Investigation of Data Pre-Processing Algorithms for Power Curve Modeling of Wind Turbines Based on ECC

Data pre-processing is the first step of using SCADA data to study performance wind turbines. However, there a lack knowledge how obtain more effective algorithms. This paper fully explores multiple algorithms for power curve modeling. A three-stage processing mode proposed, namely, preliminary filtering and compensation (Stage I), secondary II), single-valued Ⅲ). Different are selected at different stages finally merged into nine novel evaluation method based on energy characteristic consistency (ECC) proposed evaluate reliability various The influence sliding benchmark Binning has been investigated through indicators. Four turbines verify advantages disadvantages methods. result shows that same speed, rotational speed values obtained by MLE (maximum likelihood estimation) relatively high among three Among outlier methods, value KDE (kernel density large. In general, KDE-LSM (least square method) good in general. sum four evaluating index from smallest.