Hierarchical Multi-Layered Sparse Identification for Prediction of Non-Linear Dynamics of Reconfigurable Microgrids

The increasing growth of the global energy demand has become one of the key factors that necessitate the development in the field of distributed energy resources (DERs). The interconnection of the DERs with the main grid makes the system extremely sensitive to disturbances, as they decrease the overall system's inertia. These disturbances are primarily caused by variations in the loads connected to the distribution system and environmental conditions affecting DER generation. The system dynamics exhibit strong non-linearity, requiring the use of various controllers to stabilize microgrids and maintain steady-state operation. Predicting future states is challenging yet crucial in power systems, as it enables the determination of control strategies to enhance transient stability following contingencies. This paper proposes a hierarchical multi-layered sparse identification technique to comprehend the system and forecast transient dynamics in different microgrid operation modes. The developed algorithm employs a multi-layered structure, which reduces the overall computational cost while ensuring accurate model dynamics. In the primary layer, the functions influencing the system dynamics are derived from measured data. These primary layer terms are then fitted into the secondary layer to determine the precise system dynamics under different disturbances. Numerical examples have been presented in this paper to validate the effectiveness of the proposed algorithm. The developed method proves particularly valuable in re-configurable and scalable networked microgrids where the system structure and the associated controls frequently change.