Robust Scheduling for Uncertain Infrastructure Systems

94 concepts82 papers1 proposal

Traditional optimization assumes we can predict the future. In a deterministic model, a grid operator might look at a weather forecast, predict exactly 500MW of solar power at noon, and schedule other generators to fill the remaining gap. However, as Xue et al. and Lesieutre observe, the modern surge in renewable energy and electric vehicles (EVs) has made these single-point forecasts dangerously unreliable. When the sun goes behind a cloud or a fleet of EVs arrives early, a deterministic schedule can lead to grid failure or massive economic penalties.

To solve this, researchers have moved toward uncertainty-aware scheduling. Instead of optimizing for a single predicted value, these frameworks optimize for a range of possibilities, ensuring that systems remain stable even when the future deviates from the forecast.

From Points to Uncertainty Sets

The core intuition of robust optimization is to replace a single forecast number with an uncertainty set—a mathematical "envelope" that contains all likely scenarios.

Zhang et al. demonstrate that by using polyhedral uncertainty sets, energy systems can maintain reliability during extreme volatility without the economic penalties of being overly conservative. Rather than preparing for every possible disaster, the system prepares for the "worst-case" scenario within a defined boundary. This approach is not limited to power; Miao et al. apply it to taxi dispatching, treating passenger demand as an uncertainty set to reduce the mismatch between supply and demand by 45% during unpredictable urban events.

However, preparing for the worst-case across all variables simultaneously is often too expensive—a phenomenon known as the "robustness premium." To combat this, Hajiesmaili et al. and Contreras-Ocaña et al. utilize a budget of uncertainty. This allows planners to tune the level of protection, assuming that while many things might go wrong, they won't all hit their worst-case values at the exact same time.

Two-Stage and Adjustable Robustness

In many real-world problems, decisions happen in stages. You might decide to build a power plant today (Stage 1), but you can adjust its output tomorrow (Stage 2).

Two-stage robust optimization models this "wait-and-see" capability. Li et al. use this to estimate the reactive power potential of a grid: the first stage fixes hardware settings, while the second stage verifies that these settings work across all possible renewable fluctuations. Similarly, Wang et al. show that utilities can use two-stage models to set prices that incentivize data centers to shift workloads, providing a safety margin against inaccurate demand predictions.

To make these models even more flexible, Zhai et al. and Qazi et al. employ adjustable robust optimization. By using linear decision rules, they allow system outputs to adjust dynamically to real-time forecast errors. This prevents the "frozen" conservatism of static models, allowing the system to react to information as it arrives.

Probabilistic Safety and Chance Constraints

Sometimes, preparing for the absolute worst case is physically impossible or economically ruinous. In these cases, researchers use chance-constrained programming. Instead of guaranteeing success in 100% of cases, the system guarantees success with a high probability (e.g., 99%).

Roald et al. apply this to integrated gas-electric systems, ensuring that gas pipelines can handle the sudden fuel demands of power plants with a predefined reliability threshold. Qiu et al. argue that this probabilistic approach is essential for transmission expansion, as it allows planners to balance the cost of new lines against the statistical risk of load shedding. To handle cases where the underlying probability distribution is itself unknown, Zhou et al. propose distributionally robust models, which ensure safety across a family of possible distributions.

Learning to be Robust

Recent work has begun integrating machine learning to handle even more complex uncertainties. Donti frames grid security as an adversarially robust learning problem, using minimax optimization to find the "worst-case" simultaneous outages and optimizing power flow to withstand them.

Furthermore, Sun et al. highlight the challenge of decision-dependent uncertainty, where the act of building infrastructure (like hydrogen refueling stations) actually changes the demand patterns the system is trying to predict. By modeling these feedback loops, the authors show that uncertainty-aware models can capture economic opportunities that deterministic models completely miss.

Whether through interval optimization for building thermal mass (Mu et al.) or state-of-charge buffering for electric buses (Qi et al.), the shift toward robustness ensures that our most critical systems are built not just for the forecast, but for the reality of an unpredictable world.

Key Foundational Work

Stochastic unit commitment and reserve scheduling: A tractable formulation with probabilistic certificates (2013) — contributed 2 concept elements to this direction.
Monitoring and Optimization for Power Grids: A Signal Processing Perspective (2013) — contributed 1 concept elements to this direction.
Risk-aware management of distributed energy resources (2013) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_uncertainty_aware_scheduling_4ff3030b,
  title = {uncertainty-aware scheduling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/uncertainty-aware-scheduling/rdr_4ff3030b8f9544c2a578b38c6eaa737f},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-22},
}

This exact version (version 1, frozen text):

@misc{lacuna_uncertainty_aware_scheduling_4ff3030b_v1,
  title = {uncertainty-aware scheduling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/uncertainty-aware-scheduling/rdr_4ff3030b8f9544c2a578b38c6eaa737f/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-22},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake