Multidimensional Matrix Profile-Based Anomaly Detection for Power Grid with High Variable Renewable Energy Penetration

The development of various emerging artificial intelligence and big data technologies has gradually transformed traditional physical power systems into smart grids. With the high proportion of renewable energy connected to the grid in various countries becoming an irreversible trend, the increasingly frequent climate change has made data quality issues in transmission networks increasingly prominent. In order to ensure the safe operation of the power system, it is necessary to accurately and quickly detect and process abnormal data. This article comprehensively considers key issues in the field of multidimensional temporal anomaly detection and designs an anomaly detection method based on multidimensional Matrix Profile (MP) for real-time detection of multi-source heterogeneous data associated with transmission networks. It can extract key anomaly feature subsets from high-dimensional data and optimize window length selection in traditional MP methods. This article has been validated in open-source datasets and simulated synthetic power data. The experimental results show that the method based on multidimensional MP can effectively avoid the multi-parameter and data dependency problems of deep learning models, and has high scalability and flexibility.