Towards model partitioning automation of distributed real-time power systems simulations of distribution networks
Bogdanović, Milica; Monti, Antonello (Thesis advisor); Benigni, Andrea (Thesis advisor)
1. Auflage. - Aachen : E.ON Energy Research Center, RWTH Aachen University (2023)
Book, Dissertation / PhD Thesis
Dissertation, RWTH Aachen University, 2023
Modern power systems are large, complex systems with distributed dynamic loads, renewable energy sources, and active storage, and the energy transition is fundamentally changing them. Renewable energy sources production will be dominant world’s electricity production by 2050, cutting carbon emissions and helping to alleviate climate change. Integrating these sources into the power system has become an ongoing research and industry topic. With the increased number of renewable energy sources connected to the distribution grid, the power system distribution network is encountering a radical change in its dynamics. With the rising demand for the integration of new active elements in the distribution system, the need for real-time power system simulation of distribution networks is increasing. Real-time simulation of power systems is an important tool for grid analysis together with testing of the devices in the real-time environment known as Power Hardware-in-the-Loop. However, the increased number of devices and size of the grid simulated in a real-time environment pose new challenges concerning simulating the model on just one real-time capable simulation unit. As result, it is becoming more and more typical to use distributed real-time power system simulation to overcome the limits given by a single target, rack, or CPU core. Partitioning the system model and simulating the model subsystems on multiple real-time simulation units is referred to as distributed real-time power system simulation. Distributed real-time power system simulation is used when we have a real-time simulation of a system that cannot be simulated on just one target, rack, or CPU core of a real-time simulator. The key challenge in developing distributed implementation is to give criteria for performing the model partitioning and assigning the subsystems to different simulator units. The currently existing methods for model partitioning of real-time power system simulations are not addressing distribution system model partitioning for distributed real-time simulations. As matter of fact, a comprehensive analysis and partitioning simulation approach is still missing in the literature. This dissertation aims at presenting an extensive decoupling point impact analysis to show the effect of partitioning and delays occurring between subsystems on the distributed real-time simulation performance. The thesis analyses the decoupling point selection impact on the system eigenvalues of the separately discretized subsystems and shows that by observing the right and left eigenvectors of the system, the monolithic system dynamics preservation in the decoupled system can be observed. This thesis identifies challenges when developing partitioning methodology for distributed real-time simulations and shows that model partitioning regarding fidelity of the distributed simulations is difficult to generalize, and an analysis of the fidelity calculation impact on the decoupling point decision is conducted. Moreover, a non-intrusive model partitioning methodology for preserving distributed system stability as the main distributed simulation issue is developed and verified on example systems and developed realistic distributed power system model. The decoupling point impact study and stability methodology developed in the thesis represent a research step forward towards the design of automatized solutions for the partitioning of real-time simulations of distribution power systems. The contributions of the thesis are an extensive literature review of model partitioning in real-time simulations with classification missing in the literature, analysis of the fidelity calculation impact on decoupling point decision, model of the distributed system simulation in the state-space domain, analysis of the decoupling point selection impact on the system eigenvalues of the separately discretized subsystems, stability methodology of distributed simulation partitioning acting non-intrusive on critical system modes and application of the stability methodology to the developed simplified distribution system model in state-space domain.
- E.ON Energy Research Center 
- Institute for Automation of Complex Power Systems