基于马氏距离和密度聚类的电力现货报价模式分析

为体现电力的时间价值,针对供需变化的价格弹性和遵循的经济学原理,电力现货市场通常采取多段报价的申报形式,即每段均包括段容量和价格.现货市场中机组报价行为复杂,且除部分极端高价或低价外,潜在的串谋行为或离群度高的异常报价则难以区分.为有效识别现货市场中的异常报价,有效监测和分析市场行为并及时发现和控制市场风险,文中采用马氏距离消除价格和容量量纲不同的影响,以衡量不同机组报价的相似性.提出衡量报价相似性的3维现货报价对比向量,运用密度聚类分析报价模式以规避对离散点的过度分类,并设计了针对串谋行为特征中价格相似和交易结果呈现串谋的预期效益的判断指标.最后,在5段报价的背景下,文中以模拟现货市场申报数据为算例对比了马氏距离和欧氏距离在聚类分析中的特点.所提方法能消除量纲的影响,有效发现机组在价格申报、容量申报、量价申报3个维度的相似性.

Analysis of Quotation Mode for Electricity Spot Market Based on Mahalanobis Distance and Density Clustering

In order to reflect the time value of electricity, aiming at the price elasticity for changes in supply and demand and the followed economic principle, the electricity spot market usually adopts the declaration form of multi-segment quotation,i.e., each segment includes the segment capacity and the price. In the spot market, the unit quotation behavior is complicated, except for some extremely high or low prices, it is difficult to distinguish potential conspiracy or abnormal quotes with highly deviated outliers. In order to effectively identify abnormal quotations in the spot market, effectively monitor and analyze market behaviors, and timely discover and control market risks, the Mahalanobis distance is used to eliminate the impact of different prices and capacity dimensions to measure the similarity of different units quotation. The three-dimensional contrast vector of spot quotations is proposed to measure the similarity of quotations. The quotation mode of the density clustering analysis is used to avoid the excessive classification of discrete points, and the judgment indicators are designed for two conspiracy characteristics, the price similarity and the transaction results showing the expected performance of the collusion. Finally, In the context of five segment quotes, in this paper, a simulation example of declaration data in the spot market is used to compare the characteristics of Mahalanobis distance and Euclidean distance in the clustering analysis. The proposed method can eliminate the impact of the dimension, and effectively discover the similarities of the unit in the three dimensions of price declaration, capacity declaration and volume price declaration.