Predictive Maintenance and Anomaly Detection for Wind Energy
summary
embw builds Asset Radar, a proprietary system that monitors roughly 450 wind turbines using SCADA 10‑minute aggregates, high‑rate oscillation captures (~50 kHz), and InfluxDB/Azure infrastructure. predictive approach trains GBRT regressors per signal (about 20,000 models total), sets quantile thresholds on test residuals, and surfaces deviations to human diagnosticians. labeled defects are tiny (≈0.05% of data), so anomaly space normalizes per‑unit context; current method flags issues but scalability and sparse labels limit full automated diagnosis.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:40Table of Contents
- 1:39Condition Monitoring at EnBW: Asset RADAR
- 4:21Condition Monitoring at EnBW: Data Sources
- 5:06Condition Monitoring at EnBW: SCADA
- 6:49Condition Monitoring at EnBW: Oscillation
- 9:24Anomaly Detection: Some More Complications
- 11:04Predictive Modeling: Idea
- 13:51Predictive Modeling: Thresholds
- 14:30Predictive Modeling: Specific vs. Generic
- 17:07Future Plans: Automatic Diagnoses
- 18:11Challenges: Heterogeneity
- 19:40Challenges: Labels
- 21:19Solutions: Anomaly Space
- 22:53Anomaly Space: Example
- 25:09Takeaways
related talks
26mDriving towards a cleaner future through data
19mEnterprise AI-Powered Predictive Maintenance Platform
28mDetect abnormal equipment behavior by analyzing sensor data
32mPredicting Grid Faults Before They Happen: AI for Weather-Driven Grid Resilience
26mProactive Failure Prediction Through Anomaly Detection
26mAnomaly Detection System for the CERN Cloud
20mDr. June Andrews, Principal Data Scientist, Wise.io, From GE Digital
05mPredictive maintenance with sensor "BOB"