Keep an eye on wind turbine performance and detect anomalies via baseline power curve
The wind turbine power curve contains vital information regarding the turbine efficiency and health, but the simplified power curve displayed on the SCADA system loses this data information in translation.
“The Power Curve contains the key to how much revenue your assets will generate.”
Kavaken’s Power Curve Analyzer can be used to analyze the data and verify power curves generically using SCADA data with different conditions like fault, icing, and dirt on blades, down rating, pitch malfunctions, Etc. This approach is one of the ways that we keep track of the performance of a wind turbine.
The technical approach to analyzing power curve using SCADA data
We are interested in determining a baseline power curve for each wind turbine that represents the most expected behavior of that specific turbine. The easiest way would be using a theoretical power curve that a wind turbine manufacturer has provided. However, this theoric power curve does not reflect the power curve you observe from SCADA due to turbine location and geographic conditions, wind characteristics, air density, and wind measurement uncertainties. First, a wind range of the SCADA data is selected to prevent this disparity, free of faults and maintenance. Then, we filter out data that cause unusual deviations from the expected power curve, such as icing, grid failure, Etc. Filtering these samples in which the wind turbine is not producing any output power are helpful steps to achieve a more generic power curve representing the turbine’s actual behavior, which we call the baseline power curve. To follow the common standards, then, we use the IEC standard binning method for data representation:
As the above figure shows, our baseline is very close to the theoretical power curve. Consider again; sometimes, the baseline may differ slightly from the theoretical one.
How to monitor anomalies? Is something wrong?
To find odd cases, which is a deviation from the baseline, we use several time intervals like 60 days, 30 days, 15 days, and 7 days of data to plot new power curves. The data sets are continuously monitored. If there is a deviation from the baseline power curve, the period is flagged for further inspection.
An example case;
Above, you see some anomalies in the past 60 days of the new power curve, which have been derived using the described methods. Now that we identified the problem, we can look at the raw data on the plot:
Let’s watch exactly what is happening in these periods and when:
Further inspection on the low-performance period has indicated that the power curve underperformance has been related to a manual curtailment on the turbine. The issue has been identified as a commercial decision due to market pricing at given hours; service and trading departments have been aligned.
A continuous power curve analysis allows the asset manager to detect any anomaly on turbine performance and increase the asset revenues by improved performance/shorter reaction times. With the assistance of Kavaken’s Power Curve Analyzer, asset managers will have a never-ending power curve verification capability.