Abstract:

In this paper, the node tearing method is adopted to convert the discrete reactive-power optimization model of large-scale power systems into a multi-zone decomposition one, and the nonlinear primal-dual interior- point method with discrete penalty is employed to solve the decomposition model and to further obtain reduced-order linear correction equations with a block structure. In weak coupling systems, the complete coupling of the correc- tion equations is implemented by setting the off-diagonal submatrixes to zero, and the algorithm is of local linear convergence. However, in strong coupling systems, both the approximate Newton directions and the decoupled dia- gonal matrix, which are computed by the method similar to handling weak coupling systems, are respectively taken as the initial values and the preconditioner when solving the linear correction equations using the GMRES algo- rithm, thus resulting in good convergence and high calculation speed of the algorithm. The proposed algorithm is fi- nally applied to a 1062-bus and a real 538-bus systems, with its effectiveness being verified and some practical de- coupling criteria being presented.

Key words: power system, reactive power optimization, decoupling, nonlinear primal-dual interior-point method, discrete penalty, generalized minimizing residual method, approximate Newton direction