Research Of Optimal Reactive Power Compensation For Distribution Power Grid Based On Improved Genetic Algorithms

With the problem of economical running of power grid being more and more attached importance to, how to reduce the grid loss on the premise that the voltage quality is guaranteed has become an important problem which the electric power department has to urgently research and resolve. The reactive power optimization compensation is an important method to resolve it. In the light of the present situation of reactive power optimization compensation, some research has been done in the Genetic Algorithms (GA) applied in reactive power optimization compensation for distribution power grid, and some reforms have been made on the base of preceding researches which have improved its optimization ability.In this thesis, all kinds of power flow algorithms and reactive power optimization algorithms (including conventional algorithms and artificial intelligence algorithms) have been analyzed and compared and their advantages and defects have been concluded.The thesis has summarized the reactive power compensation, and has concluded the principle and the basic ways of it.Aiming at the disadvantage that the conventional power flow algorithms can't converge well in computing distribution power grid flow, I have applied the forward/backward sweep method. For the meshed distribution grid, I have used the superimposition principle to compute its flow after it is decomposed. This method is simple in principle and also has high calculating efficiency and superior convergent property.On the premise that the voltage quality is guaranteed, the mathematics model for reactive power compensation is founded on principle that the compensation power and grid loss are least and the comprehensive economic benefit is great.On the basis of Simple Genetic Algorithms (SGA), a series of improvements have been put forward. By applying decimal coding and preserving the best individual, the computing time is reduced. The nonlinear-rank select based on Roulette Wheel, which has embodied the priority-rule, has overcome the shortcoming of blindness for Roulette Wheel.The adaptable Genetic Algorithms can adjust the rates of crossover and mutation and provide the best for a specific result in the light of the specific evolution circumstances. They promote the stability and avoid to partial optimization. By introducing the Chaos Operator, it has overcome the defect of precocity for SGA, for its particularly inherent randomness and ergodicity to skip the partial optimization. At last, the results of real cases demonstrate that it is superior to standard GA.