基于改进S变换与支持向量机的电能质量扰动识别

发布时间：2018-03-29 08:10

本文选题：电能质量　切入点：扰动分类识别　出处：《华北电力大学》2017年硕士论文

【摘要】：随着科技的不断进步,大量非线性负荷接入电力系统中,造成了大量复杂的电能质量问题,因此电能质量扰动分类识别问题日益成为关注的焦点。识别各类单一及复合电能质量扰动是解决电力系统故障的首要前提,也是谐波源责任分摊、电能质量评估等后续研究的基础,有着重要的科研意义。首先,本文介绍了电能质量的基本指标,对电压暂降、谐波等常见扰动进行建模仿真,利用S变换的电能质量扰动检测分析方法,基于其多分辨率特性对信号进行时频分析,得到基频幅值和频率最大值曲线描述信号的幅值、起止时间以及谐波等时频特征。其次,本文对S变换窗函数进行改进,提出引入幅度和指数调节系数的改进S变换,针对不同扰动的时频特征,综合考虑噪声等影响因素,提出四个波形指标对时域波形进行描述,并利用主客观赋权法结合的粗糙集理论确定各波形指标权重,从而找到最优调节系数得出最优时频波形,为电能扰动信号的分类识别奠定基础。最后对最优波形进行特征量的精确提取,采用将全局与局部核函数结合的混合核函数支持向量机进行分类,通过改变混合核函数调节系数可以实现不同数据类型下分类精度的最大化,有效提高了分类器的泛化能力和分类准确率。基于Lab VIEW环境,编写电能质量扰动分类识别软件,能够直观的对电力系统扰动信号进行检测分析,然后对实际变电站各支路电压信号进行分析验证,结合工程实际情况,证明所提方法的可行性。
[Abstract]:With the development of science and technology, a large number of nonlinear loads are connected to the power system, resulting in a large number of complex power quality problems. Therefore, the problem of classification and identification of power quality disturbances has become the focus of attention day by day. Identifying all kinds of single and complex power quality disturbances is the first prerequisite to solve power system faults, and it is also the responsibility allocation of harmonic sources. Power quality evaluation and other follow-up studies have important scientific significance. Firstly, this paper introduces the basic indicators of power quality, modeling and simulation of common disturbances, such as voltage sag, harmonic, etc. Using S-transform power quality disturbance detection and analysis method, based on the multi-resolution characteristic of the signal, time-frequency analysis is carried out, and the amplitude of the fundamental frequency and the maximum frequency curve are obtained to describe the time-frequency characteristics of the signal, such as amplitude, starting and ending time and harmonic wave. Secondly, In this paper, the window function of S-transform is improved, and an improved S-transform with amplitude and exponential adjustment coefficient is proposed. Considering the time-frequency characteristics of different disturbances and considering the influence factors such as noise, four waveform indexes are proposed to describe the time-domain waveforms. The weight of each waveform index is determined by rough set theory combined with subjective and objective weighting method, and the optimal time-frequency waveform can be obtained by finding the optimal adjustment coefficient. It lays a foundation for the classification and recognition of power disturbance signals. Finally, the optimal waveform is extracted accurately, and the hybrid kernel support vector machine (SVM), which combines global and local kernel functions, is used to classify the signal. By changing the adjustment coefficient of mixed kernel function, the classification accuracy can be maximized under different data types, and the generalization ability and classification accuracy of classifier can be improved effectively. Based on Lab VIEW environment, the power quality disturbance classification recognition software is developed. It can detect and analyze the disturbance signal of power system intuitively, and then analyze and verify the voltage signal of each branch in actual substation. Combined with the actual situation of the project, the feasibility of the proposed method is proved.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM711

【参考文献】