Cyber-physical modeling, analysis, and optimization - a shipboard smartgrid reconfiguration case study

Abstract

Many physical and engineered systems (e.g., smart grid, transportation and biomedical systems) are increasingly being monitored and controlled over a communication network. These systems where sensing, communication, computation and real time control are closely integrated are referred to as cyber physical systems (CPS). Cyber physical systems present a plethora of challenges related to their design, analysis, optimization and control. In this dissertation, we present some fundamental methodologies to analyze the optimization of physical systems over a communication network. Specifically, we consider a medium voltage DC shipboard smart grid (SSG) reconfiguration problem as a test case to demonstrate our approach. The main goal of SSG reconfiguration is to change the topology of the physical power system by switching circuit breakers, switches, and other devices in the system in order to route power effectively to loads especially in the event of faults/failures. A majority of the prior work has focused on centralized approaches to optimize the switch configuration to maximize specific objectives. These methods are prohibitively complex and not suited for agile reconfiguration in mission critical situations. Decentralized solutions proposed do reduce complexity and implementation time at the cost of optimality. Unfortunately, none of the prior efforts in this arena address the cyber physical aspects of an SSG. This dissertation aims to bridge this gap by proposing a suite of methods to analyze both centralized and decentralized SSG reconfigurations that incorporate the effect of the underlying cyber infrastructure. The SSG reconfiguration problem is a mixed integer non convex optimization problem for which branch and bound based solutions have been proposed earlier. Here, optimal reconfiguration strategies prioritize the power delivered to vital loads over semi-vital and non vital loads. In this work, we propose a convex approximation to the original non convex problem that significantly reduces complexity of the SSG reconfiguration. Tradeoff between power delivered and number of switching operations after reconfiguration is discussed at steady state. Second, the distribution of end-to-end delay associated with fault diagnosis and reconfiguration in SSG is investigated from a cyber-physical system perspective. Specifically, a cross-layer total (end-to-end) delay analysis framework is introduced for SSG reconfiguration. The proposed framework stochastically models the heterogeneity of actions of various sub-systems viz., the reconfiguration of power systems, generation of fault information by sensor nodes associated to the power system, processing actions at control center to resolve fault locations and reconfiguration, and information flow through communication network to:(1) analyze the distribution of total delay in SSG reconfiguration after the occurrence of faults; and (2) propose design options for real-time reconfiguration solutions for shipboard CPS, that meet total delay requirements. Finally, the dissertation focuses on the quality of SSG reconfiguration solution with incomplete knowledge of the overall system state, and communication costs that may affect the quality (optimality) of the resulting reconfiguration. A dual decomposition based decentralized optimization in which the shipboard system is decomposed into multiple separable subsystems with agents is proposed. Specifically, agents monitoring each subsystem solve a local concave dual function of the original objective while neighboring agents share information over a communication network to obtain a global solution. The convergence of the proposed approach under varying network delays and quantization noise is analyzed and comparisons with centralized approaches are presented. Results demonstrate the effectiveness as well as tradeoffs involved in centralized and decentralized SSG reconfiguration approaches.