A Testbed‐based Approach for the Resiliency Assessment of Multi‐Microgrids

Abstract

The urgently needed rapid decarbonization of the power system and related sectors such as mobility often drastically increases volatility and imposes significant challenges on network design and operation. Microgrids are commonly treated as one lever to tightly integrate volatile generation, improve power quality, and strengthen grid resilience. However, to compensate for events such as main grid/feeder outages, often considerable local generation reserves are needed. To exploit synergies between single microgrids including economic benefits and reduced reserve needs, the concept of multi-microgrids was introduced. Multi-microgrids jointly operate the individual microgrids in a distribution system. Due to the larger extent and tight control capabilities of the participating microgrids, multi-microgrids enable new fault mitigation techniques such as partial islanding. Still, the extended scope increases the overall complexity of the system and urges for efficient validation and testing environments. This work addresses testing and validation needs by presenting an advanced simulation-based resilience testbed with a strong focus on related engineering aspects.

In contrast to related works, the presented testbed uses large-scale simulations to cover a broad variety of operating scenarios and failure cases. To assess the performance of microgrid scheduling algorithms in detail, the scheduling-time information such as forecasts and day-ahead prices is independently modeled from the real-time information including measurement-based input profiles. All scheduling algorithms operate on forecasts only. A dedicated simulation run is conducted on the measurement information and setpoints from scheduling to compute the assessment results. The testbed architecture is designed to handle the large computational workloads of simulating extensive scenario sets by massive parallelization beyond the boundaries of a single computer. At the same time, a compact design is kept by formulating vectorized tasks on the input scenarios and encoding them in a graph-based representation, first. After the fine-grained compute-graph is defined, a separate task scheduler dynamically assigns the workload. Additional efforts are presented to seamlessly integrate the assessment in an automated software development toolchain and to efficiently manage the system description including any algorithm(s) under test.

An exemplary case study demonstrates the feasibility of the testbed and shows the benefits of the proposed system design. Since the testbed first computes a common set of input scenarios, multiple algorithms can be effectively compared on the same set of input conditions without repeating input/output-intensive pre-processing steps. Due to the graph-based architecture, further performance improvements can be included, and even large workloads are transparently scaled to multiple compute nodes. Additionally, this work showcases the integration of the assessment process into a software development platform. The architectural measures such as an advanced text-based input description and the scripting-friendly interface allowed to fully automatize the simulation runs and to precisely trace the configuration that led to a specific outcome. Given the detailed architecture and the validation thereof, another step towards a more efficient development of multi-microgrids and the improvement of system resilience is made. It is expected that the detailed reflection on the software architecture provided in this work guides further efforts in implementing simulation-based assessment methods and in providing enhanced engineering workflows.