基于S变换的电能质量扰动识别算法研究

发布时间：2018-05-16 12:35

本文选题：电能质量 + 扰动识别　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：随着科学技术与工业智能化的迅猛发展,电能质量问题对于航空航天、汽车、电子制造等重要领域造成的影响日益显著,因此电能质量问题的治理得到广泛的关注和研究。电能质量问题治理的关键步骤是电能质量扰动的识别,即电能质量扰动检测和分类识别。本文总结了由电能质量问题所引起的危害,并根据电能质量相关标准以及国内外的发展现状,从电能质量扰动检测和分类识别两个方面深入研究电能质量扰动识别算法。针对电能质量扰动检测问题,依据电力系统中常见的电能质量问题,总结归纳了12种电能质量扰动数学表达式。研究了常用的时频域检测算法(短时傅里叶变换、小波变换),并利用MATLAB对常用检测算法进行仿真,并阐述了各算法的优缺点。针对常用时频域法的不足,本文着重研究了利用S变换进行电能质量扰动检测的方法,并在对S变换原理分析的基础上,提出了一种分段改进S变换检测方法,克服了差分法、有效值法、傅里叶变换无法检测所有电能质量扰动类型、小波变换受噪声影响大以及短时傅里叶变换窗宽固定的缺点。该方法将检测信号分为低频、中频和高频段,在每个频段中存在相对应的窗宽调节因子,通过改变窗宽调节因子,可以改变S变换的时频分辨率。在研究过程中,分析检测误差和分段改进S变换结果中扰动区域的峰度之间的关系,确定每个频段的最优窗宽调节因子。经仿真对比得出分段改进S变换比常用检测算法的检测能力更强。针对电能质量扰动信号分类识别问题,利用分段改进S变换对扰动信号进行检测,并对检测结果进行分析,分析得出时间-幅值和频率-幅值两种特征向量,其能够反映扰动信号在时域和频域的主要特征。从两种特征向量中,提取了7种特征量。根据7种特征量构建出具有14条分类规则的决策树分类器,实现了扰动信号的自动分类。根据每种扰动信号的扰动参数随机生成测试样本对本文所提出的识别算法进行仿真测试。测试结果表明新的识别算法具有较高的识别准确率。基于上述提出的电能质量扰动识别算法,设计了一种以DSP+ARM双核CPU芯片OMAP-L138为平台,具有基本电力参数测量、电能质量扰动识别、电能质量参数测量、历史数据存储和网络通信等功能的电能质量扰动识别装置。试验测试结果表明,该电能质量扰动识别装置基本达到所设计功能的要求。
[Abstract]:With the rapid development of science and technology and industrial intelligence, power quality problems have become increasingly significant in aerospace, automotive, electronic manufacturing and other important fields. The key step of power quality problem solving is the identification of power quality disturbance, that is, the detection and classification of power quality disturbance. In this paper, the harm caused by power quality problems is summarized. According to the relevant standards of power quality and the current situation of development at home and abroad, the power quality disturbance identification algorithm is deeply studied from two aspects: power quality disturbance detection and classification recognition. Aiming at the problem of power quality disturbance detection, 12 kinds of mathematical expressions of power quality disturbance are summarized according to the common power quality problems in power system. The common detection algorithms in time-frequency domain (short time Fourier transform, wavelet transform) are studied. The common detection algorithms are simulated by MATLAB, and the advantages and disadvantages of each algorithm are expounded. In order to overcome the shortcomings of time-frequency domain method, this paper mainly studies the method of power quality disturbance detection using S-transform. On the basis of analyzing the principle of S-transform, a piecewise improved S-transform detection method is proposed, which overcomes the difference method. Effective value method, Fourier transform can not detect all types of power quality disturbance, wavelet transform is greatly affected by noise and the window width of short time Fourier transform is fixed. In this method, the detected signals are divided into low frequency, intermediate frequency and high frequency bands. There are corresponding window width adjustment factors in each frequency band. By changing the window width adjustment factor, the time-frequency resolution of S-transform can be changed. In the research process, the relationship between the detection error and the kurtosis of the disturbed region in the modified S transform is analyzed, and the optimal window width adjustment factor for each frequency band is determined. The simulation results show that the improved S transform has better detection ability than the common detection algorithms. Aiming at the problem of classification and recognition of power quality disturbance signal, the disturbance signal is detected by using piecewise improved S transform, and the detection results are analyzed, and two characteristic vectors, time-amplitude and frequency-amplitude, are obtained. It can reflect the main characteristics of disturbance signal in time domain and frequency domain. Seven feature vectors were extracted from two kinds of feature vectors. A decision tree classifier with 14 classification rules is constructed according to the seven characteristic quantities, and the disturbance signal is classified automatically. According to the disturbance parameters of each disturbance signal, random test samples are generated to test the identification algorithm proposed in this paper. The test results show that the new recognition algorithm has high recognition accuracy. Based on the above proposed power quality disturbance identification algorithm, a DSP ARM dual-core CPU chip OMAP-L138 platform is designed, which has basic power parameter measurement, power quality disturbance identification, power quality parameter measurement, and power quality parameter measurement. Power quality disturbance recognition device for historical data storage and network communication. The test results show that the power quality disturbance identification device basically meets the requirements of the designed function.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM711

【参考文献】