A dual-sensor partial discharge localization method for transformer based on ultrasonic multipath propagation characteristic

Abstract Partial discharge (PD) ultrasonic detection and localization is an important technology for on-site operation and maintenance of transformers. Traditional ultrasonic localization techniques typically rely on the time difference of arrival (TDOA) method, which requires at least four ultrasonic sensors to simultaneously detect the ultrasonic signals generated by the discharge. However, due to factors such as the large size of transformers, rapid attenuation of ultrasonic signals, complex on-site electromagnetic and vibrational noise, and limited sensor deployment, it is often difficult to simultaneously capture effective PD signals with four sensors, particularly in online monitoring. To address this issue, this paper proposes a dual-sensor PD ultrasonic localization method for transformers based on the multipath propagation characteristic of ultrasonic waves. Firstly, based on finite element simulation, this paper analyzes the main propagation paths of the ultrasonic waves from the PD source to the sensors when the source is located inside or outside the windings, respectively. According to the simulation results, each sensor receives signals with distinct time of arrival (TOA) values corresponding to different propagation paths. By extracting the TOA values of the two earliest arriving paths for each sensor, two independent time-delay characteristic parameters can be obtained. Utilizing dual-sensor collaborative observation, the TOA information from four propagation paths could be captured. Subsequently, using the ray tracing algorithm, the simulated TDOA values between the four paths could be obtained. The multi-scale dynamic focused energy method is then employed to extract the measured TDOA values between the pulses from the four paths in the sensors recorded signals. Through the particle swarm optimization algorithm, the position of the particle swarm is iteratively adjusted based on the difference between the simulated and measured values, ultimately converging to the actual location of the PD source. To verify the accuracy of the proposed algorithm, experiments were conducted on both a scaled transformer and a converter transformer. The results demonstrate that the proposed method could reduce hardware deployment by 50% compared to traditional TDOA methods while maintaining a comparable level of localization accuracy.