M.A. Thesis:

Analyzing the behaviour of power transformer using audio signal processing and neural network

Due to the construction of power plants in remote locations, limited energy resources, energy transfer with the desired characteristics, changes the voltage level of the power transformers is done. Power transformers are critical components are considered in the transmission and distribution and power systems account for a large share of investment. Analysis of the performance and increase the useful life of the transformer, plays an important role. Since the transformer fault diagnosis and repair process is difficult and time consuming to detect and localize errors in accuracy and speed, two parameters are considered crucial in repair. Most of the methods for troubleshooting and fault detection, fault location are not able to or do not have sufficient sensitivity and accuracy Several factors have an impact on their results. Digital audio signal processing method for rapid detection and accurate location of these features exception. In this project, analyzing the behavior of power transformers transmission in neutral and pregnancy sound with digital signal processing is performed by software MATLAB. In this respect, MLP neural network to diagnose transformer behavior Backpropation architecture is used. To increase accuracy and reduce errors in detection transformer behavior, propagation neural network architecture with different algorithms is tested And the best algorithm to achieve the objective, taking into account the best performance is selected. The k-means algorithm classification in both neutral and pregnancy is detected. Digital signal processing procedure of voice, high-speed processing, timely detection of defects, including target location error is to find it.

Keywords: Power transformers, neural network MLP, digital signal processing audio, Violet