Author

Jian Shi

Advisor

Abdelwahed, Sherif

Committee Member

Follett, Randolph F.

Committee Member

Fu, Yong

Committee Member

Mazzola, Michael

Date of Degree

1-1-2014

Document Type

Dissertation - Open Access

Abstract

Medium-Voltage Direct-Current (MVDC) power system has been considered as the trending technology for future All-Electric Ships (AES) to produce, convert and distribute electrical power. With the wide employment of highrequency power electronics converters and motor drives in DC system, accurate and fast assessment of system dynamic behaviors , as well as the optimization of system transient performance have become serious concerns for system-level studies, high-level control designs and power management algorithm development. The proposed technique presents a coordinated and automated approach to determine the system adjustment strategy for naval power systems to improve the transient performance and prevent potential instability following a system contingency. In contrast with the conventional design schemes that heavily rely on the human operators and pre-specified rules/set points, we focus on the development of the capability to automatically and efficiently detect and react to system state changes following disturbances and or damages by incooperating different system components to formulate an overall system-level solution. To achieve this objective, we propose a generic model-based predictive management framework that can be applied to a variety of Shipboard Power System (SPS) applications to meet the stringent performance requirements under different operating conditions. The proposed technique is proven to effectively prevent the system from instability caused by known and unknown disturbances with little or none human intervention under a variety of operation conditions. The management framework proposed in this dissertation is designed based on the concept of Model Predictive Control (MPC) techniques. A numerical approximation of the actual system is used to predict future system behaviors based on the current states and the candidate control input sequences. Based on the predictions the optimal control solution is chosen and applied as the current control input. The effectiveness and efficiency of the proposed framework can be evaluated conveniently based on a series of performance criteria such as fitness, robustness and computational overhead. An automatic system modeling, analysis and synthesis software environment is also introduced in this dissertation to facilitate the rapid implementation of the proposed performance management framework according to various testing scenarios.

URI

https://hdl.handle.net/11668/18485

Share

COinS