Decision-Focused Learning via Neural Optimization Layers for Battery-Assisted Grid-Import Smoothing in AI Data Centers
Organizations: Department of Data Science and AI, Faculty of Information Technology, Monash University, Australia
Abstract
Rapid AI data-center power fluctuations challenge grid integration. Sustained grid-import smoothing requires allocating limited battery power and energy under uncertain demand. We formulate risk-aware BESS scheduling that coordinates advance reserve commitments and demand-responsive dispatch, accounting for operating costs, grid-ramp penalties, and their conditional value-at-risk. We develop battery scheduling decision-focused learning (BS-DFL) with a constraint-aware neural optimization layer trained directly from this objective without optimal-solution labels. The trained layer is then frozen to guide demand-predictor learning through scheduling costs and ramping risk, enabling solver-free rolling-horizon scheduling. In a 10,000-minute simulation using loads derived from measured AI workloads, BS-DFL reduces mean absolute minute-to-minute grid-import changes by 71.70% and total scheduling cost by 18.01% relative to an MSE-trained predictor coupled with a mixed-integer linear programming solver, with an average decision time of 18.77 ms.
Figures & tables
| Metric | No BESS | Perf-Info. | MSE+Gurobi | MSE+BS-NOL | BS-DFL |
|---|---|---|---|---|---|
| Mean | 34.7274 | 0.8237 | 3.5293 | 1.1626 | 0.9988 |
| P95 | 86.6182 | 0.4066 | 23.4497 | 0.0971 | 0.0104 |
| Maximum | 199.949 | 986.790 | 413.191 | 179.400 | 203.920 |
| Total cost | 439,557 | 100,832 | 128,382 | 106,500 | 105,257 |
| Reserve cost | 0 | 215 | 717 | 2,485 | 2,898 |
| Forecast MAE | — | — | 179.97 | 179.97 | 240.66 |