In traditional distributed state estimation， a subarea does not consider the network model and operating state of its neighboring subnet when making its own estimation independently while making up for the previously neglected information only during coordination. To remedy such a situation， a new distributed state estimation based on decomposition and coordination equivalent to external network concentration is presented according to the multi-agent theory. The large-scale power system is decomposed into a number of overlapping subsystems to build a 2-dimensional federated multi-agent framework suitable for distributed state estimation. By establishing the external network concentration equivalent model all subsystems are enabled to consider their neighbors influence. Thus the estimate solution in each subarea can satisfy definite precision requirements even without outside coordination. A heuristic coordination and correction mechanism based on the blackboard model is founded and the information fusion rate accelerated. And the adjacent systems are subjected to the effect from non-boundary nodes. In addition， outer iteration is performed according to the analysis of the injected powers sensitivity by the quantity of the node states to avoid repeated estimate. Test results on IEEE 14-bus， 30-bus and 118-bus system each show the correctness and effectiveness of the algorithm proposed.