Mining Method of Power Grid Monitoring Alarm Association Rules Based on Improved FP-growth Algorithm

Abstract

The core of current grid monitoring eventization technology is the discrimination method based on the alarm-event correlation rule base. However, with the continuous expansion of the grid scale, the traditional artificially refined rules can no longer meet the needs of monitoring eventization. Based on an improved FP-growth algorithm, this paper proposes a mining method for alarm association rules in power grid monitoring. First, the DBSCAN clustering model is used to replace the traditional sliding time window model to divide the original alarm siganls. Then, based on the FP-growth algorithm, a signal filter is added to eliminate the known signal pairs with strong correlation in the alarm cluster, and the signal is weighted according to the importance of the alarm signal. Next, according to the improved FP-growth algorithm, the association rules of the divided alarm clusters are mined. Finally, through the analysis and comparison of data from the power grid, the effectiveness of this method is verified, and the efficiency of rule mining is further improved, which has a certain reference value for the maintenance and improvement of the alarm-event correlation rule base.