Supervised Neural Surrogates for DC-OPF Economic Dispatch in Grid-Aware Data Center Scheduling

This project develops supervised neural surrogate models to learn the DC-OPF economic dispatch operator for 24-hour flexible data center scheduling. Trained on CVXPY-GUROBI solutions of the IEEE-14 system, the models achieve 450–700× faster inference with <0.2% cost deviation, enabling real-time grid-constrained dispatch emulation.