In order to be more applicable and scalable

distributed power flow algorithms are required to utilize only simple and fundamental information such as states of boundary nodes. In this paper

schedule equations of distributed power flows are modeled as a set of implicit nonlinear equations. To solve these equations

an adaptive preconditioned Jacobian-Free Newton-GMRES (m) method is proposed. Based on it

a new distributed power flow algorithm is constructed

which only requires exchanging states of boundary nodes during cooperative computation. Test results of IEEE standard systems and a large real power system show that the proposed distributed power flow algorithm has fast convergence rate. As using simple data interfaces

it can integrate heterogeneous computation resources easily. The continuous preconditioning method makes it more feasible for on-line distributed power flow calculations.