Usually， power system calculation requires distributed computing not only with a good performance of having quick convergence and less communication times， but also capable of protecting the privacy of different dispatch centres and coping with legacy problems， which implies that only boundary conditions should be exchanged in calculation. This paper analyses the iterative decomposing and coordinating process from the control engineering point of view. Based on the control system model and aided by the self-tuning control method， the performance of the currently used distributed computing method is improved， especially in terms of convergence and communication times. Two simulation cases are used to validate the effectiveness of the proposed method.