Energy Time-Series Features for Emerging Applications on the Basis of Human-Readable Machine Descriptions

Feature extraction from energy time series gives way to use data mining methods that require static vectors as an input. However, there is a plethora of feature extraction methods, and selecting a good set of features for energy time series is difficult. In this article, we make some strides towards the long-term vision to guide feature selection for emerging applications in the energy domain. To this end, we study the issue of extracting features from energy time series for a novel use case: Deriving human-understandable descriptions for smart-meter measurements of industrial production machines. We first categorize existing feature extraction based on their technical specifications and on their usefulness with our application. Based on it, we select features suitable for our use case to derive machine descriptions for an industrial production facility. Our experimental results show that our overview and categorization are useful to select features for a novel use case.