Microgrid system health management

• Main Author: Yeo, Wee Tat
• Other Authors: Soong Boon Hee
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/150011
• Institution: Nanyang Technological University

This project focuses on the development of a supervised machine learning model for fault detection and prognosis in a Permanent Magnet Synchronous Generator (PMSG) system. With a prolonged use of electrical machines, wear and tear in some of the components will occur, which will affect the performance of the PMSG. Early detection of faults is essential as corrective maintenance can be performed before the faulty component damages the equipment further. Following the supervised machine learning workflow, a model was selected and developed for a PMSG system to detect various health types from input data. Healthy, winding fault and bearing fault simulations were experimentally introduced into the PMSG and the data was collected for analysis and processing for model training. Features were generated from the training data with the help of MATLAB Diagnostic Feature Designer and ranked with one-way Analysis of Variance (ANOVA) in order of importance. The selected features were then exported to the Classification Learner application to train and evaluate the supervised machine learning models. In the project, a cubic Support Vector Machine (SVM) model with an accuracy of 90% was developed and used on the test data for prediction. Model testing resulted in an accuracy of 90%, which was within the range of the accuracy obtained during model training. Selection of several parameters also enabled an interpretation of the scatter plot. Separately, analysis was performed on the phase currents and vibration to find insights from the data.