United KingdomEnergy and utilities2024 – ongoingDrax Group
Machine-learning anomaly detection on power plant assets
Drax caught a rising fault gas in a generator transformer early and avoided a repeat of a four-week outage
Problem
A similar generator transformer at Drax had failed earlier and stopped the unit for more than four weeks. Plant engineers needed earlier warning of faults in large assets such as transformers and steam turbines.
Approach
Drax used AVEVA Predictive Analytics, a machine-learning model that learns normal behaviour from historical sensor data in AVEVA's PI system and then watches live data. It flags anomalies and points to the sensors behind an alert. The tool flagged a rising gas-in-oil reading on a generator transformer. Drax changed how it ran the transformer and made a short repair, then kept the unit running until the next major planned stop.
Results
Takeaway
Early warning let the operator choose a short repair on its own schedule instead of an unplanned outage. The vendor's write-up gives no cost saving.
Categories
- Industry
- Energy and utilities
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Off-the-shelf tool or platform
- Goal
- Quality and risk
- Evidence
- Vendor case study
Sources
Source summaries
A techUK case study written around AVEVA's work with Drax and SUEZ. It describes how the anomaly-detection model works and how it flagged a transformer fault at Drax. The only Drax figure is the length of the earlier outage; no savings are given.
Full details are in the original sources.