In this paper

a model of a large-scale optimal power flow (OPF) under voltage grading and network partition and its algorithm is presented. Based on the principles of open loop operations

the node injecting current method is used to divide the large-scale power grid into voltage grading and district dividing structures. The power network is further divided into a high-voltage main network and several subnets according to voltage levels of 220 kV. The subnets are connected by means of boundary nodes

and the partition model is solved using the improved approximate Newton direction method

which achieves complete dynamic decoupling simply by exchanging boundary variables between the main network and the subnets. A largescale power grid thus is decomposed into many subnets

making the solution of the problem simpler and faster while helping to protect the information of individual subnets. The system is tested for correctness and effectiveness of the proposed model

and the results obtained are matched in real-time. Finally

the algorithm is seen to have good convergence while improving calculation speed.