AC Microgrid protection based on machine learning and multi-agent systems

Uzair, Muhammad

AC Microgrid protection based on machine learning and multi-agent systems

Uzair, Muhammad

Permalink

Publication Type:: Thesis
Issue Date:: 2023

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

Adobe PDF

Download thesisAdobe PDF (2.82 MB)

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Uzair, Muhammad
dc.date.accessioned	2024-10-04T04:38:24Z
dc.date.available	2024-10-04T04:38:24Z
dc.date.issued	2023
dc.identifier.uri	http://hdl.handle.net/10453/181189
dc.description	University of Technology Sydney. Faculty of Engineering and Information Technology.	en_US.UTF-8
dc.description.abstract	The increasing demand for power, aging distribution systems, and concerns over greenhouse gas emissions have significantly increased distributed generation (DG) integration within distribution networks. This integration challenges conventional protection methods due to the bidirectional power flow and the constraints of anti-islanding protection. The microgrid concept offers a promising solution but also presents challenges, including the significant variation in fault currents between grid-connected and autonomous modes and the arbitrary output impedance of inverter-interfaced DG units during faults and current limiting modes. These issues complicate the development of a robust protection scheme. Therefore, an intelligent and adaptive protection scheme is required to protect microgrids against various faults and operational conditions. Artificial intelligence, especially supervised machine learning (ML), holds significant potential for solving microgrid protection challenges. However, the limited availability of datasets and the need for innovative feature extraction techniques have impeded progress. To address these issues, this research develops different radial and meshed AC microgrid models for collecting fault and no-fault data. Three comprehensive datasets are prepared to train various supervised ML and deep learning (DL) algorithms. The largest dataset consists of 16,000 fault cases and 432 no-fault cases. Additionally, innovative feature extraction techniques, such as Peaks Metric and Max Factor, are formulated and applied alongside investigating eight other methods to extract features that are not commonly used for microgrid fault detection and classification. Various feature ranking techniques are employed to reduce the number of predictors. A novel hybrid DL-based fault detection and classification protection method is developed and validated using unseen data to ensure robust predictive performance. Moreover, a multi-agent system (MAS) framework is established to integrate data-driven ML models within a process-driven MAS structure, enhancing coordination and adaptive protection. Simulation results show that the proposed scheme, combining ML and MAS, outperforms previous methods, achieving high fault detection and classification accuracy and exceptional protection sensitivity for both microgrid operational modes across various fault scenarios. This research develops comprehensive models of radial and meshed microgrids, formulates new feature extraction techniques, and evaluates the performance of the intelligent protection scheme under varying conditions. The result is a robust protection scheme that improves system resilience and economic benefits by providing precise fault detection, classification, and phase identification, which are essential for future intelligent grids.	en_US.UTF-8
dc.format	Thesis (PhD)
dc.language.iso	en_US	en_US.UTF-8
dc.relation	https://opus.lib.uts.edu.au/bitstream/10453/181189/1/thesis.pdf
dc.rights	info:eu-repo/semantics/openAccess
dc.rights	The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.
dc.rights	© 2023 Muhammad Uzair
dc.rights	au.edu.uts.lib/cph
dc.title	AC Microgrid protection based on machine learning and multi-agent systems	en_US.UTF-8
dc.type	Thesis
utslib.copyright.status	open_access	*

Abstract:

The increasing demand for power, aging distribution systems, and concerns over greenhouse gas emissions have significantly increased distributed generation (DG) integration within distribution networks. This integration challenges conventional protection methods due to the bidirectional power flow and the constraints of anti-islanding protection. The microgrid concept offers a promising solution but also presents challenges, including the significant variation in fault currents between grid-connected and autonomous modes and the arbitrary output impedance of inverter-interfaced DG units during faults and current limiting modes. These issues complicate the development of a robust protection scheme. Therefore, an intelligent and adaptive protection scheme is required to protect microgrids against various faults and operational conditions. Artificial intelligence, especially supervised machine learning (ML), holds significant potential for solving microgrid protection challenges. However, the limited availability of datasets and the need for innovative feature extraction techniques have impeded progress. To address these issues, this research develops different radial and meshed AC microgrid models for collecting fault and no-fault data. Three comprehensive datasets are prepared to train various supervised ML and deep learning (DL) algorithms. The largest dataset consists of 16,000 fault cases and 432 no-fault cases. Additionally, innovative feature extraction techniques, such as Peaks Metric and Max Factor, are formulated and applied alongside investigating eight other methods to extract features that are not commonly used for microgrid fault detection and classification. Various feature ranking techniques are employed to reduce the number of predictors. A novel hybrid DL-based fault detection and classification protection method is developed and validated using unseen data to ensure robust predictive performance. Moreover, a multi-agent system (MAS) framework is established to integrate data-driven ML models within a process-driven MAS structure, enhancing coordination and adaptive protection. Simulation results show that the proposed scheme, combining ML and MAS, outperforms previous methods, achieving high fault detection and classification accuracy and exceptional protection sensitivity for both microgrid operational modes across various fault scenarios. This research develops comprehensive models of radial and meshed microgrids, formulates new feature extraction techniques, and evaluates the performance of the intelligent protection scheme under varying conditions. The result is a robust protection scheme that improves system resilience and economic benefits by providing precise fault detection, classification, and phase identification, which are essential for future intelligent grids.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/181189