Good wind farm production data is vital for performance evaluations or training machine learning algorithms. But before we use the data, we must identify and remove abnormal data that can introduce unwanted biases. Instead of using a single method to determine all anomalies, Wang et al. (2023) apply specialised methods in succession (bidirectional quartile, K-means clustering, and DBSCAN methods) that each identifies specific anomaly types. The approach is validated by comparing the reconstructed data to operation data for 30 wind farms under ideal operating conditions.

Wang, Han, Ning Zhang, Ershun Du, Jie Yan, Shuang Han, Nan Li, Hongxia Li, and Yongqian Liu. 2023. “An Adaptive Identification Method of Abnormal Data in Wind and Solar Power Stations.” Renewable Energy 208 (May): 76–93. https://doi.org/10.1016/j.renene.2023.03.081.