Online high-dimensional security boundary rapid generation framework considering over-limit risk for different operating states

The Thermal Security Region is a key tool for assessing the security margin of random operating modes in power systems. However, with the large-scale integration of renewable energy and the ongoing expansion of the power grid, operating modes have become increasingly complex and diverse. As a result, low-dimensional security boundaries constructed from historical data struggle to meet the current demands in terms of adaptability and timeliness. This study proposes an online generation method for high-dimensional security boundaries based on power flow over-limit risk, while considering the impact of different operating states on the accuracy of reduced-dimensional security boundaries. Firstly, the impact of different operating states on boundary fitting in the power system is identified using unsupervised clustering. Next, a security boundary set is constructed, and a risk prediction model based on AdaBoost.M2 is employed to identify potential boundary sets that may exceed limits, using confidence scoring. Finally, these selected security boundary features are utilized for generative adversarial network training, enabling the high-dimensional security boundaries to comprehensively account for multiple scenarios. This method enhances the credibility of the decision-making process through mechanism analysis while improving the speed and accuracy of high-dimensional security boundary generation. In tests using operational data from different states within the IEEE 39-bus and HZPG systems, it achieved an accuracy of over 95%. In terms of timeliness, the risk prediction component satisfies real-time application requirements, and the boundary generation process can be performed online.