Research and Application of Power System Data Anomaly Identification Based on Time Series and Deep Learning

Abstract

Abnormal data in the power system will reduce the accuracy of system state estimation and affect the safe operation of the power dispatch system. This paper proposes a data anomaly identification model based on time series and neural network, which establishes time series for various measuring points of the control master station, creates a time series group of associated measuring points based on the network topology, and extracts the sample characteristics of the time series. The neural network model is used to realize the intelligent identification of normal data and abnormal data. The neural network recognition results are compared with normal distribution and DBSCAN density clustering methods to verify the abnormal recognition performance of the neural network. Using a provincial power grid dispatch center operating data set as a training and testing sample, it verifies the advancement of the proposed method in the comprehensive performance of anomaly detection recall rate and precision rate and its feasibility in actual system application.