Disturbances identification in renewable energy system using multi-kernel decomposition fusion method

TY - JOUR

T1 - Disturbances identification in renewable energy system using multi-kernel decomposition fusion method

AU - Xu, Guangda

AU - Wang, Bo

AU - Wang, Zhaohua

AU - Zhang, Jian

AU - Li, Guowu

N1 - Publisher Copyright: © 2026

PY - 2026/6/16

Y1 - 2026/6/16

N2 - The cumulative installed capacity of renewable energy sources (RES) such as wind energy has expanded rapidly due to the growing demand for clean energy. However, investments in equipment such as inverters and converters have exacerbated the degradation of power quality, leading to voltage fluctuations and harmonic distortion. To address complex disturbance identification, this study proposes a novel power quality disturbance (PQD) identification method based on a multi-kernel decomposition fusion (MKDF) framework. First, dynamic variational mode decomposition is employed for waveform decomposition, ensuring minimal signal decomposition energy loss. Then, an improved twin support vector machine is developed to integrate global and local characteristics. By combining decomposition and identification techniques, the MKDF framework is designed to effectively address complex power quality disturbances coupled with mechanical failure-induced specific frequency modulations in electrical signals, and thus accurately account for mechanical failure effects that are distinct from standard IEEE-1159 PQD features. Additionally, experimental evaluations, including a scenario with outer race bearing fault interference, demonstrate that the proposed method outperforms several recently developed approaches. It maintains 98.88% overall accuracy and 98.71% F1-score under mechanical failure interference, while delivering an end-to-end execution time of only 94 ms for real-time on-site monitoring.

AB - The cumulative installed capacity of renewable energy sources (RES) such as wind energy has expanded rapidly due to the growing demand for clean energy. However, investments in equipment such as inverters and converters have exacerbated the degradation of power quality, leading to voltage fluctuations and harmonic distortion. To address complex disturbance identification, this study proposes a novel power quality disturbance (PQD) identification method based on a multi-kernel decomposition fusion (MKDF) framework. First, dynamic variational mode decomposition is employed for waveform decomposition, ensuring minimal signal decomposition energy loss. Then, an improved twin support vector machine is developed to integrate global and local characteristics. By combining decomposition and identification techniques, the MKDF framework is designed to effectively address complex power quality disturbances coupled with mechanical failure-induced specific frequency modulations in electrical signals, and thus accurately account for mechanical failure effects that are distinct from standard IEEE-1159 PQD features. Additionally, experimental evaluations, including a scenario with outer race bearing fault interference, demonstrate that the proposed method outperforms several recently developed approaches. It maintains 98.88% overall accuracy and 98.71% F1-score under mechanical failure interference, while delivering an end-to-end execution time of only 94 ms for real-time on-site monitoring.

KW - Multi-kernel decomposition fusion

KW - Power quality disturbance

KW - Renewable energy system

KW - Support vector machine

KW - Variational mode decomposition

UR - https://www.scopus.com/pages/publications/105037027529

U2 - 10.1016/j.measurement.2026.121653

DO - 10.1016/j.measurement.2026.121653

M3 - Article

AN - SCOPUS:105037027529

SN - 0263-2241

VL - 278

JO - Measurement: Journal of the International Measurement Confederation

JF - Measurement: Journal of the International Measurement Confederation

M1 - 121653

ER -