Abnormal detection and early warning mechanism of power market transaction data: a practice and exploration
Yuying Gong,Xiaolei Cheng,Yanhang Liu,Kai Xu, Wudabala
DOI: https://doi.org/10.1093/ijlct/ctae147
IF: 2.3
2024-01-01
International Journal of Low-Carbon Technologies
Abstract:Abstract With the rapid evolution of the electricity market and the significant surge in power transactions, anomalies in data could wield substantial influence and pose risks to market participants, regulatory bodies, and the overall market operation. This study addresses the voluminous nature of power trading data and the imperative for real-time processing by introducing an anomaly detection method based on the local outlier factor (LOF) algorithm to identify anomalous data points in the electricity market. Simultaneously, this research enhances the LOF algorithm by incorporating the closest distance to center (CDC) algorithm and leveraging the introduction of local reachability density based on k-nearest neighbors, termed as the CDCLOF-k model. The F1 score of the two-stage identification method generally surpasses that of the single-stage method, achieving optimal effectiveness even with smaller values of k, with minimal fluctuation as k varies. Simulation experiments demonstrate that the improved algorithm not only ensures the accuracy of outlier detection but also enhances its efficiency. Future endeavors may delve into more intricate trading anomaly scenarios to further enhance the performance and applicability of this algorithm.
energy & fuels,thermodynamics