Positive Active Power Outlier Detection based on One-Class SVM

The smart grid creates a large-scale intelligent energy delivery network. It requires a much more complex and highdimensional data processing system. Thus, smart grid has become one of the most potential fields for machine learning application. This paper adopts an anomaly detection algorithm based on one-class SVM to realize anomaly detection task based on large-scale electrical energy data. One-class SVM model uses training data to train a hypersphere and calculate the distance bebveen data points and center. The algorithm is able to process a large number of data quickly and has high reliability. In this paper, large-scale positive active power data is used as training and testing datasets. As the experiment shows, the algorithm has high detection efficiency and high accuracy.