A sensitive point mining method for power angle stabilization control based on improved convolutional neural network

During the critical period of energy transition, the strategic goals of “Carbon Neutral and Peak Carbon” provide a direction for the sustainable development of new power systems. However, the increasing complexity of the grid architecture and the diversification of equipment have made traditional physical models less applicable, posing significant challenges to power system stability control under large disturbances. This paper proposes a sensitivity-based method for identifying sensitive points in rotor angle stability control. Firstly, a rotor angle stability margin evaluation metric is constructed based on the Extended Equal Area Criterion (EEAC). Secondly, a data-driven rotor angle stability margin assessment model is developed using an improved Convolutional Neural Network (CNN) to assess the effectiveness of control measures at candidate action points in improving system stability. Finally, a control sensitivity is defined, and a detailed emergency control workflow is provided. The proposed method is validated through an actual power grid, demonstrating its practical applicability.