Decentralized power systems : reference-frame theory and stability region generation

Abstract

Electricity provides the foundation for many of today's technological advances. The desire for energy security, a reduction in carbon dioxide emissions and a diversification of resources are all motivations for changes in how electricity is generated and transmitted. Recent alternatives to traditional centralized power-plants include technologies that are decentralized and intermittent, such as solar photovoltaic and wind power. This trend poses considerable challenges in the hardware making up these systems, the software that control and monitor power networks and their mathematical modelling. This thesis presents a set of contributions that address some of the aforementioned challenges. Firstly, we examine the fundamental theories used in modelling and controlling power systems. We expand previous work on reference-frame theory, by providing an alternative interpretation and derivation of the commonly used Park and Clarke transformations. We present a geometric interpretation that has applications in power quality. Secondly, we introduce a framework for producing regions of stability for power systems using conditional generative adversarial neural networks. This provides transmission and distribution operators with an accurate set of control options even as the system changes significantly.