Akademik Veri Yönetim Sistemi
Journal of Electrical Engineering and Technology, cilt.21, sa.2, ss.1873-1887, 2026 (SCI-Expanded, Scopus)
In this study, the state of energy estimation of the INR18650A28 lithium-ion battery type is performed with a random forest algorithm using hyperparameter values determined by the latest metaheuristic optimization methods. For state of energy (SOE) estimation, the hyperparameter values of the Random Forest (RF) machine learning method were determined by the Sparrow Search Algorithm (SSA) and ten other cutting-edge optimization methods and new artificial intelligence models were created. Different hyperparameter values for charge-discharge data consisting of 1C and 2C discharge levels were obtained with optimization results. The tests showed that the random forest algorithm’s SOE estimation, which relied on the sparrow search optimization technique, was the most effective. When the average of all machine learning experiments was taken, the SSA-RF method was the most successful method, reaching a value of 98.8072% according to the R2 metric. Additionally, the five other state-of-the-art artificial intelligence models were compared to the SSA-RF method’s estimate performance while predicting the lithium-ion battery’s SOE. With a value of 0.023873%, the MAE measure for the SOE estimates of the lithium-ion battery created using the SSA-RF method and the most advanced artificial intelligence models emerged as the most effective approach with the lowest estimation error. The application of the proposed approach to an additional INR21700M50LT lithium-ion battery, further substantiates its efficacy.