@article{article_1716660, title={THE ESTIMATION OF THE STATE OF CHARGE OF A LITHIUM-ION AND SUPERCAPACITOR HYBRID BATTERY MODEL BASED ON K-NEAREST NEIGHBOURS A MACHINE LEARNING APPROACH}, journal={International Journal of Energy and Smart Grid}, volume={10}, pages={58–78}, year={2025}, DOI={10.55088/ijesg.1716660}, author={Kakeci, Muhammed İkbal and Aslan, Zülfikar and Özçelik, Mehmet Ali}, keywords={KNN neural network, Electric vehicle, hybrid energy storage systems, lithium-ion batteries, supercapacitors, state of charge.}, abstract={The Hybrid Energy Storage System signifies a substantial advancement in the domain of energy storage technology, particularly within the context of electric vehicles. The system integrates batteries and supercapacitors, offering a combination that is regarded as one of the most crucial technologies in this domain. The primary advantage of Hybrid Energy Storage System lies in its ability to provide high efficiency in terms of storage capacity and the immediate availability of power when it is required. The estimation of the state of charge is of paramount importance, given its impact on enhancing the performance, efficiency, and safety of vehicles. The estimation of the state of charge is a considerable challenge due to the variable charging and discharging currents present in both the battery and the supercapacitor. In response to this challenge, researchers have developed numerous methods to estimate the state of charge. The present study proposes a novel approach to charge state estimation, underpinned by a sophisticated algorithm that aims to minimize complexity and enhance accuracy. The K-Nearest Neighbors algorithm is utilized in this study due to its simplicity and interpretability, rendering it well-suited for prediction tasks in complex and non-linear systems, such as those found in battery and supercapacitor technologies. The experimental results demonstrated an average absolute error of 0.0021 and a mean square error of 0.0031. These figures are indicative of the model’s high degree of accuracy and its capacity to closely mirror the true values. The supercapacitor also demonstrates robust performance. The correlation coefficient was measured at 0.9864. This finding suggests a strong correlation between the independent variables and the dependent variable, as well as a high degree of model fidelity. This high correlation indicates that the model predictions are consistent with the true values. The proposed study calculated the mean absolute error to be 0.0075 and the root mean square error to be 0.0835. These findings suggest that the model predictions are in close proximity to the true values, thereby demonstrating the model’s overall high performance.}, number={2}, publisher={Zülküf GÜLSÜN}