Author(s): Weichao Ma; Wei Zeng; Xu Lai; Jiandong Yang
Linked Author(s):
Keywords: Hydropower stability analysis Physics-informed neural network system identification model calibration
Abstract: Ensuring the stability of hydropower generating systems (HGSs) is important for the flexible operations of HGSs nowadays. Achieving accurate predictions of dynamic processes for HGSs remains a key challenge due to the simplification of the physics-based models. To address this issue, this paper proposes a physics-based and data-driven method for the reconstruction of the hydraulic parameter variations during dynamic processes of HGSs. The new method formulates a physics-informed neural network (PINN) by incorporating both observed data and physical functions of HGSs. This enables the PINN to learn and explore hidden information of the dynamic processes embedded in the observed data. The trained PINN can then be used to predict the complete dynamic characteristics of HGSs, even using simplified physical functions. A numerical case study showed the effectiveness and advantages of the proposed approach.
Year: 2025