| Transformer is one of the most critical electrical equipment in the power grid,and its fault will cause serious losses,among which the insulation fault accounts for the largest proportion,so it is necessary to evaluate the insulation state of transformer.Partial discharge(PD)is an important factor of insulation deterioration in transformers,and its detection is of great significance for detecting latent insulation faults in transformers.At present,the ultrasonic detection method has been widely used in the field,but there are still some problems,such as poor detection sensitivity and low positioning accuracy.In this paper,the characteristics of ultrasonic signal and the ultrasonic localization method of PD source are deeply studied.Firstly,aiming at the complex propagation process of transformer partial discharge ultrasound,the acoustic finite element simulation is carried out by COMSOL,and the acoustic time domain model and frequency domain model of PD are established by the pressure acoustic physical field.The time-space distribution of ultrasonic pressure of fixed point sound source are studied.Sound sources and observation points at different locations are set up and the propagation path of ultrasonic wave in transformer is studied.According to the simulation results,the propagation characteristics of ultrasonic are summarized and the scheme of optimizing the installation position of sensor is given.Secondly,aiming at the problem of noise interference in the ultrasonic signal received by the sensor,a threshold denoising method based on optimized variational mode decomposition(VMD)and noise estimation is proposed.Because VMD algorithm may cause under-decomposition or over-decomposition of noisy signals,the optimal mode decomposition number K is determined by spectral analysis and quartile method,and the narrow-band interference and high frequency white noise are removed by using kurtosis;The standard deviation of noise is estimated by using the two statistical characteristics of the modes obtained from the decomposition of white noise by VMD,and the interval threshold function is introduced to further denoise the residual low-frequency white noise of the PD-dominant mode.Finally,aiming at the location of PD source in transformer,a localization method based on time difference screening and artificial bee colony(ABC)secondary optimization is proposed.The PD signal is detected by the sensor placement method in this paper,and the ultrasonic signal is pre-processed by the denoising method mentioned above.Then the time difference of received signals is estimated by the generalized cross-correlation method.On this basis,the time-difference screening link based on spherical equations is added to eliminate the time-difference with large errors.The constrained optimization model of spatial positioning is set up according to the effective time-difference and solved by ABC algorithm.The K-Means clustering analysis is carried out for some positioning results.The ABC algorithm is used to optimize the minimum point of the sum of distance clustering centers,and the location result with large error is corrected to approximate the actual PD source step by step.The sensor placement scheme proposed in this paper can minimize the received reflected wave and receive direct wave as soon as possible,which is convenient for subsequent denoising and time difference estimation.The proposed denoising method can not only suppress noise more effectively,but also retain the characteristics of PD signal better,which is conducive to reducing the error of time difference estimation.The proposed localization method of partial discharge synthetically considers the propagation characteristics of ultrasonic wave and noise interference in transformer,and comprehensively improves the time difference estimation and spatial localization algorithm,which effectively improves the localization accuracy.Field experiments show that the localization error is less than 5 cm,which has certain engineering application value. |
PDF Full Text Request