Joint prediction of the state of health and remaining useful life for lithium-ion batteries based on Wrapper Cascade-Stacking
The state of health (SOH) and remaining useful life (RUL) of lithium-ion batteries are important indicators for assessing battery safety and stability. Currently, a single model is still used in most studies to predict the health status or remaining service life of lithium batteries, making it diffi...
Saved in:
| Published in: | Journal of energy storage Vol. 112; p. 115563 |
|---|---|
| Main Authors: | , , |
| Format: | Journal Article |
| Language: | English |
| Published: |
Elsevier Ltd
15.03.2025
|
| Subjects: | |
| ISSN: | 2352-152X |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The state of health (SOH) and remaining useful life (RUL) of lithium-ion batteries are important indicators for assessing battery safety and stability. Currently, a single model is still used in most studies to predict the health status or remaining service life of lithium batteries, making it difficult to comprehensively assess the aging of lithium-ion batteries. Therefore, this paper proposes a joint SOH and RUL prediction method based on Cascade-Stacking integration. The method first uses the Wrapper feature selection algorithm to select the best health feature. Then, learning algorithms such as long and short-term memory neural network and extreme gradient boosting tree (XGBoost) are stacked using the first Stacking Integration algorithm to build a lithium-ion battery aging model for SOH prediction. A second Stacking integration algorithm is used to superimpose support vector machine, XGBoost and random forest to predict the trend of the optimal health feature with increasing number of cycles, and the predictions are series-coupled with the battery aging model to achieve RUL prediction. The performance of the proposed method is verified using the NASA battery degradation dataset, and the results show that the method can simultaneously achieve RUL prediction error within 4 cycles and SOH root-mean-square error within 1.6 %, thus validating the effectiveness and accuracy of the method.
•By extracting different HFs from the two processes of charging and discharging of the battery, and then adopting the wrapper feature selection method, in which the backward elimination (BE) method is chosen as the search method as well as the gaussian process as the objective function, the feature selection is conducted and the best health feature (BHF) is selected. This reduces the complexity of the feature variables under the premise of guaranteeing the accuracy of the model;•As lithium batteries in the use of the process of capacity regeneration phenomenon, resulting in different moments in the battery SOH the same but the actual RUL is different, so only rely on the battery SOH a parameter is difficult to comprehensively and accurately evaluate the health of the battery, so this paper proposes a lithium-ion battery SOH and RUL joint prediction method, not only to get the short-term health status of lithium batteries, and also accurately predict the battery long-term health status change trend. health status change trend;•Establish two Stacking integration models, embed learning algorithms such as LSTM and XGBoost in Stacking integration model 1, and embed a variety of machine learning algorithms in Stacking integration model 2, and tandem-couple the two Stacking integration models and use the double stacking prediction of the Stacking integration model effectively integrates the advantages of multiple dissimilar machine learning algorithms, enhances the effectiveness of the overall prediction model and the ability to resist overfitting, and achieves better prediction results in the short-term trend of SOH change and the long-term trend of RUL change in lithium-ion batteries. |
|---|---|
| ISSN: | 2352-152X |
| DOI: | 10.1016/j.est.2025.115563 |