A Digital Twin Framework for Transient Stability Assessment Based on Knowledge Distillation

In the context of the digital transformation of power systems, various new sensors are capable of acquiring massive real-time data from the power grid, providing an opportunity for the development of digital twins in power systems. This paper proposes a transient stability assessment digital twin framework based on knowledge distillation. Addressing the challenges of high computational resource consumption and model deployment difficulties associated with existing high-precision complex models in transient stability assessment, the knowledge distillation method is introduced for model lightweighting. Simultaneously, the proposed continuous learning online updating strategy enables adaptive online updates for complex operating conditions and structures, enhancing the learning capability for online additions of diverse samples. Using Transformer as the transient stability assessment model, the effectiveness of the proposed approach was validated on the IEEE 39-bus system.