Title: The Machine Learning Approach to Dynamic Security Assessment
Date: December 2, 2019
Bio: Jochen Cremer is a final year PhD student in the Control and Power Group (CAP) of the Department of
Electrical and Electronic Engineering at Imperial College London. Before joining CAP group in 2017
he undertook research in mathematical optimization and control theory at Carnegie Mellon and MIT.
He is an engineer at heart and holds a M.Sc. in Chemical Engineering, a B.Sc. in Electrical Engineering
and a B.Sc. in Mechanical Engineering from RWTH Aachen University, Germany. His research interest
lies in the intersection of machine learning and mathematical optimization applied to the operation of
the power system.
Abstract: The integration of large amounts of renewable energy makes flows of electrical power more
variable and less predictable. This presents a large computational challenge for system operators, who
must anticipate and mitigate insecure operating conditions using detailed time-domain simulations of
the physical system.
In this tutorial we present a machine learning approach to addressing this challenge that is increasingly
being used in the research community. After formulating the problem of dynamic security assessment
(DSA), we show how time-consuming simulations can partially be replaced by machine-learned proxies
or emulators. This approach separates the workflow into two parts: offline training and online
application. For the offline part, we cover the process of learning classifiers, and methods to generate
training data. For the online part, we describe novel approaches to control the risk due to imperfect
proxies, and a means of embedding the proxies in the calculation of optimal control actions.