A new and efficient diakoptical static state estimation algorithm has been proposed. The main objective of developing this algorithm was to obtain an estimator with an improved performance in terms of robustness, accuracy, online CPU time, core memory occupation, intercomputer data transfer, scope for sparsity exploitation, and localisation of bad data effect. In the developed method, a system has been divided into a number of subsystems to take advantage of low dimensionality, sparsity, fast decoupled estimation and parallel processing. The co-ordination among the subsystem level estimators has been done, in each iteration, to improve accuracy of estimates at the cost of little CPU time and nominal data transfer, by the transformation of the tie line flow measurements into voltage drop between subsystems' slack busbar, and the involvement of a small-sized, sparse and constant coefficient matrix with real elements. Saving in CPU time and convergence have also been enhanced by making each iteration of the diakoptical estimator unified, within which no more iterations, either at the subsystem level or at the coordinating level, have been involved. The results of a comprehensive comparative study of the performance of the developed method against the integrated fast decoupled method have also been presented.