Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Accelerating Quasi-Static Time Series Simulations with Foundation Models

Alban Puech, François Mirallès, Jonas Weiss, Vincent Mai, Alexandre Blondin Massé, Martin de Montigny, Thomas Brunschwiler, Hendrik F. Hamann

TL;DR

This paper envision how recently introduced grid foundation models could improve the economic viability of neural power flow solvers, and calls for collaboration between the AI and power grid communities to develop and open-source these models, enabling all operators to benefit from AI without building solutions from scratch.

Abstract

Quasi-static time series (QSTS) simulations have great potential for evaluating the grid's ability to accommodate the large-scale integration of distributed energy resources. However, as grids expand and operate closer to their limits, iterative power flow solvers, central to QSTS simulations, become computationally prohibitive and face increasing convergence issues. Neural power flow solvers provide a promising alternative, speeding up power flow computations by 3 to 4 orders of magnitude, though they are costly to train. In this paper, we envision how recently introduced grid foundation models could improve the economic viability of neural power flow solvers. Conceptually, these models amortize training costs by serving as a foundation for a range of grid operation and planning tasks beyond power flow solving, with only minimal fine-tuning required. We call for collaboration between the AI and power grid communities to develop and open-source these models, enabling all operators, even those with limited resources, to benefit from AI without building solutions from scratch.

Accelerating Quasi-Static Time Series Simulations with Foundation Models

TL;DR

This paper envision how recently introduced grid foundation models could improve the economic viability of neural power flow solvers, and calls for collaboration between the AI and power grid communities to develop and open-source these models, enabling all operators to benefit from AI without building solutions from scratch.

Abstract

Quasi-static time series (QSTS) simulations have great potential for evaluating the grid's ability to accommodate the large-scale integration of distributed energy resources. However, as grids expand and operate closer to their limits, iterative power flow solvers, central to QSTS simulations, become computationally prohibitive and face increasing convergence issues. Neural power flow solvers provide a promising alternative, speeding up power flow computations by 3 to 4 orders of magnitude, though they are costly to train. In this paper, we envision how recently introduced grid foundation models could improve the economic viability of neural power flow solvers. Conceptually, these models amortize training costs by serving as a foundation for a range of grid operation and planning tasks beyond power flow solving, with only minimal fine-tuning required. We call for collaboration between the AI and power grid communities to develop and open-source these models, enabling all operators, even those with limited resources, to benefit from AI without building solutions from scratch.

Paper Structure

This paper contains 9 sections, 1 figure.

Table of Contents

  1. Introduction
  2. QSTS Simulations - Operational Relevance and Obstacles to Larger Scale Application
  3. Operational Relevance
  4. Obstacles to Larger Scale Application
  5. AI Foundation Models to Overcome QSTS Computational Bottlenecks
  6. Shifting Computational and Numerical Stability Challenges from Runtime to Training
  7. Making AI Models Economically Scalable using Foundation Models
  8. Foundation Models - One Foundation for All and All Data for One
  9. Conclusion and Future Outlook

Figures (1)

  • Figure 1: Masking and reconstruction steps for grid FM training. 1. Masking: Given a graph representation of the transmission grid $\mathcal{G}$, the function $m_\alpha$ randomly masks node variables (independently of the bus type) with masking probability $\alpha$. The resulting masked graph is $\mathcal{G}'$. 2. Reconstruction: We assume the existence of a function $f$ that, given a masked graph $\mathcal{G}'$, returns the original graph $\mathcal{G}$. Our model is trained to approximate $f$.