Multi-Level State Estimation and Optimal Dispatch Techniques

Abstract

The distributed nature of modern power systems encourages the application of multi-level computer control systems. In the present paper a number of decomposition methods are considered which utilise the sparse nature of power system problems, in order to obtain efficient multi-level computational schemes. The state estimation problem, in which the nodal voltages and phase angles are determined from noisy measurements on the system network, is solved using a modified form of network tearing. Although an exact diakoptical solution of the problem is possible, an alternative approximate method is presented which reduces the computational burden of the technique. Comparisons are made with the conventional solution of the problem by generalised inversion, and the tearing method is found to produce acceptable state estimates in less computing time and with a reduced storage requirement. The problem of optimal power dispatch which often succeeds state estimation in an on-line control scheme is also investigated. A new decomposition algorithm is proposed which solves the linearised dispatch formulation by means of sensitivities. The method is characterised by the transfer of cost and constraint information between the master and subproblems. Comparisons are made with the more conventional Dantzig-Wolfe decomposition principle, and examples are given which include a problem containing 100 generators. In the case of the very high dimensional problems occurring in power system control the process of decomposition and multi-level solution is shown to be very important, with particular advantage atcruing from the application of the algorithms described in the present paper.