Real-Time Non-Gaussianity Evaluation for Anomaly Detection in Noisy Power Grid Data Stream

In response to concerns regarding energy sustainability and increasing energy demand, a substantial number of renewable energy systems (RESs) are being integrated into the power grid. Unlike conventional power plants that can be adjusted to meet demand, RESs rely on weather conditions, introducing numerous unstable factors and abnormal events. Anomaly detection is a critical data analysis task, beneficial for pinpointing grid irregularities, failures, or intrusions. By examining anomaly patterns in integrated energy systems, a deeper comprehension of their operational status can be attained, leading to enhanced reliability and efficiency of the power grid. This paper suggests a statistical algorithm based on assessing nonGaussianity. In contrast to computationally expensive learning-based anomaly detection, the proposed algorithm can efficiently detect anomalies by evaluating kurtosis and skewness of the time series. It proves to be both time and resource-efficient, making it suitable for real-time and resource-constrained scenarios.