Tehnički vjesnik, Vol. 33 No. 2, 2026.
Izvorni znanstveni članak
https://doi.org/10.17559/TV-20250723002856
A Diode-Enhanced Equivalent Circuit and SVM-Based Framework for Accurate Simulation of Battery Discharge under Variable Currents
Houcheng Zhao
; 1) Geely University of China, Chengdu, Sichuan 641423, China 2) Ulaanbaatar Erdem University, Mongolia School of Management, 999097, Ulaanbaatar, Mongolia
Juan Li
; Sichuan Hope Automobile Vocational College, School of Automotive Engineering, Ziyang, Sichuan 641399, China
Yaorong Yang
; Geely University of China, Chengdu, Sichuan 641423, China
*
Chun Gan
; Sichuan Hope Automobile Vocational College, School of Automotive Engineering, Ziyang, Sichuan 641399, China
Yanxi Ou
; Geely University of China, Chengdu, Sichuan 641423, China
B. Ochgerel
; Ulaanbaatar Erdem University， Mongolia School of Management, 999097, Ulaanbaatar, Mongolia
Y. Oyuntugalag
; Ulaanbaatar Erdem University， Mongolia School of Management, 999097, Ulaanbaatar, Mongolia
* Dopisni autor.
Sažetak
Accurate simulation of battery discharge under variable current conditions is essential for improving the performanceand safety of new energy vehicles. Conventional equivalent circuit models struggle to capture voltage hysteresis, leading to cumulative errors in state of charge (SOC) estimation. This paper proposes a diode-incorporated dual-polarization equivalent circuit model that simulates voltage delay behaviour during charge-discharge transitions. A support vector machine (SVM) model is employed to estimate open-circuit voltage (OCV), which is combined with the ampere-hour integration method for SOC prediction. The proposed framework is further validated against the pseudo two-dimensional (P2D) electrochemical model using ion and charge conservation laws. Simulation experiments on a lithium-ion battery model under representative driving conditions demonstrate that the diode-enhanced model reduces hysteresis-induced errors by 32.7%, with SOC estimation errors within ±1.5%. The discharge duration at 75% initial capacity is shortened by 35% compared to full capacity, and discharge efficiency varies linearly with driving speed. These results confirm that the proposed hybrid electrochemical-circuit-machine learning framework provides a reliable method for analyzing battery discharge behaviour under dynamic currents.
Ključne riječi
converter discharge; converter working conditions; discharge efficiency; new energy vehicles; power batteries; support vector machine
Hrčak ID:
345002
URI
Datum izdavanja:
28.2.2026.
Posjeta: 163 *