基于改进能量集中度的S变换与随机森林的电能质量扰动识别  

S-transform based on modified energy concentration and identification of power quality disturbance in random forest

作　　者：高健 崔雪[1] 邹晨露 刘洋[1] Gao Jian;Cui Xui;Zou Chenlu;Liu Yang(School of Electrical Engineering,Wuhan University,Wuhan 430000,China)

机构地区：[1]武汉大学电气工程学院,武汉430000

出　　处：《电测与仪表》2019年第1期8-14,21,共8页Electrical Measurement & Instrumentation

摘　　要：鉴于S变换的窗口函数对不同频带信号的自适应能力差,提出一种新型的改进S变换(Modified S-Transform,MST),该方法通过引入四个辅助参数,优化高斯窗函数尺度因子的自适应能力,使改进S变换的能量集中度最大化,获得了更出色的时频分辨能力。建立了基于扰动信号幅值和相位的特征值评价体系,采用随机森林(Random Forest,RF)算法对包括标准信号和电压暂降、电压暂升、高次谐波、暂态振荡等10种扰动信号共11类电能质量信号分类识别。与已有文献采用的决策树、支持向量机和神经网络分类结果进行了对比分析,仿真试验结果表明,该方法分类准确率高,抗干扰能力强,且在训练样本少、低信噪比(Signal-to-Noise Radio,SNR)条件下分类结果优势明显。In view of the poor adaptability of the S-transform window function to different frequency band signals,a new modified S-transform(MST)is proposed in this paper.This method optimizes the Gaussian window function scale factor by introducing four auxiliary parameters.Adaptability,it maximizes the energy concentration of the improved S-transformation,and achieves better time-frequency resolution.The eigenvalue evaluation system based on the amplitude and phase of the disturbance signal is established.The random forest(RF)algorithm is used to include 10 kinds of disturbance signals such as standard signal and voltage sag,voltage spurt,higher harmonic and transient oscillation.A total of 11 types of power quality signals are classified and identified.Compared with the decision tree,support vector machine and neural network classification results are used in the existing literature,the simulation results show that the method has high classification accuracy,strong anti-interference ability,and few training samples and low signal-to-noise ratio.Signal-to-noise radio(SNR)conditions have obvious advantages in classification results.

关 键 词：改进S变换 能量集中度 随机森林 电能质量 扰动识别 

分 类 号：TM743[电气工程—电力系统及自动化]