Modeling Energy Consumption of Industrial Processes with Seq2Seq Machine Learning

Abstract

Energy considerations in production planning are gaining importance due to concerns over the climate change, but also because of the explosion of energy costs in the recent past. With increasing share of renewables, energy prices are changing over the day, and it thus seems evident to include the cost of energy as an optimization parameter in production planning. However, this requires to estimate the load profile of a given production plan, which in turn requires knowledge of the energy consumption of individual process steps. As the energy consumption of machines often depends on the concrete sequence of states, a straightforward modeling is not possible. In this paper, we investigate the use of machine learning to derive a black-box model from accessible energy measurement data and a known production plan. Preliminary results show that the Seq2Seq method is a promising candidate.