【作者】 陈春玲；

【导师】 杨勇；

【作者基本信息】 沈阳农业大学 ， 农业电气化与自动化， 2009， 博士

【摘要】 随着电力负荷的日趋复杂,电能质量问题给现代电力系统以及电力用户造成的危害越来越突出。谐波、闪变等稳态扰动产生的危害普遍存在,电压骤升骤降等暂态扰动造成的危害则在不断加大。对电网实现综合全面的电能质量在线监测将成为保障其高质量运行的必要手段,而实现在线监测的基础是合理有效的电能质量检测分析方法。本文重点研究了稳态谐波检测的改进方法、基于小波、希尔伯特变换的暂态电能质量扰动检测与定位方法、基于支持向量机的电能质量扰动分类方法。以上述理论方法为基础,研究虚拟仪器环境下电能质量在线监测系统的设计与实现。为了解决电力系统中不可避免的噪声对稳态谐波检测的影响,在分析了小波变换和傅立叶变换各自的优势,提出了基于小波去噪的加窗FFT谐波检测方法。该方法采用db5软阈值4层Daubechies小波去噪方法对采集的含噪信号进行去噪处理,再采用加汉宁窗的傅立叶变换对去噪信号进行检测,即能比较准确的反映总谐波畸变率(THD),又能比较精确的检测各次谐波含有率,为谐波治理提供有利的、准确的谐波分量数据。提出两种暂态电能质量扰动检测方法:(1)基于提升db4小波变换的暂态电能质量扰动检测与定位方法,在电力系统低噪声情况下,所构造的提升db4小波对多种暂态电能质量扰动信号不仅能进行扰动突变点检测,而且能进行扰动起止时刻的准确定位。(2)基于希尔伯特变换的移相电能质量扰动检测方法,在高噪声情况下,仍能出色完成电能质量扰动检测,而且原理简单、实时性好,集成于虚拟仪器电能质量检测平台,验证了其有效性和实用性。鉴于傅立叶变换优秀的幅频特性,小波变换优秀的时频特性和支持向量机优秀的统计学习能力,提出了基于FFT、小波变换和多类支持向量机的电能质量扰动识别方法。利用傅立叶变换和小波变换提取电能质量扰动的特征向量;构建多类支持向量机分类器对电力系统常见的八种电能质量扰动信号进行识别分类。在含噪声情况进行仿真,结果表明该方法分类准确率高,对噪声不敏感,训练样本少、训练时间短,是电能质量扰动识别的有效方法。采用NI PXI架构虚拟仪器平台完成了电能质量监测系统硬件设计,在NI LabView环境中设计并集成了所研究的电能质量扰动检测分析算法,构成基于PC资源的电能质量柔性监测系统。采用标准功率源6100A输出的扰动信号进行检测,证明该平台对稳态电能质量扰动(电压允许偏差、谐波、三相电压允许不平衡度和频率偏差)和暂态电能质量事件(电压骤升、电压骤降、电压波动、暂态振荡等)均可进行实时检测,测试结果准确、界面友好、性能稳定,同时具有高可靠性、功能可扩展性等优点。能够综合、全面地反映电网中的电能质量问题,为提高电能质量提供科学准确的数据支持。更多还原

【Abstract】 With the power loads being more and more complex, the disturbance of power quality harms power system and the electricity users more seriously. The harm of steady-state disturbance is widespread such as harmonics, flicker, etc. Therefore sags, swells and other transient disturbances harm power quality badly, and the comprehensive online monitoring of power quality detection is becoming a necessary means in running a high-quality power network, and the realization fundation of online monitoring which is based on reasonable and efficient methods of power quality analysis. This research work focuses on the improvement of steady-state harmonic detection method and the transient power quality disturbance detection and location based on wavelet and Hilbert transform and the classification of power quality disturbances based on SVM. In view of the above-mentioned theoretical approaches we researched the design and implementation of the power quality online monitoring System under the circumstance of Virtual Instrument (in short VI).In order to resolve the inevitable impact of noise in the steady-state harmonic detection, and after the analysis of the wavelet and Fourier transform, one method of harmonic detection was put forward by using wavelet de-noising added windowed FFT. This method uses firstly db4 wavelet with db5 layer soft-threshold in de-noising, and then detects the noise based on Fourier transform added Hanning window. It is able to reflect the total harmonic distortion (THD) more accurately, and also to detect more precise rate of each harmonic contains. It can produce more accurate and useful harmonic data to control harmonics.Two transient power quality disturbance detection methods were put forward: (1) The method in transient power quality disturbance detection and location based on db4 lifting wavelet transform. Under this method, the power system in case of low-noise, the db4 lifting wavelet not only to detect mutation point of transient power quality disturbance signals, but also to locate the beginning and ending time of the disturbance. (2) The method in phase-shifting power quality disturbance detection based on Hilbert transformation. In case of high-noise, the power quality disturbances can be detected. According to the principle of the simple, real-time, and integrated in the virtual instrument platform for power quality detection, the effectiveness and practicality can be verified. After studying excellent amplitude-frequency characteristics of Fourier transform, and excellent time-frequency characteristics of wavelet transform and excellent statistical learning ability of SVM, the method in power quality disturbances recognition based on FFT, wavelet and SVM was created. The multi-class SVM classifier was constructed to recognize and classify the eight common power quality disturbances by applying the Fourier and wavelet transform to extract power quality disturbances eigenvector. The simulation was done under the noise situation. The simulation results showed that the classification accuracy was high, and not sensitive to noise. It has also the characters of less training samples and sort training time. The method is therefore an effective way in recognition of power quality disturbances.The monitoring system hardware design was completed by using NI PXI virtual instrument platform. Under the environment of NI LabView we designed and integrated all the algorithms in power quality disturbance analysis and detection, and also developed the soft power quality monitoring system based on PC resources. Using 6100A standard power source output signal in detection to the disturbance proved that the platform of power quality disturbances on the steady-state (voltage deviation allowed, harmonic, three-phase voltage imbalance and frequency deviation) and the transient power quality events (swell, sags, voltage fluctuations, transient oscillation ) can be real-time detected. The tests showed some advantages such as results accuracy, user-friendly interface, stable performance, high reliability and scalability, etc. The method can reflect the power quality status of distribution network integrated and comprehensively, and provide scientific and accurate data support for improving the operation quality of power distribution network.更多还原