Completed
Climate change warrants rapid action
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Optimization-in-the-Loop AI for Energy and Climate - IPAM at UCLA
Automatically move to the next video in the Classroom when playback concludes
- 1 Climate change warrants rapid action
- 2 Climate & energy problems involve physics, hard constraints, and decision-making
- 3 Machine learning methods struggle with physics, hard constraints, and decision-making
- 4 Optimization-in-the-loop ML
- 5 Talk outline
- 6 Overview: Differentiable optimization
- 7 Background: Deep learning
- 8 Differentiating through optimization problems
- 9 Follow-on work in differentiable optimization
- 10 Overview: Enforcing hard control constraints
- 11 Deep reinforcement learning vs. robust control
- 12 Differentiable projection onto stabilizing actions
- 13 Details: Finding a set of stabilizing actions Insight: Find a set of actions that are guaranteed to satisfy relevant Lyapunov stability criteria at a given state, even under worst-case conditions
- 14 Illustrative results: Synthetic NLDI system
- 15 Energy-efficient heating and cooling
- 16 Differentiable projection onto feasible actions
- 17 Results on realistic-scale building simulator
- 18 Summary: Enforcing hard control constraints
- 19 Overview: Incorporating downstream decision-making
- 20 Decision-cognizant demand forecasting
- 21 Decision-cognizant approach can dramatically improve generation scheduling outcomes
- 22 Approximating AC optimal power flow
- 23 Approximate robust power system optimization
- 24 Summary: Incorporating downstream decision-making
