An Adaptive Spectral Clustering Approach to Dynamic Partitioning and Weak Node Identification of a Hybrid AC/DC Grid

Introduction: In a hybrid AC/DC receiving-end grid, the proportion of external power injection continues to increase. The receiving-end AC grid may develop new weak nodes in grid operation. To investigate the voltage violation problem caused by the high penetration of external power injection in multi-infeed hybrid AC/DC receiving-end grids, this study proposes an adaptive spectral clustering approach for dynamic partitioning and weak node identification in hybrid AC/DC grids. Methods: Considering the degree of coupling between the AC grid and DC grid, the indices of short-circuit ratio (SCR) and multi-infeed short-circuit ratio (MISCR) are used to describe the electrical distance. To reduce the voltage control complexity, the zonal voltage regulation capability is integrated into the improved similarity matrix. Based on the spectral clustering algorithm, an improved similarity matrix is proposed, incorporating the similarity of operational state, electrical distance, and zonal voltage regulation capability, to classify the grid into several partitions. The optimal number of partitions can be determined automatically by identifying the maximum eigengap. Finally, the node with the largest singular value entropy in each partition is identified as the weak node, allowing for the quantification of its impact on system stability under disturbances. Results: The modified IEEE 118 bus system yields results that validate the feasibility of the proposed method. Discussion: This work aims to automatically determine the partitions of the hybrid AC/DC grid and identify the weak node within each partition, considering the mutual influence of the AC and DC grids and the zonal voltage regulation capability, in order to facilitate further operation and control. Conclusion: The proposed method incorporates the SCR, MISCR, and zonal reactive power compensation capability. The SCR and MISCR are used to describe the correlation mechanism between AC and DC grids, while the zonal reactive power compensation capability is utilized to enhance the accuracy of partitioning for hybrid AC/DC grids. By incorporating the eigengap in the spectral clustering method, the weak nodes in each partition are identified for further operation and control of the hybrid grid.