Research On Characteristic Simulation And Diagnosis Method Of Stator Winding Inter-turn Short Circuit Fault In Synchronous Generator

Synchronous generators are one of the important mechanical equipment,occupying an increasingly important position in all industries and departments in modern society.Research has shown that stator winding inter-turn short circuit fault is the main type of fault in synchronous generators,and the harm to various aspects of the machine after the fault is significant.Therefore,the fault characteristics of synchronous generator stator windings before and after inter-turn short circuit are studied,and it is very necessary to detect and diagnose faults in synchronous machines.Most of the existing research on the simulation of stator inter-turn short circuit fault characteristics in synchronous generators focuses on analyzing the electromagnetic and temperature before and after the fault,and there are also differences in the force conditions of the winding before and after the fault,which may lead to further development of the fault.In terms of fault detection and diagnosis technology,some existing methods have certain limitations,such as complex detection processes,only indicating the presence of faults and changes in the approximate degree of faults,and low accuracy.Based on this current situation,this article studies the mechanical response characteristics of synchronous generator stator inter-turn short circuit before and after,and applies sweep frequency response analysis(SFRA)to the detection of this fault,achieving simple,fast,non-destructive,and economical detection.Finally,deep learning algorithms are used to achieve more accurate diagnosis of this fault.The main work of this article is as follows:(1)Simulation was conducted on the characteristics of inter-turn short circuit faults in the stator winding of synchronous generators.Using finite element simulation software,a "field-circuit" coupling model was built for a salient pole synchronous generator.The internal magnetic field of the machine during no-load operation,as well as the distribution patterns of current,voltage,loss,and winding electromagnetic force before and after stator inter-turn short circuit faults,were analyzed.The results were then imported into the simulation platform,and the temperature and mechanical response(overall deformation,equivalent strain,and equivalent stress)changes of the machine before and after the fault were obtained using the coupling methods of "electromagnetic-heat" and "electromagnetic-structure".It was analyzed that the temperature rise of the machine stator winding after the fault had an impact on the temperature of the nearby slot winding.The dangerous parts of the machine that are prone to insulation wear and damage are the nose end of the stator winding and other parts.This reveals the harm of inter-turn short circuits in synchronous generators and also demonstrates the necessity and importance of fault diagnosis in synchronous generators.(2)Detected the inter-turn short circuit fault of synchronous generator stator winding.Firstly,the variation of winding inductance parameters after a machine fault was simulated,and the reduction of inductance after the fault was obtained.Based on this characteristic,the fault detection method used in this paper,SFRA method,was introduced.The influence of rotor position on this method was analyzed.Then,this method was used to detect faults in a 7.5 k W synchronous generator,obtaining data at different levels and locations of faults,and drawing SFRA curves for analysis.It was found that after the fault,the winding curve showed varying degrees of deviation compared to the normal curve in the low,medium,and high frequency bands.In order to more intuitively obtain the degree of deviation of the curve as the degree of fault increases,this article also uses statistical indicators to analyze SFRA data.The experimental and analytical results have verified the feasibility of SFRA method in detecting machine winding faults,and the obtained data provides a foundation for fault diagnosis.(3)Using deep learning to process SFRA data for diagnosing inter-turn short circuit faults in synchronous generator stator windings.The sequence data of SFRA was input into the Multivariate Time Series Transformer(MTST)algorithm model,achieving accurate diagnosis.In addition,Res CNN,fully convolutional networks(FCN),long short-term memory(LSTM)and random forest algorithms are introduced,and the same fault data is input into the above four algorithms.The fault diagnosis accuracy obtained is compared with MTST,which verifies the performance of this algorithm in fault diagnosis.In addition,this article conducted SFRA experiments on stator inter-turn short circuit of another 5 k W synchronous generator in the laboratory.The obtained data was also inputted into several models in the article.Through comparison,it was found that the MTST still has the highest fault classification accuracy.It is finally proved that the MTST algorithm used in this paper has better performance in terms of computation,feature extraction of SFRA response,accuracy of fault classification and generalizability.