电力系统电能质量的扰动检测与识别方法研究

发布时间：2018-01-08 19:12

本文关键词：电力系统电能质量的扰动检测与识别方法研究　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：电能作为清洁环保,经济高效,易于控制和转换的能源,广泛的应用到生产和生活的各个领域。电能质量的优劣不仅影响电能用户利益,同样会影响电网的安全运行,所以对电能质量问题的检测具有重要的意义。针对目前主要的电能质量问题进行分类,并具体分析了各类型的电能质量问题的发生原因及其危害。根据常见的暂态和稳态电能质量扰动问题的特点,建立扰动数学模型,并分析了目前电能质量扰动的检测方法和分类方法的研究现状。本文主要工作内容如下:1)采用小波包变换对电能质量扰动信号进行分析,根据扰动信号的特点研究小波包变换的采样频率、小波基函数和分解层数等参数的选取,分析得到各扰动类型的小波包节点的归一化能量分布,提出对能量分布进行处理得到具有明显变化的新小波包能量分布的方法。在新的能量分布中可以看出,含谐波的扰动和暂态振荡扰动在对应节点上的能量分布较大,可以提取对应节点的新小波包节点能量作为表征这些扰动的特征向量,并通过仿真对利用小波包变换提取扰动特征向量的可行性进行研究。2)首先,采用HHT变换对各种类型的电能质量扰动信号进行分析,利用分析结果中的瞬时频率对不同扰动类型发生的起止时刻进行估计,并根据瞬时幅值和边际谱提取扰动信号的幅值特征和频率特征;由于HHT变换后特征提取效果不明显,为寻求更有效的提取特征值的方法,又采用S变换对各种类型的电能质量扰动信号进行分析,利用分析结果中的最高频率幅值变化对扰动发生的起止时刻进行估计,并利用基频幅值变化和时间幅值平方和均值变化提取扰动信号的幅值特征和频率特征,具体分析了S变换提取扰动特征向量的过程。之后,通过对比两种方法的扰动时刻估计和特征提取的效果可以看出,利用S变换的扰动起止时刻定位较准确,对应特征变化明显,阈值选取方便,方法更易实现。3)针对单一特征向量不能有效的表征所有的电能质量扰动信号的差异性,分析了不同检测方法特征提取的特点,对电能质量扰动的多特征组合逻辑进行了研究。根据小波包能量分布中谐波信号的明显差异,以及S变换中的频率特征对频率特征明显的含谐波和暂态振荡进行分类,并利用S变换得到的其它幅值特征和奇异性特征进行后续的分类。研究了利用多个概率神经网络构造多特征组合的分类器实现扰动的分类的方法。仿真结果表明,在噪声强度为40dB时平均分类准确度可达99.25%,噪声强度为20dB时平均分类准确度也可达88.38%。
[Abstract]:Electric energy as clean and environmental protection, high efficiency, easy control and conversion of energy, widely used in various fields of production and life. The power quality of the electric energy quality not only affect the interests of users, will also affect the safe operation of power grid, so the detection of power quality problems has important significance. In view of the current classification the main power quality problems, and analyzes the causes and hazards of various types of power quality problems. According to the common transient and steady state characteristics of power quality disturbance, a disturbance model, and analyzes the current research status of power quality disturbance detection and classification. The main work of this paper the contents are as follows: 1) by using the wavelet packet transform of power quality disturbance signal is analyzed according to the disturbance signal characteristics of wavelet packet transform sampling frequency, wavelet function and decomposition level Select the number of parameters, analysis of normalized energy distribution of each disturbance type of wavelet packet, the energy distribution could be obtained by processing the energy distribution method has obvious changes of the new wavelet packet. The energy distribution in the novel can be seen in the disturbance of harmonic and transient oscillations in the corresponding node energy distribution large, node energy can be extracted from the corresponding node of the new wavelet packet as the characterization of these perturbations of the feature vector, and the simulation of using wavelet packet transform to extract the feature vector of disturbance feasibility study.2) firstly, mining quality disturbance signals by analyzing various types of power with HHT transform, the instantaneous frequency by using the analysis results of time to estimate the different disturbance types, and according to the instantaneous amplitude and amplitude of marginal spectrum extraction characteristics and frequency characteristics of disturbance signal due to the HHT transform; After feature extraction, the effect is not obvious, for feature extraction of more effective value, and using S transform to various types of power quality disturbance signal analysis, the highest frequency and amplitude changes with the result of analysis of the disturbance occurred in the starting and ending time estimation, and using the fundamental frequency amplitude change and amplitude and time square the mean changes in amplitude and frequency characteristics of feature extraction of disturbance signal, and detailed analysis of the S transform to extract the feature vectors. After the process disturbance, the disturbance moment estimation and feature extraction of two methods to contrast effect can be seen, using S transform to disturbance time accurate positioning, the corresponding characteristic changes obviously, the threshold is convenient the method is easier to realize.3) for effective characterization of single feature vector can't be all the difference power quality disturbance signal, analyzes the characteristics of different detection methods of extraction. Point of power quality disturbance logic combination of multiple features is studied. According to the difference of wavelet packet energy distribution in the harmonic signal, and S transform in frequency characteristics of frequency characteristic with harmonic and transient oscillation amplitude classification, other features and singularity for subsequent classification and use of S transform. Method of classification using a probabilistic neural network to construct multi feature combination classifier can achieve disturbance. The simulation results show that the noise intensity is 40dB the average classification accuracy of 99.25%, noise intensity is 20dB the average classification accuracy is up to 88.38%.

【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM711;TM930

【参考文献】