Compensation of The Approximation Error of Support Vector Regression Using Semi-Parametric Model: Application to Lithium-Ion Battery State of Charge Estimation

Kiani, Maryam; Khorashadizade, Saeed; Shamsi Nejad, Mohammad Ali

doi:10.48311/mme.2025.117205.82876

Compensation of The Approximation Error of Support Vector Regression Using Semi-Parametric Model: Application to Lithium-Ion Battery State of Charge Estimation

Articles in Press

Document Type : Original Article

Authors

maryam kiani ¹

saeed khorashadizade ²

mohammad ali shamsi nejad ²

¹ Ph.D. Student, Department of Electrical and Computer Engineering, University of Birjand

² University of Birjand, Faculty of Electrical Engineering and Computer, power group

10.48311/mme.2025.117205.82876

Abstract

An accurate estimation of the state of charge (SOC) is necessary for optimal management of the energy in electric vehicles (EV) and protection of the battery from going to the deep discharge or overcharge conditions. Nowadays, many different artificial intelligence methods have been broadly developed and applied to this problem. In this paper, support vector regression (SVR) and semi-parametric models are combined to improve the accuracy of estimation. The motivation of this idea stems from the similarity of the optimal hyper-plain of SVR and the equation used in parametric regression models. However, to have a more flexible and accurate model, semi-parametric model is used. In fact, semi-parametric model plays the role of compensation for the approximation error of the SVR. For validation of the proposed method, various profiles of currents are used. Comparison of SVR, Radial basis function (RBF) neural network, multilayer perceptron (MLP) neural network with the proposed method, shows that the proposed method is more accurate in SOC estimation. In addition, the proposed method shows high accuracy and fast convergence in real-world implementations. The results based on Urban Dynamometer Driving Schedule (UDDS) shows the superiority of the proposed approach under realistic operating conditions.

Keywords

Lithium-ion battery

State of Charge Estimation

Support Vector Regression

Semi-Parametric Model

Subjects