基于分阶段离群点检测的电力市场异常辨识

为了监管电力市场中存在的各类违规行为,保证市场的公平竞争,提出了一种基于分阶段离群点检测算法的电力市场异常行为辨识方法.梳理不同交易阶段异常行为的特征,提取相应的特征指标,采用主成分分析法对其进行降维,分阶段地进行异常行为的检测.同时利用平均距离改进局部离群因子算法,显著地提升了算法的检测效果.通过某地区电力市场提供的交易数据进行实验分析,实验结果表明,该方法能有效识别市场中的异常行为,为市场监管人员利用海量数据进行有效监管提供了新思路.

Identification of abnormal behavior in power market based on phased outlier detection

To ensure the effective supervision on illegal behaviors and fair market competition in power market, a method for identifying abnormal behaviors in the power market based on phased outlier detection algorithm is proposed. The characteristics of abnormal behaviors in stages were firstly reviewed and their index were extracted. Characteristics dimensions were then reduced by the method of principal component analysis, and abnormal behaviors were detected by stage. At the same time, the average distance was applied to local outlier factor algorithm optimization and its detection effect improvement. Experimental analysis was performed in the transaction data from a power market, the experimental results indicate that market abnormal behaviors would be effectively identified by this method. It would contribute to new ideas for market supervisors to achieve effective supervision by massive data.