摘要

This paper proposes a two-stage decentralized optimal reactive power dispatch (D-ORPD) framework, which considers the network partitioning influence in distributed optimization. The first stage is to divide the distribution networks (DNs) into several high-intra-cohesion and low-coupling sub-networks while the mathematical model is constructed based on partition indexes. This optimal partition designation reduces information exchange between adjacent sub-networks. Then, with the reduced information exchange, the distributed computation would be accelerated. Based on the network optimal partition, the D-ORPD model is constructed in a "decomposition-coordination" pattern during the second stage. To enhance the consistency between distributed and centralized optimization, the D-ORPD is improved by the proposed virtual load, which simulates the load characteristic of sub-networks. The alternating directions method of the multipliers (ADMM) algorithm is utilized to solve the mathematical model effectively. Moreover, case studies on the PG&E 69 bus test system and the IEEE 123 bus test system are executed by the MATLAB platform to demonstrate the validity and effectiveness of the proposed method.