Explainable Grid Informed Model For Power System Transient Stabilisation

Demand response represents a cost-effective strategy for transient corrective control in power systems. However, evaluating transient scenarios that involve demand response programmes using traditional physical simulators is often computationally intensive, potentially resulting in suboptimal solutions. This paper proposes a data-driven grid informed approach to approximate transient stability based on the system initial conditions, fault information, and demand response control actions. First, a third-order structure-preserving model is introduced to capture transient dynamics. Second, the grid informed constraints incorporate both differential and algebraic constraints. Third, Shapley additive explanations are employed to interpret the contributions of different factors to final stability. Case studies conducted on the IEEE 9-bu, 39-bus, and 68-bus systems validate the proposed approach, demonstrating its high prediction accuracy and interpretability.