Transformer Partial Discharge Identification Method Based On Multidimensional Ultrasonic Characteristic Parameters

As one of the core equipment in power grid system and traction power supply system,its insulation reliability determines whether the whole system can operate safely and stably.Partial discharge is not only an important characterization of insulation deterioration,but also a key object of transformer operation condition monitoring.The location and type of partial discharge are tied closely to the electrical equipment accidents caused by insulation deterioration.Therefore,it is quiet essential to locate and identify partial discharge.Ultrasonic detection method can not only locate partial discharge,but also identify partial discharge types.Scholars have done lots of research on the location and identification of partial discharge,but there are still some problems not yet definitely settled as follows.1)most of the existing research uses different detection ways to locate and identify independently,which greatly increases cost in the maintenance process.2)The ultrasonic signal is easy to change the propagation path due to the influence of the transformer structure.However,the non-direct wave cannot accurately reflect the spatial position information of the partial discharge power supply,which will lead to excessive positioning error of the partial discharge power supply.3)For the partial discharge type recognition based on ultrasonic detection,the feature parameters extracted in the existing research often contain only one dimension of features,which are not comprehensive and the amount of coverage information is insufficient.4)The influence of the position of partial discharge source on the partial discharge recognition rate is not considered.Considering the above reasons,and experimental tests and method innovations are carried out in this paper.The specific research work are as follows.Aiming at the typical partial discharge defects of oil immersed transformer,four partial discharge models were made in this paper,and a multi-functional partial discharge experimental platform was build,which including partial discharge high voltage experimental platform,ultrasonic positioning / identification integrated detection system and pulse current method detection and comparison system.The denoising effects of different methods were compared,and the best denoising method was selected.The ultrasonic signal characteristics of four kinds of partial discharge were analyzed,and the ultrasonic characteristic parameters were extracted from multiple dimensions(distribution characteristic parameters,envelope characteristic parameters and image characteristic parameters).Then,the Pearson correlation coefficient method was used to reduce the dimension and optimize the parameters,and the multi-dimensional ultrasonic characteristic parameters were obtained.Using CS-SVM algorithm,a partial discharge recognition model based on multi-dimensional ultrasonic characteristic parameters was established,and the recognition effects of different algorithms before and after dimensionality reduction of ultrasonic characteristic parameters were compared.The direct wave and non-direct wave of ultrasonic signal were identified,the selfcorrection factor of non-direct wave was calculated according to the identification results,and a location method of partial discharge in transformer considering non direct wave error was proposed.According to the location of partial discharge source,a partial discharge identification method of transformer considering the location of partial discharge was proposed.