The SynErgie project, funded through the German federal research programme, aimed to increase the flexibility of air‑separation units by developing new radial compressor stages equipped with innovative aerodynamic control elements. The overarching objective was to enable the compressors to adapt rapidly to fluctuating energy supply while maintaining high efficiency. The project was carried out from 1 November 2019 to 30 June 2023 by a consortium of the Technical University of Berlin and the industrial partner MAN Energy Solutions, under the umbrella of the SynErgie consortium. The first funding phase focused on concept development and simulation, whereas the second phase, which is reported here, involved the construction of a single‑stage prototype, experimental validation, and the refinement of the process‑control concept for multi‑stage compressors.

Experimentally, the first series of tests on the single‑stage prototype was postponed by roughly one year due to necessary modifications at the MAN ES test field and pandemic‑related delays. The second series, originally scheduled for project month 27, was shifted to month 35. During data acquisition, several key measurement signals were corrupted or missing, preventing the planned classical approach of parameter identification, cross‑validation, and subsequent controller synthesis. To overcome this, a model‑based monitoring methodology was employed, allowing the compensation of missing signals and the automatic validation of the dynamic models. Based on these validated models, the controller was redesigned and successfully tested in the delayed second experimental series. The controller architecture combined an integral regulator that adjusts the mass flow through coordinated changes of inlet guide vanes and diffuser vanes with an extremum‑seeking regulator that adds a command to the diffuser vanes to minimise the energy consumption of the compressor. The extremum‑seeking strategy, originally a well‑known model‑free technique, was extended to a quasi‑Newton variant to achieve faster response times suitable for highly flexible operations. Although the quasi‑Newton extremum controller had only been evaluated in simulation during the first phase, its integration into the experimental setup demonstrated its feasibility.

Following the experimental validation, the process‑control concept was revised for multi‑stage, flexible compressors. The revised concept incorporated additional cooling units between stages and adapted the control logic to the increased number of actuators. A dynamic simulation that included a throttle valve was developed and used to test the updated control strategy, confirming improved robustness and operational feasibility in an industrial environment. The real‑time implementation of the controller was carried out on a Speedgoat Performance Target Machine equipped with an Intel Core i7‑7700K @ 4.2 GHz CPU, 32 GB RAM, an FPGA‑based I/O module, 16 analog input channels ±10 V, 8 analog output channels ±10 V, and 16‑bit 500 kHz ADCs per channel. The controller code was automatically generated from the Simulink model and deployed to the real‑time hardware, enabling rapid iteration and fine‑tuning during the experimental campaigns.

The project’s funding was fully allocated to the employment of a scientific researcher and project‑related travel expenses. The collaboration between TU Berlin and MAN Energy Solutions facilitated the transfer of academic research into a practical prototype, while the SynErgie consortium provided the necessary industrial infrastructure and test facilities. The final report documents the successful achievement of the project’s objectives, the demonstration of a flexible, energy‑efficient compressor control strategy, and the readiness of the process‑control concept for industrial deployment.