Lithium-ion battery state of health estimation using artificial intelligence
Nowadays, with the continuous development of lithium-ion batteries (LIBs) technology, LIBs are ubiquitous, finding application in a wide array of devices ranging from smartphones and laptops to electric cars and exploration satellites. However, as LIBs undergo cycles, their capacity gradually decrea...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Master by Coursework |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/179910 |
| _version_ | 1826119985651515392 |
|---|---|
| author | Zhang, Zichao |
| author2 | Xu Yan |
| author_facet | Xu Yan Zhang, Zichao |
| author_sort | Zhang, Zichao |
| collection | NTU |
| description | Nowadays, with the continuous development of lithium-ion batteries (LIBs) technology, LIBs are ubiquitous, finding application in a wide array of devices ranging from smartphones and laptops to electric cars and exploration satellites. However, as LIBs undergo cycles, their capacity gradually decreases, leading to a range of issues. Therefore, it is particularly important to measure the state of health (SOH) of LIBs. This dissertation uses a hybrid network constructed by convolutional neural network (CNN) and bidirectional long-short-term memory (BiLSTM) neural networks, another hybrid network constructed by CNN, an active-state-tracking long-short-term memory (ASTLSTM) neural network, and a vision transformer with muse attention (MaViT), to estimate SOH. Three data formats are used for SOH estimation, namely, estimation using all data, estimation using a single health indicator (HI), and estimation using integrated HIs. In this experiment, two datasets were selected to test the effectiveness of using HIs for prediction. In the NASA dataset, the best result using all data for estimation is achieved by MaViT, with RMSE of 0.004 and MAPE of 0.00405; MaViT uses all data and integrated HIs for prediction, and CNN-BiLSTM uses a single HI for prediction, achieving excellent prediction results; In the CALCE dataset, CNN-ASTLSTM achieves excellent prediction results by utilizing a single HI, while CNN-BiLSTM also attains outstanding outcomes through the use of integrated HIs. |
| first_indexed | 2024-10-01T05:09:07Z |
| format | Thesis-Master by Coursework |
| id | ntu-10356/179910 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T05:09:07Z |
| publishDate | 2024 |
| publisher | Nanyang Technological University |
| record_format | dspace |
| spelling | ntu-10356/1799102024-09-06T15:43:50Z Lithium-ion battery state of health estimation using artificial intelligence Zhang, Zichao Xu Yan School of Electrical and Electronic Engineering xuyan@ntu.edu.sg Engineering Nowadays, with the continuous development of lithium-ion batteries (LIBs) technology, LIBs are ubiquitous, finding application in a wide array of devices ranging from smartphones and laptops to electric cars and exploration satellites. However, as LIBs undergo cycles, their capacity gradually decreases, leading to a range of issues. Therefore, it is particularly important to measure the state of health (SOH) of LIBs. This dissertation uses a hybrid network constructed by convolutional neural network (CNN) and bidirectional long-short-term memory (BiLSTM) neural networks, another hybrid network constructed by CNN, an active-state-tracking long-short-term memory (ASTLSTM) neural network, and a vision transformer with muse attention (MaViT), to estimate SOH. Three data formats are used for SOH estimation, namely, estimation using all data, estimation using a single health indicator (HI), and estimation using integrated HIs. In this experiment, two datasets were selected to test the effectiveness of using HIs for prediction. In the NASA dataset, the best result using all data for estimation is achieved by MaViT, with RMSE of 0.004 and MAPE of 0.00405; MaViT uses all data and integrated HIs for prediction, and CNN-BiLSTM uses a single HI for prediction, achieving excellent prediction results; In the CALCE dataset, CNN-ASTLSTM achieves excellent prediction results by utilizing a single HI, while CNN-BiLSTM also attains outstanding outcomes through the use of integrated HIs. Master's degree 2024-09-03T01:33:15Z 2024-09-03T01:33:15Z 2024 Thesis-Master by Coursework Zhang, Z. (2024). Lithium-ion battery state of health estimation using artificial intelligence. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/179910 https://hdl.handle.net/10356/179910 en application/pdf Nanyang Technological University |
| spellingShingle | Engineering Zhang, Zichao Lithium-ion battery state of health estimation using artificial intelligence |
| title | Lithium-ion battery state of health estimation using artificial intelligence |
| title_full | Lithium-ion battery state of health estimation using artificial intelligence |
| title_fullStr | Lithium-ion battery state of health estimation using artificial intelligence |
| title_full_unstemmed | Lithium-ion battery state of health estimation using artificial intelligence |
| title_short | Lithium-ion battery state of health estimation using artificial intelligence |
| title_sort | lithium ion battery state of health estimation using artificial intelligence |
| topic | Engineering |
| url | https://hdl.handle.net/10356/179910 |
| work_keys_str_mv | AT zhangzichao lithiumionbatterystateofhealthestimationusingartificialintelligence |