A decomposition approach to optimal reactive power dispatch in large-scale power systems

Power system operation is aimed at reducing system losses and minimizing the operational cost of large scale systems while satisfying network performance requirements, both in normal states and under contingency situations. In this project, the procedure for the reactive power optimization consists of the solutions for investment and operation subproblems. The global solution is an iterative process between these two subproblems using the Bender decomposition method. In the investment subproblem decisions for the capacity and location of new reactive sources are made. These decisions are used in the optimization of the system operation. The outstanding features of the proposed method are represented by the fact that it does not require any matrix inversion, will save computation time and memory space, and hence can be implemented on very large scale power systems. The method employs a linearized objective function and constraints, and its approach is based on adjusting control variables which are tap positions of transformers and reactive power injections. Linear programming is used to calculate voltage increments which would minimize transmission losses, and adjustments of control variables would be obtained by a modified Jacobian matrix. This approach would greatly simplify the application of Dantzig-Wolfe decomposition method for solving the operation subproblem. According to the mathematical features of the Dantzig-Wolfe method, a multi-area approach is implemented and system equations are decomposed into a master problem and several subproblems. The master problem is formed by constraints which represent linking transmission lines between areas. Two updated techniques are incorporated in the method to enhance the optimization process which would save additional computation time and memory space. The proposed method is applied to the IEEE-30 bus system, a 60-bus system, a 180-bus system and a 1200-bus system, and numerical results are presented. These results verify the superiority of the proposed method over the conventional ones.