Research on Power Grid Fault Diagnosis Based on a Quantitative Representation of Alarm Information

Power grid operation is trending toward digitalization and intelligence, and data-driven power grid fault diagnosis has become a realistic demand for smart grid dispatching. At present, power grid fault diagnosis is based on text content analysis and alarm information feature extraction. With a lack of models and methods for alarm information quantitative representation, current models cannot adapt to complex and changeable engineering field practices. Based on the quantity and time-series distribution characteristics of alarm information during faults, in this article, we propose a quantitative representation method for alarm information and use the quantified alarm information as classification features. The classification algorithm is used to obtain the classification results based on a power grid fault diagnosis framework. A fault diagnosis strategy is designed to integrate fault classification and faulty device discrimination model results, and fault types and faulty devices are determined. Finally, the simulation system-generated fault cases and actual fault cases collected by a power grid are used to verify the calculation example. The experiments and model comparisons show that the method, with the advantages of simple modeling, stronger practicability, and better diagnosis accuracy than previous methods, can handle different forms of alarm information and perform power grid fault diagnosis based on the data.