Transformer explosion risk dynamic bayesian assessment method and system based on multi-modal feature fusion

Abstract

The application provides a transformer explosion risk dynamic Bayesian evaluation method and system based on multi-modal feature fusion, relates to the technical field of transformer explosion risk evaluation, and comprises the following steps: collecting the temperature, humidity and air pressure parameters of the environment where the transformer of an outdoor transformer substation is located, fusing the collected parameters with regional meteorological data of a satellite cloud map to obtain fusion data; acquiring temperature field distribution data through a distributed fiber grating sensor array on a transformer winding; analyzing the spatio-temporal correlation of the two through a spatio-temporal graph neural network, and then identifying abnormal temperature rise data; and finally performing multi-modal fusion on the quantitative values of the environmental influence parameters, the heat resistance parameters of the transformer winding and the abnormal temperature rise data through a Bayesian network to obtain a dynamic evaluation result of the transformer explosion risk. The application realizes dynamic and accurate evaluation of the transformer explosion risk by fusing multi-source data and using a spatio-temporal graph neural network and a Bayesian network.

Images