Randomized trial demonstrates accurate state of health estimation in lithium-ion batteries, indicating enhanced safety and reliability.
Key Points
The goal is to create a more accurate method for estimating the state of health of lithium-ion batteries using partial charging data and advanced image techniques.
Collected voltage and temperature data during specific state of charge intervals after wavelet denoising.
Transformed partial charge data into multimodal image data using various techniques.
Implemented an adaptive graph channel attention mechanism within a residual network architecture and used a bidirectional LSTM for feature extraction.
Achieved mean absolute error and root mean square error both below 1.3% for SOH estimation.
The method demonstrates strong accuracy and generalization capability when tested on NASA and Oxford datasets.