{"id":1158675,"date":"2026-01-06T09:43:03","date_gmt":"2026-01-06T17:43:03","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-project&#038;p=1158675"},"modified":"2026-02-02T15:22:51","modified_gmt":"2026-02-02T23:22:51","slug":"gridfm","status":"publish","type":"msr-project","link":"https:\/\/www.microsoft.com\/en-us\/research\/project\/gridfm\/","title":{"rendered":"GridFM"},"content":{"rendered":"<section class=\"mb-3 moray-highlight\">\n\t<div class=\"card-img-overlay mx-lg-0\">\n\t\t<div class=\"card-background  has-background- card-background--full-bleed\">\n\t\t\t<img loading=\"lazy\" decoding=\"async\" width=\"1536\" height=\"1024\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Designer.png\" class=\"attachment-full size-full\" alt=\"AI backkground giving a sense of power grids and foundtaional models\" style=\"object-position: 66% 43%\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Designer.png 1536w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Designer-300x200.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Designer-1024x683.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Designer-768x512.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Designer-240x160.png 240w\" sizes=\"auto, (max-width: 1536px) 100vw, 1536px\" \/>\t\t<\/div>\n\t\t<!-- Foreground -->\n\t\t<div class=\"card-foreground d-flex mt-md-n5 my-lg-5 px-g px-lg-0\">\n\t\t\t<!-- Container -->\n\t\t\t<div class=\"container d-flex mt-md-n5 my-lg-5 \">\n\t\t\t\t<!-- Card wrapper -->\n\t\t\t\t<div class=\"w-100 w-lg-col-5\">\n\t\t\t\t\t<!-- Card -->\n\t\t\t\t\t<div class=\"card material-md-card py-5 px-md-5\">\n\t\t\t\t\t\t<div class=\"card-body \">\n\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\n\n<h1 class=\"wp-block-heading\" id=\"gridfm\">GridFM<\/h1>\n\n\n\n<p>Small foundation models for the electric grid<\/p>\n\n\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\t<\/div>\n<\/section>\n\n\n\n\n\n<p>GridFM is a Microsoft Research initiative to build a <strong>foundation model (FM) for electric power grids<\/strong>, applying modern AI methods\u2014similar to large language\/weather models\u2014to complex grid physics.<\/p>\n\n\n\n<p>Traditional power\u2011flow solvers (like AC\u2011OPF) are <strong>accurate but extremely slow<\/strong>, taking minutes to hours on real-world grids with tens of thousands of components. As power systems grow more volatile due to datacenter expansion, renewable variability, electrification, and extreme weather, grid operators need <strong>fast, scalable, and generalizable models<\/strong> to evaluate thousands of scenarios in real time.<\/p>\n\n\n\n<div class=\"wp-block-media-text has-video  has-vertical-margin-small  has-vertical-padding-none  has-media-on-the-right is-stacked-on-mobile is-vertically-aligned-top\"><div class=\"wp-block-media-text__content\">\n<p>GridFM aims to deliver exactly that:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Robust tools for planning, reliability analysis, and emergency management<\/strong><\/li>\n\n\n\n<li><strong>Rapid inference<\/strong> for operational decision\u2011making<\/li>\n\n\n\n<li><strong>Physics\u2011informed modeling<\/strong> with high numerical fidelity<\/li>\n\n\n\n<li><strong>Generalized representations<\/strong> that can be fine\u2011tuned to specific grid topologies<\/li>\n<\/ul>\n<\/div><figure class=\"wp-block-media-text__media video-wrapper\"><iframe class=\"media-text__video\" src=\"https:\/\/www.youtube-nocookie.com\/embed\/k-fTkg3x-_U?enablejsapi=1&rel=0\" frameborder=\"0\" allow=\"accelerometer; autoplay; encrypted-media; gyroscope; picture-in-picture\" allowfullscreen><\/iframe><\/figure><\/div>\n\n\n\n<div style=\"height:30px\" aria-hidden=\"true\" class=\"wp-block-spacer\"><\/div>\n\n\n\n\n\n<h2 class=\"wp-block-heading\" id=\"motivation-behind-the-research\">Motivation behind the research<\/h2>\n\n\n\n<div class=\"wp-block-media-text has-vertical-margin-small  has-vertical-padding-none  is-stacked-on-mobile is-vertically-aligned-top\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"904\" height=\"581\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-16-110540.png\" alt=\"example model of Texas power grid (ERCOT)\" class=\"wp-image-1158695 size-full\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-16-110540.png 904w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-16-110540-300x193.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-16-110540-768x494.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-16-110540-240x154.png 240w\" sizes=\"auto, (max-width: 904px) 100vw, 904px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p>Modern electric grids are becoming increasingly complex. Rapid growth in hyperscale datacenters, variable renewable generation, transportation electrification, and more frequent extreme weather events are placing unprecedented stress on grid operations. These pressures require operators to evaluate many more scenarios far faster than traditional planning tools can support.<\/p>\n\n\n\n<p>Today\u2019s backbone method for grid analysis, AC Optimal Power Flow (AC OPF), offers high physical accuracy but is computationally complex and very slow. On realistic, large-scale networks, a single OPF solve can take minutes to hours. This makes it impossible to analyze the thousands of contingencies needed during fast-moving operational conditions.<\/p>\n<\/div><\/div>\n\n\n\n<div class=\"wp-block-media-text has-vertical-margin-small  has-vertical-padding-none  has-media-on-the-right is-stacked-on-mobile is-vertically-aligned-top\" style=\"grid-template-columns:auto 40%\"><div class=\"wp-block-media-text__content\">\n<p>GridFM is motivated by a clear insight: foundation models have transformed fields such as language, vision, and weather forecasting, and can create the same step\u2011change for power systems. By learning from large corpora of grid states and physics-based simulations, GridFM captures the structure and dynamics of the electric grid. It can then be rapidly fine\u2011tuned to any specific network using minimal labeled data.<\/p>\n\n\n\n<p>This enables grid operators, utilities, and researchers to perform accurate, physics\u2011aware inference in milliseconds, unlocking real\u2011time contingency analysis, adaptive control, and large-scale planning studies. The speed and generality of GridFM open the door to entirely new reliability and operational workflows.<\/p>\n\n\n\n<p>GridFM bridges the gap between the demands of a modern electric grid and the limitations of classical tools. It provides a unified, fast, and generalizable modeling capability that supports grid resilience, flexibility, and sustainability for the decades ahead.<\/p>\n<\/div><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"924\" height=\"1024\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-Topology-edited-924x1024.png\" alt=\"GridFM Topology diagram\" class=\"wp-image-1158697 size-full\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-Topology-edited-924x1024.png 924w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-Topology-edited-271x300.png 271w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-Topology-edited-768x851.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-Topology-edited-1386x1536.png 1386w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-Topology-edited-162x180.png 162w\" sizes=\"auto, (max-width: 924px) 100vw, 924px\" \/><\/figure><\/div>\n\n\n\n<p><\/p>\n\n\n\n<p><\/p>\n\n\n\n\n\n<h2 class=\"wp-block-heading\" id=\"gridfm-features\">GridFM features<\/h2>\n\n\n\n<p>GridFM brings a new class of modeling capabilities to power\u2011system planning and operations by combining physics, data, and foundation\u2011model architectures. Instead of relying on slow, case\u2011by\u2011case numerical solvers, GridFM provides a fast, generalizable representation of the electric grid that can be adapted to different networks, operating conditions, and study objectives<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"a-foundation-model-trained-on-grid-physics\">A foundation model trained on grid physics<\/h3>\n\n\n\n<p>At its core, GridFM learns the structure and behavior of power networks from large collections of grid states and physics\u2011based simulations. This enables the model to deliver <strong>accurate, physically consistent inferences in milliseconds<\/strong>, supporting workflows that previously required minutes or hours of computation.<br>The model learns fundamental relationships among:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Voltage magnitudes<\/li>\n\n\n\n<li>Phase angles<\/li>\n\n\n\n<li>Line flows & thermal limits<\/li>\n\n\n\n<li>Generator dispatch<\/li>\n\n\n\n<li>Load variations<\/li>\n\n\n\n<li>Topological or cost perturbations<\/li>\n<\/ul>\n\n\n\n<p>This FM becomes a <strong>general-purpose, reusable base model<\/strong> for many downstream tasks.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"fine-tuning-for-specialized-applications\">Fine\u2011tuning for specialized applications<\/h3>\n\n\n\n<p>Because the model is trained as a general foundation model, GridFM can be <strong>rapidly fine\u2011tuned<\/strong> to reflect local topologies, operational practices, or proprietary datasets.  Organizations can adapt the model to their proprietary networks using <strong>limited labeled data<\/strong>, enabling:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Custom OPF solvers<\/li>\n\n\n\n<li>Islanding analysis<\/li>\n\n\n\n<li>Dispatch optimization<\/li>\n\n\n\n<li>Planning studies<\/li>\n\n\n\n<li>Emergency operations support<\/li>\n\n\n\n<li>Real-time reliability evaluation<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"fast-and-accurate-inference\">Fast and accurate inference<\/h3>\n\n\n\n<p>These capabilities make GridFM a unified modeling layer for the modern electric grid.  GridFM targets <em>orders\u2011of\u2011magnitude faster<\/em> performance than classical solvers, enabling:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Online optimization under uncertainty<\/li>\n\n\n\n<li>Real-time contingency ranking<\/li>\n\n\n\n<li>Rapid n\u20111\/n\u2011k analysis<\/li>\n\n\n\n<li>Multi-scenario planning<\/li>\n<\/ul>\n\n\n\n<div style=\"height:30px\" aria-hidden=\"true\" class=\"wp-block-spacer\"><\/div>\n\n\n\n\n\n<h2 class=\"wp-block-heading\" id=\"technical-innovations\">Technical innovations<\/h2>\n\n\n\n<p>The development of GridFM has led to the following technical innovations:<\/p>\n\n\n\n<div class=\"wp-block-columns are-vertically-aligned-top is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-vertically-aligned-top is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:66.66%\">\n<h3 class=\"wp-block-heading h4\" id=\"hybrid-sl-rl-training-1\">Hybrid SL + RL training<\/h3>\n\n\n\n<p>The GridFM team uses a hybrid pipeline that includes:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Supervised Learning (SL)<\/strong> using classical solver ground truth<\/li>\n\n\n\n<li><strong>Reinforcement Learning (GRPO)<\/strong> to optimize performance under physics constraints<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading h4\" id=\"physics-informed-loss-functions\">Physics\u2011informed loss functions<\/h3>\n\n\n\n<p>Experiments incorporate constraints such as:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Kirchhoff\u2019s Current Law (KCL)<\/li>\n\n\n\n<li>Thermal limits<\/li>\n\n\n\n<li>Flow constraints<\/li>\n\n\n\n<li>Economic objectives<\/li>\n\n\n\n<li>Feasibility metrics (PQE)<\/li>\n<\/ul>\n\n\n\n<p>These physics-informed loss functions allow the model to <em>learn<\/em> physics and how the laws impact optimal power<\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-vertically-aligned-top is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:33.33%\">\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1084\" height=\"1203\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-RL-Modeling.png\" alt=\"GridFM diagram showing the physical process\" class=\"wp-image-1158700\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-RL-Modeling.png 1084w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-RL-Modeling-270x300.png 270w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-RL-Modeling-923x1024.png 923w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-RL-Modeling-768x852.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/GridFM-RL-Modeling-162x180.png 162w\" sizes=\"auto, (max-width: 1084px) 100vw, 1084px\" \/><\/figure>\n<\/div>\n<\/div>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"integration-with-established-ac-opf-solvers\">Integration with established AC-OPF solvers<\/h3>\n\n\n\n<p>The output of GridFM is not a guaranteed optimal solution, but it provides a great &#8220;warm start&#8221; for existing solvers.  This means that utility providers can continue to rely on their certified OPF solvers but improve the solving speed by using a GridFM solution as a starting point.  <\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"power-grid-topology-modeling\">Power grid topology modeling<\/h3>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:60%\">\n<p>Accurate, up-to-date grid topology is largely unavailable or privately controlled.  Our approach to develop accurate data for testing and training GridFM was to fuse multi-modal public and proprietary data in order to reconstruct the grid with as high fidelity as possible.<\/p>\n\n\n\n<p>Utilizing GIS, generation and demand time-series, datacenter metadata, demographics\/zoning, and satellite imagery from OpenStreetMap, and U.S. EIA we construct a heterograph with busses, lines, transformers, generators, and loads for training and simulation.<\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:40%\">\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1141\" height=\"954\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/gridfm-topology-modeling-pipeline.png\" alt=\"diagram\" class=\"wp-image-1158699\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/gridfm-topology-modeling-pipeline.png 1141w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/gridfm-topology-modeling-pipeline-300x251.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/gridfm-topology-modeling-pipeline-1024x856.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/gridfm-topology-modeling-pipeline-768x642.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/12\/gridfm-topology-modeling-pipeline-215x180.png 215w\" sizes=\"auto, (max-width: 1141px) 100vw, 1141px\" \/><\/figure>\n<\/div>\n<\/div>\n\n\n","protected":false},"excerpt":{"rendered":"<p>Small foundation models for the electric grid GridFM is a Microsoft Research initiative to build a foundation model (FM) for electric power grids, applying modern AI methods\u2014similar to large language\/weather models\u2014to complex grid physics. Traditional power\u2011flow solvers (like AC\u2011OPF) are accurate but extremely slow, taking minutes to hours on real-world grids with tens of thousands [&hellip;]<\/p>\n","protected":false},"featured_media":1158679,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","footnotes":""},"research-area":[13556,13548],"msr-locale":[268875],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-1158675","msr-project","type-msr-project","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-research-area-economics","msr-locale-en_us","msr-archive-status-active"],"msr_project_start":"2025-08-01","related-publications":[],"related-downloads":[],"related-videos":[],"related-groups":[901101],"related-events":[],"related-opportunities":[],"related-posts":[],"related-articles":[],"tab-content":[],"slides":[],"related-researchers":[{"type":"user_nicename","display_name":"Andrea Britto Mattos Lima","user_id":42393,"people_section":"Related people","alias":"andreabri"},{"type":"user_nicename","display_name":"Spencer Fowers","user_id":33581,"people_section":"Related people","alias":"sfowers"},{"type":"user_nicename","display_name":"Kate Lytvynets","user_id":38073,"people_section":"Related people","alias":"kalytv"},{"type":"user_nicename","display_name":"Thiago Vallin Spina","user_id":42246,"people_section":"Related people","alias":"tvallinspina"},{"type":"user_nicename","display_name":"Weiwei Yang","user_id":40138,"people_section":"Related people","alias":"weiwya"}],"msr_research_lab":[199565,1161007],"msr_impact_theme":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1158675","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-project"}],"version-history":[{"count":22,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1158675\/revisions"}],"predecessor-version":[{"id":1161274,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1158675\/revisions\/1161274"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1158679"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1158675"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1158675"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1158675"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1158675"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=1158675"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}