Thermal Management of Liquid-Cooled 21700 Lithium-Ion Batteries: A Comparative Evaluation of LQR and Reinforcement Learning Control

                @article{CRSIJ26000390,
                  author = {Kelfin Munene Njagi, Benjamin Aidoo, Robert Sowah},
                  title = {Thermal Management of Liquid-Cooled 21700 Lithium-Ion Batteries: A Comparative Evaluation of LQR and Reinforcement Learning Control},
                  journal = {Cosmo Research and Science International Journal},
                  year = {2026},
                  volume = {2},
                  number = {2},
                  pages = {447-468},
                  issn = {3108-1584},
                  url = {https://cosmorsij.com/published/CRSIJ26000390.pdf},
                  abstract = {Effective thermal management is essential for lithium-ion battery safety and performance. This study develops a control-oriented five-node lumped thermal model of a liquid-cooled cylindrical 21700 cell incorporating Bernardi heat generation, state-of-charge dynamics, and temperature-dependent internal resistance. The model compares Linear Quadratic Regulator (LQR) and Reinforcement Learning (RL) controllers under identical conditions. Simulation results show that LQR maintains the cathode temperature closer to the operating point and reduces cooling energy by 39.13% compared with continuous pump operation. At an initial temperature of 45 °C, LQR requires less cooling energy than RL.},
                  keywords = {Battery thermal management, 21700 lithium-ion battery, liquid cooling, Linear Quadratic Regulator, reinforcement learning, optimal control},
                  month = {September}
        }