Research On The Fault Diagnosing Strategy For Partial Discharge In Power Transformer

In the insulation system of electric equipment, electric field intensity of each part is often not equal, when the breakdown field intensity local area reached the region medium, the region will appear discharge, but such a discharge does not penetrate the applied voltage between two conductors, namely the whole insulation system and no breakdown, still keep the insulation performance, this phenomenon is called partial discharge. The emergence of Transformer PD on-line monitoring technology, which makes real-time understanding of partial discharge in transformer condition known as possible, while the fault diagnosis algorithm as the fault of transformer partial discharge discrimination and prediction method, which can provide effective reference for the maintenance of the transformer operation.Firstly, the elementary knowledge of partial discharge, the method of online monitoring of uhf partial discharge, and the digitalized anti-interference method of high frequency signals are introduced in partial discharge. In addition, the graph analysis method is applied to introduce the five fault types and the corresponding characteristics of electrical signals.Afterwards, the basic principle of genetic algorithm is explained, with fault types and electrical signal amount of partial discharge being analyzed according to the genetic algorithm. The oil wedge discharge, creeping discharge, suspended discharge, fixed metal particles group of insulator surface discharge, and discharge spikes for are defined as the fault set, and their corresponding phase, amplitude, discharge times sign as symptom set. The intensity of causes and effects are counted, and a model of genetic algorithm in the diagnosis of partial discharge is then established with examples illustrating the feasibility of the method. After that, genetic algorithm is adopted to optimize the neural network method. A new mathematical model is established, being compared with the LM neural network to test and verify the feasibility of the GA neural network diagnosis of partial discharge.Finally, a partial discharge fault diagnosis expert system is established based on MATLAB, to realize the function of transformer partial discharge current monitoring data query, and historical data query, etc. The result of partial discharge fault graph analysis can also be observed, with the calculation result of the two fault diagnosis methods being directly queried. In addition, the monitoring data and the diagnostic results are generated in a report form, to provide original data for further modification.