GIS Status Early Warning Method Based on GRA-ELMAN-Fusion Model

Condition assessment of gas-insulated combined electrical appliances is an important decision-making approach to improve the operational reliability of power systems. The current GIS state parameter prediction models are often based on a single or a small number of state parameters for analysis and judgment, and the prediction stability and scientificity need to be improved. This paper combines GIS internal status information with external power grid operation and environmental meteorological data to propose a combined model for GIS status parameter prediction that considers deep mining of complex correlations. First, gray relational analysis (GRA) is performed on the collected data to screen out state variables that are highly correlated with parameters such as temperature; then, the state parameters of GIS are predicted through the ELMAN neural network, and the absolute residual error between the actual value and the predicted value is calculated. value, and use the probability distribution fitting method to set the alarm threshold, based on which early GIS faults are warned, and fault samples are extracted; finally, the extracted GIS fault samples are accurately diagnosed through the Fusion fusion algorithm framework. Taking the GIS data of a certain substation as an example, the results show that the GRA-ELMAN -Fusion diagnostic model proposed in this article can identify early faults of the GIS in advance and accurately diagnose the fault type.