A new research paper evaluates machine learning models for fault detection in power systems, focusing on protection-oriented metrics beyond aggregate scores. The study analyzed Gradient Boosted Trees, a convolutional neural network, and a logistic regression model using the PROTECT-90 dataset. While these models achieved high AUC-ROC values, they exhibited significant rates of missed trips and false trips, with missed trips correlating with higher fault resistance. AI
IMPACT Highlights potential risks of deploying ML in critical infrastructure, emphasizing the need for robust safety evaluations beyond standard accuracy metrics.
RANK_REASON Research paper published on arXiv detailing an assessment of machine learning models. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- AUC-ROC
- CatalyzeX Code Finder for Papers
- convolutional neural network
- CORE Recommender
- DagsHub
- Gotit.pub
- Gradient Boosted Trees
- Hugging Face
- IArxiv Recommender
- Influence Flower
- logistic regression model
- machine learning
- PROTECT-90
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →