Predictive Analysis for Safe and Optimal Operation of the CoBra High-Temperature Heat Pump

Kurzfassung

High-temperature heat pumps are a key technology enabling the decarbonization of conventional heating processes. The CoBra (Cottbus Brayton) high-temperature heat pump is a novel pilot plant, based on the reverse Brayton cycle, developed within DLR which achieves desired temperatures of more than 250 °C. However, during experiments, the pilot CoBra could no longer operate properly once the temperature of hydraulic oil, which is required to lubricate the compressor, exceeded the safety-relevant temperature limit. Emergency shutdowns were triggered to prevent compressor and components failure. Notably, the oil temperature rise occurs within a few minutes and shows no clear prior indication. As a consequence, the experimental objectives could not be fulfilled. This study develops a predictive model for the compressor oil temperature based on historical experimental data such as operating pressure and compressor speed, using deep learning techniques. The model aims to provide operators with early warnings during experiments and to recommend appropriate countermeasures that can prevent system shutdowns. Furthermore, this work investigates long-term operational strategies for the CoBra under varying input conditions through dynamic simulation. The oil temperature prediction model and process simulation model are combined into a unified framework, and a dynamic optimization problem, with rolling horizon approach, is formulated to simultaneously address process performance and machinery safety constraints. IPOPT is selected as the solver for this optimization problem as a proof of concept. The proposed method enables safer and more efficient operation, improves cost efficiency and extends equipment lifetime.