Intelligent O&M Platform for Offshore Wind Farms
- Victor Lih Jong,
- June 30 – Aug 23, 2024
- Green Energy
1. Executive Summary
A leading European renewable energy giant, managing over 12 GW of offshore wind assets across Northern Europe, sought to improve operational efficiency, asset longevity, and safety.
The Challenge:
-Fragmented telemetry
-Manual analysis delays
-Siloed enterprise systems
The Solution:
A unified, intelligent data and analytics platform, powered by a cloud-native lakehouse architecture on Azure and Databricks, designed to:
Ingest & process IoT/enterprise data (SCADA, SAP, GIS, MetOcean)
Enable predictive maintenance for critical components (cables, gearboxes, substructures, aviation lights)
Provide real-time operational dashboards
Use a hybrid modeling approach (Data Vault 2.0 + Kimball) for historical lineage + business analytics
2. Objectives
The key objectives of the program were:
Unify SCADA, ERP, and sensor data into a centralized data lakehouse
Enable real-time asset monitoring and condition-based alerts
Deploy predictive maintenance models for critical components
Improve asset availability and reduce unplanned downtime
Deliver actionable dashboards for technicians and executives
Ensure governance, security, and data lineage across all domains
3. Data Sources & Domains
The platform aggregated streaming and batch data from diverse sources:
| Domain | Source System | Data Type | Frequency |
|---|---|---|---|
| SCADA Systems | Turbine & substation PLCs | Vibration, speed, temperature | Streaming |
| CMS (Condition Monitoring) | Lubrication sensors | Grease & Oil level, quality | Streaming |
| ERP/CMMS | SAP S/4HANA, IBM Maximo | Maintenance, inventory | Daily batch |
| CRM | Salesforce, ServiceNow | Service tickets | Batch |
| MetOcean | Weather & ocean sensors | Wind, wave, icing data | Streaming/API |
| GIS & Asset Models | Esri ArcGIS, BIM models | Geospatial, topology | Weekly batch |
| Aviation Lights | Sensor networks | Status, alarms | Streaming |
| Icing Sensors | Hub-mounted IoT devices | Humidity, temperature | Streaming |
4. Architecture Overview
4.1 Platform Stack
| Layer | Technology Stack |
|---|---|
| Ingestion | Kafka, Azure IoT Hub, Event Hubs, Azure Data Factory |
| Processing | Azure Databricks (Structured Streaming, Delta Live Tables) |
| Storage | ADLS Gen2 with Delta Lake (bronze, silver, gold zones) |
| Data Modeling | Kimball (for BI) + Data Vault 2.0 (for integration/history) |
| Analytics/ML | MLflow, PySpark, TensorFlow, Feature Store on Databricks |
| Reporting | Power BI, Azure Analysis Services |
| Governance | Unity Catalog, Purview, RBAC, GDPR tagging |
5. Data Modeling Approach
-Hubs:
Turbine_Hub,Component_Hub,MaintenanceTicket_Hub-Links:
Turbine_Component_Link,Ticket_Turbine_Link-Satellites:
Sensor_Telemetry_Sat,Lubrication_Sat,Condition_Sat
5.2 Kimball Star Schema
Used in the gold layer for dashboarding & self-service analytics:
FactTables:
Fact_Turbine_Health,Fact_Lubrication_Events,Fact_Cable_Faults,Fact_Maintenance_ActionsDimensions:
Dim_Turbine,Dim_Component,Dim_Location,Dim_Time,Dim_FaultType
6. Key Analytics Use Cases
6.1 Predictive Maintenance
| Component | Analytics Focus | Methodology |
|---|---|---|
| Cables | Thermal load & degradation | Time-series clustering, anomaly detection |
| Sub-structures | Condition Monitoring (SSCM) | Vibration signatures, FFT analysis |
| Gearboxes | Vibration + temperature | Predictive ML (XGBoost) |
| Bearings | Lubrication analytics | Predict wear rate via regression |
| Aviation Lights | Fault prediction & compliance | Real-time alerting & threshold detection |
| Icing | Forecasting turbine icing events | LSTM-based weather model |
| Power | Power forecast | LSTM-based power/weather model |
6.2 Real-Time Operational Dashboards
Built using Power BI + DirectQuery to Delta Lake, featuring:
Live turbine health status Icing risk map (by wind park) Cable degradation risk heatmaps SSCM alerts dashboard Work order backlog & MTTR analytics Inventory & spare part availability Power trends
Alerts delivered via Teams, mobile, and email using Azure Logic Apps.
7. Implementation Timeline
| Phase | Key Activities |
|---|---|
| Phase 1: Platform Foundation | Kafka & Azure provisioning, lakehouse zones, initial SCADA/SAP ingestion |
| Phase 2: Data Integration | Data Vault & Kimball layers, streaming pipelines, SAP model integration |
| Phase 3: Advanced Analytics | ML model training (historical failures), MLflow deployment, cable/gearbox models |
| Phase 4: Visualisation & Alerting | Power BI dashboards, SSCM/icing alerts, Unity Catalog rollout |
8. Business Outcomes
| Metric | Before Implementation | After Platform Launch |
|---|---|---|
| Unplanned Downtime (avg/month) | 18 hours | < 5 hours |
| Cable Failure Incidents (yearly) | ~12 | 3 |
| MTTR (Mean Time to Repair) | >72 hours | <28 hours |
| Maintenance Cost (per turbine) | >€12,000/year | <€7,000/year |
| Data Access Time (analytics) | Days | Minutes |
| Regulatory Response Time | Manual, slow | <1 hour (auditable) |
9. Challenges & Solutions
| Challenge | Resolution |
|---|---|
| Massive SCADA data volumes | Delta Lake optimisations (Z-Ordering, OPTIMISE, Auto Compaction) |
| Disparate systems (SAP, GIS, CRM) | Unified schema via Data Vault, ADF pipelines |
| Real-time + historical harmonisation | Bronze-silver-gold zoning + time-windowed joins |
| GDPR & asset access security | Row/column-level security via Unity Catalog & Purview |
| Offshore network latency | Edge buffering + Azure IoT Edge + deduplication logic |
10. Future Roadmap
-Digital Twin Integration – Full-park simulations & virtual commissioning
-Reinforcement Learning – Optimise turbine behavior under failure scenarios
-3D GIS Integration – Overlay BIM/geospatial models for immersive insights
-AI Co-Pilot – NLP-based field support with telemetry lookups
-Sustainability KPIs – CO₂ offset, biodiversity impact, seabed disturbance models
11. Conclusion
This intelligent O&M platform marks a shift from reactive to predictive operations, leveraging Azure + Databricks to transform data into a strategic asset—enabling smarter, safer, and more sustainable energy production.