Fast Relax-and-Round Unit Commitment with Economic Horizons

arXiv:2603.15934 · math.OC, cs.MS, cs.SY, eess.SY · Submitted 2026-03-16 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: We're looking at the paper: "Fast Relax-and-Round Unit Commitment with Economic Horizons".

Dev: Here's the gist of the paper. The gist The US energy system is increasingly under pressure to serve expanding data loads and to accommodate a larger number of generating units with varying technologies and ownership structures, necessitating new unit commitment methods that include ramping constraints and long-horizon planning How it works The paper introduces a novel computational method for unit commitment (UC) that expands the formulation to include ramping constraints and long-horizon planning, providing a theoretical bound on its error.

Rosa: The method includes a fast novel algorithm specifically designed to commit hydro-generators.

Dev: It is shown that the method can solve UC problems with over 20,000 generators in approximately 10 seconds on commodity hardware and that an increased planning horizon leads to sizable operational cost savings.

Rosa: The mathematical formulation for long-horizon planning UC expands previous work to explicitly include minimum run time, maximum daily starts, ramping constraints and generation costs in subsequent time periods.

Title and authors: Rosa: So we're talking about "Fast Relax-and-Round Unit Commitment with Economic Horizons". This paper tackles the growing pressure on US energy systems to handle more generators and different types of power plants.

Dev: Yeah, it looks like they're expanding a known method for unit commitment by adding ramping constraints and long-horizon planning to get a tighter bound on the error. It’s about making those horizon decisions more robust.

Taro: And they introduce this fast novel algorithm specifically designed to commit hydro generators before all the traditional units. That seems like it could really speed things up in practice, Rosa.

Rosa: Exactly, Taro. They show that this method can solve problems with over twenty thousand generators in about ten seconds on regular hardware, and they point out that a longer planning horizon leads to sizable operational cost savings <ref:2603.15934#pg1,planning horizon leads to sizable operational cost savings>.

Dev: That's the big payoff there—getting those cost savings from the extended planning period. I see they’re expanding the mathematical formulation to explicitly include minimum run time, maximum daily starts, ramping constraints, and generation costs in subsequent time periods.

Taro: And they define these constraints very precisely for each generator i: max(Pmin,i, Pt-1,i - rd,i) less than or equal to Pi less than or equal to min(Pmax,i, Pt-1,i + ru,i) for generator i Gd <ref:2603.15934#pg1>.

Rosa: Right. So it’s not just about turning things on and off now; it’s about planning what those units can actually do over the whole horizon while respecting their physical limits and ramping requirements.

Dev: And they introduce a fast novel algorithm to adaptively commit hydro units before traditional units, which they test empirically in Section III of this paper. I'm interested in how that works when the world misbehaves, Taro.

Taro: Well, the core of Algorithm one involves temporarily relaxing those commitment status variables ui from zero or one to yi between zero and one to solve Eq <ref:2603.15934#pg1>. (one) in a way similar to branch-and-bound methods <ref:2603.15934#pg1>. They reorder the discretionary generators Gd so that y1 is greater than or equal to...

Rosa: So they find the first m˜ discretionary generators such that Pmax,i plus X one less than or equal to i in Gd Pmax,i is greater than or equal to R up arrow <ref:2603.15934#pg1>. This gives them a minimum number of generators needed to operate, defined as m equals the size of Gm plus tilde-m.

Title and authors: Dev: And then they solve the economic dispatch for all options of generators that can supply the load, meaning for every k between m and mmax, they set yj less than or equal to k equals one and yj greater than k equals zero.

Taro: They solve min Pj X k j is one a P2 i plus b Pi plus c i ui at t minus one i plus X i in Gd Ki(ut minus 1,i(one minus ui) plus (one minus ut-1,i)ui at t with gamma sigma over delta X i in Gm union Gd aiP2 delta,i plus bi Pdelta,i plus ci ui <ref:2603.15934#pg1>.

Rosa: That formulation shows they are accounting for the on/off cost terms in line three of Eq. (1a) within the rounding scheme and adding them to the objective of their respective economic dispatches in Eq <ref:2603.15934#pg1>. (four) before deciding the minimal objective <ref:2603.15934#pg3,4) before deciding the minimal objective>.

Dev: It seems like they’ve made a big step by showing that this approach, called RRUC, has a relative error that vanishes as the problem size increases and gives a tighter bound on the deviation from the solution to the relaxed problem. That's strong for scalability.

Taro: And appendix B shows that the original problem proposed in Section III is NP-Hard and that their proposed algorithm runs in log-linear time, which is quite fast for a problem of this complexity.

Rosa: So, they’ve got a provably scalable algorithm here that runs in O n four over three n where n is the number of units. That’s very interesting when you consider the size of these problems.

Dev: They also have a specific method for hydro-generator commitment based on shifting predicted demands to commit them, which they test using Algorithm two <ref:2603.15934#pg2>. This approach runs in log-linear time, O n (log n plus T log T) where T is the number of periods over which to optimize.

Taro: And the simulation results show that this Algorithm two retains accuracy in its objective and its run time scales almost linearly <ref:2603.15934#pg2>. They even showed that Algorithm two can commit one thousand seven hundred ninety-two generators for an entire year in twenty seconds.

Rosa: That speed is impressive when you think about committing units for a whole year. It also suggests that the correlation between marginal cost and demand pairs is highly positive, which motivates using hydro generators to flatten load or peak shave by smoothing out the demand D bar t equals Dt minus Pn j equals one u j t kappa j.

Title and authors: Dev: So, what does this mean for operators? It means they can make these horizon-aware economic decisions for large systems with hydro units, which is something we’ve been talking about.

Taro: It means they have a tool that handles the complexity of ramping constraints and long planning horizons without getting bogged down in intractable computation.

Rosa: And they aren't just building a model; they are providing an implementable method for real-world grid operations, which is what I look for when I run trials outside the lab.

Dev: So, looking at the paper "Fast Relax-and-Round Unit Commitment with Economic Horizons", we see they’ve improved the unit commitment formulation by incorporating mechanical constraints and future cost considerations into that original model.

Taro: They also introduced a provably scalable algorithm to commit hydro-generators that runs in O n log n plus T log T time, where n is the number of hydro units and T is the number of periods over which to optimize.

Rosa: And they specifically introduce a fast novel algorithm for committing hydro generators, showing it can solve UC problems with over twenty thousand generators in approximately ten seconds on commodity hardware <ref:2603.15934#pg1,can solve UC problems with over 20,000 generators in approximately 10>.

Dev: Their conclusion is that this paper extends the RRUC formulation to Eq. (one), incorporating mechanical constraints and future cost considerations into UC <ref:2603.15934#pg1>. They also introduce a provably scalable algorithm to commit hydro-generators.

Taro: And they introduce a fast novel algorithm specifically designed to commit hydro units before traditional units, which they test empirically in Section III of this paper.

Rosa: So, the big implication is that it allows operators to make horizon-aware economic decisions for large systems with hydro units.

Dev: I think the main thing is that they’ve given us a tool that handles ramping constraints and long-horizon planning effectively, which we need for these expanding US energy systems.

Taro: It’s about having a method that can handle the sheer size of these problems without needing supercomputers just to get a reasonable solution.

Rosa: So, for those of us listening who might be driving or cooking and not deep into power systems, it means there's a way to make better economic choices for the grid by using this fast and accurate unit commitment method.

Dev: We’ll wrap up here on "Fast Relax-and-Round Unit Commitment with Economic Horizons." Next up, we’re looking at something completely different.

The paper's summary: Rosa: Basically, they take that old model and expand it to include those ramping constraints—meaning generators can't change output too fast—and they add long-horizon planning so we can make decisions over a much longer time period. This gives operators a better look at operational costs down the line.

Dev: That’s the main point, yeah, expanding the math to include minimum run times, maximum daily starts, and all that generation cost stuff across subsequent time periods. It’s about making sure we're not just looking at today but planning for next month or even next year.

Taro: But the big part I’m interested in is how they tackle committing hydro generators specifically. They introduce a fast novel algorithm that works before the traditional units, which is crucial because hydro units are different—they recharge naturally, so we need to treat them differently than conventional plants.

Rosa: Right, and this algorithm runs in log-linear time for committing those hydro units. They show it can commit over twenty thousand generators in about ten seconds on regular hardware. That speed is what makes it practical for real operations outside the lab, Dev.

Dev: It’s fast enough that the increased planning horizon actually leads to sizable operational cost savings, which is exactly what we need for large US energy systems dealing with a lot of different generation types.

Taro: And they have this method based on shifting predicted demands to commit hydro units before traditional units, optimizing that demand over a specific period. It helps flatten the load or shave peaks by smoothing out those high-demand periods.

Rosa: So, the implication for me is that we’ve got a provably scalable algorithm now for committing hydro generators without needing supercomputers just to get a reasonable solution. It moves this from theory into something operators can actually use.

Dev: That scalability is key because the problems they are solving with these constraints and long-term planning are huge, and they need tools that run efficiently in real-time loops.

Taro: They also showed that their approach retains accuracy in its objective even when scaling up, which means we can trust the decisions being made with this method.

Rosa: So the big picture is that this gives operators a tool to make these horizon-aware economic decisions for large systems with hydro units, which is what we’ve been talking about needing.

Dev: It’s about having a method that handles ramping constraints and long planning horizons effectively, which we need for these expanding US energy systems.

Taro: It’s about having a method that can handle the sheer size of these problems without needing supercomputers just to get a reasonable solution.

Dev: We’ll wrap up here on "Fast Relax-and-Round Unit Commitment with Economic Horizons." Next up, we’re looking at something completely different.

The paper's improvements: Taro: So, we’re looking at how they suggest making this whole unit commitment formulation even better than what they first proposed. They aren't just stopping at adding ramping constraints and long planning horizons, though that was a big step for me too.

Rosa: Yeah, they talk about incorporating mechanical constraints directly into the mathematical model, not just as a condition you check later. This means the solution is necessarily feasible with respect to those physical limits right from the start of the calculation.

Dev: That’s important for my control loop concerns because it means we aren't dealing with solutions that might look mathematically optimal but are physically impossible to run on a power system. We need that safety case built in earlier.

Taro: And they also bring in future cost considerations into the planning, meaning you're not just optimizing for today’s price, but thinking about costs in subsequent time periods too. That’s what makes it a long-horizon planner.

Rosa: Exactly. It shifts the focus from just minimizing immediate operational costs to making decisions that are better across the entire planning window they set up. It’s about making horizon-aware economic decisions for large systems with hydro units, which is what I mean.

Dev: So, the improvement isn't just adding more variables; it's changing how the whole system is structured so that feasibility and future cost are baked into every decision variable they solve.

Taro: It makes sense because when you’re dealing with a lot of units, those constraints pile up fast, so defining them upfront saves a lot of headaches down the line during simulation or real-world testing.

Rosa: And I think the main implication is that this refined formulation gives operators a much more robust framework for making these long-term economic bets on how they run their power plants.

Dev: It suggests that for complex systems, we need to go beyond simple reactive control and build in these forward-looking constraints into the core commitment logic.

Taro: The next thing I’m wondering is if this entire structure can handle even more complexity, like transmission constraints, which they mention looking into in their future work.

Conclusion: Rosa: So to wrap up, this paper on "Fast Relax-and-Round Unit Commitment with Economic Horizons" shows they’ve built a method that handles ramping constraints and long planning horizons really well, especially for committing hydro units. This means operators can actually make horizon-aware economic decisions for large systems.

Dev: Yeah, the main implication is having a tool that handles those ramping constraints and long planning horizons without getting bogged down in intractable computation. It’s about making sure the safety case holds up when you scale up to thousands of generators.

Taro: I think what really stands out is that they introduced a provably scalable algorithm for hydro-generator commitment that runs in O n log n plus T log T time, which is pretty fast for a problem of this complexity.

Rosa: And the speed they achieved, committing one thousand seven hundred ninety-two generators in twenty seconds, shows it’s an implementable method for real-world grid operations. It proves the model works outside the lab.

Dev: It confirms that we can build models that handle these complex constraints and future cost considerations into unit commitment calculations effectively. That’s a big step for controls engineers looking at real-time loops.

Taro: I just think it gives us a solid foundation for testing how these systems recover when the world misbehaves, since they’ve shown the math is robust even with those long planning horizons included.

Rosa: Exactly. So we have this fast and accurate unit commitment method that incorporates mechanical constraints and future cost considerations into the original model.

Dev: It really gives us a much better framework for making those long-term economic bets on how we run our power plants in the next few months or years.

Taro: I’m curious to see what they do next, especially if they can incorporate transmission constraints into this formulation, which is where I think the real complexity lies.

Rosa: Right, so for those of us listening who might be driving or cooking and not deep into power systems, it means there's a way to make better economic choices for the grid by using this fast and accurate unit commitment method.

Dev: We’ll wrap up here on "Fast Relax-and-Round Unit Commitment with Economic Horizons." Next up, we’re looking at something completely different.

Shijie Pan, Zaint A. Alexakis, Charalambos Konstantinou

PJM

math.OC, cs.MS, cs.SY, eess.SY

Submitted: 2026-03-16

Updated: 2026-10-20

Comments: 19 pages, 5 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 84/100

The gist: The gist The US energy system is increasingly under pressure to serve expanding data loads and to accommodate a larger number of generating units with varying technologies and ownership structures,

Key concepts

Unit Commitment (UC) Formulation: This is the mathematical model used to decide which power generators should be on or off at specific times. The paper expands this model to include complex rules like ramping constraints (how fast a generator can change output) and long-horizon planning, allowing operators to make decisions that look far into the future rather than just for the next hour.
Ramping Constraints: These are physical limitations on generators, specifying the maximum rate at which a power plant's output can increase or decrease over a short period. Including these constraints makes the commitment model more realistic by ensuring that planned changes in power generation are physically achievable.
Hydro-Generator Commitment (Algorithm 2): This is a specialized, fast method designed to commit hydro generators. Instead of treating them like traditional units, this approach uses predicted demand shifts to decide when to turn them on. It runs very quickly and helps smooth out load peaks by using hydropower strategically.

Terminology

Generator — 2× in this paper · explained in 2 episodes

Summary

The gist The US energy system is increasingly under pressure to serve expanding data loads and to accommodate a larger number of generating units with varying technologies and ownership structures, necessitating new unit commitment methods that include ramping constraints and long-horizon planning

How it works

The paper introduces a novel computational method for unit commitment (UC) that expands the formulation to include ramping constraints and long-horizon planning, providing a theoretical bound on its error. The method includes a fast novel algorithm specifically designed to commit hydro-generators. It is shown that the method can solve UC problems with over 20,000 generators in approximately 10 seconds on commodity hardware and that an increased planning horizon leads to sizable operational cost savings.

The mathematical formulation for long-horizon planning UC expands previous work to explicitly include minimum run time, maximum daily starts, ramping constraints and generation costs in subsequent time periods. The decision variables include power output of must-run or discretionary generator i (Pi), production level by generator i at demand level δ (Pδ,i), and commitment status of discretionary generator i (ui). Constraints are defined to ensure feasibility, such as max(Pmin,i, Pt-1,i − rd,i) ≤ Pi ≤ min(Pmax,i, Pt-1,i + ru,i) for generator i (Gd).

Algorithm Design and Performance

The paper introduces a fast novel algorithm to adaptively commit hydro units before traditional units and tests its performance empirically. The core algorithm, Algorithm 1, temporarily relaxes ui ∈ [0, 1] to yi ∈ [0, 1] to solve Eq. (1) in a similar fashion to branch-and-bound methods. It re-orders discretionary generators Gd so that y1 ≥... ≥ yGd and finds the first m˜ discretionary generators such that X i∈Gm Pmax,i + X 1⩽i≤m˜ i∈Gd Pmax,i ≥ R↑. This leads to a minimum number of generators to operate defined as m.= Gm + ˜m.

The algorithm then solves economic dispatch for all options of generators that can supply the load, meaning ∀k: m ≤ k ≤ mmax, we set yj≤k = 1 and yj>k = 0 and solve: min Pj X k j=1 ajP2 j + bjPj + cj s.t. max(Pmin,i, Pt-1,i − rd,i) ≤ Pi ≤ min(Pmax,i, Pt-1,i + ru,i) (4b). The runtime complexity of RRUC is estimated to be O(n 4/3), where n is the number of units.

Hydro-Generator Commitment Method

To commit hydro generators, the approach decouples them from traditional units, justified because hydro units recharge by natural water flow and cannot indefinitely hoard water due to the dam’s capacity. The paper introduces a method based on shifting predicted demands to commit hydro generators. This involves defining the marginal cost of the system as ui(2aiPi +bi) for one such unit i.

Algorithm 2 approximates a solution by “greedily” assigning a hydro generator which still has remaining periods to the period with the highest remaining demand. This approach runs in log-linear time, O(n(log n + T log T)), where n is the number of hydro units and T is the number of periods over which to optimize. The simulation results show that Algorithm 2 retains accuracy in its objective and its run time scales almost linearly.

Simulation Results

Simulations use historical PJM data for load and generating unit bids, scaling the system using a 20% cold reserve margin for resource adequacy. The model is implemented in Julia with the MadNLP nonlinear programming solver and Gurobi linear program solver. The results show that RRUC produces numerically indistinguishable solutions from an exact algorithm, though RRUC is twice as fast on the largest problem tested.

The analysis of demand-marginal cost pairs shows that marginal cost and demand are highly positively correlated, with a few outliers. This correlation motivates using hydro generators to flatten load or peak shave by smoothing out D¯t = Dt − Pn j=1 uj,tκj. The final results show that Algorithm 2 can commit 1792 generators for an entire year in 20 seconds.

Conclusion and Future Work

The paper extends the RRUC formulation to Eq. (1), incorporating mechanical constraints and future cost considerations into UC. It introduces a provably scalable algorithm to commit hydro-generators. Future research intends to incorporate transmission constraints in UC, leveraging insights from Fig. 6. Additionally, the authors plan to attempt to leverage parallelism, including GPUs, to accelerate their solver and further examine RRUC’s speed-accuracy trade-offs.

The paper is a scientific study on developing a fast and accurate unit commitment method for power systems that incorporates ramping constraints and long-horizon planning, which is important because it allows operators to make horizon-aware economic decisions for large systems with hydro units. The US energy system is increasingly under pressure to serve expanding data loads and to accommodate a larger number of generating units with varying technologies and ownership structures, necessitating new unit commitment methods that include ramping constraints and long-horizon planning.

Improvements for AI systems

Improve unit commitment planning by introducing long-horizon planning to allow for sizable operational cost savings. This enables operators to make horizon-aware economic decisions for large systems with hydro units.
Develop a fast novel algorithm tailored specifically for committing hydro-generators that runs in log-linear time, allowing the system to commit 1792 generators for an entire year in 20 seconds.
Implement a method to commit hydro generators based on shifting predicted demands by optimizing the demand remaining after hydro-generator commitment at time t optimized over [T]. This allows for flattening load or peak shave by using hydropower to reduce demand from high demand periods.
Enhance the overall unit commitment formulation by incorporating ramping constraints and mechanical constraints into the mathematical model, which ensures the solution is necessarily feasible with respect to ramping constraints.
Create a provably scalable algorithm for hydro-generator commitment that runs in O(n log n + T log T) time, where n is the number of hydro units and T is the number of periods over which to optimize.

Sources

A Fast Relax-and-Round Approach to Unit Commitment for Data Center Own Generation

Listen

Transcript

Key concepts

Terminology

Summary

Improvements for AI systems

Sources

Related papers