KNN Regression as Geo-Imputation Method for Spatio-Temporal Wind Data

The shift from traditional energy systems to distributed systems of energy suppliers and consumers and the power volatileness in renewable energy imply the need for e↵ective short-term prediction models. These machine learning models are based on measured sensor information. In practice, sensors might fail for several reasons. The prediction models cannot naturally cannot work properly with incomplete patterns. If the imputation method, which completes the missing data, is not appropriately chosen, a bias may be introduced. The objective of this work is to propose the k-nearest neighbor (kNN) regression as geo-imputation preprocessing step for pattern-labelbased short-term wind prediction of spatio-temporal wind data sets. The approach is compared to three other methods. The evaluation is based on four turbines with neighbors of the NREL Western Wind Data Set and the values are missing uniformly distributed. The results show that kNN regression is the most superior method for imputation.