{"id":{"repo_id":"mit","oai_identifier":"oai:dspace.mit.edu:1721.1/152673"},"canonical_url":"https://search.dev.ndltd.org/etd/mit/oai:dspace.mit.edu:1721.1/152673","repository":{"repo_id":"mit","name":"MIT","base_url":"https://dspace.mit.edu/oai/request"},"display":{"title":"Power Failure Cascade Prediction using Machine Learning","abstract":"We consider the problem of predicting power failure cascades due to branch failures. We propose several flow-free models using machine learning techniques like support vector machines, naive Bayes classifiers, and logistic regression. These models predict the grid states at every generation of a cascade process given the initial contingency. Further, we also propose a model based on graph neural networks (GNNs) that predicts cascades from the initial contingency and power injection values. We train the proposed models using a cascade sequence data pool generated from simulations. We then evaluate our models at various levels of granularity. We present several error metrics that gauge the models’ ability to predict the failure size, the final grid state, and the failure time steps of each branch within the cascade. We benchmark the proposed models against the influence model proposed in the literature. We show that the proposed machine learning models outperform the influence models under every metric. We also show that the graph neural network model, in addition to being generic over randomly scaled power injection values, outperforms multiple influence models that are built specifically for their corresponding loading profiles. Finally, we show that the proposed models reduce the computational time by almost two orders of magnitude.","abstract_html":"We consider the problem of predicting power failure cascades due to branch failures. We propose several flow-free models using machine learning techniques like support vector machines, naive Bayes classifiers, and logistic regression. These models predict the grid states at every generation of a cascade process given the initial contingency. Further, we also propose a model based on graph neural networks (GNNs) that predicts cascades from the initial contingency and power injection values. We train the proposed models using a cascade sequence data pool generated from simulations. We then evaluate our models at various levels of granularity. We present several error metrics that gauge the models’ ability to predict the failure size, the final grid state, and the failure time steps of each branch within the cascade. We benchmark the proposed models against the influence model proposed in the literature. We show that the proposed machine learning models outperform the influence models under every metric. We also show that the graph neural network model, in addition to being generic over randomly scaled power injection values, outperforms multiple influence models that are built specifically for their corresponding loading profiles. Finally, we show that the proposed models reduce the computational time by almost two orders of magnitude.","abstract_has_math":false,"creators":["Chadaga, Sathwik P."],"institution":"Massachusetts Institute of Technology","degree_name":"Master","degree_level":null,"degree_discipline":null,"degree_department":"Massachusetts Institute of Technology. Department of Aeronautics and Astronautics","school":null,"contributors":[],"advisors":["Modiano, Eytan H."],"committee_chairs":[],"committee_members":[],"year":2023,"date_issued":"2023-09","date_published":"2023-09","updated_at":"2026-07-22T22:22:31Z","subjects":[],"languages":[],"rights":["In Copyright - Educational Use Permitted","Copyright retained by author(s)"],"rights_urls":["https://rightsstatements.org/page/InC-EDU/1.0/"],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/1721.1/152673","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Modiano, Eytan H."]},{"key":"dc:contributor.department","label":"Department","values":["Massachusetts Institute of Technology. Department of Aeronautics and Astronautics"]},{"key":"dc:creator","label":"Author","values":["Chadaga, Sathwik P."]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2023-11-02T20:07:30Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2023-11-02T20:07:30Z"]},{"key":"dc:date.issued","label":"Date","values":["2023-09"]},{"key":"dc:publisher","label":"Institution","values":["Massachusetts Institute of Technology"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master","Master of Science in Aeronautics and Astronautics"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["In Copyright - Educational Use Permitted","Copyright retained by author(s)"]},{"key":"dc:rights.uri","label":"Rights URI","values":["https://rightsstatements.org/page/InC-EDU/1.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1721.1/152673"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["We consider the problem of predicting power failure cascades due to branch failures. We propose several flow-free models using machine learning techniques like support vector machines, naive Bayes classifiers, and logistic regression. These models predict the grid states at every generation of a cascade process given the initial contingency. Further, we also propose a model based on graph neural networks (GNNs) that predicts cascades from the initial contingency and power injection values. We train the proposed models using a cascade sequence data pool generated from simulations. We then evaluate our models at various levels of granularity. We present several error metrics that gauge the models’ ability to predict the failure size, the final grid state, and the failure time steps of each branch within the cascade. We benchmark the proposed models against the influence model proposed in the literature. We show that the proposed machine learning models outperform the influence models under every metric. We also show that the graph neural network model, in addition to being generic over randomly scaled power injection values, outperforms multiple influence models that are built specifically for their corresponding loading profiles. Finally, we show that the proposed models reduce the computational time by almost two orders of magnitude."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["S.M."]},{"key":"dc:title","label":"Title","values":["Power Failure Cascade Prediction using Machine Learning"]}]}],"canonical_facts":{"dc:contributor.advisor":["Modiano, Eytan H."],"dc:contributor.department":["Massachusetts Institute of Technology. Department of Aeronautics and Astronautics"],"dc:creator":["Chadaga, Sathwik P."],"dc:date.accessioned":["2023-11-02T20:07:30Z"],"dc:date.available":["2023-11-02T20:07:30Z"],"dc:date.issued":["2023-09"],"dc:description.abstract":["We consider the problem of predicting power failure cascades due to branch failures. We propose several flow-free models using machine learning techniques like support vector machines, naive Bayes classifiers, and logistic regression. These models predict the grid states at every generation of a cascade process given the initial contingency. Further, we also propose a model based on graph neural networks (GNNs) that predicts cascades from the initial contingency and power injection values. We train the proposed models using a cascade sequence data pool generated from simulations. We then evaluate our models at various levels of granularity. We present several error metrics that gauge the models’ ability to predict the failure size, the final grid state, and the failure time steps of each branch within the cascade. We benchmark the proposed models against the influence model proposed in the literature. We show that the proposed machine learning models outperform the influence models under every metric. We also show that the graph neural network model, in addition to being generic over randomly scaled power injection values, outperforms multiple influence models that are built specifically for their corresponding loading profiles. Finally, we show that the proposed models reduce the computational time by almost two orders of magnitude."],"dc:description.degree":["S.M."],"dc:identifier.uri":["https://hdl.handle.net/1721.1/152673"],"dc:publisher":["Massachusetts Institute of Technology"],"dc:rights":["In Copyright - Educational Use Permitted","Copyright retained by author(s)"],"dc:rights.uri":["https://rightsstatements.org/page/InC-EDU/1.0/"],"dc:title":["Power Failure Cascade Prediction using Machine Learning"],"dc:type":["Thesis"],"thesis:degree_name":["Master","Master of Science in Aeronautics and Astronautics"]},"updated_at":"2026-07-22T22:22:31Z"}