A transformer winding state detection method based on temperature field distribution

Abstract

This invention provides a method for detecting the state of transformer windings based on temperature field distribution, belonging to the field of transformer operation monitoring technology. The method includes: establishing a parameterized three-dimensional simulation model of the transformer; using the parameterized model to simulate the transformer temperature field and obtain the global temperature field distribution of the transformer windings; performing principal component analysis on the obtained global temperature field distribution to obtain a dimensionality-reduced global temperature field distribution; collecting the transformer winding temperature using finite-point fiber optic temperature sensors; reconstructing the transformer winding temperature field distribution based on a tensor encoder-decoder network; and comparing the reconstructed transformer winding temperature field distribution with a set threshold to determine the operating state of the transformer windings. By combining the data from finite-point fiber optic temperature sensors with the transformer winding temperature field distribution reconstructed by a deep learning network, accurate determination of the transformer winding state and inter-turn short-circuit faults can be achieved.

Images