Power factor anomaly detection using data stream summaries

Anomaly detection in power energy is crucial for energy efficiency. This aspect of electricity management assures for smooth functioning of vital sectors such as health, education and communications, particularly in developing countries. However, it is very difficult to detect anomalies directly in data streams generated by the power grid's equipment. This study focuses on the analysis of anomalies related to electrical power factor by exploiting data stream summaries. These summaries were created on the basis of real data from electrical measurements. An experiment was conducted using the Isolation Forest (IF) machine learning algorithm on both initial data and data stream summaries. The results show that the summaries reach conclusions similar to those provided by the original data streams. This study demonstrates the ability of data stream summaries to serve as a basis for anomaly detection, while ensuring adequate storage and processing resources.