In this article:
Introduction
The Turbine Classifier takes the work out of monitoring power curves by automatically identifying the most relevant performance issues across your wind turbines. There's no need to track KPIs or analyze data for every turbine—this feature scans the data for you, filters out the noise, and highlights only the actionable insights.
Each insight comes with a clear dashboard that includes relevant graphs for easy understanding, along with key metrics to help you prioritize what needs attention first.
The Classification flag evaluates the operational state of a wind turbine using time-series data. It identifies various conditions such as full performance, performance degradations, downtime, and more.
This flag can be used to monitor turbine performance and detect potential issues. It integrates with Unity APM, allowing you to generate events via the Automation service. While the flag provides valuable insights, it does not consider existing OEM fault code events—so a complete assessment is still required to verify those events.
To streamline this process, the Turbine Classifier automates the analysis by using the Classification flag as an input, monitoring turbine states via OEM fault codes, and evaluating the duration of an issue. It then raises alarms only for meaningful, actionable issues—helping you track what's important across every turbine in your site or portfolio, every single day.
Each day, the classifier fetches and assesses the last week of data
The algorithm checks if the classification flag is reporting a sufficient number of timestamps with these primary categories:
Degraded Unknown
Nacelle Wind Speed Anomaly
Derate
The algorithm then discards any anomaly already explained by OEM faults or events when events have the state
Downtimes, Curtailment, or Degraded.Next, it checks the severity of the deviation between expected power and power for the flagged points within that data period. If, over the course of one week, the deviation is severe enough (greater than a predefined threshold), and not explained the above mentioned OEM faults or events, an insight will open with a defined start date and a metric.
The Turbine Classifier continuously monitors existing insights and keeps metrics up to date as long as abnormal turbine states persist.
If the turbine returns to normal operation, the Turbine Classifier will automatically close the insight, adding an end date and updating the related metrics. To confirm recovery, the feature verifies that normal turbine states are sustained across a broad range of wind speeds. Specifically, an insight will only close if normal behavior has been observed at wind speeds of at least 10 m/s.
The Turbine Classifier generates specific types of insights to help you understand and act on turbine performance issues. Each insight includes relevant metrics, visualizations, and tailored recommendations to support informed decision-making.
The table below outlines the types of insights currently available through this feature, along with the associated metrics and recommended actions.
Insight type | Description | Metric | Recommendations |
|---|---|---|---|
Wind Speed Anomaly | Power curve to the left of expected behavior | Duration (days) | Validate anemometer settings, parameters, software updates and data issues. |
Unexpected Power Limitation | Power curve shows derated or curtailed behavior without an associated OEM fault code | Potential Energy Loss (kWh) | Validate turbine settings, parameters and software updates and data issues, or misclassified/missing OEM fault codes. |
Potential Power Degradation | Power curve appears degraded without an associated OEM fault code or icing conditions | Potential Energy Loss (kWh) | Validate turbine and anemometer settings, parameters and software updates, data issues, or investigate potential sources of underperformance. |
You’ll find the Insight Log under the Analyze module, shown here:
 |
Clicking on any insight opens a dedicated dashboard with visualizations that highlight the source of the anomaly and its impact, making it easy to interpret the results at a glance. The visualization types and its purpose are listed below:
Power Curve: highlights an anomaly
RPM vs Power scatter plot: shows impact of anomaly on control algorithm
Insight log: in case other insights exist on the same asset
Timeseries of expected vs actual power: evolution of the issue over time
Gantt chart of events: see if warnings, downtimes or other events are occurring simultaneously
Loss walk: impact of anomaly with regards to expected, actual production, and loss categories
In this example site, the Turbine Classifier highlighted 2 out of the 42 turbines having power limitations that are not taken into account by an OEM fault code. Each line includes information on dates and potential energy loss related to the issue:
 |
Clicking on an insight opens a side panel with additional information, including Recommendations on troubleshooting:
 |
The Turbine Classifier Dashboard presents key information through a set of visual widgets, allowing instant access to the most relevant data for the selected asset and time period. Below is an example showing the power curve, one of the core visualizations used to assess turbine performance. Here, turquoise and blue dots show Classification flag categories (Derate, Degraded Unknown):
 |
Most sites will occasionally have turbines showing signs of degradation in the power curve, which can impact performance, metrics, and key KPIs. The Turbine Classifier automatically detects these issues and provides complementary information to help you take appropriate action.
This first example shows a power degradation in the power curve (left) accompanied by a deviation in the relation between Active Power, RPM and Pitch (right). This tells you there's a performance issue that needs to be addressed:
 |
 |
The second example shows a shift to the right of the power curve scatter plot, with the next graph in the Insight Dashboard also showing that the relation between Active Power, RPM, and Pitch is not impacted. The RPM slowly increases until rated power, which is normal behavior. This visualization shows you that the issue lies in the wind speed reading (sensor malfunction or replacement, parameter change, etc), and not in performance. The Gantt Chart included in the dashboard lets you check if there were any wind speed sensor related faults.
 |
 |
 |
Device Type | Signal | Measurement Type | Calculation Type | Unit | Calculation Period | Optional? |
|---|---|---|---|---|---|---|
WindTurbine | ActivePower | Power | Final | kW | siteTR | |
WindTurbine | ActivePowerExpected | Power | Final | kW | siteTR | |
WindTurbine | WindSpeedAdjusted | Wind Speed | Calculated | m/s | siteTR | |
WindTurbine | ClassificationFlag | Flag | Calculated | - | siteTR |
All OEM events at the turbine level which are in the categories Downtime, Degraded or Curtailment:
Timestamps
Category
Event ID
Was this article helpful?
That’s Great!
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
Feedback sent
We appreciate your effort and will try to fix the article