With the ongoing digitalization of industrial production, an increasing number of energy measuring points are installed in manufacturing environments which enable promising use cases for machine learning applications. This paper presents a generic machine learning approach to forecast the very short term load of production machines which can be utilized as decision support basis for control schemes and measures to decrease energy costs. The presented approach is developed and evaluated on production machines of the ETA research factory at the Technische Universität Darmstadt. The results indicate that the developed approach is feasible and creates a precise very short term load forecasting model for different production machines.
Keywords Load forecasting, machine learning, feature selection, feature engineering, artificial neural networks