Lithium battery fault diagnosis by integrating improved EMD decomposition algorithm and 2DCNN.

Area of Science:

• Energy Storage
• Artificial Intelligence
• Signal Processing

Background:

• Lithium batteries are crucial energy storage devices requiring robust fault diagnosis.
• Traditional methods struggle with complex signals and feature extraction in battery diagnostics.
• Developing advanced diagnostic models is essential for lithium battery safety and performance.

Purpose of the Study:

• To develop a high-performance fault diagnosis model for lithium batteries.
• To overcome limitations of traditional fault diagnosis techniques.
• To improve the accuracy and efficiency of lithium battery fault detection.

Main Methods:

• Utilized an optimized empirical mode decomposition (EMD) algorithm for processing lithium battery voltage data.
• Employed a two-dimensional convolutional neural network (2DCNN) for feature extraction, training, and classification.
• Integrated EMD's decomposition stability with 2DCNN's high-dimensional representation capabilities.

Main Results:

• Achieved 98.7% fault feature information consistency at 600 iterations.
• Maintained 99.2% feature recognition accuracy with 70 features.
• Demonstrated significantly lower running times (below 6ms) compared to other methods across 7 data groups.

Conclusions:

• The developed lithium battery fault diagnosis model enhances accuracy and efficiency.
• This integrated approach offers new technological solutions for battery diagnostics.
• The findings support the advancement and application of lithium battery technology.