{"id":1171470,"date":"2026-05-13T09:00:20","date_gmt":"2026-05-13T16:00:20","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?p=1171470"},"modified":"2026-05-13T09:58:05","modified_gmt":"2026-05-13T16:58:05","slug":"gridsfm-a-new-small-foundation-model-for-the-electric-grid","status":"publish","type":"post","link":"https:\/\/www.microsoft.com\/en-us\/research\/blog\/gridsfm-a-new-small-foundation-model-for-the-electric-grid\/","title":{"rendered":"GridSFM: A new, small foundation model for the electric grid"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"576\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-1024x576.jpg\" alt=\"Three white line icons\u2014a transmission tower, a lightning bolt, and a stopwatch\u2014displayed on a teal-to-green gradient background with a subtle textured pattern.\" class=\"wp-image-1171570\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-1280x720.jpg 1280w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1.jpg 1400w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<h2 class=\"wp-block-heading h4\" id=\"microsoft-releases-a-lightweight-foundation-model-that-can-predict-ac-optimal-power-flow-in-milliseconds-boosting-efficiency-and-unlocking-cost-savings-in-grid-analysis\">Microsoft releases a lightweight foundation model that can predict AC optimal power flow in milliseconds, boosting efficiency and unlocking cost savings in grid analysis.<\/h2>\n\n\n\n<div style=\"padding-bottom:0; padding-top:0\" class=\"wp-block-msr-immersive-section alignfull row wp-block-msr-immersive-section\">\n\t\n\t<div class=\"container\">\n\t\t<div class=\"wp-block-msr-immersive-section__inner wp-block-msr-immersive-section__inner--narrow\">\n\t\t\t<div class=\"wp-block-columns mb-10 pb-1 pr-1 is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\" style=\"box-shadow:var(--wp--preset--shadow--outlined)\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<h2 class=\"wp-block-heading h3\" id=\"at-a-glance\">At a glance<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Microsoft introduces GridSFM, a small foundation model that approximates AC optimal power flow in milliseconds, unlocking decisions that can directly impact up to $20B\/year in congestion losses and 3.4\u202fTWh of renewable curtailment.<\/li>\n\n\n\n<li>Beyond estimating generator dispatch and costs, GridSFM produces full AC system states, giving operators direct visibility into congestion, stability, and overall system health.<\/li>\n\n\n\n<li>It provides a foundation for the community to build advanced power grid simulators and planning tools without recreating data or models from scratch.<\/li>\n<\/ul>\n<\/div>\n<\/div>\t\t<\/div>\n\t<\/div>\n\n\t<\/div>\n\n\n\n<p>Microsoft introduces <strong>GridSFM<\/strong>, a small foundation model for solving AC optimal power flow (AC-OPF) problems in transmission power grids. This follows our <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/blog\/building-realistic-electric-transmission-grid-dataset-at-scale-a-pipeline-from-open-dataset\/\">earlier release<\/a> of a U.S.-based open transmission-topology dataset that powers GridSFM.<\/p>\n\n\n\n<p>Power grids face increasing strain from surging demand, the need to integrate renewable energy sources, transportation electrification, and extreme weather events. Across all these challenges, the core question is the same: what are the optimal operating points that keep the grid functioning under each new condition?<\/p>\n\n\n\n<p>Answering this requires solving AC optimal power flow (AC\u2011OPF), a complex, non-convex optimization problem that computes the cheapest generator dispatch (how much each generator produces) that meets demands while respecting power flow physics, voltage limits, thermal constraints, and stability requirements, and underpins core power system operations including reliability, real-time dispatch, market clearing, and contingency analysis. These decisions directly govern outcomes at the scale of up <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/gridstrategiesllc.com\/wp-content\/uploads\/GS_Transmission-Congestion-for-2024.pdf\" target=\"_blank\" rel=\"noopener noreferrer\"><em>$20\u202fbillion per year in congestion costs<\/em><span class=\"sr-only\"> (opens in new tab)<\/span><\/a> and <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.eia.gov\/todayinenergy\/detail.php?id=65364\" target=\"_blank\" rel=\"noopener noreferrer\"><em>multi\u2011terawatt\u2011hour renewable curtailment<\/em><span class=\"sr-only\"> (opens in new tab)<\/span><\/a> (lost renewable energy due to congestion), making both economic efficiency and grid reliability highly sensitive to how well these operating points are found. However, AC\u2011OPF is computationally expensive: power utility scale grid can take up to hours solve, forcing a trade-off between solving a small number of carefully selected scenarios or relying on approximations that ignore critical physics, which can misestimate power flows and binding constraints and lead to suboptimal dispatch and degraded reliability under stressed conditions.<\/p>\n\n\n\n\t<div class=\"border-bottom border-top border-gray-300 mt-5 mb-5 msr-promo text-center text-md-left alignwide\" data-bi-aN=\"promo\" data-bi-id=\"1144028\">\n\t\t\n\n\t\t<p class=\"msr-promo__label text-gray-800 text-center text-uppercase\">\n\t\t<span class=\"px-4 bg-white display-inline-block font-weight-semibold small\">PODCAST SERIES<\/span>\n\t<\/p>\n\t\n\t<div class=\"row pt-3 pb-4 align-items-center\">\n\t\t\t\t\t\t<div class=\"msr-promo__media col-12 col-md-5\">\n\t\t\t\t<a class=\"bg-gray-300 display-block\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/story\/the-ai-revolution-in-medicine-revisited\/\" aria-label=\"The AI Revolution in Medicine, Revisited\" data-bi-cN=\"The AI Revolution in Medicine, Revisited\" target=\"_blank\">\n\t\t\t\t\t<img decoding=\"async\" class=\"w-100 display-block\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/06\/Episode7-PeterBillSebastien-AIRevolution_Hero_Feature_River_No_Text_1400x788.jpg\" alt=\"Illustrated headshot of Bill Gates, Peter Lee, and S\u00e9bastien Bubeck\" \/>\n\t\t\t\t<\/a>\n\t\t\t<\/div>\n\t\t\t\n\t\t\t<div class=\"msr-promo__content p-3 px-5 col-12 col-md\">\n\n\t\t\t\t\t\t\t\t\t<h2 class=\"h4\">The AI Revolution in Medicine, Revisited<\/h2>\n\t\t\t\t\n\t\t\t\t\t\t\t\t<p id=\"the-ai-revolution-in-medicine-revisited\" class=\"large\">Join Microsoft\u2019s Peter Lee on a journey to discover how AI is impacting healthcare and what it means for the future of medicine.<\/p>\n\t\t\t\t\n\t\t\t\t\t\t\t\t<div class=\"wp-block-buttons justify-content-center justify-content-md-start\">\n\t\t\t\t\t<div class=\"wp-block-button\">\n\t\t\t\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/story\/the-ai-revolution-in-medicine-revisited\/\" aria-describedby=\"the-ai-revolution-in-medicine-revisited\" class=\"btn btn-brand glyph-append glyph-append-chevron-right\" data-bi-cN=\"The AI Revolution in Medicine, Revisited\" target=\"_blank\">\n\t\t\t\t\t\t\tListen now\t\t\t\t\t\t<\/a>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t<\/div><!--\/.msr-promo__content-->\n\t<\/div><!--\/.msr-promo__inner-wrap-->\n\t<\/div><!--\/.msr-promo-->\n\t\n\n\n<p>To address this limitation, we introduce <strong>GridSFM, a single neural network that approximates AC\u2011OPF in milliseconds across grids ranging from 500 to 80,000 buses<\/strong>. It takes standard AC\u2011OPF inputs (grid topology, generator and load specifications, transmission line constraints) and produces an operating point and a feasibility verdict (whether the system satisfies all physical and operational constraints). By removing the compute bottleneck, GridSFM makes it possible to evaluate orders of magnitude more scenarios in real time, enabling more informed decisions and shifting grid operations from reactive response to proactive optimization.<\/p>\n\n\n\n<p>In this initial release we offer two tiers:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>GridSFM-Open<\/strong> for research-scale grids up to 4,000 buses.<\/li>\n\n\n\n<li><strong>GridSFM-Premier<\/strong> for production-scale systems up to 80,000 buses.<\/li>\n<\/ul>\n\n\n\n<p>The model is built as a block-structured discrete neural operator (Figure 1), representing each grid as a directed graph, with buses (connection points in the grid) and generators as vertices, and transmission and AC lines as edges. It is trained using both <strong>solver supervision<\/strong>, where reference solutions are generated using the AC-OPF solver (<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/lanl-ansi.github.io\/PowerModels.jl\/stable\/quickguide\/\" target=\"_blank\" rel=\"noopener noreferrer\"><em>IPOPT in PowerModels.jl<\/em><span class=\"sr-only\"> (opens in new tab)<\/span><\/a>), and <strong>physics-based constraints<\/strong> that penalize violations of fundamental physical laws such as Kirchhoff\u2019s voltage and current laws, as well as operating constraints like thermal limits. This enables the model to learn from both feasible and infeasible regimes. Most learning-based AC-OPF surrogates train one model per grid on a <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/abs\/2406.07234\" target=\"_blank\" rel=\"noopener noreferrer\"><em>narrow distribution<\/em><span class=\"sr-only\"> (opens in new tab)<\/span><\/a>.&nbsp;GridSFM&nbsp;takes the opposite approach: in this release a single model trained across&nbsp;<strong>150+&nbsp;base grid topologies<\/strong>&nbsp;(network structures)<strong>&nbsp;and&nbsp;roughly&nbsp;half a million&nbsp;scenarios<\/strong>&nbsp;spanning varying load profiles, multi-element outages, line-rating derates, voltage-bound tightening, and&nbsp;different&nbsp;generator cost coefficients, so the model is forced to generalize rather than memorize. Across the 54-grid mix&nbsp;test scenarios&nbsp;for&nbsp;GridSFM-Open,&nbsp;our model&nbsp;achieves a&nbsp;<strong>median cost gap of&nbsp;2.23% vs solver ground truth labels<\/strong>&nbsp;(mean 3.41%; <5% gap on&nbsp;83&nbsp;%&nbsp;of scenarios). When&nbsp;more&nbsp;precision&nbsp;is needed,&nbsp;GridSFM\u2019s&nbsp;prediction also serves as a&nbsp;<strong>warm start<\/strong>&nbsp;seed for traditional numerical solvers,&nbsp;GridSFM-seeded-warm beats cold solve by&nbsp;1.66\u00d7&nbsp;geometric&nbsp;mean&nbsp;across the same&nbsp;test&nbsp;scenarios&nbsp;and beats the industry-standard DC-OPF warm-start by&nbsp;1.59\u00d7&nbsp;geomean (per-grid breakdown and full white-paper analysis to follow).&nbsp;&nbsp;Geometric&nbsp;mean, otherwise known as&nbsp;the multiplicative&nbsp;average,&nbsp;is&nbsp;used here&nbsp;since it is more robust to outliers.&nbsp;Our model also&nbsp;demonstrates&nbsp;the ability to adapt to new grids with just a handful of fine-tune scenarios.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"804\" height=\"551\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-2.jpg\" alt=\"diagram\" class=\"wp-image-1171471\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-2.jpg 804w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-2-300x206.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-2-768x526.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-2-240x164.jpg 240w\" sizes=\"auto, (max-width: 804px) 100vw, 804px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 1.<\/strong> GridSFM architecture. Bus, generator, and branch features are embedded into a shared latent space, then refined by a stack of attention blocks operating directly on the grid topology. Output heads decode the latent state into (i) a full AC-OPF operating point, bus voltages and angles, generator dispatch, branch flows, and (ii) a per-scenario feasibility score.<\/figcaption><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"what-it-enables\">What it enables<\/h2>\n\n\n\n<p>A common pattern in grid operations and planning is having to choose between solving a small, hand-picked set of scenarios accurately with full AC-OPF or running thousands of scenarios through a faster approximation that drops parts of the physics. For example, a commonly used tool is the DC-OPF approximation, a linearized version that assumes flat voltage magnitudes and small angle differences and ignores reactive power and losses. DC-approximation solves in seconds what takes full AC minutes to hours, which is why most contingency screens, market-clearing pre-stages, and planning sweeps run on DC-approximation today. The cost is real: DC-approximation ignores voltage and reactive constraints entirely, and its dispatch cost can run >10% off the AC optimum on stressed scenarios (with worst-case grids out past 20% in our test benchmark).<\/p>\n\n\n\n<p>GridSFM is designed as a drop-in alternative to DC-approximation in that fast approximation slot, and unlike most existing AC-OPF neural surrogates, which require a fresh training run for every new topology, GridSFM generalizes across grids in its supported size range without per-topology retraining, so it slots in as universally as DC-approximation. Especially when compared with DC-OPF, GridSFM has three concrete advantages:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Same accuracy class as DC-approximation on standalone dispatch cost.<\/strong> GridSFM and DC fall within the same per-scenario cost-gap distribution (\u00a72 \/ Figure 6), with complementary failure modes: DC fails on grids where its no-loss \/ no-reactive linearization is structurally wrong; GridSFM fails on grids outside its training distribution. The two limitations close along orthogonal axes. DC\u2019s ceiling is fixed by the linearization, whereas GridSFM\u2019s tail closes with more training data.<\/li>\n\n\n\n<li><strong>1,000\u00d7 faster than a full AC solver and approximately 100\u00d7 faster than DC-approximation at the inference step<\/strong>, fast enough to sweep thousands of contingencies (e.g., line or generator outages) in minutes on a single commodity GPU.<\/li>\n\n\n\n<li><strong>A real AC operating point, not a linear approximation.<\/strong> GridSFM produces voltages and reactive power, so the same prediction can be handed to a traditional numerical solver as an <strong>AC warm-start<\/strong>, opening a workflow DC-approximation cannot.<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"1-feasibility-screening-stress-score-triage\">1. Feasibility screening: stress-score triage<\/h3>\n\n\n\n<p>A scenario is <strong>infeasible<\/strong> when no dispatch satisfies all constraints simultaneously: the requested load cannot be served within voltage bounds, thermal limits or generator capacities. Operationally, infeasibility is the most consequential failure signal: the requested operating condition cannot be served at all, and the response is intervention (load shedding, redispatch, relaxing thermal limits). It is also the most expensive class of scenario to screen, because the solver only learns a scenario is infeasible after iterating to non-convergence: each infeasible case costs a full solver run, often longer than a feasible one. Sweeping thousands of contingencies or stress cases to identify the infeasible ones is therefore one of the worst-case budgets in any planning workflow.<\/p>\n\n\n\n<p>GridSFM addresses this with a per-scenario stress score trained jointly with the dispatch head. We evaluate the score on three classes of scenarios on each grid: real-feas are scenarios the AC-OPF solver successfully converged on (i.e., genuinely feasible operating points), real-infeas are scenarios the solver failed to converge on (genuinely infeasible operating points), and synth-infeas are feasible base points we deliberately perturbed to violate a specific constraint (voltage squeeze, thermal bottleneck, angle tightening, or DC-thermal congestion). Across the 54-grid test scenarios, the stress score&#8217;s per-grid binary accuracy is broadly uniform across classes: real-feas (green) mean 94.5%, real-infeas (red) mean 96.1%, synth-infeas (orange) mean 90.4%. Most grids cluster within a few points of the means; outliers below 80% are the same hard grids that show up in cost-gap analysis below.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"936\" height=\"517\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-11.png\" alt=\"chart, line chart\" class=\"wp-image-1171486\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-11.png 936w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-11-300x166.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-11-768x424.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-11-240x133.png 240w\" sizes=\"auto, (max-width: 936px) 100vw, 936px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 2.<\/strong> GridSM per-grid feasibility prediction accuracy across the 54-grid test scenarios, broken out by class (real-feas, real-infeas, synth_infesible). Filled KDE + per-grid dots, with mean (\u2013) and median (:) light dashed lines. The three distributions overlap heavily, the model\u2019s quality is broadly uniform across classes, with a small failing tail of structurally hard grids.<\/figcaption><\/figure>\n\n\n\n<p><strong>Drilling into a case study.<\/strong> Let\u2019s zoom into a single representative grid, the <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/electricgrids.engr.tamu.edu\/electric-grid-test-cases\/activsg2000\/\" target=\"_blank\" rel=\"noopener noreferrer\"><em>Texas2k summer-peak grid<\/em><span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, to show how the learned representation separates feasibility and ROC for predicting.<\/p>\n\n\n\n<p><strong>Representation.<\/strong>&nbsp;Figure 3 visualizes the model&#8217;s learned representation of each&nbsp;Texas2k&nbsp;scenario. We project the per-graph representation (128-dimensional)&nbsp;onto two axes (LD1, LD2) chosen to maximally separate the scenario&nbsp;classes: real-feasible, real-infeasible, and synthetic-infeasible.&nbsp;Squeezing 128 dimensions into 2 inevitably loses information, so this&nbsp;view exaggerates&nbsp;apparent&nbsp;overlap:&nbsp;classes that look mixed here may&nbsp;still be cleanly separable in the full 128-dimensional space the&nbsp;model&nbsp;uses. The shaded cloud shows where graphs of each&nbsp;class concentrate, and the cross at the center of each cloud marks&nbsp;the class centroid, the average position of all graphs of that&nbsp;class. Centroids that sit far apart mean the model treats those&nbsp;classes as clearly distinguishable. Where two shaded clouds overlap, the model&nbsp;is producing&nbsp;similar embeddings for graphs with different&nbsp;labels.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"936\" height=\"673\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-12.png\" alt=\"diagram\" class=\"wp-image-1171487\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-12.png 936w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-12-300x216.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-12-768x552.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-12-240x173.png 240w\" sizes=\"auto, (max-width: 936px) 100vw, 936px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 3.<\/strong> Linear discriminant projection of grid embeddings on the Texas2k scenarios. Real feasibles (green), real infeasibles (red), and synthetic infeasibles (orange), projected onto two axes (LD1, LD2) chosen to maximize between-class separation. Crosses mark class centroids; shaded clouds show where each class concentrates. Overlap between clouds means the model produces similar embeddings for graphs in those classes; in the full 128-dimensional space the model may still separate them along directions not shown.<\/figcaption><\/figure>\n\n\n\n<p><strong>Operation and ROC.<\/strong> The score itself is continuous and ranking-calibrated. Figure 4 shows the ROC over its test mix: <strong>AUC = 0.986<\/strong>. At the natural operating point the same score, thresholded as a binary classifier, yields <strong>95.5% accuracy<\/strong>. Per-mode detection at that threshold is 99\u2013100% on the three perturbation modes that drive a constraint cleanly past its limit.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"590\" height=\"489\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-10.png\" alt=\"chart, line chart\" class=\"wp-image-1171472\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-10.png 590w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-10-300x249.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-10-217x180.png 217w\" sizes=\"auto, (max-width: 590px) 100vw, 590px\" \/><figcaption class=\"wp-element-caption\"><strong><em>Figure 4.<\/em><\/strong><em>&nbsp;ROC curve of the&nbsp;GridSFM&nbsp;stress score&nbsp;for feasibility&nbsp;on the Texas2k summer-peak test mix (real&nbsp;feasibles&nbsp;+ solver-labeled&nbsp;infeasibles&nbsp;+ synthetic perturbation modes that drive a constraint past its limit). Area under the curve =&nbsp;0.986,&nbsp;binary accuracy&nbsp;95.5%&nbsp;at the natural operating point. The score is calibrated for ranking; where to draw the binary cutoff is an&nbsp;operator&nbsp;choice.&nbsp;<\/em><\/figcaption><\/figure>\n\n\n\n<p><strong>Triage cutoff.<\/strong> For routing scenarios into action buckets, Figure 5 shows the stress-score distribution per population. Operators pick the cutoff that matches their workflow: very-confident feasibles pass through to indicative dispatch; very-confident-stressed scenarios are flagged for engineering review; the borderline middle band is sent to the solver for verification. The cutoff sets the balance between solver budget and screening miss-rate.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"935\" height=\"330\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-3.jpg\" alt=\"chart, histogram\" class=\"wp-image-1171476\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-3.jpg 935w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-3-300x106.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-3-768x271.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-3-240x85.jpg 240w\" sizes=\"auto, (max-width: 935px) 100vw, 935px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 5.<\/strong> Distribution of the model&#8217;s feasibility logit on the same Texas2k test scenarios, split by population: real-feasibles (green), real-infeasibles (red), and synth-infeasibles (orange). The dashed vertical line is the decision boundary where logit=0. Samples to the right are predicted feasible. At this operating threshold, real-feasible pass through at 99.5%, real-infeas are correctly flagged at 90.4%, and the synthetic perturbation are caught at 88-100%.<\/figcaption><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"2-gridsfm-as-a-fast-approximation\">2. GridSFM as a fast approximation<\/h3>\n\n\n\n<p>GridSFM\u2019s prediction can be used in two ways without producing an exact AC-OPF solution from scratch: as a standalone dispatch and cost estimate, or as the initial guess (warm-start) for an exact numerical solver. We compare both against the same two reference points throughout: full AC-OPF (the ground-truth optimum) and DC-approximation (the established fast baseline). All numbers below come from the same test set of 54 grids scenarios GridSFM-Open, with solver solve_time measured per scenario under single-core CPU pinning.<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"standalone-cost-estimate\">Standalone cost estimate<\/h4>\n\n\n\n<p>When an exact solver round-trip is not required, GridSFM\u2019s predicted dispatch can be costed directly. In our test set, <strong>GridSFM-Open and DC-approximation fall in the same accuracy class<\/strong>: comparable means (DC 2.80%, GridSFM 3.41%), comparable medians (DC 1.81% vs GridSFM 2.23%), and overlapping per-scenario distributions across two decades of cost gap (Figure 6). They have complementary failure modes rather than one dominating the other.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"936\" height=\"487\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-13.png\" alt=\"chart\" class=\"wp-image-1171488\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-13.png 936w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-13-300x156.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-13-768x400.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-13-240x125.png 240w\" sizes=\"auto, (max-width: 936px) 100vw, 936px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 6.<\/strong> Per-scenario cost-gap distribution from AC-OPF ground truth: DC-approximation (blue) and GridSFM (green) across the 54-grid GridSFM-Open benchmark. Filled KDE + per-scenario dots underneath; light dashed lines mark mean (\u2013) and median (:). DC: mean 2.8%, median 1.81%, <5% gap on 90% of scenarios. GridSFM: mean 3.41%, median 2.23%, <5% gap on 90% of scenarios. The two distributions overlap heavily in the body \u2014 methods are in the same accuracy class with complementary failure modes. Reference dashed line at 5%.<\/figcaption><\/figure>\n\n\n\n<p>Both distributions look the same in shape: a single peak in the 2\u20133% gap range, with the bulk of scenarios under 5% and a small tail of outliers extending out into the >25% range. The outlier tails come from different sources: DC fails on grids where its no-reactive linearization is structurally wrong (case1803_snem and a handful of meshed transmission grids); GridSFM\u2019s outliers are concentrated on a few of our open sourced grids whose AC-OPF reference itself required <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/abs\/2605.04289\" target=\"_blank\" rel=\"noopener noreferrer\"><em>additional constraint relaxation to become feasible<\/em><span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, so the ground-truth target on those grids is noisier and the gap partly reflects reference-side instability. The two limitations close along orthogonal axes: DC\u2019s ceiling is fixed by the linearization and does not improve with more data or compute; GridSFM\u2019s tail closes with cleaner reference labels and more training data on those grid families.<\/p>\n\n\n\n<p>The differentiating value of&nbsp;GridSFM&nbsp;is therefore not the standalone cost number,&nbsp;but that&nbsp;GridSFM&nbsp;produces a full AC operating point including voltages and reactive power. This allows operators to directly assess the state of the grid.&nbsp;This is important since the feasibility and security of a system is often&nbsp;determined&nbsp;by the voltage and reactive power limits,&nbsp;but neither are&nbsp;considered&nbsp;in&nbsp;DC-OPF.&nbsp;&nbsp;At the same time, the operating point&nbsp;also&nbsp;enables the warm-start workflow, as we&nbsp;describe&nbsp;next.<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"warm-start-handoff\">Warm-start handoff<\/h4>\n\n\n\n<p>An AC-OPF solver works by iteratively refining an initial guess of the operating point until the optimality conditions are satisfied, and the number of&nbsp;refinement iterations it needs depends directly on how close the initial guess starts to the true optimum: a poor starting point can require thousands&nbsp;of iterations, a near-optimal one only a couple. A&nbsp;<strong>cold start<\/strong>&nbsp;(also known as a flat start)&nbsp;sets&nbsp;voltage&nbsp;magnitude&nbsp;to&nbsp;1.0 per unit&nbsp;and angle to&nbsp;zero&nbsp;&nbsp;on&nbsp;every bus,&nbsp;so&nbsp;the solver does the full amount of work. A&nbsp;<strong>warm&nbsp;start<\/strong>&nbsp;replaces&nbsp;that generic value with a closer estimate&nbsp;to make the solver converge faster.&nbsp;DC-approximation warm-start solves the&nbsp;linearized DC-OPF version of the problem first and seeds the AC solver with that solution.&nbsp;Whereas,&nbsp;GridSFM&nbsp;warm-start&nbsp;runs a single forward pass through the model&nbsp;and seeds the solver with its predicted voltage angles and active dispatch. The absolute ceiling on how much any warm-start can help is what we call&nbsp;the&nbsp;<strong>GT (ground-truth) ceiling<\/strong>: we run the full AC-OPF solve once at high precision to find the true optimum, then re-run the solver with that exact&nbsp;solution as the&nbsp;warm start seed.&nbsp;This&nbsp;is the practical&nbsp;limit&nbsp;on&nbsp;solving&nbsp;time and therefore the&nbsp;ceiling on&nbsp;speedup.&nbsp;<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"936\" height=\"486\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-14.png\" alt=\"diagram\" class=\"wp-image-1171490\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-14.png 936w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-14-300x156.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-14-768x399.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/image-14-240x125.png 240w\" sizes=\"auto, (max-width: 936px) 100vw, 936px\" \/><figcaption class=\"wp-element-caption\"><strong><em>Figure 7.<\/em><\/strong><em>&nbsp;Warm-start speedup over AC-OPF cold start, across the 54-grid test set (log-scale&nbsp;x axis).&nbsp;GridSFM&nbsp;(green,&nbsp;sits cleanly right of the cold-start reference)&nbsp;achieves a&nbsp;geomean&nbsp;speedup of&nbsp;1.66\u00d7,&nbsp;<\/em><strong><em>and&nbsp;outperforms cold start&nbsp;<\/em><\/strong><em>on&nbsp;41&nbsp;of&nbsp;54 grids&nbsp;; DC-approximation (blue) achieves a&nbsp;geomean&nbsp;speedup of&nbsp;1.04\u00d7&nbsp;and improves performance&nbsp;on 34&nbsp;of&nbsp;54 grids; the GT ceiling (gold,&nbsp;&nbsp;geomean&nbsp;2.72\u00d7)&nbsp;is the upper bound on warm-start headroom. Each&nbsp;method&#8217;s&nbsp;ratio is computed within&nbsp;<\/em><strong><em>the same<\/em><\/strong><em>&nbsp;Julia process to remove cross-run timing noise.&nbsp;<\/em><\/figcaption><\/figure>\n\n\n\n<p>Our profile showed that GridSFM warm-start is <strong>1.66\u00d7 faster than cold start<\/strong> and <strong>1.59\u00d7 faster than DC-approximation warm-start<\/strong> (geometric means across the 54 grids test scenarios) and is faster than both baselines on 41 of 54 grids. The largest per-grid speedups exceed 7\u00d7 over cold on the meshed transmission grids (Texas2k summer-peak, case2742_goc). DC-approximation warm-start, by contrast, is a wash on average across this broader grid mix (geomean 1.04\u00d7 vs cold), DC saves on AC iterations on some grids and spends them rebuilding voltage\/reactive on others.<\/p>\n\n\n\n<p>The gap between the GridSFM distribution in Figure 7 and the GT-ceiling distribution (2.72\u00d7 geomean) can be closed by improving GridSFM\u2019s residual reactive-power and voltage prediction error, both targeted by the next release.<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"generalization\">Generalization<\/h4>\n\n\n\n<p>We tested whether GridSFM-Open acts like a true foundation model by running it on a grid it had never seen before: the 6,470-bus case6470_rte from <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/abs\/2406.07234\" target=\"_blank\" rel=\"noopener noreferrer\"><em>OPFData<\/em><span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, about 1.4\u00d7 larger than any grid in training.<\/p>\n\n\n\n<p>In a zero-shot setting, performance drops as expected. Cost error increases from 3.35% in-sample to about 14% on the new grid. Voltage predictions capture only about 27% of the true variation and appear nearly flat. The feasibility classifier flags every scenario as infeasible. Even so, the model still preserves the correct ordering of costs across scenarios.<\/p>\n\n\n\n<p>With light fine-tuning, performance recovers quickly. After 10 epochs on 1,000 scenarios, cost error falls to 1.12%, voltage variation reaches 91% of the true signal, and feasibility detection becomes nearly perfect. An N-1 contingency split that was fully held out during fine-tuning matches the full-topology results within 0.2 percentage points on all metrics, showing that adaptation transfers across contingencies.<\/p>\n\n\n\n<p>The model adapts even with very limited data. With just 10 scenarios, cost errors are 1.76% and feasibility detection exceeds 90%, with strong results already on cost and active power dispatch. Voltage magnitude is slower to recover and needs closer to 1,000 scenarios (see Table 1).<\/p>\n\n\n\n<p>This test showed that GridSFM-Open already captures AC-OPF physics during pre-training. Adapting to a new grid is mostly a matter of calibration rather than relearning. The released checkpoint can therefore serve as a practical starting point for users to fine-tune on their own topology and tasks.<\/p>\n\n\n\n<figure class=\"wp-block-table aligncenter\"><table class=\"has-fixed-layout\"><thead><tr><th>Fine-tune scenarios<\/th><th>Cost error<\/th><th>Feasibility Detection<\/th><\/tr><\/thead><tbody><tr><td>0 (0-shot)<\/td><td>14%<\/td><td>0 (Collapsed)<\/td><\/tr><tr><td>10<\/td><td>1.76%<\/td><td>92%<\/td><\/tr><tr><td>100<\/td><td>0.88%<\/td><td>97%<\/td><\/tr><tr><td>1000<\/td><td>1.12%<\/td><td>99%<\/td><\/tr><\/tbody><\/table><figcaption class=\"wp-element-caption\">Table 1: Few-shot fine-tuning of GridSFM-Open on case6470_rte (held-out test split, 10 epochs per row): even ~10 scenarios already give useful cost and feasibility predictions.<\/figcaption><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"looking-ahead\">Looking ahead<\/h2>\n\n\n\n<p>Active directions for the next release:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Generalization.<\/strong> Tighter accuracy on grids and operating conditions outside the training mix. The current out-of-distribution analysis is in the white paper.<\/li>\n\n\n\n<li><strong>Continued accuracy improvements<\/strong> across all prediction channels, narrowing the residual gap between Figure 7\u2019s GridSFM distribution and the gold GT-ceiling.<\/li>\n\n\n\n<li><strong>Multi-snapshot extensions.<\/strong> Unit commitment (discrete on\/off generator decisions across time), weather-conditioned scenario generation, dynamic-stability surrogates.<\/li>\n<\/ul>\n\n\n\n<p>We&nbsp;previously&nbsp;released the&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/huggingface.co\/datasets\/microsoft\/GridSFM_US_power_grid\" target=\"_blank\" rel=\"noopener noreferrer\"><em>GridSFM_US&nbsp;_Powergrid_dataset<\/em><span class=\"sr-only\"> (opens in new tab)<\/span><\/a>.&nbsp;This release adds the first open AC-OPF model that supports multiple grid topologies, completing a stack of open topology data, open code, and open weights for ML-driven grid simulation and planning. We see it as a starting point for the community to build richer simulators, planning workflows, and decision-support tools without re-creating the data or the model from scratch. The applications we expect to see most leverage from are the ones where the cost of a single solve has historically forced cherry-picking: contingency screening, transmission expansion planning, demand-siting analysis, and resilience studies under extreme weather.&nbsp;<\/p>\n\n\n\n<p>Everything in the GridSFM-Open tier is released for research use today:<\/p>\n\n\n\n<div class=\"wp-block-buttons is-content-justification-center is-content-justification-center is-layout-flex wp-container-core-buttons-is-layout-16018d1d wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button is-style-fill-github\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"http:\/\/github.com\/microsoft\/gridSFM\">GitHub<\/a><\/div>\n\n\n\n<div class=\"wp-block-button is-style-fill\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/huggingface.co\/microsoft\/GridSFM_Open\">Hugging Face<\/a><\/div>\n\n\n\n<div class=\"wp-block-button is-style-outline is-style-outline--1\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/gridsfm-a-foundation-model-for-ac-optimal-power-flow\/\">White Paper<\/a><\/div>\n\n\n\n<div class=\"wp-block-button is-style-fill\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/project\/gridfm\/\">Project Page<\/a><\/div>\n<\/div>\n\n\n\n<p><strong>A note on GridSFM-Premier.<\/strong> The larger production-scale tier is not part of this open release. If you are interested in evaluating it, collaborating with us, or otherwise getting access, please contact us at <a href=\"mailto:gridFM@microsoft.com\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>gridFM@microsoft.com<\/strong><\/a>.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Introducing GridSFM, a small foundation model that can predict AC optimal power flow in milliseconds, boosting efficiency and unlocking cost savings. Learn how GridSFM gives grid operators direct visibility into congestion, stability, and system health.<\/p>\n","protected":false},"author":44124,"featured_media":1171570,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":null,"msr_hide_image_in_river":0,"footnotes":""},"categories":[1],"tags":[],"research-area":[13556,13548,13547],"msr-region":[],"msr-event-type":[],"msr-locale":[268875],"msr-post-option":[243984],"msr-impact-theme":[],"msr-promo-type":[],"msr-podcast-series":[],"class_list":["post-1171470","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-research-blog","msr-research-area-artificial-intelligence","msr-research-area-economics","msr-research-area-systems-and-networking","msr-locale-en_us","msr-post-option-blog-homepage-featured"],"msr_event_details":{"start":"","end":"","location":""},"podcast_url":"","podcast_episode":"","msr_research_lab":[199565,1161007],"msr_impact_theme":[],"related-publications":[],"related-downloads":[],"related-videos":[],"related-academic-programs":[],"related-groups":[],"related-projects":[1158675],"related-events":[],"related-researchers":[{"type":"user_nicename","value":"Weiwei Yang","user_id":40138,"display_name":"Weiwei Yang","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/weiwya\/\" aria-label=\"Visit the profile page for Weiwei Yang\">Weiwei Yang<\/a>","is_active":false,"last_first":"Yang, Weiwei","people_section":0,"alias":"weiwya"},{"type":"user_nicename","value":"Andrea Britto Mattos Lima","user_id":42393,"display_name":"Andrea Britto Mattos Lima","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/andreabri\/\" aria-label=\"Visit the profile page for Andrea Britto Mattos Lima\">Andrea Britto Mattos Lima<\/a>","is_active":false,"last_first":"Britto Mattos Lima, Andrea","people_section":0,"alias":"andreabri"},{"type":"user_nicename","value":"Thiago Vallin Spina","user_id":42246,"display_name":"Thiago Vallin Spina","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/tvallinspina\/\" aria-label=\"Visit the profile page for Thiago Vallin Spina\">Thiago Vallin Spina<\/a>","is_active":false,"last_first":"Vallin Spina, Thiago","people_section":0,"alias":"tvallinspina"},{"type":"user_nicename","value":"Spencer Fowers","user_id":33581,"display_name":"Spencer Fowers","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/sfowers\/\" aria-label=\"Visit the profile page for Spencer Fowers\">Spencer Fowers<\/a>","is_active":false,"last_first":"Fowers, Spencer","people_section":0,"alias":"sfowers"},{"type":"guest","value":"baosen-zhang","user_id":"1171051","display_name":"Baosen   Zhang","author_link":"<a href=\"https:\/\/people.ece.uw.edu\/zhang_baosen\/index.html\" aria-label=\"Visit the profile page for Baosen   Zhang\">Baosen   Zhang<\/a>","is_active":true,"last_first":"Zhang, Baosen  ","people_section":0,"alias":"baosen-zhang"}],"msr_type":"Post","featured_image_thumbnail":"<img width=\"960\" height=\"540\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-960x540.jpg\" class=\"img-object-cover\" alt=\"Three white line icons\u2014a transmission tower, a lightning bolt, and a stopwatch\u2014displayed on a teal-to-green gradient background with a subtle textured pattern.\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1-1280x720.jpg 1280w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/05\/GridSFM-BlogHeroFeature-1400x788-1.jpg 1400w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/>","byline":"","formattedDate":"May 13, 2026","formattedExcerpt":"Introducing GridSFM, a small foundation model that can predict AC optimal power flow in milliseconds, boosting efficiency and unlocking cost savings. Learn how GridSFM gives grid operators direct visibility into congestion, stability, and system health.","locale":{"slug":"en_us","name":"English","native":"","english":"English"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/1171470","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/44124"}],"replies":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/comments?post=1171470"}],"version-history":[{"count":27,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/1171470\/revisions"}],"predecessor-version":[{"id":1171860,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/1171470\/revisions\/1171860"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1171570"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1171470"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/categories?post=1171470"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/tags?post=1171470"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1171470"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=1171470"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=1171470"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1171470"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1171470"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1171470"},{"taxonomy":"msr-promo-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-promo-type?post=1171470"},{"taxonomy":"msr-podcast-series","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-podcast-series?post=1171470"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}