基于均值聚类算法的电力现货市场报价行为分析

Analysis of Quotation Behavior in Power Spot Market Based on Mean Clustering Algorithm

随着电力市场建设的不断深入与完善,发电商在现货市场中的报价行为监管问题已经成为电力市场化改革中的关键问题。为此,提出一种基于均值聚类算法的现货市场报价行为分析模型。首先,系统性地介绍了现有的发电商报价行为分析指标,在此基础上提出机组簇报价曲线相似度及机组簇报价持留水平等指标。然后,使用新指标与均值聚类算法对现货市场报价行为进行分析,同时使用报价行为分析结果对均值聚类算法进行改进,克服了传统均值聚类法需要人为指定聚类簇心的缺陷,减少了算法在初值及参数设置上的随机性,提高了算法的效率。最后,以我国某地区电力现货市场的模拟运营数据为例,有效地识别出了其中具有市场力操纵行为的发电商,验证了模型的可行性和适用性。

With the continuous deepening and improvement of power market construction,the supervision of the quotation behavior of power generators in the spot market has become a key issue in the reform of power market. In this paper,an analysis model of quotation behavior in the spot market based on the mean clustering algorithm is proposed.First,the existing analysis indicators for the quotation behavior of power generators are systematically introduced,based on which the indicators such as generator set cluster quotation curve similarity and generator set cluster quotation withholding level are put forward. Then,the new indicators and the mean clustering algorithm are used to analyze the quotation behavior in the spot market. At the same time,the mean clustering algorithm is improved by using the analysis results of quotation behavior,which overcomes the defect that the traditional mean clustering method needs to manually specify the clustering center,reduces the randomness of the algorithm in the initial value and parameter settings,and improves the algorithm’s efficiency. Finally,the simulated operation data of the power spot market in one area of China is taken as an example,and results show that the generators with market power manipulation can be effectively identified,which verifies the feasibility and applicability of the analysis model.