An Adaptive Clustering-Based Partitioning Method of Large-Scale Power Grid for Data-Driven Dynamic Security Assessment

The accuracy of the data-driven dynamic security assessment (DSA) of power systems considering fault locations is closely related to the system scale. The dynamic security assessment of the large-scale power grid has the problems of strong nonlinearity, high model complexity, and insufficient assessment accuracy. To improve the data-driven DSA accuracy of large-scale power grids, this paper proposes an adaptive clustering-based power grid partitioning method. Considering the electrical connection tightness between nodes, the power grid is partitioned based on the mean shift clustering algorithm. The node location is represented with electrical coordinates, which are input features of the mean shift clustering algorithm. To mitigate the imbalance of partitioning results, the mean shift clustering algorithm is improved by adjusting the high-dimensional sphere region. The effectiveness of the proposed power grid partitioning method is verified with a regional power grid of China.