Distributed-Computing-Based Reacvtive Power Optimization Of Power System

In a power system, reactive power optimization is not only a critical problem in voltage stability and economical operation but also an important measure to improve the voltage quality. According to the fact that the mathematical model of a real power system includes both discrete and continuous variables, a method to get the exact solution of reactive power optimization is presented. By converting discrete variables into continuous variables, applying the primal-dual interior method to global optimization and the branch-and-bound method based on BFS (breadth-first-search) to deal with the discrete variables, the exact solution obtained is more reasonable than the conventional result. With the increasing of system scale, the calculation time will badly increase, this paper proposes distributed computing method to reduce the execution time so that the global optimal solution could be found with high efficiency. The paper proposes distributed computing program based on a PC-cluster which has a high performance and price ratio with Matlab distributed computing engine (MDCE) and distributed computing toolbox (DCT).By simulating in Matlab distributing environment with PC-cluster on three test system, large-scale of optimal reactive power flow optimization problem is solved successfully. Comparing with serial computing and traditional method, the results verify that the proposed distributed algorithm can save much computing time and it is also very effective. The results of simulation also indicate that distributed computing PC-cluster system based on MDCE and DCT is a promising distributed parallel computing platform for large-scale power system computing problems.On the other hand, this paper shows the exact solution on the unit commitment which is solved by interior method associate with primal-dual interior method.The exiting calculation result on an 10 units for 24 hours case compared with a lot of methods such as Semi-definite Programming, Lagrangian Relaxation Method, Genetic Method verifies that the method in use has better effect when deals with mixed-integer non-linear programming.