Machine Learning Aided Efficient Tools for Risk Evaluation and Operational Planning of Multiple Contingencies

In power system reliability assessment, the system security limits and adequacy indices depend on the set of contingencies analyzed. Consequently the final solution strategies for short term operational and long term investment planning studies also depend on the set of contingencies considered for planning. Generally, planning is done for the most critical contingency, with the assumption that the solution strategy for the most constraining contingency will also perform well on the contingencies that have lower severity. But this is not always true. In reality, under highly stressed and uncertain nature of power system conditions, the operational rules for the most constraining contingency may not be effective for all other contingencies. In fact some contingencies, which are generally less severe, may have more pronounced ill-effect during certain other operating conditions. Therefore, it is important to perform a comprehensive contingency analysis of many contingencies under several operating conditions (a computationally burdensome task), screen the most important ones among them that may violate the probabilistic reliability criteria, and devise effective solution strategies. Thus, the focus of this chapter is to devise a computationally efficient operational planning strategy against voltage stability phenomena for many critical contingencies. The chapter accomplishes this with the help of a hybrid approach that combines the strength of modelbased analytical indicators and data driven techniques to design two important aspects of planning for multiple contingencies, namely: risk based contingency ranking and contingency grouping. Utilizing realistic probability distributions of operating conditions together with machine learning techniques makes the risk assessment process of multiple contingencies credible and computationally tractable. In order to group the contingencies efficiently for devising a common solution strategy, the chapter introduces a novel graphical index, termed as progressive entropy that captures the degree of overlap among post-contingency performances of various contingencies. The objective of the proposed contingency grouping method is to strike a balance between producing simple and accurate operational V. Krishnan (&) Department of Electrical and Computer Engineering, Iowa State University, Ames, IA 50014, USA e-mail: [email protected] © Springer International Publishing Switzerland 2015 A.T. Azar and S. Vaidyanathan (eds.), Chaos Modeling and Control Systems Design, Studies in Computational Intelligence 581, DOI 10.1007/978-3-319-13132-0_12 291 guidelines for multiple contingencies, while reducing the operational complexity in terms of the total number of guidelines that operators handle.