A Balanced Partitioning Method and Its Application on Distributed State Estimation in Power Grids

The performance of grid partitioning methods has a significant impact on the results of distributed state estimation (DSE). Most existing grid partitioning methods struggle to balance node distribution while maintaining partition quality and network connectivity, which may lead to low resource utilization and reduced accuracy in DSE. This paper proposes a novel balanced grid partitioning method, aiming to promote effective DSE. To ensure partition balance and grid connectivity, a multiobjective fitness model for partitioning is formulated based on the correlation of electrical modularity, node distribution balance factor and partition density influence factor. The grid partitioning problem is a non-convex discrete optimization problem. To obtain a solution, an improved particle swarm optimization algorithm is employed, incorporating graph theory and spectral clustering to enhance solution diversity and accelerate convergence speed. To address the vulnerability of large-scale grid partitioning to topological connectivity, a partition quality assurance mechanism is proposed to automatically detect and reassign outlier nodes. Numerical analysis validates the method's effectiveness, including multidimensional performance evaluation and comparative experiments with multiple power grid partitioning cases. Further simulation results when applied to a DSE scenario demonstrate significant advantages in the accuracy of the proposed scheme by comparing to existing schemes in the literature.