Author(s): Hong Wang; Osman Ahmed; Kyle Desomber; Colin Sasthav; Pal-Tore Selbo Storli; Ole Gunna Dahlhaug; Hans Ivar Skjelbred; Ingrid Vilberg

Linked Author(s): Ingrid Vilberg

Keywords: Daptive learning; Digital twin; Hydropower systems; Modeling and simulation; Recursive least squares

Abstract: This paper summarizes the dynamic modeling of hydropower systems for the development of digital twin (DT) for hydropower systems. The obtained modeling suite covers the penstock dynamics, turbine and generator dynamics, and linkages to the grid, where linearized models have been developed for various components in the NTNU testing system. In this context, a discretized input and output model for the turbine shaft speed control has been obtained as a starting point to build the adaptively learned models representing the relationship between the guide vane opening, shaft speed, and water head. This allows the establishment of adaptive learning strategy where the data from any reference hydropower generation unit can be used to learn the model parameters. To enhance the robustness of the online learning of model parameters, a modeling error dead-zone based recursive least squares algorithm has been developed. In terms of the synchronous generator, a standard dynamic model has been used. Both the real-time data driven modeling and synchronous generator simulation have been performed and desired results have been obtained.

DOI: https://doi.org/10.3850/978-90-833476-1-5_iahr40wc-p1289-cd

Year: 2023