基于负荷量测数据的电力负荷聚类方法研究

发布时间：2018-10-30 10:00

【摘要】：电力负荷聚类分析是需求侧管理、负荷建模、电力系统规划等工作的重要基础,对电力系统的分析、运行、规划都具有重要意义。随着电力系统信息化程度的不断提高,配用电侧不断生产并记录着海量数据,同时电力公司的营配一体化实现了多种不同业务系统的集成,这为更有效的挖掘负荷数据奠定了基础。精确化的电力负荷聚类能提炼出负荷的共性特征,对于用户层的负荷聚类分析能提取其用电模式,帮助深刻把握用户用电规律;对于变电站层的负荷聚类分析则能在很大程度上反映系统的运行状态,能有效解决负荷的时变性及区域分散性问题。针对负荷聚类研究已经取得的成果和存在的不足,本文分别对用户负荷与变电站负荷进行聚类方法研究,主要工作如下:(1)总结了负荷聚类的基本步骤,介绍了现有电力负荷聚类研究涉及的经典算法以及常见的几种聚类有效性评价指标,并对不同算法的优缺点做了评价;(2)对于用户负荷聚类,通过充分的算法比较分析,对用户聚类算法的选取提供了参考意见,为提高聚类算法计算效率,研究分析了常见的降维方法的效果;(3)现有关于变电站负荷的聚类研究基本都以行业构成比例为分析数据,而变电站下层的用户行业属性通常是人为划分的,并不以其实际用电特性为划分标准,为此分析了传统基于行业属性考虑构成的不足,并提出基于用户层聚类的变电站负荷构成分析;(4)对于变电站负荷的聚类提出了综合考虑变电站总负荷特性与构成成分两方面的聚类模型,提出使用权重自动更新的k-means算法予以求解,在变电站聚类结果评价方面提出了综合评价指标。基于某电网的用户与变电站负荷等数据构造算例,结果证明了所提方法的优越性,并通过负荷数据加噪声的方式,验证了算法具有较好的鲁棒性。
[Abstract]:Power load clustering analysis is an important foundation of DSM, load modeling, power system planning and so on. It is of great significance to power system analysis, operation and planning. With the improvement of power system informatization, the distribution side is producing and recording massive data. At the same time, the integration of operation and distribution of power company realizes the integration of many different business systems. This lays the foundation for more effective mining load data. The accurate power load clustering can extract the common characteristics of the load, and the load clustering analysis of the user layer can extract the power consumption mode and help to grasp the law of the user's electricity consumption deeply. The load clustering analysis of substation layer can reflect the running state of the system to a great extent and can effectively solve the problems of time-varying load and regional dispersion. Aiming at the achievements and shortcomings of load clustering research, this paper studies the clustering methods of user load and substation load respectively. The main work is as follows: (1) the basic steps of load clustering are summarized. This paper introduces the classical algorithms involved in the research of power load clustering and some common evaluation indexes of clustering effectiveness, and evaluates the merits and demerits of different algorithms. (2) for the user load clustering, through the sufficient comparison and analysis of the algorithms, it provides the reference for the selection of the user clustering algorithm. In order to improve the computing efficiency of the clustering algorithm, the effects of the common dimensionality reduction methods are studied and analyzed. (3) the existing research on substation load clustering is based on the analysis data of industry composition ratio, but the industry attribute of the lower layer of substation is usually classified artificially, and not according to its actual power consumption characteristics. In this paper, the shortcomings of traditional industry attribute consideration are analyzed, and the substation load composition analysis based on user layer clustering is put forward. (4) for the clustering of substation load, a clustering model considering the characteristics and components of the total load of substation is proposed, and the k-means algorithm of automatic updating of the weight is used to solve the problem. The comprehensive evaluation index is put forward in the evaluation of substation clustering results. Based on the example of user and substation load data in a power network, the results show the superiority of the proposed method, and the robustness of the proposed method is verified by adding noise to load data.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM714

【相似文献】