Loading...
Enhanced DC Microgrid Protection: a Neural Network and Wavelet Transform Approach
El Gohary, Youssef Hesham
El Gohary, Youssef Hesham
Date
2024-05
Author
Advisor
Type
Thesis
Degree
Citations
Altmetric:
Files
35.232-2024.73a Youssef Hesham El Gohary.pdf
Adobe PDF, 1.66 MB
- Embargoed until 2026-04-22
Description
A Master of Science thesis in Electrical Engineering by Youssef Hesham El Gohary entitled, “Enhanced DC Microgrid Protection: a Neural Network and Wavelet Transform Approach”, submitted in May 2024. Thesis advisor is Dr. Ahmed Osman. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).
Abstract
This thesis introduces an advanced protection scheme for DC microgrids, focusing on enhancing fault detection, classification, and localization while ensuring real-time operation. Leveraging the wavelet transform algorithm and neural networks' pattern recognition capabilities, the proposed system integrates modern techniques for achieving its objectives. The protection coordination scheme encompasses two settings: the primary coordination scheme, activated when the ANN accurately identifies the fault's location, and the backup coordination scheme, activated in the event of inaccuracies or errors in the neural-based algorithm. In this scenario, an optimization model is deployed to ensure that protective devices operate with predefined operation times and parameter settings, aiming to minimize the total operation time of all relays, including primary and backup. This ensures fault isolation regardless of the neural-based algorithm's status, with the optimization problem modeled as an NLP programming problem and solved using the optimization software GAMS. The optimization model acts as a duplicate protection, enhancing the protection system's reliability by providing an additional layer of defense. Furthermore, an innovative inductor injection mechanism is introduced to enhance the protection scheme's effectiveness. By injecting an inductor into the system after fault detection, the rate of fault current rise is significantly reduced, allowing for an expanded SFV (spatial feature vector) size without compromising fault detection accuracy. The inductor injection mechanism enables the SFV to encompass additional time slots, facilitating more comprehensive data input to the neural network for improved fault classification and localization. Additionally, the inductor injection mechanism is carefully selected to balance current damping with fault detection requirements, ensuring optimal system performance under various fault conditions. Simulations using MATLAB Simulink validate the proposed protection scheme's effectiveness, demonstrating high accuracy and reliability with real-time operation and robust error handling mechanisms. This research advances protection systems in DC microgrids, offering improved fault detection, classification, coordination, and localization capabilities.