【作者】 王继东；

【导师】 王成山；

【作者基本信息】 天津大学 ， 电力系统及其自动化， 2005， 博士

【摘要】 现代社会中,电能是一种最为广泛使用的能源,其应用程度成为一个国家发展水平的主要标志之一。随着计算机、电力电子和信息技术等高新产业的发展和普及,对电能质量提出了越来越高的要求。暂态电能质量扰动带来的问题已日益受到电力部门与用户的关注。为了采取适当的措施降低扰动带来的影响,改善电能质量,首先要对电能质量问题进行评估与分析。本文针对暂态电能质量问题,主要在电能质量扰动检测与扰动源定位、扰动分类、扰动数据压缩与信号去噪等方面做了一些工作,取得了如下成果:在小波变换局部模极大值理论的基础上,针对连续小波变换的计算量大,存在较大冗余的缺点,采用二进小波变换对电能质量扰动进行检测。小波变换的局部模极大值对应信号的突变点,可以用来检测电能质量扰动。仿真结果表明在分解尺度一上可以实现较为准确的检测。在扰动功率定义的基础上,根据Parseval定理,采用小波变换计算扰动发生期间三相瞬时功率信号能量的变化。由能量变化极性的正负,并结合扰动功率的第一个峰值的极性,提出了判断电容器投切相对位置的新判据,仿真结果表明了该方法的有效性。在电能质量扰动信号小波包分解的基础上,提出了用熵特征向量法提取电能质量扰动特征,并与能量特征向量比较,分别用Fisher分段线性分类器和贝叶斯分类器对典型电能质量扰动进行分类,仿真结果表明,采用熵特征向量法提取特征具有较高的识别正确率。在对扰动信号进行多分辨率分析的基础上,提出了一种能量阈值法,通过计算需要保留的能量来减少小波系数,并结合自适应算术编码方法进一步压缩数据,通过对典型电能质量扰动数据压缩与重构的仿真,并与小波系数阈值法相比较,表明该方法进一步改进了压缩效果。在小波变换软阈值法去噪的基础上,考虑到电力系统中的噪声一般为服从正态分布的高斯白噪声,在正态分布3σ原则的基础上,提出了一种改进的软阈值去噪方法,仿真结果表明与传统的软阈值去噪方法相比,该方法具有更好的去噪性能。更多还原

【Abstract】 Electrical energy is the broadest used energy in modern society. Its application levelis a main development level symbol of a country. With the development andpopularization of computer, power electronic and information technology, higher andhigher request is proposed about power quality. More and more attentions are paid totransient power quality disturbance by power unit and consumers. Analyzing andevaluating power quality is the premise of adopting appropriate measures to improvepower quality and reduce the influences brought by disturbances.This paper aims at transient power quality problems, has done some researches onpower quality disturbance detection, sources locating of disturbance, disturbanceclassification, disturbance data compression and signal de-noising. Someimprovements have been achieved.Because the calculation quantity and redundancy of continuous wavelet transformis big. Binary wavelet transform is adopted to detect power quality disturbance basedon wavelet transform local modulus maximum theory, which corresponding to thesingular point of signal. Simulation results indicate that relative exact detection can beacquired at decomposition scale one. A New criterion for capacitor switching disturbance sources is proposed. Theenergy of three-phase instantaneous power is calculated based on the disturbancepower and Parseval’s theorem. With the polarity of energy change during thedisturbance and the polarity of initial disturbance power peak, a new criterion isproposed, which can determine the relative location of the capacitor switching.Simulation results confirm the effectiveness of the method. A new power quality disturbance classification method is proposed based onwavelet packet decomposition, using entropy feature vectors to acquire power qualitydisturbances feature, combining with pattern recognition technology to classifydisturbances. Fisher piecewise linear classifier and Bayes classifier are usedrespectively. Simulation results show that the classification method, in which theentropy is used as feature vector, possesses higher classification correctnesscomparing with energy feature vector. An energy threshold method is proposed based on Multi-resolution Analysistheorem. Computing the energy need save to reduce wavelet coefficients, and thencombining with adaptive arithmetic encoding to compress disturbance data further.Typical power quality disturbances are used to test the proposed method. Simulationresults show that the compression effectiveness is improved comparing with waveletcoefficients threshold method.Based on wavelet transform soft-threshold de-noising method, an improvedsoft-threshold de-noising method to de-noise power quality disturbance signal isproposed. Considering the noise of power system is Gauss white noise commonly,which obey normal distribution. 3σ principle of normal distribution is used in theimproved method. Comparing with universal soft-threshold de-noising method,simulation results show that de-noising effectiveness is improved by this method.更多还原