<![CDATA[This paper presents an adaptive charging system for lithium-ion batteries using an Elman backpropagation neural network (EBNN) integrated with a PID controller and a ZETA converter. The system dynamically identifies the battery type and adjusts the charging voltage accordingly. The EBNN model was trained using 1441 samples of initial current and voltage data, achieving a mean squared error (MSE) of 7.64×10⁻¹⁴. A ZETA converter enables both step-up and step-down voltage regulation, while the PID controller ensures stability toward the predicted setpoint. Simulations in Simulink were conducted on four lithium-ion battery types with setpoints of 4.4 V, 8.8 V, 14.4 V, and 21.6 V. The results show that the PID-regulated output voltage closely matches the target with a maximum deviation of ±0.05V and an average voltage error of 0.1725%. The system achieves fast response times between 0.015 and 0.033 seconds. Extended testing through 24 randomized trials confirmed consistent identification and regulation across varying battery types. These findings validate the proposed EBNN-PID-based charging system as a highly accurate, flexible, and efficient solution for managing lithium-ion battery charging in real-time embedded applications.]]>
Copyrights © 2025
