Design of Online Monitoring Method for Distribution IoT Devices Based on DBSCAN Optimization Algorithm

Abstract

In response to the data mutation problem caused by equipment failures in the distribution Internet of Things, this study proposes a density-based clustering optimization algorithm for online monitoring of equipment data anomalies. This method considers the local and global similarity of high-dimensional measurement data, and constructs a composite time series similarity measurement criterion. Improvements are made to the density-based clustering algorithm, which combines with preprocessed device historical measurement data to adaptively generate global density parameters. Through clustering training, core data points are obtained to detect abnormal changes in data. The experiment showed that compared to traditional density-based clustering algorithms, the improved algorithm had good clustering performance, with standardized mutual information and adjusted mutual information increased by about 2%. Compared to anomaly detection algorithms, the density-based clustering optimization algorithm for anomaly detection of equipment data in the distribution Internet of Things has increased the detection rate by 38% and reduced the false detection rate by 65%. Therefore, the proposed online monitoring method for data anomalies can improve the data detection rate and has high practical value for the reliable operation of distribution IoT systems.