A Transformer Fault Diagnosis Method Based on Particle Swarm Optimization Optimization Random Forest Model

Abstract

The invention discloses a transformer fault diagnosis method based on a particle swarm algorithm to optimize a random forest model. First, the non-coded ratio of dissolved gas analysis data in transformer oil is input as a feature vector, and a training set and a test set are divided; then a random forest model is constructed. , and optimize the random forest model through the particle swarm optimization algorithm to obtain two optimal parameters; finally, the optimal parameters obtained re-establish the random forest model to identify the fault type of the transformer. This method effectively improves the accuracy of transformer fault diagnosis and provides a reliable basis for operation and maintenance personnel to correctly judge the operating status of the transformer.

Description

technical field [0001] The invention relates to the technical field of power equipment monitoring, in particular to a transformer fault diagnosis method based on a particle swarm algorithm optimization random forest model. Background technique [0002] At present, the power system has developed into a cross-regional interconnected large power grid. As the energy conversion hub of the network, the transformer will seriously affect the stable operation of the power grid if it fails. The analysis of dissolved gas in oil can identify latent faults inside and outside the transformer and their development trend, which is a feasible method for diagnosing transformer faults recognized by the power industry. Therefore, the DGA data of the dissolved gas concentration in the transformer oil is the most intuitive and effective characteristic parameter of the transformer, which can provide a basis for diagnosing the state of the transformer. [0003] Based on the characteristics of diss...

AI Technical Summary

Problems solved by technology

[0003] Based on the characteristics of dissolved gas in transformer oil, so far, researchers have proposed many fault diagnosis methods, mainly including two: one is the traditional diagnosis method, such as IEC three-ratio method, Rogers four-ratio method, uncoded ratio method etc. These ratio discrimination methods are simple to operate, but they often show defects such as imperfect coding and absolute fault boundary distinction; the second is a machine learning model that uses the concentration ratio of dissolved gas in oil or the proportion of components as feature quantities to mine, and is commonly used There are artificial neural network (BPNN), support vector machine (SVM), etc. These machine learning models have effectively improved the accuracy of fault diagnosis and achieved certain results, but there are also certain defects.

For example, BPNN takes a long time to train, it is easy to fall into the local optimum, and it is difficult to obtain the global optimal solution; SVM is not sensitive to the selection of kernel functions, and it needs to combine multiple binary classifiers to solve the multi-classification situation, and it is difficult to obtain more accurate classification results.

Images