电能计量数据聚类分析与窃电检测研究

发布时间：2018-01-11 18:26

本文关键词：电能计量数据聚类分析与窃电检测研究　出处：《昆明理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着电网的智能化发展,各种智能电量计量装置的广泛使用,电能计量自动化系统中积累了大量的电能计量数据,采用数据挖掘的方法对这些数据进行分析,可以帮助电力部门掌握用户的用电行为,对负荷预测、负荷控制、用电异常检测等具有重要的意义。窃电作为影响电网发展的主要问题之一,不仅给国家经济造成了重大的损失,而且威胁着电网的安全运行。随着科技的发展,窃电手段越来越多样化,而且隐蔽性更强,这给窃电检测带来了更大的困难。同时对大量的用户进行窃电检测存在着效率问题。针对上述问题,本文首先采用了聚类的方法完成了对用户用电负荷曲线的分类,得到了用户的负荷特征曲线并根据负荷特征曲线对用户用电行为进行了分析,然后采用将用户的负荷曲线与其负荷特征曲线进行匹配筛选的方法,以此对疑似窃电用户进行确定,缩小了窃电检测的用户范围,最后通过基于粒子群(Particle Swarm Optimization,PSO)优化的支持向量机回归(Support Vector Regression,SVR)算法对疑似窃电用户的窃电行为进行进一步检测。本文的主要研究内容如下:(1)基于聚类分析与支持向量机回归的相关技术,提出基于负荷、电流、电压、电量等电能计量数据的窃电检测模型,并采用了云南省某供电局的电能计量数据对模型进行了实验验证。(2)详细分析了用户负荷曲线分类过程,对分类过程中的数据抽取、数据预处理、单一用户典型日负荷曲线的提取和聚类数确定进行了具体研究,利用模糊C均值(FuzzyC-means,FCM)聚类算法得到用户的日负荷特征曲线,并对用户用电行为进行了分析。这为疑似窃电用户的判别提供了依据。(3)针对传统的支持向量机回归算法在准确性方面的不足,本文采用了粒子群算法对SVR算法进行了优化,将优化后的SVR算法应用到窃电检测模型中,建立了疑似窃电用户窃电行为的检测模型。实验表明,相比于传统的SVR算法,经过优化后的SVR算法的准确性较高,为窃电检测提供了一种新方法。
[Abstract]:With the intelligent development of power grid and the wide use of various intelligent metering devices, a large number of electric energy measurement data have been accumulated in the electric energy measurement automation system, and these data are analyzed by the method of data mining. It can help the power department to master the user's behavior of electricity consumption, and it has important significance to load forecasting, load control, abnormal detection of electricity consumption, etc. As one of the main problems affecting the development of power grid, electricity theft is one of the main problems. It not only causes great loss to the national economy, but also threatens the safe operation of power grid. With the development of science and technology, the means of stealing electricity are more and more diversified and more concealment. This brings more difficulties to the detection of electricity theft. At the same time, there are efficiency problems in detecting a large number of users. In this paper, we first use the clustering method to complete the classification of the user load curve, get the user load characteristic curve and analyze the user power consumption behavior according to the load characteristic curve. Then the method of matching the load curve and the load characteristic curve of the user is adopted to determine the suspected power theft user and to reduce the range of the users of the electricity theft detection. Finally, Particle Swarm Optimization based on particle swarm optimization was used. PSO-optimized support Vector Regression. SVR (SVR) algorithm is used to further detect the power theft behavior of suspected users. The main research contents of this paper are as follows: 1) based on clustering analysis and support vector machine regression technology, a load-based approach is proposed. Current, voltage, electricity and other electrical energy measurement data detection model. Using the electric energy metering data of a power supply bureau in Yunnan Province, the model is verified by experiments. (2) the classification process of user load curve is analyzed in detail, and the data extraction and data preprocessing in the process of classification are analyzed in detail. The typical daily load curve of a single user is extracted and the clustering number is determined. The fuzzy C-means clustering algorithm is used to obtain the daily load characteristic curve of the user. It provides a basis for the discrimination of the suspected users of electricity theft. 3) aiming at the lack of accuracy of the traditional support vector machine regression algorithm. In this paper, particle swarm optimization (PSO) algorithm is used to optimize the SVR algorithm, and the optimized SVR algorithm is applied to the detection model of electricity theft. Compared with the traditional SVR algorithm, the accuracy of the optimized SVR algorithm is higher, which provides a new method for the detection of electricity theft.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13;TM933.4

【参考文献】