题　目：Differentiable Robust Optimization for Electricity Dispatch: A Gradient-Based Framework for Managing Uncertainty

嘉　宾：Owen Shen, PhD Candidate, MIT Operations Research Center

主持人：何斯迈  教授 上海交通大学安泰经济与管理学院

时　间：2025年5月28日（周三）14:00-15:30

地　点：上海交通大学 徐汇校区安泰浩然楼308室

内容简介:

Rising uncertainty in power systems—driven by renewable energy integration and extreme weather—challenges the effectiveness of traditional optimal dispatch methods.   This work presents a differentiable robust optimization framework for electricity dispatch with energy storage, leveraging duality theory, optimization theory, and gradient-based methods.  The approach offers three key advantages: First, it accommodates time-varying uncertainty sets, allowing a more accurate representation of renewable generation volatility than fixed-set methods. Second, it supports fast re-optimization via warm-starting, enabling responsiveness to real-time forecast updates. Third, it leverages linear programming duality for efficient gradient computation, yielding economically meaningful shadow prices across uncertainty scenarios. A numerical case study illustrates the framework's scalability and adaptability, highlighting its potential to improve grid reliability and economic performance amid increasing uncertainty. Furthermore, we establish theoretical guarantees for our method and extend our analysis to provide a generalizable framework applicable to a broader class of multi-stage resource allocation problems under uncertainty.

演讲人简介：

Owen Shen is a third-year PhD candidate at MIT Operations Research Center. He holds an honors B.S. in Mathematics and B.A. in Philosophy from Stanford University. His research focuses on structure-aware integrative modeling and optimization that bridges classic parametric models with modern ML/AI frameworks. His work has been implemented across multiple industries, with sponsorships from leading companies in the hospitality and wholesale sectors.

欢迎广大师生参加!