Master's Thesis Tim Bartikowski


Development of a model-based, programmable logic controller concept for a heat pump hardware-in-the-loop test rig

smith predictor Copyright: EBC Block diagram of the Smith-Predictor

The Institute for Energy Efficient Buildings and Indoor Climate tests heat pump systems under reproducible, dynamic boundary conditions on a Hardware-In-The-Loop test bench. The energy system supplies or receives heat during the experiments. This is achieved by admixing cold or hot water from supply pipelines via mixing valves. Currently a PLC controls the mixing with PI controllers. Changing volume flows and temperature ranges as well as the coupling of the hydraulic circuits cause disturbances in the regulation. This leads to insufficient results regarding control deviation.

The present work identifies the hydraulic properties and disturbances of the test bench control. Based on this system knowledge, controller concepts are developed on the digital twin of the Hardware-In-The-Loop test bench. These concepts take into account the physical structure and the hydraulic properties of the test bench. The first developed controller concept reacts to the disturbances of the controlled system by means of feedforward control. In addition, a model of the controlled system is developed in order to improve the dynamic control behavior by means of a variable dead-time compensation with a Smith-Predictor. Finally, a model-predictive control of the temperature mixing is designed.

Instead of using structured text, the controllers are implemented in Python and connected to the PLC via an ADS-communicator. This higher-level programming language makes the more complex model-predictive control algorithms possible. A comparison of the controllers on the digital twin of the test bench shows that the Python controllers improve the control quality through the feedforward control. Compared to the original PI controller, the control accuracy increases by 50 %. The addition of the model-based dead time compensation leads to an additional improvement. The model predictive control needs significantly more computation time and cannot compensate for a permanent control deviation due to inaccuracies in the map. Tests on the real heat pump test bench confirm that the feedforward control increases the control accuracy. Finally, an outlook gives further improvement approaches of the test bench control.