【作者】 胡为兵；

【导师】 李开成；

【作者基本信息】 华中科技大学 ， 电力电子与电力传动， 2008， 博士

【摘要】 近年来,随着电力系统容量不断扩大,各种分布式发电单元逐步接入电网,同时各种大功率的非线性负载,如各种整流设备,电弧炉和开关电源使用不断增加,造成公用电网电能质量日益恶化。而各种不同的电子设备对电压扰动敏感性有增无减,造成了整个电力系统的电能质量越来越受到关注。基于此提出了综合电能质量测试平台的构架,前端的电能质量监测仪可以实现电能质量稳态指标的计算和显示,并且具有扰动触发单元,一旦监测到电能质量扰动就可以触发记录单元,记录扰动期间的数据,并将其通过变电站通讯系统传输到电能质量分析中心,以便利用傅立叶变换、小波变换、模糊专家系统、谱估计等信号处理方法进行分析,得到更多关于电能质量扰动的信息。本文针对配电系统电能质量扰动监测中波形特征参数的计算、电能质量扰动自动分类以及投切电容的自动辨识和定位进行了全面而深入的研究,取得了如下成果:本文提出了一种基于滑动窗RMS值和小波变换的电压骤降起止时刻的检测方法。电压骤降是配电系统最普遍的电能质量扰动问题,而电压骤降的主要特征通过骤降的电压RMS值和骤降的时间间隔表征其特性。电压RMS值可以通过直接计算得到,而电压骤降起止时刻的精确检测由于在骤降结束区间波形变化平缓,很难实现。提出的方法充分考虑了电压骤降实际波形的特征,采用滑动窗RMS方法确定电压骤降起止时刻的范围,然后在小波分解尺度d1中采用正态分布三倍标准偏差阈值确定相关的奇异点,而电压骤降的起止时刻对应着小波变换的奇异点,因此通过在确定的范围内搜索小波变换奇异点可以精确地检测骤降的起止时刻。本文提出了基于小波变换和改进型模糊专家系统的电能质量扰动分类方法。电能质量扰动的检测、定位和分类是电能质量分析的重要方面,特别是扰动分类可以辨识电能质量扰动类型,便于统计配电系统的主要扰动,为电网改善提供实时数据。提出的方法首先利用傅立叶变换、小波变换以及RMS值等计算电能质量扰动信号的特征量,将计算的特征矢量输入到模糊专家系统,通过if-then实现输入到输出的转化,从而可以有效地实现振荡暂态、短时断电、骤降、骤升、电压波动和闪变、谐波、脉冲和陷波等电能质量扰动的自动分类,识别率达到99%。模糊专家系统基于统计研究和专家知识得到特征量的成员函数和规则基,采用模糊推理实现电能质量扰动分类。考虑噪声和谐波对已有模糊专家分类系统的影响,本文通过对隶属函数的改进和增加模糊推理规则基,改善了模糊专家系统在噪声和谐波两种扰动情况的分类识别率,达到了95.25%。针对电能质量振荡暂态特性难以提取的问题,提出了基于模极大值小波域和总体最小二乘法-旋转不变技术的信号参数估计(TLS-ESPRIT)算法,可以实现投切电容的振荡辨识和投切电容的方位确定。小波变换的模极大值对应信号的奇异点实现振荡起止时刻确定,从而得到振荡区间的持续时间。TLS-ESPRIT算法是一种基于子空间的高分辨率信号分析方法,直接以振荡区间的数据构成的数据矩阵为基础,把信号空间分解为信号子空间和噪声子空间,能够高精度地估计振荡信号各分量的频率、阻尼比、振幅和相位等信息。根据估计的主导振荡频率分量的电压和电流的相位角关系就可以确定投切电容相对于测量点的方位。更多还原

【Abstract】 Power quality is paid more attention due to the prolification of power system volume and the interaction of grid with all kinds of distributed generation in recent year. At the same time the power grid is deteriorated by massive utilization of high-power nonlinear load, such as rectifier equipment, arc furnace and switching power supply. However, electronics instruments and equipments desensitize against power quality disturbance increasingly.A framework about power quality monitoring platform based wavelet transform is proposed in this paper. The power quality testing instrument is setup in the site to measure the voltage and current waveform and calculate the concerned stable parameters and display the results, the trigger unit is designed in the instrument to record the disturbance data once the detected power quality disturbance. The recorded disturbance waveform data will transmit to the power quality analysis center to be processed furthermore by Fourier, wavelet transform, fuzzy expert sytem and spectrum esimation method, then with the calculated result to detect, location, classify the power quality problem and draw the characteristic parameter of different type power quality disturbance, even to performing equipment sensitivity study during power quality events and so on. The main work and achievements is following:Voltage sag as the most common power quality disturbances are characterized by several parameters, including but not limited to sag duration and depth. The root mean square (RMS) can calculate directly from waveform but an accurate determination of fault inception and clearing times is difficult due to the sag ends are generally smoother transitions than the fault inceptions. The proposed method use the windowed RMS calculation method determine the scope of the sag start and end times initially, then detect the singular point use the three times standard deviations of mean value as thresholds in the level of d1 of the wavelet transform, after all the singular points of wavelet transform are corresponding to power quality disturbance points. So the proposed method is reasonable and correct to detect the start and end times of sags accurately in the specified scope. A recommendation is made for the best overall wavelet choice. The procedure is fully described, and the results are compared with other method for determining sag duration, such as the specified RMS voltage and fractal method.An automated detection and classification method for transient power quality disturbances in which the wavelet transform is integrated with fuzzy expert system is proposed in the fourth chapter. The detection, location and classification of power quality disturbances are the important task for power quality research, especially the power quality disturbances recognition is useful for providing the real time data for improvement of distribution power system based on the statistics power quality disturbances character. The RMS calculation, Fourier transform and wavelet transform are utilized to obtain unique features for the disturbance signal. The extracted features vector are input the fuzzy expert system based on fuzzy theory by applying theory of artificial intelligent, database and fuzzy theory for making a decision regarding the types of the disturbances. The classified accuracy rate are 99% for various transients events such as voltage sags, swell, interuptions, switching transient, impulse, flicker, harmonics, and notches and so on. The proposed method modified the membership function and increased the fuzzy-rule base under the condition of noise and harmonics. The improved classification method can attain the classified accuracy rate 95.25% for the power quality disturbances signal polluted by noises and harmonics.Power system oscillatory transient is one of the most common power quality disturbance problems in transimmision and distribution grid. The modulus maximum wavelet domain and total least squares-estimation of signal parameter via rotational invariance techniques (TLS-ESPRIT) are applied to oscillatory transient waveforms to extract the relevant parameters and distinguish the direction of switching capacitor. The beginning and ending points of the oscillatory transients can be located by the modulus maximum wavelet domain, then TLS-ESPRIT which is one of the model-based spectrum estimation approach decomposed into signal subspace and noise subspace from data matrix that is directly make up of measured data within disturbance intervals by singular value decomposition (SVD), and truncate the eigenvectors so that the oscillatory transients frequency components, damping factors, amplitude and phase can be obtained by eigenvalue calculation under the condition of TLS. By evaluating the phase difference of voltage and current at the time and frequency of interest and at different spatial points, one can identify the direction of transient disturbance energy flow in order to pinpoint the location of the switched capacitor.更多还原