How Turbit works
General Alarm settings
In general, Turbit Detection KPIs measure the difference between simulated data from neural networks and actual data from the wind turbine. When this difference reaches a certain threshold an alarm will be created.
The thresholds are set dynamically from Turbit depending on the neural network (NN) performance and prediction certainty per analytics module per turbine.
Alarm settings for Turbit Modules
On a high-level Turbit uses different Detection KPIs for AI Monitoring Modules. While Detection KPIs for Power Monitoring analyze shorter periods (last 30 Minutes). Detection KPIs for the Main Components look at a longer time frames (5-10 days) to ensure high alarm relevance.
For power monitoring and wildlife monitoring, we add detailed filters based on status codes and additional information from the operation software like Bazefield to ensure high alarm relevance. We can set filters per turbine and module.
Increase model performance
To improve NN performance and prediction certainty, we manage training data sets per neural network while onboarding and monitoring the wind turbine.
With every confirmed alarm, Turbit automatically adds and excludes data for the training data set. Our retraining schedules ensure regular retraining to bring better-performing neural networks into production.
Each automatic alert comes with an automatically generated report. The report is built up to interpret alerts quickly. The report is called Event Card. The Event Card displays relevant plots, benchmarks of nearby turbines, and overlapping status codes of the anomaly.
Smart Dashboards and Alerts
Overview of data quality, model performance, and anomalies sent per Turbit Module booked. The Dashboard and the Event Card are the central design pattern of the Turbit Platform.
Workflows with Turbit
Providing in-depths root cause analysis
Preparing video analysis for better communications
Dashboards for fleet health, data quality and model performance
Turbit & Customer
Verify root causes
Prepare preventive measure proposals for OEM or service
Reports and KPIs for internal and external communication
Act on early diagnosis of potential component damage
Communicates with service partners and Turbit
Provides direct feedback to Turbit after maintenance
Deploy Turbit Monitoring
Technical Set Up
For each turbine, component and each site, Turbit AI learns the normal performance behavior from historical SCADA data. Turbit models use physically relevant input data such as wind speed, temperature, wind direction and turbulence intensity.
Continuous simulation of the normal behavior (measured values). Turbit AI detects deviations from the dynamic normal behavior and provides this anomaly as an automatic report. Each report is accessible via link or API.
Evaluate every anomaly detection. This feedback continuously improves the Turbit AI:
Communication per wind turbine
Precise anomaly detection
Root cause predtiction