Systematic Comparison of Imputation Models for Automatized Gap Filling on Electrical Load Data of Compressor Composites in the Industrial Sector

With increasing digitization for constructing intelligent energy systems, automated data processing is moving more and more into focus. Gaps in the recorded data pose a central problem for further processing instances. This work systematically investigates which methods are suitable for the imputation of data gaps of different sizes. It tackles the imputation performance’s influence on overlying applications, such as load forecasting and total energy determination. The presented method is applied to four datasets of compressors of industrial. Based on these Use Case’s evaluation results, recommendations for action are derived. Gap sizes should be considered when choosing an imputation method to minimize imputation error. For load forecasting, the prediction error correlates with the imputation error in certain missingness scenarios. Energy consumption analysis on the imputed data yields good results due to a balanced ratio of over-and undershooting of the imputation error.