Improving Alarm Data Management in Power Communication Networks: A Hybrid Approach Using DBSCAN Clustering and Sliding Window Techniques

Abstract

The original power communication network alarm data is characterized by issues such as discreteness, redundancy, and temporal asynchrony. To address these challenges, this paper introduces a method for processing alarm data that integrates DBSCAN clustering with a sliding window technique. Initially, the DBSCAN clustering algorithm is leveraged to effectively mitigate the issue of discreteness. Subsequently, the sliding window method is employed to tackle the problems of alarm redundancy and temporal asynchrony. Simulations indicate that the proposed method significantly reduces the number of alarms, achieving an average reduction of up to 47.68%.