ABSTRACT The high penetration of renewable energy exacerbates stochastic fluctuations on both the generation and demand sides. It results in spatiotemporally frequent static security violations and requires high‐frequency assessment and adjustment of the operation mode. However, conventional offline verification and manual adjustment methods fail to meet this requirement. Therefore, this paper proposes an interpretable machine learning‐based intelligent high‐frequency static security adjustment method for power system operations. First, a multitask XGBoost collaborative framework‐based static security analysis model is developed by utilising power flow characteristics under the base case ( N state) and various N − 1 contingency scenarios ( N − 1 states) as input–output features. It can efficiently and accurately evaluate the power flow violation status under both the N state and various N − 1 states. Second, when operational modes exhibit power flow violation status, an intelligent adjustment methodology based on the SHAP (SHapley Additive exPlanations) interpretability framework is developed to quantify input feature contributions to security risks and identify key factors and their causality‐driven mechanisms. Then, it generates transparent decision logic and effective adjustment strategies. Finally, case studies on the IEEE 30‐bus system, a practical 341‐bus grid and a practical 1834‐bus grid validate that the proposed method not only achieves efficient and accurate static security assessment but also provides interpretable adjustment schemes for diverse operating states, thereby offering operators clear decision‐making guidance.
Duan et al. (Thu,) studied this question.