Three-level hierarchical microgrid control—model development and laboratory implementation
Problem and Research Question
A grid-connected microgrid must coordinate economic scheduling with faster frequency support and device-level setpoint tracking. The central question is: can these functions be integrated across their different timescales and demonstrated on a physical laboratory microgrid?
Method
This paper developed a three-level hierarchical control architecture:
- Energy-management system — economic dispatch at a 15-minute resolution.
- Frequency controller — MPC with an observer, using a 20-second prediction horizon and updating commands on a faster control cycle.
- Local controllers — classical tracking of the power setpoints issued to individual units.
The architecture was implemented on DTU’s physical SYSLAB grid-connected microgrid and coordinated through SCADA and MOSAIK. This is a laboratory implementation, not hardware-in-the-loop with a wholly simulated real-time microgrid.
Personal Contribution
Frederik Banis was a co-author. He contributed to the laboratory implementation and experimental validation, working on the integration of control layers and laboratory setup. The first author (Mateo Beus, University of Zagreb) led the model development and overall architecture design. Niels Kjølstad Poulsen (DTU) and Hrvoje Pandžić (University of Zagreb) provided supervision.
Validation and Key Results
- The three optimization and control layers were integrated on a physical, grid-connected laboratory microgrid.
- The experiment showed frequency-responsive rescheduling followed by device-level setpoint tracking.
- It did not demonstrate islanded operation, voltage restoration, inverter droop control, source switching, or a comparison with simulation-only control.
The method was published in Electric Power Systems Research (2020), a peer-reviewed journal.
Related Outputs
- PhD thesis: Efficient Operation of Energy Grids, DTU, 2020. DOI: 10.11581/DTU.00000334
- Related work: Load Frequency Control in Microgrids using target adjusted MPC, IET RPG, 2019. DOI: 10.1049/iet-rpg.2019.0487
Collaborators and Institutions
- DTU Compute, Technical University of Denmark — Niels Kjølstad Poulsen (supervisor)
- University of Zagreb — Mateo Beus (first author), Hrvoje Pandžić (supervisor)
Status and Next Steps
Status: Published (2020). A key contribution is the integration of different optimization and control timescales on a laboratory microgrid.
Transfer to current research: The closed-loop experimentation architecture — measure, infer, decide, act, with physical validation — is the same pillar that now drives human-supervised autonomous bioassay platforms. The shift is from energy hardware to biomedical hardware, but the closed-loop validation methodology is shared.
See the related project page for the broader research context and the Methods page for the closed-loop experimentation pillar.