{"id":378,"date":"2024-04-30T10:38:18","date_gmt":"2024-04-30T08:38:18","guid":{"rendered":"https:\/\/websites.fraunhofer.de\/rl4ces\/?page_id=378"},"modified":"2025-05-27T09:08:43","modified_gmt":"2025-05-27T07:08:43","slug":"publikationen","status":"publish","type":"page","link":"https:\/\/rl4ces.de\/en\/publikationen\/","title":{"rendered":"Publications"},"content":{"rendered":"<div id=\"trailimageid\"><img decoding=\"async\" id=\"ttimg\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/loading.gif\"><\/div> <div class=\"shariff shariff-align-flex-start shariff-widget-align-center\" style=\"display:none\"><ul class=\"shariff-buttons theme-round orientation-horizontal buttonsize-medium\"><li class=\"shariff-button linkedin shariff-nocustomcolor\" style=\"background-color:#1488bf;border-radius:10%\"><a href=\"https:\/\/www.linkedin.com\/sharing\/share-offsite\/?url=https%3A%2F%2Frl4ces.de%2Fen%2Fpublikationen%2F\" title=\"Bei LinkedIn teilen\" aria-label=\"Bei LinkedIn teilen\" role=\"button\" rel=\"noopener nofollow\" class=\"shariff-link\" style=\";border-radius:10%; background-color:#0077b5; color:#fff\" target=\"_blank\"><span class=\"shariff-icon\" style=\"\"><svg width=\"32px\" height=\"20px\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewbox=\"0 0 27 32\"><path fill=\"#0077b5\" d=\"M6.2 11.2v17.7h-5.9v-17.7h5.9zM6.6 5.7q0 1.3-0.9 2.2t-2.4 0.9h0q-1.5 0-2.4-0.9t-0.9-2.2 0.9-2.2 2.4-0.9 2.4 0.9 0.9 2.2zM27.4 18.7v10.1h-5.9v-9.5q0-1.9-0.7-2.9t-2.3-1.1q-1.1 0-1.9 0.6t-1.2 1.5q-0.2 0.5-0.2 1.4v9.9h-5.9q0-7.1 0-11.6t0-5.3l0-0.9h5.9v2.6h0q0.4-0.6 0.7-1t1-0.9 1.6-0.8 2-0.3q3 0 4.9 2t1.9 6z\"\/><\/svg><\/span><\/a><\/li><li class=\"shariff-button xing shariff-nocustomcolor\" style=\"background-color:#29888a;border-radius:10%\"><a href=\"https:\/\/www.xing.com\/spi\/shares\/new?url=https%3A%2F%2Frl4ces.de%2Fen%2Fpublikationen%2F\" title=\"Bei XING teilen\" aria-label=\"Bei XING teilen\" role=\"button\" rel=\"noopener nofollow\" class=\"shariff-link\" style=\";border-radius:10%; background-color:#126567; color:#fff\" target=\"_blank\"><span class=\"shariff-icon\" style=\"\"><svg width=\"32px\" height=\"20px\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewbox=\"0 0 25 32\"><path fill=\"#126567\" d=\"M10.7 11.9q-0.2 0.3-4.6 8.2-0.5 0.8-1.2 0.8h-4.3q-0.4 0-0.5-0.3t0-0.6l4.5-8q0 0 0 0l-2.9-5q-0.2-0.4 0-0.7 0.2-0.3 0.5-0.3h4.3q0.7 0 1.2 0.8zM25.1 0.4q0.2 0.3 0 0.7l-9.4 16.7 6 11q0.2 0.4 0 0.6-0.2 0.3-0.6 0.3h-4.3q-0.7 0-1.2-0.8l-6-11.1q0.3-0.6 9.5-16.8 0.4-0.8 1.2-0.8h4.3q0.4 0 0.5 0.3z\"\/><\/svg><\/span><\/a><\/li><li class=\"shariff-button facebook shariff-nocustomcolor\" style=\"background-color:#4273c8;border-radius:10%\"><a href=\"https:\/\/www.facebook.com\/sharer\/sharer.php?u=https%3A%2F%2Frl4ces.de%2Fen%2Fpublikationen%2F\" title=\"Bei Facebook teilen\" aria-label=\"Bei Facebook teilen\" role=\"button\" rel=\"nofollow\" class=\"shariff-link\" style=\";border-radius:10%; background-color:#3b5998; color:#fff\" target=\"_blank\"><span class=\"shariff-icon\" style=\"\"><svg width=\"32px\" height=\"20px\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewbox=\"0 0 18 32\"><path fill=\"#3b5998\" d=\"M17.1 0.2v4.7h-2.8q-1.5 0-2.1 0.6t-0.5 1.9v3.4h5.2l-0.7 5.3h-4.5v13.6h-5.5v-13.6h-4.5v-5.3h4.5v-3.9q0-3.3 1.9-5.2t5-1.8q2.6 0 4.1 0.2z\"\/><\/svg><\/span><\/a><\/li><li class=\"shariff-button twitter shariff-nocustomcolor\" style=\"background-color:#595959;border-radius:10%\"><a href=\"https:\/\/twitter.com\/share?url=https%3A%2F%2Frl4ces.de%2Fen%2Fpublikationen%2F&text=Publikationen\" title=\"Bei X teilen\" aria-label=\"Bei X teilen\" role=\"button\" rel=\"noopener nofollow\" class=\"shariff-link\" style=\";border-radius:10%; background-color:#000; color:#fff\" target=\"_blank\"><span class=\"shariff-icon\" style=\"\"><svg width=\"32px\" height=\"20px\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewbox=\"0 0 24 24\"><path fill=\"#000\" d=\"M14.258 10.152L23.176 0h-2.113l-7.747 8.813L7.133 0H0l9.352 13.328L0 23.973h2.113l8.176-9.309 6.531 9.309h7.133zm-2.895 3.293l-.949-1.328L2.875 1.56h3.246l6.086 8.523.945 1.328 7.91 11.078h-3.246zm0 0\"\/><\/svg><\/span><\/a><\/li><li class=\"shariff-button whatsapp shariff-nocustomcolor shariff-mobile\" style=\"background-color:#5cbe4a;border-radius:10%\"><a href=\"https:\/\/api.whatsapp.com\/send?text=https%3A%2F%2Frl4ces.de%2Fen%2Fpublikationen%2F%20Publikationen\" title=\"Bei Whatsapp teilen\" aria-label=\"Bei Whatsapp teilen\" role=\"button\" rel=\"noopener nofollow\" class=\"shariff-link\" style=\";border-radius:10%; background-color:#34af23; color:#fff\" target=\"_blank\"><span class=\"shariff-icon\" style=\"\"><svg width=\"32px\" height=\"20px\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewbox=\"0 0 32 32\"><path fill=\"#34af23\" d=\"M17.6 17.4q0.2 0 1.7 0.8t1.6 0.9q0 0.1 0 0.3 0 0.6-0.3 1.4-0.3 0.7-1.3 1.2t-1.8 0.5q-1 0-3.4-1.1-1.7-0.8-3-2.1t-2.6-3.3q-1.3-1.9-1.3-3.5v-0.1q0.1-1.6 1.3-2.8 0.4-0.4 0.9-0.4 0.1 0 0.3 0t0.3 0q0.3 0 0.5 0.1t0.3 0.5q0.1 0.4 0.6 1.6t0.4 1.3q0 0.4-0.6 1t-0.6 0.8q0 0.1 0.1 0.3 0.6 1.3 1.8 2.4 1 0.9 2.7 1.8 0.2 0.1 0.4 0.1 0.3 0 1-0.9t0.9-0.9zM14 26.9q2.3 0 4.3-0.9t3.6-2.4 2.4-3.6 0.9-4.3-0.9-4.3-2.4-3.6-3.6-2.4-4.3-0.9-4.3 0.9-3.6 2.4-2.4 3.6-0.9 4.3q0 3.6 2.1 6.6l-1.4 4.2 4.3-1.4q2.8 1.9 6.2 1.9zM14 2.2q2.7 0 5.2 1.1t4.3 2.9 2.9 4.3 1.1 5.2-1.1 5.2-2.9 4.3-4.3 2.9-5.2 1.1q-3.5 0-6.5-1.7l-7.4 2.4 2.4-7.2q-1.9-3.2-1.9-6.9 0-2.7 1.1-5.2t2.9-4.3 4.3-2.9 5.2-1.1z\"\/><\/svg><\/span><\/a><\/li><li class=\"shariff-button mailto shariff-nocustomcolor\" style=\"background-color:#a8a8a8;border-radius:10%\"><a href=\"mailto:?body=https%3A%2F%2Frl4ces.de%2Fen%2Fpublikationen%2F&subject=Publikationen\" title=\"Per E-Mail versenden\" aria-label=\"Per E-Mail versenden\" role=\"button\" rel=\"noopener nofollow\" class=\"shariff-link\" style=\";border-radius:10%; background-color:#999; color:#fff\"><span class=\"shariff-icon\" style=\"\"><svg width=\"32px\" height=\"20px\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewbox=\"0 0 32 32\"><path fill=\"#999\" d=\"M32 12.7v14.2q0 1.2-0.8 2t-2 0.9h-26.3q-1.2 0-2-0.9t-0.8-2v-14.2q0.8 0.9 1.8 1.6 6.5 4.4 8.9 6.1 1 0.8 1.6 1.2t1.7 0.9 2 0.4h0.1q0.9 0 2-0.4t1.7-0.9 1.6-1.2q3-2.2 8.9-6.1 1-0.7 1.8-1.6zM32 7.4q0 1.4-0.9 2.7t-2.2 2.2q-6.7 4.7-8.4 5.8-0.2 0.1-0.7 0.5t-1 0.7-0.9 0.6-1.1 0.5-0.9 0.2h-0.1q-0.4 0-0.9-0.2t-1.1-0.5-0.9-0.6-1-0.7-0.7-0.5q-1.6-1.1-4.7-3.2t-3.6-2.6q-1.1-0.7-2.1-2t-1-2.5q0-1.4 0.7-2.3t2.1-0.9h26.3q1.2 0 2 0.8t0.9 2z\"\/><\/svg><\/span><\/a><\/li><\/ul><\/div> \n <div class=\"bibsonomycsl_jump_list\">[<a class=\"bibsonomycsl_publications-headline-jumplabel\" href=\"#jmp_2026\" title=\"Goto 2026\">2026<\/a>] [<a class=\"bibsonomycsl_publications-headline-jumplabel\" href=\"#jmp_2025\" title=\"Goto 2025\">2025<\/a>] [<a class=\"bibsonomycsl_publications-headline-jumplabel\" href=\"#jmp_2024\" title=\"Goto 2024\">2024<\/a>] [<a class=\"bibsonomycsl_publications-headline-jumplabel\" href=\"#jmp_2023\" title=\"Goto 2023\">2023<\/a>]<\/div><ul class=\"bibsonomycsl_publications\">\n<\/ul>\n<a class=\"bibsonomycsl_publications-headline-anchor\" name=\"jmp_2026\"><\/a><h3 class=\"bibsonomycsl_publications-headline\" style=\"font-size: 1.1em; font-weight: bold;\">2026<\/h3>\n<ul class=\"bibsonomycsl_publications\"><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/inproceedings.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Hassouna, Mohamed, Clara Holzh\u00fcter, Malte Lehna, Matthijs de Jong, Jan Viebahn, Bernhard Sick, and Christoph Scholz. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">Learning Topology Actions for\u00a0Power Grid Control: A Graph-Based Soft-Label Imitation Learning Approach<\/span>\u201d<\/span>. In <i>Machine Learning and Knowledge Discovery in Databases. Applied Data Science Track and Demo Track<\/i>, edited by In\u00eas Dutra, Mykola Pechenizkiy, Paulo Cortez, Sepideh Pashami, Arian Pasquali, Nuno Moniz, Al\u00edpio M. Jorge, Carlos Soares, Pedro H. Abreu, and Jo\u00e3o Gama, 129\u2013146. Cham: Springer Nature Switzerland, 2026.<\/div>\n<\/div><span class=\"bibsonomycsl_export bibsonomycsl_abstract\"><a rel=\"abs-97f2bb4f3776c68bdaf9c7415370d699\"  href=\"#\">Abstract<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-97f2bb4f3776c68bdaf9c7415370d699\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-97f2bb4f3776c68bdaf9c7415370d699\" href=\"#\">EndNote<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_abstract\" style=\"display:none;\" id=\"abs-97f2bb4f3776c68bdaf9c7415370d699\">The rising proportion of renewable energy in the electricity mix introduces significant operational challenges for power grid operators. Effective power grid management demands adaptive decision-making strategies capable of handling dynamic conditions. With the increase in complexity, more and more Deep Learning (DL) approaches have been proposed to find suitable grid topologies for congestion management. In this work, we contribute to this research by introducing a novel Imitation Learning (IL) approach that leverages soft labels derived from simulated topological action outcomes, thereby capturing multiple viable actions per state. Unlike traditional IL methods that rely on hard labels to enforce a single optimal action, our method constructs soft labels that capture the effectiveness of actions that prove suitable in resolving grid congestion. To further enhance decision-making, we integrate Graph Neural Networks (GNNs) to encode the structural properties of power grids, ensuring that the topology-aware representations contribute to better agent performance. Our approach significantly outperforms its hard-label counterparts as well as state-of-the-art Deep Reinforcement Learning (DRL) baseline agents. Most notably, it achieves a 17% better performance compared to the greedy expert agent from which the imitation targets were derived.<\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-97f2bb4f3776c68bdaf9c7415370d699\"><p>@inproceedings{10.1007\/978-3-032-06129-4_8,<br\/>  abstract = {The rising proportion of renewable energy in the electricity mix introduces significant operational challenges for power grid operators. Effective power grid management demands adaptive decision-making strategies capable of handling dynamic conditions. With the increase in complexity, more and more Deep Learning (DL) approaches have been proposed to find suitable grid topologies for congestion management. In this work, we contribute to this research by introducing a novel Imitation Learning (IL) approach that leverages soft labels derived from simulated topological action outcomes, thereby capturing multiple viable actions per state. Unlike traditional IL methods that rely on hard labels to enforce a single optimal action, our method constructs soft labels that capture the effectiveness of actions that prove suitable in resolving grid congestion. To further enhance decision-making, we integrate Graph Neural Networks (GNNs) to encode the structural properties of power grids, ensuring that the topology-aware representations contribute to better agent performance. Our approach significantly outperforms its hard-label counterparts as well as state-of-the-art Deep Reinforcement Learning (DRL) baseline agents. Most notably, it achieves a 17% better performance compared to the greedy expert agent from which the imitation targets were derived.},<br\/>  address = {Cham},<br\/>  author = {Hassouna, Mohamed and Holzh\u00fcter, Clara and Lehna, Malte and de Jong, Matthijs and Viebahn, Jan and Sick, Bernhard and Scholz, Christoph},<br\/>  booktitle = {Machine Learning and Knowledge Discovery in Databases. Applied Data Science Track and Demo Track},<br\/>  editor = {Dutra, In\u00eas and Pechenizkiy, Mykola and Cortez, Paulo and Pashami, Sepideh and Pasquali, Arian and Moniz, Nuno and Jorge, Al\u00edpio M. and Soares, Carlos and Abreu, Pedro H. and Gama, Jo\u00e3o},<br\/>  keywords = {rl4ces},<br\/>  pages = {129--146},<br\/>  publisher = {Springer Nature Switzerland},<br\/>  title = {Learning Topology Actions for\u00a0Power Grid Control: A Graph-Based Soft-Label Imitation Learning Approach},<br\/>  year = 2026<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-97f2bb4f3776c68bdaf9c7415370d699\"><p>%0 Conference Paper<br\/>%1 10.1007\/978-3-032-06129-4_8<br\/>%A Hassouna, Mohamed<br\/>%A Holzh\u00fcter, Clara<br\/>%A Lehna, Malte<br\/>%A de Jong, Matthijs<br\/>%A Viebahn, Jan<br\/>%A Sick, Bernhard<br\/>%A Scholz, Christoph<br\/>%B Machine Learning and Knowledge Discovery in Databases. Applied Data Science Track and Demo Track<br\/>%C Cham<br\/>%D 2026<br\/>%E Dutra, In\u00eas<br\/>%E Pechenizkiy, Mykola<br\/>%E Cortez, Paulo<br\/>%E Pashami, Sepideh<br\/>%E Pasquali, Arian<br\/>%E Moniz, Nuno<br\/>%E Jorge, Al\u00edpio M.<br\/>%E Soares, Carlos<br\/>%E Abreu, Pedro H.<br\/>%E Gama, Jo\u00e3o<br\/>%I Springer Nature Switzerland<br\/>%P 129--146<br\/>%T Learning Topology Actions for\u00a0Power Grid Control: A Graph-Based Soft-Label Imitation Learning Approach<br\/>%X The rising proportion of renewable energy in the electricity mix introduces significant operational challenges for power grid operators. Effective power grid management demands adaptive decision-making strategies capable of handling dynamic conditions. With the increase in complexity, more and more Deep Learning (DL) approaches have been proposed to find suitable grid topologies for congestion management. In this work, we contribute to this research by introducing a novel Imitation Learning (IL) approach that leverages soft labels derived from simulated topological action outcomes, thereby capturing multiple viable actions per state. Unlike traditional IL methods that rely on hard labels to enforce a single optimal action, our method constructs soft labels that capture the effectiveness of actions that prove suitable in resolving grid congestion. To further enhance decision-making, we integrate Graph Neural Networks (GNNs) to encode the structural properties of power grids, ensuring that the topology-aware representations contribute to better agent performance. Our approach significantly outperforms its hard-label counterparts as well as state-of-the-art Deep Reinforcement Learning (DRL) baseline agents. Most notably, it achieves a 17% better performance compared to the greedy expert agent from which the imitation targets were derived.<br\/>%@ 978-3-032-06129-4<br\/><\/p><\/div><\/div><\/li>\n<\/ul>\n<a class=\"bibsonomycsl_publications-headline-anchor\" name=\"jmp_2025\"><\/a><h3 class=\"bibsonomycsl_publications-headline\" style=\"font-size: 1.1em; font-weight: bold;\">2025<\/h3>\n<ul class=\"bibsonomycsl_publications\"><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/inproceedings.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Holzh\u00fcter, Clara, Pawel Lytaev, Marcel Dipp, Mohamed Hassouna, Kurt Brendlinger, Jan Viebahn, Wiktor Gegelman, and Christian Merz. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">Graph Neural Networks for Grid Control: Prospects in AI-Assisted Transmission Grid Operation<\/span>\u201d<\/span>. In <i>ETG Kongress 2025<\/i>.<\/div>\n<\/div><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-85f35485589b70dba6f63a6bd5920293\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-85f35485589b70dba6f63a6bd5920293\" href=\"#\">EndNote<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-85f35485589b70dba6f63a6bd5920293\"><p>@inproceedings{holzhueter2025gnn4gc,<br\/>  author = {Holzh\u00fcter, Clara and Lytaev, Pawel and Dipp, Marcel and Hassouna, Mohamed and Brendlinger, Kurt and Viebahn, Jan and Gegelman, Wiktor and Merz, Christian},<br\/>  booktitle = {ETG Kongress 2025},<br\/>  keywords = {rl4ces},<br\/>  note = {Accepted for presentation, not yet published},<br\/>  title = {Graph Neural Networks for Grid Control: Prospects in AI-assisted Transmission Grid Operation},<br\/>  year = 2025<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-85f35485589b70dba6f63a6bd5920293\"><p>%0 Conference Paper<br\/>%1 holzhueter2025gnn4gc<br\/>%A Holzh\u00fcter, Clara<br\/>%A Lytaev, Pawel<br\/>%A Dipp, Marcel<br\/>%A Hassouna, Mohamed<br\/>%A Brendlinger, Kurt<br\/>%A Viebahn, Jan<br\/>%A Gegelman, Wiktor<br\/>%A Merz, Christian<br\/>%B ETG Kongress 2025<br\/>%D 2025<br\/>%T Graph Neural Networks for Grid Control: Prospects in AI-assisted Transmission Grid Operation<br\/><\/p><\/div><\/div><\/li><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/misc.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Hassouna, Mohamed, Clara Holzh\u00fcter, Malte Lehna, Matthijs de Jong, Jan Viebahn, Bernhard Sick, and Christoph Scholz. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">Learning Topology Actions for Power Grid Control: A Graph-Based Soft-Label Imitation Learning Approach<\/span>\u201d<\/span>. https:\/\/arxiv.org\/abs\/2503.15190.<\/div>\n<\/div><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/arxiv.org\/abs\/2503.15190\" target=\"_blank\">URL<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-8f85a07def2612a5989e18ad3c8a02f2\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-8f85a07def2612a5989e18ad3c8a02f2\" href=\"#\">EndNote<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-8f85a07def2612a5989e18ad3c8a02f2\"><p>@misc{hassouna2025,<br\/>  author = {Hassouna, Mohamed and Holzh\u00fcter, Clara and Lehna, Malte and de Jong, Matthijs and Viebahn, Jan and Sick, Bernhard and Scholz, Christoph},<br\/>  keywords = {rl4ces},<br\/>  note = {Accepted at ECML 2025. Preprint available at https:\/\/arxiv.org\/abs\/2503.15190.},<br\/>  title = {Learning Topology Actions for Power Grid Control: A Graph-Based Soft-Label Imitation Learning Approach},<br\/>  year = 2025<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-8f85a07def2612a5989e18ad3c8a02f2\"><p>%0 Generic<br\/>%1 hassouna2025<br\/>%A Hassouna, Mohamed<br\/>%A Holzh\u00fcter, Clara<br\/>%A Lehna, Malte<br\/>%A de Jong, Matthijs<br\/>%A Viebahn, Jan<br\/>%A Sick, Bernhard<br\/>%A Scholz, Christoph<br\/>%D 2025<br\/>%T Learning Topology Actions for Power Grid Control: A Graph-Based Soft-Label Imitation Learning Approach<br\/>%U https:\/\/arxiv.org\/abs\/2503.15190<br\/><\/p><\/div><\/div><\/li><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/misc.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Leyli abadi, Milad, Ricardo J. Bessa, Jan Viebahn, Daniel Boos, Clark Borst, Alberto Castagna, Ricardo Chavarriaga, et al. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">A Conceptual Framework for AI-Based Decision Systems in Critical Infrastructures<\/span>\u201d<\/span>. https:\/\/arxiv.org\/abs\/2504.16133.<\/div>\n<\/div><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/arxiv.org\/abs\/2504.16133\" target=\"_blank\">URL<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-5942b5295dd172e02e3f9ebb838ce6ee\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-5942b5295dd172e02e3f9ebb838ce6ee\" href=\"#\">EndNote<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-5942b5295dd172e02e3f9ebb838ce6ee\"><p>@misc{leyliabadi2025conceptualframeworkaibaseddecision,<br\/>  author = {Leyli abadi, Milad and Bessa, Ricardo J. and Viebahn, Jan and Boos, Daniel and Borst, Clark and Castagna, Alberto and Chavarriaga, Ricardo and Hassouna, Mohamed and Lemetayer, Bruno and Leto, Giulia and Marot, Antoine and Meddeb, Maroua and Meyer, Manuel and Schiaffonati, Viola and Schneider, Manuel and Waefler, Toni},<br\/>  keywords = {rl4ces},<br\/>  title = {A Conceptual Framework for AI-based Decision Systems in Critical Infrastructures},<br\/>  year = 2025<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-5942b5295dd172e02e3f9ebb838ce6ee\"><p>%0 Generic<br\/>%1 leyliabadi2025conceptualframeworkaibaseddecision<br\/>%A Leyli abadi, Milad<br\/>%A Bessa, Ricardo J.<br\/>%A Viebahn, Jan<br\/>%A Boos, Daniel<br\/>%A Borst, Clark<br\/>%A Castagna, Alberto<br\/>%A Chavarriaga, Ricardo<br\/>%A Hassouna, Mohamed<br\/>%A Lemetayer, Bruno<br\/>%A Leto, Giulia<br\/>%A Marot, Antoine<br\/>%A Meddeb, Maroua<br\/>%A Meyer, Manuel<br\/>%A Schiaffonati, Viola<br\/>%A Schneider, Manuel<br\/>%A Waefler, Toni<br\/>%D 2025<br\/>%T A Conceptual Framework for AI-based Decision Systems in Critical Infrastructures<br\/>%U https:\/\/arxiv.org\/abs\/2504.16133<br\/><\/p><\/div><\/div><\/li><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/inproceedings.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Holzh\u00fcter, Clara Juliane, Pawel Lytaev, Marcel Dipp, Mohamed Hassouna, Kurt Brendlinger, Jan Viebahn, Wiktor Gegelman, and Christian Merz. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">Graph Neural Networks for Grid Control: Prospects in AI-Assisted Transmission Grid Operation<\/span>\u201d<\/span>. In <i>Energietechnische Gesellschaft (ETG Kongress) 2025<\/i>, 874\u2013881. https:\/\/www.etg-kongress.com\/resource\/blob\/2385346\/9a8b59f27822512a0884d3cd0ec8e20e\/etg-kongress-2025-proceedings-link-data.pdf.<\/div>\n<\/div><span class=\"bibsonomycsl_export bibsonomycsl_abstract\"><a rel=\"abs-fdc7f4fbce01cd3f0309bd5cede9804c\"  href=\"#\">Abstract<\/a><\/span><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/www.etg-kongress.com\/resource\/blob\/2385346\/9a8b59f27822512a0884d3cd0ec8e20e\/etg-kongress-2025-proceedings-link-data.pdf\" target=\"_blank\">URL<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-fdc7f4fbce01cd3f0309bd5cede9804c\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-fdc7f4fbce01cd3f0309bd5cede9804c\" href=\"#\">EndNote<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_abstract\" style=\"display:none;\" id=\"abs-fdc7f4fbce01cd3f0309bd5cede9804c\">Transmission grid congestion management and outage planning are critical tasks in modern grid operation due to thenon-linear nature of power flows and the large-scale optimization challenges faced by operators. Traditionally, overloadsare addressed through generator redispatch, a costly and therefore suboptimal measure. In the project \"Graph NeuralNetworks for Grid Control\" (GNN4GC), we investigate alternative strategies, focusing on topological remedial actionsthat could minimize or even completely eliminate redispatch costs. Topology optimization, a core aspect of this project,presents significant challenges due to its combinatorial nature, requiring extensive computational resources for powerflow calculations. To address this, GNN4GC is split into three stages. In the first stage, we explore the use of GraphNeural Networks (GNNs) to accelerate these calculations and benchmark their performance against established tools likepandapower and a DC power flow solver developed by 50Hertz Transmission GmbH and TenneT TSO GmbH. In thesecond stage, we use Reinforcement Learning and other heuristics to select suitable topologies and solve the topologyoptimization problem. As a third stage, we test the respective agent on real-life grids to benchmark the methodology. Theaim of the final stage is to build a recommender system that can be used in a control room in the future.<\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-fdc7f4fbce01cd3f0309bd5cede9804c\"><p>@inproceedings{holzhuter2025graph,<br\/>  abstract = {Transmission grid congestion management and outage planning are critical tasks in modern grid operation due to thenon-linear nature of power flows and the large-scale optimization challenges faced by operators. Traditionally, overloadsare addressed through generator redispatch, a costly and therefore suboptimal measure. In the project \"Graph NeuralNetworks for Grid Control\" (GNN4GC), we investigate alternative strategies, focusing on topological remedial actionsthat could minimize or even completely eliminate redispatch costs. Topology optimization, a core aspect of this project,presents significant challenges due to its combinatorial nature, requiring extensive computational resources for powerflow calculations. To address this, GNN4GC is split into three stages. In the first stage, we explore the use of GraphNeural Networks (GNNs) to accelerate these calculations and benchmark their performance against established tools likepandapower and a DC power flow solver developed by 50Hertz Transmission GmbH and TenneT TSO GmbH. In thesecond stage, we use Reinforcement Learning and other heuristics to select suitable topologies and solve the topologyoptimization problem. As a third stage, we test the respective agent on real-life grids to benchmark the methodology. Theaim of the final stage is to build a recommender system that can be used in a control room in the future.},<br\/>  author = {Holzh\u00fcter, Clara Juliane and Lytaev, Pawel and Dipp, Marcel and Hassouna, Mohamed and Brendlinger, Kurt and Viebahn, Jan and Gegelman, Wiktor and Merz, Christian},<br\/>  booktitle = {Energietechnische Gesellschaft (ETG Kongress) 2025},<br\/>  keywords = {rl4ces},<br\/>  pages = {874-881},<br\/>  title = {Graph Neural Networks for Grid Control: Prospects in AI-assisted Transmission Grid Operation},<br\/>  year = 2025<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-fdc7f4fbce01cd3f0309bd5cede9804c\"><p>%0 Conference Paper<br\/>%1 holzhuter2025graph<br\/>%A Holzh\u00fcter, Clara Juliane<br\/>%A Lytaev, Pawel<br\/>%A Dipp, Marcel<br\/>%A Hassouna, Mohamed<br\/>%A Brendlinger, Kurt<br\/>%A Viebahn, Jan<br\/>%A Gegelman, Wiktor<br\/>%A Merz, Christian<br\/>%B Energietechnische Gesellschaft (ETG Kongress) 2025<br\/>%D 2025<br\/>%P 874-881<br\/>%T Graph Neural Networks for Grid Control: Prospects in AI-assisted Transmission Grid Operation<br\/>%U https:\/\/www.etg-kongress.com\/resource\/blob\/2385346\/9a8b59f27822512a0884d3cd0ec8e20e\/etg-kongress-2025-proceedings-link-data.pdf<br\/>%X Transmission grid congestion management and outage planning are critical tasks in modern grid operation due to thenon-linear nature of power flows and the large-scale optimization challenges faced by operators. Traditionally, overloadsare addressed through generator redispatch, a costly and therefore suboptimal measure. In the project \"Graph NeuralNetworks for Grid Control\" (GNN4GC), we investigate alternative strategies, focusing on topological remedial actionsthat could minimize or even completely eliminate redispatch costs. Topology optimization, a core aspect of this project,presents significant challenges due to its combinatorial nature, requiring extensive computational resources for powerflow calculations. To address this, GNN4GC is split into three stages. In the first stage, we explore the use of GraphNeural Networks (GNNs) to accelerate these calculations and benchmark their performance against established tools likepandapower and a DC power flow solver developed by 50Hertz Transmission GmbH and TenneT TSO GmbH. In thesecond stage, we use Reinforcement Learning and other heuristics to select suitable topologies and solve the topologyoptimization problem. As a third stage, we test the respective agent on real-life grids to benchmark the methodology. Theaim of the final stage is to build a recommender system that can be used in a control room in the future.<br\/><\/p><\/div><\/div><\/li>\n<\/ul>\n<a class=\"bibsonomycsl_publications-headline-anchor\" name=\"jmp_2024\"><\/a><h3 class=\"bibsonomycsl_publications-headline\" style=\"font-size: 1.1em; font-weight: bold;\">2024<\/h3>\n<ul class=\"bibsonomycsl_publications\"><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/article.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Lehna, Malte, Clara Holzh\u00fcter, Sven Tomforde, and Christoph Scholz. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">HUGO \u2013 Highlighting Unseen Grid Options: Combining Deep Reinforcement Learning With a Heuristic Target Topology Approach<\/span>\u201d<\/span>. <i>Sustainable Energy, Grids and Networks<\/i> 39 (September 2024): 101510. doi:10.1016\/j.segan.2024.101510.<\/div>\n<\/div><span class=\"bibsonomycsl_url\"><a href=\"http:\/\/dx.doi.org\/10.1016\/j.segan.2024.101510\" target=\"_blank\">URL<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-b443c48bccedd7b808ed46b2295b8061\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-b443c48bccedd7b808ed46b2295b8061\" href=\"#\">EndNote<\/a><\/span><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/dx.doi.org\/10.1016\/j.segan.2024.101510\" target=\"_blank\">DOI<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-b443c48bccedd7b808ed46b2295b8061\"><p>@article{Lehna_2024,<br\/>  author = {Lehna, Malte and Holzh\u00fcter, Clara and Tomforde, Sven and Scholz, Christoph},<br\/>  journal = {Sustainable Energy, Grids and Networks},<br\/>  keywords = {rl4ces},<br\/>  month = {09},<br\/>  pages = 101510,<br\/>  publisher = {Elsevier BV},<br\/>  title = {HUGO \u2013 Highlighting Unseen Grid Options: Combining deep reinforcement learning with a heuristic target topology approach},<br\/>  volume = 39,<br\/>  year = 2024<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-b443c48bccedd7b808ed46b2295b8061\"><p>%0 Journal Article<br\/>%1 Lehna_2024<br\/>%A Lehna, Malte<br\/>%A Holzh\u00fcter, Clara<br\/>%A Tomforde, Sven<br\/>%A Scholz, Christoph<br\/>%D 2024<br\/>%I Elsevier BV<br\/>%J Sustainable Energy, Grids and Networks<br\/>%P 101510<br\/>%R 10.1016\/j.segan.2024.101510<br\/>%T HUGO \u2013 Highlighting Unseen Grid Options: Combining deep reinforcement learning with a heuristic target topology approach<br\/>%U http:\/\/dx.doi.org\/10.1016\/j.segan.2024.101510<br\/>%V 39<br\/><\/p><\/div><\/div><\/li><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/techreport.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Mussi, Marco, Gianvito Losapio, Alberto Maria Metelli, Marcello Restelli, Ricardo Bessa, Antoine Marot, Daniel Boos, et al. <span class=\"csl-title\"><span class=\"csl-title\"><i>Position Paper on AI for the Operation of Critical Energy and Mobility Network Infrastructures<\/i><\/span><\/span>. Porto: AI4REALNET, 2024. doi:https:\/\/irf.fhnw.ch\/handle\/11654\/49377.<\/div>\n<\/div><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/irf.fhnw.ch\/handle\/11654\/49377\" target=\"_blank\">URL<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-41f4e774eb96613d1def461ada5720af\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-41f4e774eb96613d1def461ada5720af\" href=\"#\">EndNote<\/a><\/span><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/dx.doi.org\/https:\/\/irf.fhnw.ch\/handle\/11654\/49377\" target=\"_blank\">DOI<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-41f4e774eb96613d1def461ada5720af\"><p>@techreport{11654\/49377,<br\/>  address = {Porto},<br\/>  author = {Mussi, Marco and Losapio, Gianvito and Metelli, Alberto Maria and Restelli, Marcello and Bessa, Ricardo and Marot, Antoine and Boos, Daniel and Borst, Clark and Castagna, Alberto and Dias, Duarte and Egli, Adrian and Eisenegger, Andrina and Manyari, Yassine El and Fuxj\u00e4ger, Anton and Hamouche, Samira and Hassouna, Mohamed and Lemetayer, Bruno and Liessner, Roman and Lundberg, Jonas and Schneider, Manuel and Sturm, Irene and Usher, Julia and Van Hoof, Herke and Viebahn, Jan and W\u00e4fler, Toni and di Milano, Politecnico},<br\/>  institution = {AI4REALNET},<br\/>  keywords = {rl4ces},<br\/>  note = {Porto},<br\/>  title = {Position paper on AI for the operation of critical energy and mobility network infrastructures},<br\/>  year = 2024<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-41f4e774eb96613d1def461ada5720af\"><p>%0 Report<br\/>%1 11654\/49377<br\/>%A Mussi, Marco<br\/>%A Losapio, Gianvito<br\/>%A Metelli, Alberto Maria<br\/>%A Restelli, Marcello<br\/>%A Bessa, Ricardo<br\/>%A Marot, Antoine<br\/>%A Boos, Daniel<br\/>%A Borst, Clark<br\/>%A Castagna, Alberto<br\/>%A Dias, Duarte<br\/>%A Egli, Adrian<br\/>%A Eisenegger, Andrina<br\/>%A Manyari, Yassine El<br\/>%A Fuxj\u00e4ger, Anton<br\/>%A Hamouche, Samira<br\/>%A Hassouna, Mohamed<br\/>%A Lemetayer, Bruno<br\/>%A Liessner, Roman<br\/>%A Lundberg, Jonas<br\/>%A Schneider, Manuel<br\/>%A Sturm, Irene<br\/>%A Usher, Julia<br\/>%A Van Hoof, Herke<br\/>%A Viebahn, Jan<br\/>%A W\u00e4fler, Toni<br\/>%A di Milano, Politecnico<br\/>%C Porto<br\/>%D 2024<br\/>%R https:\/\/irf.fhnw.ch\/handle\/11654\/49377<br\/>%T Position paper on AI for the operation of critical energy and mobility network infrastructures<br\/>%U https:\/\/irf.fhnw.ch\/handle\/11654\/49377<br\/><\/p><\/div><\/div><\/li><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/misc.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Hassouna, Mohamed, Clara Holzh\u00fcter, Pawel Lytaev, Josephine Thomas, Bernhard Sick, and Christoph Scholz. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">Graph Reinforcement Learning for Power Grids: A Comprehensive Survey<\/span>\u201d<\/span>. https:\/\/arxiv.org\/abs\/2407.04522.<\/div>\n<\/div><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/arxiv.org\/abs\/2407.04522\" target=\"_blank\">URL<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-3e3fdfb7f37a4c2b9090ac8f530c6430\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-3e3fdfb7f37a4c2b9090ac8f530c6430\" href=\"#\">EndNote<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-3e3fdfb7f37a4c2b9090ac8f530c6430\"><p>@misc{hassouna2024graphreinforcementlearningpower,<br\/>  author = {Hassouna, Mohamed and Holzh\u00fcter, Clara and Lytaev, Pawel and Thomas, Josephine and Sick, Bernhard and Scholz, Christoph},<br\/>  keywords = {rl4ces},<br\/>  title = {Graph Reinforcement Learning for Power Grids: A Comprehensive Survey},<br\/>  year = 2024<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-3e3fdfb7f37a4c2b9090ac8f530c6430\"><p>%0 Generic<br\/>%1 hassouna2024graphreinforcementlearningpower<br\/>%A Hassouna, Mohamed<br\/>%A Holzh\u00fcter, Clara<br\/>%A Lytaev, Pawel<br\/>%A Thomas, Josephine<br\/>%A Sick, Bernhard<br\/>%A Scholz, Christoph<br\/>%D 2024<br\/>%T Graph Reinforcement Learning for Power Grids: A Comprehensive Survey<br\/>%U https:\/\/arxiv.org\/abs\/2407.04522<br\/><\/p><\/div><\/div><\/li><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/inproceedings.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Lehna, Malte, Mohamed Hassouna, Dmitry Degtyar, Sven Tomforde, and Christoph Scholz. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">Fault Detection for Agents in Power Grid Topology Optimization: A Comprehensive Analysis<\/span>\u201d<\/span>. In <i>Machine Learning for Sustainable Power Systems (ML4SPS), ECML<\/i>. Springer, 2024.<\/div>\n<\/div><span class=\"bibsonomycsl_export bibsonomycsl_abstract\"><a rel=\"abs-a74d49337f80b0ae9a3ddae1c71ad748\"  href=\"#\">Abstract<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-a74d49337f80b0ae9a3ddae1c71ad748\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-a74d49337f80b0ae9a3ddae1c71ad748\" href=\"#\">EndNote<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_abstract\" style=\"display:none;\" id=\"abs-a74d49337f80b0ae9a3ddae1c71ad748\">Optimizing the topology of transmission networks using Deep Reinforcement Learning (DRL) has increasingly come into focus. Various DRL agents have been proposed, which are mostly benchmarked on the Grid2Op environment from the Learning to Run a Power Network (L2RPN) challenges. The environments have many advantages with their realistic chronics and underlying power flow backends. However, the interpretation of agent survival or failure is not always clear, as there are a variety of potential causes. In this work, we focus on the failures of the power grid simulation to identify patterns and detect them in advance. We collect the failed chronics of three different agents on the WCCI 2022 L2RPN environment, totaling about 40k data points. By clustering, we are able to detect five distinct clusters, identifying common failure types. Further, we propose a multi-class prediction approach to detect failures beforehand and evaluate five different prediction models. Here, the Light Gradient-Boosting Machine (LightGBM) shows the best failure detection performance, with an accuracy of 86%. It also correctly identifies in 91% of the time failure and survival observations. Finally, we provide a detailed feature importance analysis that identifies critical features and regions in the grid.<\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-a74d49337f80b0ae9a3ddae1c71ad748\"><p>@inproceedings{lehna2024fault,<br\/>  abstract = {Optimizing the topology of transmission networks using Deep Reinforcement Learning (DRL) has increasingly come into focus. Various DRL agents have been proposed, which are mostly benchmarked on the Grid2Op environment from the Learning to Run a Power Network (L2RPN) challenges. The environments have many advantages with their realistic chronics and underlying power flow backends. However, the interpretation of agent survival or failure is not always clear, as there are a variety of potential causes. In this work, we focus on the failures of the power grid simulation to identify patterns and detect them in advance. We collect the failed chronics of three different agents on the WCCI 2022 L2RPN environment, totaling about 40k data points. By clustering, we are able to detect five distinct clusters, identifying common failure types. Further, we propose a multi-class prediction approach to detect failures beforehand and evaluate five different prediction models. Here, the Light Gradient-Boosting Machine (LightGBM) shows the best failure detection performance, with an accuracy of 86%. It also correctly identifies in 91% of the time failure and survival observations. Finally, we provide a detailed feature importance analysis that identifies critical features and regions in the grid.},<br\/>  author = {Lehna, Malte and Hassouna, Mohamed and Degtyar, Dmitry and Tomforde, Sven and Scholz, Christoph},<br\/>  booktitle = {Machine Learning for Sustainable Power Systems (ML4SPS), ECML},<br\/>  keywords = {rl4ces},<br\/>  note = {(accepted)},<br\/>  publisher = {Springer},<br\/>  title = {Fault Detection for Agents in Power Grid Topology Optimization: A Comprehensive Analysis},<br\/>  year = 2024<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-a74d49337f80b0ae9a3ddae1c71ad748\"><p>%0 Conference Paper<br\/>%1 lehna2024fault<br\/>%A Lehna, Malte<br\/>%A Hassouna, Mohamed<br\/>%A Degtyar, Dmitry<br\/>%A Tomforde, Sven<br\/>%A Scholz, Christoph<br\/>%B Machine Learning for Sustainable Power Systems (ML4SPS), ECML<br\/>%D 2024<br\/>%I Springer<br\/>%T Fault Detection for Agents in Power Grid Topology Optimization: A Comprehensive Analysis<br\/>%X Optimizing the topology of transmission networks using Deep Reinforcement Learning (DRL) has increasingly come into focus. Various DRL agents have been proposed, which are mostly benchmarked on the Grid2Op environment from the Learning to Run a Power Network (L2RPN) challenges. The environments have many advantages with their realistic chronics and underlying power flow backends. However, the interpretation of agent survival or failure is not always clear, as there are a variety of potential causes. In this work, we focus on the failures of the power grid simulation to identify patterns and detect them in advance. We collect the failed chronics of three different agents on the WCCI 2022 L2RPN environment, totaling about 40k data points. By clustering, we are able to detect five distinct clusters, identifying common failure types. Further, we propose a multi-class prediction approach to detect failures beforehand and evaluate five different prediction models. Here, the Light Gradient-Boosting Machine (LightGBM) shows the best failure detection performance, with an accuracy of 86%. It also correctly identifies in 91% of the time failure and survival observations. Finally, we provide a detailed feature importance analysis that identifies critical features and regions in the grid.<br\/><\/p><\/div><\/div><\/li><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/misc.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Lehna, Malte, Clara Holzh\u00fcter, Sven Tomforde, and Christoph Scholz. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">HUGO -- Highlighting Unseen Grid Options: Combining Deep Reinforcement Learning With a Heuristic Target Topology Approach<\/span>\u201d<\/span>. doi:https:\/\/doi.org\/10.1016\/j.segan.2024.101510.<\/div>\n<\/div><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/arxiv.org\/abs\/2405.00629\" target=\"_blank\">URL<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-fd55b8b8e7816082ca1665b9a5555aff\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-fd55b8b8e7816082ca1665b9a5555aff\" href=\"#\">EndNote<\/a><\/span><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/doi.org\/10.1016\/j.segan.2024.101510\" target=\"_blank\">DOI<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-fd55b8b8e7816082ca1665b9a5555aff\"><p>@misc{lehna2024hugohighlightingunseen,<br\/>  author = {Lehna, Malte and Holzh\u00fcter, Clara and Tomforde, Sven and Scholz, Christoph},<br\/>  keywords = {rl4ces},<br\/>  title = {HUGO -- Highlighting Unseen Grid Options: Combining Deep Reinforcement Learning with a Heuristic Target Topology Approach},<br\/>  year = 2024<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-fd55b8b8e7816082ca1665b9a5555aff\"><p>%0 Generic<br\/>%1 lehna2024hugohighlightingunseen<br\/>%A Lehna, Malte<br\/>%A Holzh\u00fcter, Clara<br\/>%A Tomforde, Sven<br\/>%A Scholz, Christoph<br\/>%D 2024<br\/>%R https:\/\/doi.org\/10.1016\/j.segan.2024.101510<br\/>%T HUGO -- Highlighting Unseen Grid Options: Combining Deep Reinforcement Learning with a Heuristic Target Topology Approach<br\/>%U https:\/\/arxiv.org\/abs\/2405.00629<br\/><\/p><\/div><\/div><\/li><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/misc.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Braun, Martin, Christian Gruhl, Christian A. Hans, Philipp H\u00e4rtel, Christoph Scholz, Bernhard Sick, Malte Siefert, Florian Steinke, Olaf Stursberg, and Sebastian Wende von Berg. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">Predictions and Decision Making for Resilient Intelligent Sustainable Energy Systems<\/span>\u201d<\/span>. https:\/\/arxiv.org\/abs\/2407.03021.<\/div>\n<\/div><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/arxiv.org\/abs\/2407.03021\" target=\"_blank\">URL<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-0b5495c3a7bfd81c1fd77c76728f5cf7\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-0b5495c3a7bfd81c1fd77c76728f5cf7\" href=\"#\">EndNote<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-0b5495c3a7bfd81c1fd77c76728f5cf7\"><p>@misc{braun2024predictionsdecisionmakingresilient,<br\/>  author = {Braun, Martin and Gruhl, Christian and Hans, Christian A. and H\u00e4rtel, Philipp and Scholz, Christoph and Sick, Bernhard and Siefert, Malte and Steinke, Florian and Stursberg, Olaf and von Berg, Sebastian Wende},<br\/>  keywords = {rl4ces},<br\/>  title = {Predictions and Decision Making for Resilient Intelligent Sustainable Energy Systems},<br\/>  year = 2024<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-0b5495c3a7bfd81c1fd77c76728f5cf7\"><p>%0 Generic<br\/>%1 braun2024predictionsdecisionmakingresilient<br\/>%A Braun, Martin<br\/>%A Gruhl, Christian<br\/>%A Hans, Christian A.<br\/>%A H\u00e4rtel, Philipp<br\/>%A Scholz, Christoph<br\/>%A Sick, Bernhard<br\/>%A Siefert, Malte<br\/>%A Steinke, Florian<br\/>%A Stursberg, Olaf<br\/>%A von Berg, Sebastian Wende<br\/>%D 2024<br\/>%T Predictions and Decision Making for Resilient Intelligent Sustainable Energy Systems<br\/>%U https:\/\/arxiv.org\/abs\/2407.03021<br\/><\/p><\/div><\/div><\/li>\n<\/ul>\n<a class=\"bibsonomycsl_publications-headline-anchor\" name=\"jmp_2023\"><\/a><h3 class=\"bibsonomycsl_publications-headline\" style=\"font-size: 1.1em; font-weight: bold;\">2023<\/h3>\n<ul class=\"bibsonomycsl_publications\"><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/article.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Lehna, Malte, Jan Viebahn, Antoine Marot, Sven Tomforde, and Christoph Scholz. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">Managing Power Grids through Topology Actions: A Comparative Study Between Advanced Rule-Based and Reinforcement Learning Agents<\/span>\u201d<\/span>. <i>Energy and AI<\/i> 14 (2023): 100276. doi:10.1016\/j.egyai.2023.100276.<\/div>\n<\/div><span class=\"bibsonomycsl_export bibsonomycsl_abstract\"><a rel=\"abs-e8949e7700da1eb974fe51c94877666e\"  href=\"#\">Abstract<\/a><\/span><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S2666546823000484?via%3Dihub\" target=\"_blank\">URL<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-e8949e7700da1eb974fe51c94877666e\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-e8949e7700da1eb974fe51c94877666e\" href=\"#\">EndNote<\/a><\/span><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/dx.doi.org\/10.1016\/j.egyai.2023.100276\" target=\"_blank\">DOI<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_abstract\" style=\"display:none;\" id=\"abs-e8949e7700da1eb974fe51c94877666e\">The operation of electricity grids has become increasingly complex due to the current upheaval and the increase in renewable energy production. As a consequence, active grid management is reaching its limits with conventional approaches. In the context of the Learning to Run a Power Network (L2RPN) challenge, it has been shown that Reinforcement Learning (RL) is an efficient and reliable approach with considerable potential for automatic grid operation. In this article, we analyse the submitted agent from Binbinchen and provide novel strategies to improve the agent, both for the RL and the rule-based approach. The main improvement is a N-1 strategy, where we consider topology actions that keep the grid stable, even if one line is disconnected. More, we also propose a topology reversion to the original grid, which proved to be beneficial. The improvements are tested against reference approaches on the challenge test sets and are able to increase the performance of the rule-based agent by 27%. In direct comparison between rule-based and RL agent we find similar performance. However, the RL agent has a clear computational advantage. We also analyse the behaviour in an exemplary case in more detail to provide additional insights. Here, we observe that through the N-1 strategy, the actions of both the rule-based and the RL agent become more diversified.<\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-e8949e7700da1eb974fe51c94877666e\"><p>@article{lehna2023managing,<br\/>  abstract = {The operation of electricity grids has become increasingly complex due to the current upheaval and the increase in renewable energy production. As a consequence, active grid management is reaching its limits with conventional approaches. In the context of the Learning to Run a Power Network (L2RPN) challenge, it has been shown that Reinforcement Learning (RL) is an efficient and reliable approach with considerable potential for automatic grid operation. In this article, we analyse the submitted agent from Binbinchen and provide novel strategies to improve the agent, both for the RL and the rule-based approach. The main improvement is a N-1 strategy, where we consider topology actions that keep the grid stable, even if one line is disconnected. More, we also propose a topology reversion to the original grid, which proved to be beneficial. The improvements are tested against reference approaches on the challenge test sets and are able to increase the performance of the rule-based agent by 27%. In direct comparison between rule-based and RL agent we find similar performance. However, the RL agent has a clear computational advantage. We also analyse the behaviour in an exemplary case in more detail to provide additional insights. Here, we observe that through the N-1 strategy, the actions of both the rule-based and the RL agent become more diversified.},<br\/>  author = {Lehna, Malte and Viebahn, Jan and Marot, Antoine and Tomforde, Sven and Scholz, Christoph},<br\/>  journal = {Energy and AI},<br\/>  keywords = {rl4ces},<br\/>  pages = 100276,<br\/>  publisher = {Energy and AI},<br\/>  title = {Managing power grids through topology actions: A comparative study between advanced rule-based and reinforcement learning agents},<br\/>  volume = 14,<br\/>  year = 2023<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-e8949e7700da1eb974fe51c94877666e\"><p>%0 Journal Article<br\/>%1 lehna2023managing<br\/>%A Lehna, Malte<br\/>%A Viebahn, Jan<br\/>%A Marot, Antoine<br\/>%A Tomforde, Sven<br\/>%A Scholz, Christoph<br\/>%D 2023<br\/>%I Energy and AI<br\/>%J Energy and AI<br\/>%P 100276<br\/>%R 10.1016\/j.egyai.2023.100276<br\/>%T Managing power grids through topology actions: A comparative study between advanced rule-based and reinforcement learning agents<br\/>%U https:\/\/www.sciencedirect.com\/science\/article\/pii\/S2666546823000484?via%3Dihub<br\/>%V 14<br\/>%X The operation of electricity grids has become increasingly complex due to the current upheaval and the increase in renewable energy production. As a consequence, active grid management is reaching its limits with conventional approaches. In the context of the Learning to Run a Power Network (L2RPN) challenge, it has been shown that Reinforcement Learning (RL) is an efficient and reliable approach with considerable potential for automatic grid operation. In this article, we analyse the submitted agent from Binbinchen and provide novel strategies to improve the agent, both for the RL and the rule-based approach. The main improvement is a N-1 strategy, where we consider topology actions that keep the grid stable, even if one line is disconnected. More, we also propose a topology reversion to the original grid, which proved to be beneficial. The improvements are tested against reference approaches on the challenge test sets and are able to increase the performance of the rule-based agent by 27%. In direct comparison between rule-based and RL agent we find similar performance. However, the RL agent has a clear computational advantage. We also analyse the behaviour in an exemplary case in more detail to provide additional insights. Here, we observe that through the N-1 strategy, the actions of both the rule-based and the RL agent become more diversified.<br\/><\/p><\/div><\/div><\/li><li class=\"bibsonomycsl_pubitem\"><div class=\"bibsonomycsl_preview_border bibsonomycsl_preview_thumb\">\n                                        <span>\n                                            <img decoding=\"async\" class=\"bibsonomycsl_preview\" style=\"z-index: 1;\" src=\"https:\/\/rl4ces.de\/wp-content\/plugins\/bibsonomy-csl\/img\/entrytypes\/article.jpg\" \/>\n                                        <\/span>\n                                 <\/div><div class=\"bibsonomycsl_entry\"><div class=\"csl-bib-body\">\n  <div class=\"csl-entry\">Heinrich, Ren\u00e9, Christoph Scholz, Stephan Vogt, and Malte Lehna. <span class=\"csl-title\">\u201c<span class=\"\u2018csl-title\u2019\">Targeted Adversarial Attacks on Wind Power Forecasts<\/span>\u201d<\/span>. <i>Machine Learning<\/i> 113, no. 2 (2023): 863\u2013889. doi:10.1007\/s10994-023-06396-9.<\/div>\n<\/div><span class=\"bibsonomycsl_export bibsonomycsl_abstract\"><a rel=\"abs-c85f3b4d92e02dde845c55776a4fa4e6\"  href=\"#\">Abstract<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_bibtex\"><a rel=\"bib-c85f3b4d92e02dde845c55776a4fa4e6\" href=\"#\">BibTeX<\/a><\/span><span class=\"bibsonomycsl_export bibsonomycsl_endnote\"><a rel=\"end-c85f3b4d92e02dde845c55776a4fa4e6\" href=\"#\">EndNote<\/a><\/span><span class=\"bibsonomycsl_url\"><a href=\"https:\/\/dx.doi.org\/10.1007\/s10994-023-06396-9\" target=\"_blank\">DOI<\/a><\/span><div style=\"clear: left\"> <\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_abstract\" style=\"display:none;\" id=\"abs-c85f3b4d92e02dde845c55776a4fa4e6\">In recent years, researchers proposed a variety of deep learning models for wind power forecasting. These models predict the wind power generation of wind farms or entire regions more accurately than traditional machine learning algorithms or physical models. However, latest research has shown that deep learning models can often be manipulated by adversarial attacks. Since wind power forecasts are essential for the stability of modern power systems, it is important to protect them from this threat. In this work, we investigate the vulnerability of two different forecasting models to targeted, semi-targeted, and untargeted adversarial attacks. We consider a Long Short-Term Memory (LSTM) network for predicting the power generation of individual wind farms and a Convolutional Neural Network (CNN) for forecasting the wind power generation throughout Germany. Moreover, we propose the Total Adversarial Robustness Score (TARS), an evaluation metric for quantifying the robustness of regression models to targeted and semi-targeted adversarial attacks. It assesses the impact of attacks on the model's performance, as well as the extent to which the attacker's goal was achieved, by assigning a score between 0 (very vulnerable) and 1 (very robust). In our experiments, the LSTM forecasting model was fairly robust and achieved a TARS value of over 0.78 for all adversarial attacks investigated. The CNN forecasting model only achieved TARS values below 0.10 when trained ordinarily, and was thus very vulnerable. Yet, its robustness could be significantly improved by adversarial training, which always resulted in a TARS above 0.46.<\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_bibtex\" style=\"display:none;\" id=\"bib-c85f3b4d92e02dde845c55776a4fa4e6\"><p>@article{heinrich2023targeted,<br\/>  abstract = {In recent years, researchers proposed a variety of deep learning models for wind power forecasting. These models predict the wind power generation of wind farms or entire regions more accurately than traditional machine learning algorithms or physical models. However, latest research has shown that deep learning models can often be manipulated by adversarial attacks. Since wind power forecasts are essential for the stability of modern power systems, it is important to protect them from this threat. In this work, we investigate the vulnerability of two different forecasting models to targeted, semi-targeted, and untargeted adversarial attacks. We consider a Long Short-Term Memory (LSTM) network for predicting the power generation of individual wind farms and a Convolutional Neural Network (CNN) for forecasting the wind power generation throughout Germany. Moreover, we propose the Total Adversarial Robustness Score (TARS), an evaluation metric for quantifying the robustness of regression models to targeted and semi-targeted adversarial attacks. It assesses the impact of attacks on the model's performance, as well as the extent to which the attacker's goal was achieved, by assigning a score between 0 (very vulnerable) and 1 (very robust). In our experiments, the LSTM forecasting model was fairly robust and achieved a TARS value of over 0.78 for all adversarial attacks investigated. The CNN forecasting model only achieved TARS values below 0.10 when trained ordinarily, and was thus very vulnerable. Yet, its robustness could be significantly improved by adversarial training, which always resulted in a TARS above 0.46.},<br\/>  author = {Heinrich, Ren\u00e9 and Scholz, Christoph and Vogt, Stephan and Lehna, Malte},<br\/>  journal = {Machine Learning},<br\/>  keywords = {rl4ces},<br\/>  number = 2,<br\/>  pages = {863--889},<br\/>  publisher = {Springer},<br\/>  title = {Targeted Adversarial Attacks on Wind Power Forecasts},<br\/>  volume = 113,<br\/>  year = 2023<br\/>}<br\/><\/p><\/div><div class=\"bibsonomycsl_collapse bibsonomycsl_pub_endnote\" style=\"display:none;\" id=\"end-c85f3b4d92e02dde845c55776a4fa4e6\"><p>%0 Journal Article<br\/>%1 heinrich2023targeted<br\/>%A Heinrich, Ren\u00e9<br\/>%A Scholz, Christoph<br\/>%A Vogt, Stephan<br\/>%A Lehna, Malte<br\/>%D 2023<br\/>%I Springer<br\/>%J Machine Learning<br\/>%N 2<br\/>%P 863--889<br\/>%R 10.1007\/s10994-023-06396-9<br\/>%T Targeted Adversarial Attacks on Wind Power Forecasts<br\/>%V 113<br\/>%X In recent years, researchers proposed a variety of deep learning models for wind power forecasting. These models predict the wind power generation of wind farms or entire regions more accurately than traditional machine learning algorithms or physical models. However, latest research has shown that deep learning models can often be manipulated by adversarial attacks. Since wind power forecasts are essential for the stability of modern power systems, it is important to protect them from this threat. In this work, we investigate the vulnerability of two different forecasting models to targeted, semi-targeted, and untargeted adversarial attacks. We consider a Long Short-Term Memory (LSTM) network for predicting the power generation of individual wind farms and a Convolutional Neural Network (CNN) for forecasting the wind power generation throughout Germany. Moreover, we propose the Total Adversarial Robustness Score (TARS), an evaluation metric for quantifying the robustness of regression models to targeted and semi-targeted adversarial attacks. It assesses the impact of attacks on the model's performance, as well as the extent to which the attacker's goal was achieved, by assigning a score between 0 (very vulnerable) and 1 (very robust). In our experiments, the LSTM forecasting model was fairly robust and achieved a TARS value of over 0.78 for all adversarial attacks investigated. The CNN forecasting model only achieved TARS values below 0.10 when trained ordinarily, and was thus very vulnerable. Yet, its robustness could be significantly improved by adversarial training, which always resulted in a TARS above 0.46.<br\/><\/p><\/div><\/div><\/li><\/ul>","protected":false},"excerpt":{"rendered":"<p>[2026] [2025] [2024] [2023] 2026 Hassouna, Mohamed, Clara Holzh\u00fcter, Malte Lehna, Matthijs de Jong, Jan Viebahn, Bernhard Sick, and Christoph Scholz. \u201cLearning Topology Actions for\u00a0Power Grid Control: A Graph-Based Soft-Label Imitation Learning Approach\u201d. In Machine Learning and Knowledge Discovery in Databases. Applied Data Science Track and Demo Track, edited by In\u00eas Dutra, Mykola Pechenizkiy, Paulo [&hellip;]<\/p>","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-gradient":""}},"footnotes":""},"class_list":["post-378","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/rl4ces.de\/en\/wp-json\/wp\/v2\/pages\/378","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/rl4ces.de\/en\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/rl4ces.de\/en\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/rl4ces.de\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/rl4ces.de\/en\/wp-json\/wp\/v2\/comments?post=378"}],"version-history":[{"count":5,"href":"https:\/\/rl4ces.de\/en\/wp-json\/wp\/v2\/pages\/378\/revisions"}],"predecessor-version":[{"id":590,"href":"https:\/\/rl4ces.de\/en\/wp-json\/wp\/v2\/pages\/378\/revisions\/590"}],"wp:attachment":[{"href":"https:\/\/rl4ces.de\/en\/wp-json\/wp\/v2\/media?parent=378"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}