Research topics

1. Learn to linearize hard constraints (Constraint learning)

• Train a neural network (NN) that maps decisions to constraint violations (e.g., bus voltage or branch flow violation)
• Equivalently reformulate the trained NN into mixed-integer linear constraints to replicate original hard constraints (e.g., power flow constraints)
• Can be interpreted as a piecewise linearization method
• [Link] for replicating deterministic constraints, [Link] for replicating probabilistic constraints, and [Link] for its piecewise linearization-based intepretation

2. End-to-end optimization proxy for power dispatch

• Design a data-driven formulation that can learn a decision rule from day-ahead observable information to cost-effective dispatch decisions for the future delivery interval [Link]
• Discuss the choice of loss functions for training end-to-end dispatch proxies [Link]

3. Power system operations under uncertainty

• A mixture model-based convexification for chance constraints with non-Gaussian uncertainty [Link]
• A fast distributionally robust chance-constrained method that can quantify the impact of the uncertainty offline [Link]
• A SVM-based approximation for chance constraints [Link]