1. The relevance of this study stems from the fact that generative AI (GenAI) tools have become routine in undergraduate education. Alongside convenience, however, they reshape the structure of aca...
23 September 2026 in Journal of problems in computer science and information technologies
Lithium-ion batteries are widely used in electric vehicles, portable electronics, and grid energy systems. The constant current–constant voltage (CC–CV) method shows clear limits at high state of charge when polarization increases, temperature rises, and charging time is extended. These effects become stronger in the high SOC region, where the charging rate decreases. Consequently, to address this challenge, in this study we presents a hybrid charging method that combines physics-informed deep learning with an auxiliary hardware circuit. The circuit is placed between the charger and the battery without changes to the charger structure. It enables real-time control of pulse frequency and duty cycle. A one-dimensional convolutional neural network is applied during short relaxation periods for SOC estimation. The proposed method reduces polarization and thermal stress and improves charging efficiency at high SOC. It also improves open-circuit voltage estimation accuracy. Experimental tests are conducted on lithium-ion cells under controlled temperature conditions. The results confirm improved performance in fast charging applications.