Double normalization recognition method for directly forecasting and recognizing transformer winding fault type

Abstract

The invention discloses a double normalization recognition method for directly forecasting and recognizing a transformer winding fault type, and relates to a novel method for directly forecasting and recognizing the transformer winding fault type based on a hidden markov model and fuzzy recognition. The method comprises the following steps of: step 1, establishing a databank about various mechanical types of transformer windings; step 2, establishing a fuzzy recognition bank, and providing membership degrees and recognition accuracy of each type of fault; step 3, establishing a hidden markov model recognition bank, providing a similar probability and recognition accuracy of each fault recognition; step 4, obtaining a weighting parameter through a hidden markov model recognition probability value; step 5, determining a weighting parameter of the hidden markov model and fuzzy recognition; step 6, obtaining a double normalization recognition index for recognizing the transformer winding fault; and step 7, directly recognizing the state and type of the transformer winding fault through the calculated dg recognition index. The method cannot be influenced by anthropic factors, is strong in environment interference resistance and has no errors caused by obtained conversion.

Description

technical field [0001] The present invention relates to a power transformer, and more specifically, relates to a new method for directly identifying and predicting fault types of transformer windings. Background technique [0002] Accurate identification of transformer winding state is of great value in engineering. For long-term power transformers, reliability issues are crucial. At the same time, large-scale power transformers are also one of the most important and expensive electrical equipment in power systems. Therefore, transformers Fault diagnosis has extraordinary significance. Traditional detection methods mainly include electrical measurement methods such as short-circuit impedance method and frequency response analysis method. The electrical measurement method can accurately judge the obvious deformation of the winding, but for faults such as the decrease of the axial preload force of the winding and slight deformation, the electrical measurement method will lose...

AI Technical Summary

Problems solved by technology

But this method has the following disadvantages: first, when the test data is changed by external interference, the method proposed in this patent cannot accurately give the state of the transformer winding; second, the measured experimental data is double frequency Acceleration, while the vibration intensity uses energy to characterize the state of the winding, which requires intermediate conversion, and the conversion process is bound to produce errors; third, the experimental data measured by the vibration intensity method must be used in the same time period. Identification brings certain difficulties; fourth, this method cannot identify the type of transformer winding fault; fifth, the use is affected by the surrounding environment

Images