AutoMoG: Automated data-driven Model Generation of multi-energy systems using piecewise-linear regression
Andreas Kämper, Ludger Leenders, Björn Bahl, André Bardow
Abstract
Operational optimization of multi-energy systems requires a mathematical model that is accurate and computationally efficient. A model can be generated in a data-driven way if measured data is available. Commonly, data is then used to model each component of the multi-energy system independently. However, independent modeling of each component may lead to models that are unnecessarily complicated and, thus, inefficient in practice. In this work, we propose the method AutoMoG for Automated data-driven Model Generation of multi-energy systems using piecewise-linear regression. AutoMoG provides Mixed-Integer Linear Programming models of multi-energy systems. To accurately model the overall multi-energy system, AutoMoG balances the errors caused by each component. Model accuracy is measured in terms of operating cost. In a case study, AutoMoG provides a multi-energy system model with less linear sections than single-component regression Still, AutoMoG retains high accuracy. Thereby, AutoMoG enables efficient data-driven modeling as the basis for multi-energy system optimization.