When
Where
Speakers: Ayrton Almada-Jimenez, Ben Stilin, Sheila Whitman, Jackson Zariski
Title: Summer Projects and Summer Internships
Speaker: Ayrton Almada-Jimenez, Applied Math 4th Year Student
Title: Real-Time Stochastic Assessment of Dynamic N-1 Grid Contingencies
Abstract: System operators increasingly require tools that enable rapid, real-time, counterfactual assessments of grid security in the face of fast-evolving operating conditions. While N-1 contingency analysis is routinely performed, it typically omits dynamic evaluations --particularly frequency swings triggered by common disturbances such as single-phase faults. This paper addresses this gap by introducing a methodology for a real-time dashboard tool to assist power system operators in screening dynamic contingencies. Our framework assumes: (a) the grid is initially in a balanced operating state; (b) a fault can occur randomly on any transmission line, causing it to be temporarily de-energized and then reconnected within one second (or a shorter system-dependent interval); (c) a contingency is flagged if, during the post-fault transient (either before or after fault clearance), the power flow on any line exceeds a predefined safety threshold. The main technical contributions of this work are: (1) Overload Indicator – a novel system-wide metric that quantifies integrated N-1 dynamic risk from any given balanced initial state; (2) Scalable Fault Evaluation Algorithm – an efficient computational scheme for assessing the dynamic consequences of faults without resorting to brute-force simulations, scaling linearly with system size; (3) MCMC-based Risk Estimation– a Markov Chain Monte Carlo approach that estimates the probability of high overload indicator values, paired with visualizations of representative extreme events. We demonstrate the approach on an open-source model of the Israeli transmission grid.
Speaker: Ben Stilin, Applied Math 5th Year Student
Title: The Lightning Algorithm as a Tool for Approximating Invariant Densities
Abstract: The invariant densities of a family of intermittently chaotic interval maps contain a power law singularity. As a result, it is difficult to use common techniques such as collocation or the Galerkin method to numerical compute the density as typical basis function struggle to approximate these singularities. In this presentation I will discuss how the recently developed lightning method provides an elegant solution to this problem.
Speaker: Sheila Whitman, Applied Math 5th Year Student
Title: Machine Learning–Driven Permanent Magnet Discovery
Abstract: How can machine learning speed up the search for materials that power our electric vehicles, wind turbines, and electronics? This short talk explores how data-driven models are accelerating magnet discovery by quickly screening thousands of candidate structures. This talk focuses on my work from my summer internship experience at Oak Ridge National Lab.
Speaker: Jackson Zariski, Applied Math 5th Year Student
Title: Learning to Create Railway Disposition Concepts
Abstract: While a set, strict timetable for train operations procedures on a daily basis is a perfect idea scenario, delays and cancellations inject a deal of uncertainty to the mix. Multi-integer programming allows us to simultaneously solve for both the corrected starting times of events as well as introduce and select binary variables relating to conflict resolution. First, I will describe our initial work with the Dutch intercity network, a simpler representation of the dispatch problem and a good test set for the flow-variable and "shortest" path processes. Next, I'll expand to the Deutsche Bahn Berlin network, examining an alternative graph model for dynamic conflict resolution. Finally, I will discuss our work with applying reinforcement learning to this process.