摘要
针对电能质量扰动导致电压、电流波形畸变,并且不同形式的畸变造成电能质量信号在时域和频域上表现出复杂的特征,增加了信号分析和处理的难度问题,以多特征融合为基本手段,提出电能质量扰动信号调制算法,使信号更易高效地实现检测和识别。从S变换、小波变换得到的所有电能质量扰动信号特征中,利用分类回归树和基尼重要度,选择出具有表征性的时域信号特征和频域信号特征,并通过主成分分析法完成多特征融合。根据LSTM(Long Short-Term Memory)基于融合特征给出的电能质量扰动信号类别,由信号生成器输出调制后的电能质量扰动信号。实验结果表明,选取的信号特征的信噪比值均超过90 dB,具有较强的表征能力;该算法调制的信号具有较强的易识别性,单一和复杂类型均能实现准确识别;频率偏差在±0.1 Hz范围内小幅波动,电能质量有显著提高。
Power quality disturbances cause distortion of voltage and current waveforms,and different forms of distortion result in complex characteristics of power quality signals in both time and frequency domains,increasing the difficulty of signal analysis and processing.Therefore,based on multi feature fusion,a modulation algorithm for signals of power quality disturbance is proposed to make signal detection and recognition easier and more efficient.From all the characteristics of power quality disturbance signals obtained from S-transform and wavelet transform,using classification regression tree and Gini importance,representative time-domain signal features and frequency-domain signal features are selected,and multi feature fusion is completed through principal component analysis.According to the LSTM(Long Short-Term Memory)based fusion feature,the category of power quality disturbance signal is given,and the modulated power quality disturbance signal is output by the signal generator.The experimental results show that the signal-to-noise ratio of the selected signal features exceeds 90 dB,indicating strong representational ability.The signal modulated by this algorithm has strong recognizability,and both single and complex types can be accurately identified.The frequency deviation fluctuates slightly within the range of±0.1 Hz,indicating a significant improvement in power quality.
作者
田野
TIAN Ye(Medical Engineering Department,Hainan Hospital of Chinese PLA General Hospital,Sanya 572013,China)
出处
《吉林大学学报(信息科学版)》
2025年第6期1237-1243,共7页
Journal of Jilin University(Information Science Edition)
基金
海南省教育厅基金资助项目(Hnjgzc2023-93)。
关键词
电能质量扰动
扰动信号
分类回归树
主成分分析
多特征融合
信号调制
power quality disturbance
disturbance signal
classification regression tree
principal component analysis
multi feature fusion
signal modulation
