Evaluation of prediction accuracy for energy-efficient switching of automation facilities

Unproductive phases of automated manufacturing systems bear a large and rarely used potential for energy savings when shifting down or even completely switching off plant components and devices. To exploit this potential, we build on an approach to the automated calculation and execution of energy-efficient switching strategies based on representing plant behavior in terms of priced timed automaton (PTA) models annotated with energy cost information for different unproductive modes. Due to simplifications concerning timing and approximate power values, however, these models are idealistic and can predict a plant's energy demand only approximately. In this paper, we evaluate the prediction accuracy of PTA-derived switching strategies based on experiments carried out on a test bed of typical automation components. We can show that measured and predicted energy demand deviate only slightly for reasonable situations of production pause times.