Open Access

Marcel Graus, Philipp Niemietz, Mohammad Touhidur Rahman, Michaela Hiller, Markus Pahlenkemper

• The integration of renewable energies in a local industrial environment is an urgent task to reduce greenhouse gas emissions. Their energy intensive processes and local energy generation make waste management companies to optimal areas to analyze micro grids. The combination of the main task to process arriving waste and the reaction on micro grid needs without disregarding user preferences is the challenge that is focused with the following approach applying machine learning techniques. First, the amount of waste is predicted with an artificial neural network. Then, the waste processing is optimized via an augmented Lagrangian algorithm regarding the energy costs that are based on volatile energy prices influenced from renewable energies. In addition, the optimization regards user preferences, which are learned from a user feedback with a support vector machine. For the user interaction, an active learning paradigm is used. The approach is applied on biological waste treatment process in the waste management company of the district of Warendorf. The results show that the energy consumptions can be controlled in a micro grid context within the frame of user preference.