Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Electrical Engineering

First Advisor

Safwan Wshah


Scientific advancements based on the wide-area measurements as a way to monitor systems, are fundamental in reliable operation of different types of complex networks. These advanced measurement units capable of real-time wide-area monitoring, which enables capture system dynamic behavior. Therefore, advanced technology is urgently necessary to analyze substantial streaming data from these networks and handle system uncertainties. As an example, uncertainties in power systems due to renewable energy and demand response. Power system operation, and planning have become more complex and vulnerable to extreme weather and natural disasters. Thus, increasing power system resilience has gained more attention.Machine Learning (ML), and Deep Learning (DL) have seen tremendous adoption in power and communication fields applications over the past few decades. Still under development, nowadays the Deep Learning recorded remarkable success using increasingly complex models. Recently, DL-based applications are gaining popularity, which have the following advantages over conventional mathematical methods: (1) better adaption to system uncertainties; (2) better robustness against different system configurations as a result of data-driven nature; (3) less dependent on the modeling accuracy and validity of assumptions. However, due to the unique physics of electrical applications i.e. power grid, many problems cannot be directly solved using current DL algorithms. Consequently, significant efforts are required to improve the adaptabilities of DL approaches for such applications. Consequently, this dissertation aims to fully investigate the effectiveness and capabilities of DL in broad range of electrical and communication systems problems and provide a different viewpoint on DL. This work demonstrates that by improving the existing DL models and reformulating electrical system problems, we can not only expand the DL applications but also significantly improve the performance in comparison with other conventional methods, and traditional machine learning methods.



Number of Pages

157 p.