Distributionally Robust Chance-Constrained Optimal Transmission Switching for Renewable Integration

Yuqi Zhou, Hao Zhu, Grani A. Hanasusanto

As we race to decarbonize the power grid, we face a frustrating paradox: sometimes we have too much clean energy. When wind or solar production spikes beyond what the local transmission lines can handle, grid operators are forced to "curtail" (waste) that green power to prevent the system from overloading.

A new paper by Yuqi Zhou, Hao Zhu, and Grani A. Hanasusanto, recently accepted in IEEE Transactions on Sustainable Energy, proposes a sophisticated way to solve this: Optimal Transmission Switching (OTS) using distributionally robust optimization.

TL;DR

The researchers developed a mathematical framework that allows grid operators to reroute power by strategically turning transmission lines on or off, specifically designed to handle the unpredictable nature of renewable energy. By using "distributionally robust" chance constraints, their method ensures the grid remains reliable across a wide range of possible weather scenarios while significantly reducing the amount of clean energy wasted.

What Problem Does This Paper Address?

The traditional power grid was designed for steady, predictable fossil fuel plants. Renewables are the opposite—highly variable and hard to forecast. This uncertainty leads to two major issues:

  1. Grid Congestion: Power lines hit their thermal limits, creating bottlenecks.
  2. Renewable Curtailment: To keep the grid safe during these bottlenecks, operators simply shut down wind farms.

Optimal Transmission Switching (OTS) is a "grid-edge" solution. Think of the power grid like a series of pipes with valves. By opening or closing specific valves (switching lines), you can change the "shape" of the grid to bypass bottlenecks.

However, OTS is a "Mixed-Integer" problem—it’s computationally "heavy" because you are making binary (on/off) decisions. When you add the massive uncertainty of wind power into the mix, standard optimization tools either become too slow for real-time use or too "conservative" (meaning they waste money to be overly safe).

Methods

The authors tackle this using Distributionally Robust Chance-Constrained (DRCC) optimization.

  • The "Ambiguity Set": Instead of assuming they know the exact probability distribution of wind power (which is impossible), they define an "ambiguity set"—a family of plausible distributions that match known data (like mean and deviation). The system is optimized to perform well even in the "worst-case" distribution within that set.
  • Two-Stage Recourse: The model makes a "here-and-now" decision (which lines to switch) and accounts for "wait-and-see" adjustments (how generators will react once the wind actually blows).
  • Linear Decision Rules (LDR): A key technical contribution is proving that for this specific grid model, the complex adjustments can be represented as simple linear functions of the uncertainty. This allows them to turn a terrifyingly complex problem into a Mixed-Integer Linear Program (MILP), which modern computers can solve efficiently.
  • Two Flavors of Robustness: They tested two types of ambiguity sets:
  • Moment-based (Mean-MAD): Fast and "scenario-free," making it ideal for real-time operations.
  • Distance-based ($\infty$-Wasserstein): A data-driven approach that looks at how far a distribution is from historical samples.

Datasets

The researchers validated their models using:

  • IEEE 14-bus and 118-bus systems: Standard benchmarks used in power systems engineering to simulate small and medium-sized grids.
  • ERCOT Wind Data: Real-world hourly wind power data from the Texas market (2018–2020) to ensure the "uncertainty" in the model reflected actual weather patterns.

Key Insights and Ideas

  • LDR is Exact, Not Just an Approximation: Usually, assuming a linear response (LDR) is a simplification to make math easier. Here, the authors analytically established that under standard "DC power flow" rules, the grid's response to uncertainty is actually linear. This theoretical "aha" moment justifies why their simplified model remains highly accurate.
  • The Scalability Win: By avoiding "scenario-based" methods (which get slower as you add more data points), their moment-based approach maintains a fixed problem size. This is the difference between a tool that works in a lab and a tool that can actually be used by a grid operator in a control room.
  • Multimodality Matters: Wind data often has "modes" (e.g., it's either very windy or very calm). The paper shows that incorporating this "multimodal" info makes the results less conservative and more cost-effective.

Key Results

  • Reliability: In "Out-of-Sample" tests (testing the model on data it hadn't seen), the DRCC approach consistently met reliability targets (e.g., staying within limits 95% of the time), whereas standard Gaussian or sample-based methods often failed.
  • Efficiency: On the 118-bus system, the proposed method achieved a 1.0% to 1.8% reduction in total system costs compared to robust benchmarks. While 1% sounds small, in the multi-billion dollar utility industry, this represents massive savings.
  • Reduced Curtailment: The DRCC-OTS approach significantly lowered the amount of wind energy that had to be thrown away compared to traditional grid management.
  • Speed: The moment-based MILP solved in under a second for the 14-bus system and remained scalable for the larger 118-bus system, confirming its readiness for real-time grid dispatch.

Limitations and Future Work

  • DC vs. AC: The study uses the "DC power flow" model, which is a common linear approximation. Real grids follow "AC" physics, which are non-linear and involve voltage limits. The authors suggest that post-processing checks or "AC relaxations" would be the next step for deployment.
  • Switching Limits: The model limits the number of lines that can be switched ($L_o$) to maintain system stability.
  • Future Directions: The authors point toward using machine learning to further accelerate the Wasserstein-based models and extending the framework to include "N-1" contingency analysis (ensuring the grid stays up even if a line fails unexpectedly).

Summary updated

Abstract

Increasing integration of renewable generation poses significant challenges to ensure robustness guarantees in real-time energy system decision-making. This work aims to develop a robust optimal transmission switching (OTS) framework that can effectively relieve grid congestion and mitigate renewable curtailment. We formulate a two-stage distributionally robust chance-constrained (DRCC) problem that assures limited constraint violations for any uncertainty distribution within an ambiguity set. Here, the second-stage recourse variables are represented as linear functions of uncertainty, yielding an equivalent reformulation involving linear constraints only. We utilize moment-based (mean-mean absolute deviation) and distance-based ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\infty$</tex-math></inline-formula> -Wasserstein distance) ambiguity sets that lead to scalable mixed-integer linear program (MILP) formulations. Numerical experiments on the IEEE 14-bus and 118-bus systems have demonstrated the performance improvements of the proposed DRCC-OTS approaches in terms of guaranteed constraint violations and reduced renewable curtailment. In particular, the computational efficiency of the moment-based MILP approach, which is scenario-free with fixed problem dimensions, has been confirmed, making it suitable for real-time grid operations.
BibTeX
@article{zhou2022distributionally,
  author = {Yuqi Zhou and Hao Zhu and Grani A. Hanasusanto},
  title = {{Distributionally Robust Chance-Constrained Optimal Transmission Switching for Renewable Integration}},
  journal = {IEEE Transactions on Sustainable Energy},
  year = {2022},
  doi = {10.1109/tste.2022.3203669},
}

Loading related content...

Figure