Intelligent Robust Control Strategies and Estimation Techniques for Smart Power Girds

点击翻译按钮获取中文摘要

The size and complexity of the smart grid poses significant computational challenges on system control through dynamic state estimation (DSE). In addition, smart grid comes with many cyber security concerns. As far back as 2002, 70% of energy companies experienced some kind of cyber-attack. Cyber-attacks are intelligently designed to bypass existing fault identification approaches. In case of a successful attack, a domino effect may follow which will result in irreparable damage in different parts of the system.******A long-term research goal of this program is to develop new robust and intelligent controllers that adapt to different operating modes and faults, modeling uncertainties, and attacks. Such a system can successfully predict the occurrence of cyber-attack and take appropriate control actions in advance. In short-term, this research program aims to address a number of important questions: How to take advantages of high performance computing to avoid tradeoff between accuracy and speed? How to leverage online information to identify/predict the cyber-attack in highly complex systems? How to achieve optimal control solutions without sacrificing stability?******The novelty of the proposed work lies in the unique approaches in combining massive parallelization, robust control theory and intelligent estimation strategies. My research group seek to advance the use of machine learning, probability-based estimation techniques, and massively parallel programming on control, security, and estimation problems in complex large-scale systems. New massively parallel estimation strategies utilizing Byesian and composite likelihood estimators will be derived to overcome the limitations of traditional DSE such as numerical instability and slow response rate. Cyber-physical models that enable identification and prediction of a coordinated cyber-attack by pattern and time series analysis (e.g. Granger causality) will be developed. An adaptive algorithm based on feedback control and state-of-the-art machine learning techniques that leverages multi-view fuzzy consensus clustering will be developed to assess different weighted views to a variety of attacks, and estimate source of the attack. ******The expected outcome of this work is to develop new techniques and strategies that yield improved robustness to uncertainties and cyber intrusions, increased accuracy of the estimation, and reduced cycle time in faults clearance. This offers a number of benefits to society (safety) as well as the economy (reduced maintenance costs and fewer catastrophic events). Aside from their application in smart grid, these methods can be applied to a number of different engineering systems, including automotive, defense, and aerospace. In addition to technical advancements, my research group trains at least 7 HQP who help meet Canada's demand for engineers and will contribute to its economic success.**