Power System Short-Term Voltage Stability Assessment and Margin Estimation via Neural Barrier Function

Short-term voltage stability (STVS) has emerged as a critical concern in modern power systems due to increasingly complex transient post-fault dynamics. Effectual STVS assessment and margin estimation are essential. Existing data-driven STVS analysis approaches tackle the flexibility issue; however, they often lack a strong mechanism basis. To address these issues, this paper introduces a novel data-mechanism-integrated framework utilizing neural network-based barrier functions (NBFs) for STVS assessment and margin estimation. In particular, we propose a neural barrier certification for STVS analysis that is trained from post-fault response trajectories and constrained by LaSalle’s invariance principle, ensuring theoretical validity without requiring explicit dynamic models. To enhance the applicability of the proposed methodology in large-scale systems with limited observability, a state expansion approach by reconstructing latent system dynamics via system energy is introduced in this paper. Additionally, we raise an NBF-based stability margin estimation approach via a fast binary search method, realizing flexible and timely STVS margin estimation. The proposed methodology preserves the interpretability of mechanism-based methods while maintaining flexibility and computational efficiency. The flexibility and effectiveness of the proposed NBF-based framework are demonstrated via numerous tests on the two-bus system and the IEEE 39-bus system. Note to Practitioners—This paper is motivated by the problem of short-term voltage stability (STVS) analysis in power systems, but it also can be applied to other safety-critical assessment or decision-making tasks in practical systems. STVS is a critical concern in modern power systems since the enlarging scales bring more complex post-fault dynamics to the power system. Existing STVS analysis approaches either rely on accurate system modeling and simulation that limits their flexibility or employ data-driven techniques that weaken their reliability and interpretability. This paper develops a novel data-mechanism-integrated framework utilizing neural network-based barrier functions (NBFs) for STVS assessment and stability margin estimation. In our work, the post-fault power system is modeled as a nonlinear system with a forward-invariant safe set defined in state variable space. We then introduce a neural network to approximate the barrier function of the post-fault system, which is essential for determining the critical safety status of the post-fault system. To enhance the applicability of the proposed methodology in practical power systems with limited observability, we raise a state variable expansion approach based on system energy. We also develop a fast binary search method for estimating the stability margin, i.e., the degree of system stability. Experimental results demonstrate the effectiveness of the proposed framework. This work provides a fast, reliable, and flexible STVS analysis approach for practical power systems. In future research, we will address the NBF-based stability analysis and safe decision-making for emergency control in power systems.