Synchronous generator fault diagnosis based on neural network
Date of Issue2018
School of Electrical and Electronic Engineering
The fault diagnosis of synchronous generators has been a popular research topic due to its wide usage in industry, agriculture, transportation and so on. Failure of generator not only damages the generator itself, but also causes the production line to collapse and resulting in huge economic losses. Therefore, accurately detection and diagnosing faults in generator during operation is of great importance in industrial production. With the help of artificial intelligent methods, the way to detect the faults becomes much smarter and more efficient. This dissertation proposes a Backpropagation Neural Network approach to diagnose and classify the generator fault type and severity. The required data for training and testing the Neural Network is experimentally obtained from a laboratory three-phase brushless synchronous generator under different interturn short-circuit faults. Sequential Forward Selection (SFS) and Principal Components Analysis (PCA) methods are introduced to improve the performance. It is found that the prediction accuracy based on PCA is effectively improved with less computational complexity. Future work will focus on identifying more effective features from other physical signals (magnetic flux and vibration). The method will also be extended to diagnose other common generator faults, such as bearing and air gap eccentricity faults.
DRNTU::Engineering::Electrical and electronic engineering