Reactive Power Planning Based On Grid Partitioning And Genetic Algorithm Optimization

Reactive power optimization is of great importance to voltage quality guarantee, power loss minimization and thus security and economic operation of power system. Detail discussion and partition of it is given according to the research contents, methods, time span and objective function etc. The method applied here, Genetic Algorithm (GA) for reactive power planning problem is summarized fully and comprehensively. Hence, optimal reactive power compensation planning is studied in detail in this thesis with intelligent method, such as Bender's cut, etc., integrated. The research includes following contents:(1) According to the characteristic of reactive power planning problem of power system, A algorithm based Bender's cut and SGA (Simple Genetic Algorithm ) are used in the optimization method. It is the typical mixed and integral non-linear planning problem that there are many control variables and the complexity is high. It is suitable for solving mixed and integral non-linear planning problem with the Genetic Algorithm; at the same time, for simplifying algorithm , the reactive power planning problem is divided into two sub-problem: planning sub-problem and operation sub-problem based Bender's cut. The object function of planning sub-problem is the minimum of the reactive power equipment and the loss of power, this problem use the SGA algorithm to resolve the position, size , type of composition equipment; Operation sub-problem is actually a optimal load flow problem when the load is fixed.(2) The operation voltage is mainly controlled by the balanced of reactive of this area, Conveying the reactive power will cause greater voltage difference on the first and end of the circuitry, the power loss will increase caused by the transmission of the reactive power. So " power system have work voltage technology lead" point out that the reactive power should balanced by its area. So, this principle of the foundation of this text, Firstly the electric power network is divided some sub-area, then accordingto some principles, compensate candidate nodes are selected as initial population of SGA algorithm in each district. Each sub area have enough reactive power store.(3) The accuracy of the result of genetic algorithm depends on whether the initial population can include all information of the result of the space, and does not lose the fine properties among them during the process of evolving. Initial population should scatter among solving space try one's best. This text chooses the initial population with two kinds of methods : One is key node of each district that divided according to electric distance, this node can not merely reflect all nodal voltage levels in this district, and it is the electric closest node of other nodes in this district too, so it is right that key node as compensate candidate; Another kind is arranged in an order according to the sensitivity in each district, the node chosen, in the real power system runs , the most effective voltage regulation is generally adopted in the control point of the supreme sensitivity on the spot. If it is insufficient to control on the spot, choose the control point with high sensitivity once to assistant control. So it is right choosing the node with highest sensitivity in each district as compensate candidate node. To sum up, combining the node that two kinds of methods elect together, genetic algorithm initial population is selected.(4) As a basic electric calculate of studying the stable state running situation of power system, the result of power flow, no matter the analysis and research of the existing systematic operation way, the analysis and compare of power supply, is very important. Equally, the power flow calculation is used many times in the optimal reactive power compensation planning problem, for improving the speed of power flow calculation, this text compares and analyses preconditioning methods using MATLAB, the result of simulation indicates P-Q method is the most effective method so far; and the more the system bigger, the better of the effect of the P-Q method.