基于鹦鹉优化多层极限学习机的电能质量扰动识别

打开文本图片集

中图分类号：TM711 文献标志码：A

A parrot optimizer-based multi-layer extreme learning machine for power quality disturbance identification

QIANWeijin，LAI Wenhao

（School of Electrical and Information Engineering，Anhui University of Science & Technology， Huainan 232001，China）

Abstract： With the widespread integration of new energy sources，power quality issues in electrical systems have become increasingly prominent. This study introduces a kernel mapping mechanism into an improved multi-layer extreme learning machine（ELM）to accurately identify power quality disturbances.It employs the parrot optimizer（PO），a novel heuristic algorithm，for parameter optimization.Based on IEEE standards，typicaldisturbancesignals are first constructed in MATLAB to colect relevant data.Theoriginal data arethen dimensionally reduced using stochastic neighbor embedding （SNE）to retain the effective key features. Finally，PO is applied to optimize the multi-layer kernel extreme learning machine（ML-KELM） parameters， enabling precisedisturbance identification andexploration of identification performance underdiferent dimensionality reductions. The proposed method achieves an identification accuracy of at least 94.49% for common single disturbances such as voltage sag，harmonic distortion，and voltage flicker，outperforming the traditional multi layer extreme learning machine （ML-ELM） methods by approximately 10% . The results confirm itsrobustnessand adaptability，providing an efective technical solution for identifying power quality disturbances.

Key words： power quality disturbance; parot optimizer; multi-layer kernel extreme learning machine; stochastic neighbor embedding

随着电力系统中电力电子设备和非线性负载的增加，电能质量问题日益复杂，电压暂降、闪变、谐波畸变及暂态扰动等电能质量扰动（powerqualitydisturbance，PQD）频发，对电网稳定性和可靠性构成威胁[1-3]。（剩余8799字）