Master's Thesis Laurin Oberkirsch
Development of modular simulation models to support real-time operation optimization of energy systemsCopyright: EBC
For real-time operation optimization, fast and sufficiently accurate models are necessary. The Aixlib, a Modelica library established by the Institute for Energy Efficient Buildings and Indoor Climate of the RWTH Aachen University, comprises several models of heating, ventilation, air conditioning and refrigeration (HVACR) systems. Those models are based on the physics in a grey-box manner. Components of HVACR systems usually exhibit hybrid system behaviour. Thus, continuous states are coupled with discrete state changes. Even though the Aixlib models represent the components in detail, the precision is time-consuming. Therefore, a more data-based three-layered hybrid model concept is developed throughout this thesis. On the top layer, a hybrid automaton is selected depending on the inputs. The state is determined on the middle layer. On the third layer, the outputs are computed according to stored data tables associated with certain transfer functions. Exemplary Aixlib models are translated into the model concept by the two steps: System identification and model recreation. During system identification, the step responses of the Aixlib models are classified for different inputs by detected events. For each classification, transfer functions are identified by several identification methods. Afterwards, the hybrid three-layered models are rebuilt in Modelica as well as in Python. As a result, run rime reductions of up to factor 220 are achieved compared to the original models. Concluding, the models are employed in an model predictive control (MPC) strategy. The MPC provides the control signal for boiler and heat pump of a single-family house equipped with a photovoltaic system. The objectives are minimising costs and switching operations. Thereby, energy costs amounting to 14.6 % can be conserved in comparison to the conventional hysteresis control.