Life degradation monitoring using generated constant-current charging voltage curves for lithium-ion batteries

Lithium-ion batteries are a critical component of modern energy storage systems, and accurate monitoring of their life degradation is essential. Many methods rely on constant-current (CC) voltage curves obtained under controlled laboratory conditions for precise life monitoring. However, diverse charging strategies in practical applications hinder the battery management system from acquiring complete CC voltage curves. To address this, we develop a model based on the Pix2pixGAN deep learning network to generate complete CC voltage curves using non-CC multi-stage charging data. We validate the model using charging data from battery cells at different stages of degradation. The results show that the root mean square error (RMSE) of the generated CC voltage curves is less than 6 mV. Considering potential inaccuracies in state of charge (SOC) estimation by actual vehicle BMS, we further validate the model using two sets of non-CC multi-stage charging data with initial SOC errors of 3 % and 5 %. The RMSE remains below 8 mV, confirming the model's reliability and accuracy. This work enhances the use of voltage data under diverse charging strategies, promoting the practical application of battery life degradation monitoring algorithms that rely on CC voltage curves.