A Study On Reactive Power Optimization Based On Improved Primal-Dual Interior Point Method And Branch-And-Bound Method

Reactive power optimization in power system is not only an effective means to ensure power system operates securely and economically, but also one of the most important measures to reduce the transmission loss and improve the voltage quality. Theoretically and practically, the study of reactive power optimization in power system is of great significance. Due to the discreteness of transformer taps and the switching of shunt capacitance/reactance banks, the rigorous reactive power optimization is the problem of Mixed-Integer Nonlinear Programming (MINLP) involving both discrete and continuous variables. A mature and effective way to solve the problem hasn't been found yet till now. In the past, the math model of reactive power optimization was simplified from different aspects. A conventional method is to take discrete variables as continuous variables to deal with; when optimal results drawn, integers are obtained by making discrete variables to the nearest integer values. Consequently, unavoidable discrepancy exists between the optimal values obtained in this way and the optimal values in the real sense.A new algorithm for rigorous optimal power flow is presented in this thesis, which is based on Primal-Dual Interior Point Method (PDIPM) under perturbed Karush-Kuhn-Tucker (KKT) conditions and Branch-and-Bound Method (BBM ). The new algorithm fully takes the discrete variables' characteristics into consideration; seeks optimal solution by adopting PDIPM and gets the approximate integers of discrete variables by using BBM; expands the idea of conventional interior point method in solving continuous variables nonlinear programming; successfully gets the precise results of Mixed-Integer Nonlinear Programming.BBM can get global optimality by dividing the feasible region, or dividing original problem and sub-problem. The process is to take discrete variables as continuous variables, then solve the sub-problem by PDIPM. PDIPM can solve continuous variables problem speedily. BBM 's cut-tree can solve the discrete variables problem precisely. So the mixed algorithm is effective to reactive power optimization.