Event Date:
Monday, February 10, 2025 - 3:30pm to 4:30pm
Event Location:
- Sobel seminar room (South Hall 5607F)
Short talks by 4 CFMAR PhD students on their research. In alphabetical order:
Speaker: Thiha Aung
Title: Optimal Dispatch of Wind-Battery Storage Systems: From stochastic control to experiments on a synthetic grid
Abstract:
This talk presents a model for a co-located wind-battery asset for firming wind generation to dispatch target level over the operation horizon. We present a comprehensive study of the battery storage control for different firming objectives and wind generation dynamics. The optimal control of the battery storage is derived as the viscosity solution of a second-order Hamilton-Jacobi-Bellman (HJB) equation. We provide an analytical solution under mild assumptions on the dynamics of the storage. We also provide numerical results via a novel adaptation of Regression Monte Carlo (RMC), Stochastic Hybrid Asset Dispatch Optimization with Gaussian Process (SHADOw-GP) algorithm.
Speaker: Cosmin Borsa
Title: Deep Learning for Optimal Stopping Problems
Abstract:
Optimal stopping problems have been algorithmically solved by approximating continuation values on a time grid using a sequence of regression operators. When solving these problems on dense time grids, the approximation errors made by regression emulators backpropagate and eventually become so large that they impair classical algorithms. We construct an aggregate deep neural network on top of a coarse sequence of regression operators and improve it in a reinforcement learning framework to perform stopping decisions in continuous time. This technique expands the field of tractable optimal stopping problems by enabling us to stop arbitrarily close to the optimal stopping boundary; thus, obtaining a higher reward.
Speaker: Hezhong Zhang
Title: A Mean Field Analysis of Climate Change Uncertainty
Abstract:
Climate-economic models are usually highly nonlinear and complex, in addition to the challenges posed by model uncertainty and heterogeneity among the population. These features make it difficult to design effective policies that balance economic growth, climate mitigation, and social equity. In this talk, we present a quantitative computational analysis of socially optimal climate policy, and the expected discounted values of the social payoffs that determine these optimal policies, in the face of regional heterogeneity and model uncertainty.
We design a deep reinforcement learning algorithm to solve the proposed climate-economic model, which we formulate as a mean-field control problem. This algorithm is used to evaluate key model mechanisms and quantify the uncertainty channels that drive our social valuations. We observe numerical convergence and demonstrate that, in our current setting, uncertainty aversion shifts optimal abatement and R&D investment policies to mitigate climate change damage inequities. This is joint work with Michael Barnett (ASU), Lars Peter Hansen (UChicago), and Ruimeng Hu (UCSB).
Speaker: Haosheng Zhou
Title: A Stochastic Control Model for Strategic Misdirection via Sequential Hypothesis Testing
Abstract:
Concealing intentions and sending misleading signals to confuse opponents is an interesting yet challenging problem in multi-player games due to its inherently ill-posed nature. In this talk, we introduce a well-interpretable linear-quadratic stochastic control framework that models such strategic interactions between two opposing teams, red and blue. A key novelty of our approach is the incorporation of sequential hypothesis testing to model the inference of intentions.
Our work consists of two main parts: firstly, we derive optimal misdirection strategies for the blue team, who aims to achieve a primary objective while minimizing the extent to which its intentions are revealed to the red team. In addition, we extend the model to a Stackelberg game, where the red team, aware of the blue team's optimal behavior, strategically leaks information to deceive the blue team into inadvertently exposing its true intentions. We demonstrate through numerical experiments that our model effectively captures the essence of strategic misdirection. This is joint work with Daniel Ralston, Xu Yang and Ruimeng Hu.
January 18, 2025 - 6:20pm