATIONAL JOURNAL OF ENERGY RESEARCH
Volume 46, Issue 15, Pages 23730-23745Publisher
WILEY-HINDAWI
DOI: 10.1002/er.8671
Keywords
BiLSTM; lithium-ion battery; SOH; wavelet analysis
Categories
Funding
- Science and Technology Project of State Grid Zhejiang Electric Power Co., Ltd [5211JY220001]
- National Natural Science Foundation of China [52007170]
- Zhejiang Province Natural Science Foundation [LY22E070007]
- Key Projects of Shandong Province Natural Science Foundation [ZR2020KF020]
- Youth Fund of Shandong Province Natural Science Foundation [ZR2020QE212]
Ask authors/readers for more resources
This paper proposes a SOH estimation method based on the double bi-directional long short-term memory (DBiLSTM) model, which can accurately estimate recovered capacity and improve the accuracy of SOH estimation.
At present, the rapid development of new energy sources makes lithium-ion batteries (LIBs) widely used, but LIBs will inevitably age during using. State of health (SOH) is a direct reflection to the aging of LIBs, so it is necessary to estimate the SOH. During the aging process of the LIBs, the phenomenon of capacity recovery will occur if the battery is standing for too long. Existing SOH estimation methods based on neural network do not propose countermeasures for the phenomenon, but in fact, capacity recovery is inevitable and it has a great impact on SOH estimation. According to this vacancy, this paper proposes a SOH estimation method based on double bi-directional long short-term memory (DBiLSTM) model, which can accurately estimate recovered capacity and improve accuracy of SOH estimation. First, the capacity of LIB is decomposed at multiple scales using wavelet analysis, and the smooth and fluctuating components are obtained. Then six features are proposed based on the changes in the battery after aging. The proposed features are decomposed into new features suitable for the two components. Finally, the smooth component and the fluctuation component are estimated synchronously, and the estimated results are reconstructed to obtain the final estimated SOH. The method proposed in this paper is verified in the NASA dataset and compared with the bi-directional long short-term memory (BiLSTM) model. Comparing with the direct estimation by BiLSTM, the root mean square error (RMSE) is reduced by at least 0.0084 and the mean absolute percentage error (MAPE) is reduced by at least 0.52% when the battery capacity fluctuates greatly. The experimental results show that the proposed method can significantly improve the accuracy of SOH estimation when the capacity fluctuates greatly.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available