Data Confidence Applied to Wind Turbine Power Curves

This paper addresses the problem of reducing Wind Turbines Power Curve modeling error and false-positive classifications of incoming wind speed and respective power generation data with a real-time approach based on a confidence assignment algorithm. The approach builds upon an IoT Platform with support to the execution of domain-specific workflows to process incoming data, removing outliers and executing an algorithm that operates based on the difference between the data read by the sensors and the values predicted by an Artificial Neural Network (ANN). These values are used to calculate a confidence level, that can be used to identify a defective sensor, as well as to ignore wrong values that can lead to wrong diagnostics. The processed data is either used to build a model or is compared to an existing model to check its validity. The proposed approach achieved an average increase of 2.96% on the model coverage and 14.96% average reduction on the false-positive rates.