ModEst – Software-supported model identification based on measured data in the field of industrial energy systems

 

Background

The purpose of an energy system is a cost-efficient and environmentally friendly supply of energy. Due to the complexity of most energy systems, mathematical optimization is required to ensure optimal operation. The application of mathematical optimization methods is useful, especially in sector coupling. Integrated energy systems supply demands of different forms of energy, such as electricity, heating, and cooling. Energy systems consist of complex components (e.g., combined heat and power units, absorption chillers, etc.). Different forms of energy are required to run an integrated energy system, such as gas or electricity. In addition, regenerative energies such as wind power, solar thermal energy, or photovoltaic units can be considered. Due to the inherent complexity of highly integrated energy systems, mathematical optimization methods are essential to ensure optimal operation.

The basis for the application of optimization methods is a mathematical model of the energy system that adequately represents the components and their interactions. Today, the major drawback in practice is the high effort that is required to model complex energy systems: Commonly, a new mathematical model is required for each energy system. However, there are no systematic methods available for the generation and parameterization of energy system models. However, in modern energy systems, more and more measured data from energy management systems are available that can be used for modeling.

 

Objective

Copyright: © Currenta

The ModEst research project aimed to develop methods that automate the modeling of energy systems. In particular, data from energy management systems were to be integrated. The developed methods fall into the research area of model identification and enable the user (e.g. planning engineer or energy consultant) to perform a model identification based on measured data. An important aspect of the developed methods is that the user does not need deep mathematical knowledge in the field of model identification. The methods for model identification were implemented in a prototype of an easy-to-use software tool. The software tool is suitable for users in energy technology, energy economics, and energy consulting. The methods for model identification were developed, validated, and tested for applicability in practice with the help of challenging practical examples from the industrial partner CURRENTA. The developed methods are generally applicable to industrial energy systems.

The Chair of Technical Thermodynamics acted as a coordinating research unit in the ModEst research project and also handled the main parts of the scientific aspects.

 

Results

Within the ModEst project, a detailed requirements analysis and work process definition for the modeling of energy systems was carried out together with all partners. Based on this, methods were identified that can significantly support the defined work process of modeling. At LTT, methods for piecewise linear regression were adapted and further developed to create models of components of an energy system based on the input and output data of the components. Methods for both one-dimensional and multidimensional piecewise linear regression have been further developed so that all typical components of energy systems can be modeled. Care was taken in the selection and further development of the regression methods to ensure that the complexity of the resulting models remained easily scalable. With the developed methods "AutoMoG" and "AutoMoG 3D" the models of the components of the energy system as well as the optimization model of the entire system with all balance equations are created automatically. The resulting software tool is freely accessible via the Git repository.

 

Relevant Publications

• Kämper, A., Leenders, L., Bahl, B., and Bardow, A. (2021). AutoMoG: Automated Data-Driven Model Generation of Multi-Energy Systems Using Piecewise-Linear Regression. Comput. Chem. Eng. 145, 107162. doi: https://doi.org/10.1016/j.compchemeng.2020.107162
• Kämper, A., Holtwerth, A., Leenders, L., and Bardow, A. (2021). AutoMoG 3D: Automated Data-Driven Model Generation of Multi-Energy Systems Using Hinging Hyperplanes. Front. Energ. Res. 9, 719658. doi: https://doi.org/10.3389/fenrg.2021.719658

 

Sponsor

Copyright: © BMWi

This project was funded by the German Federal Ministry of Economics and Energy.