Application Of The Improved Genetic Algorithm In Optimal Configuration Of The Filter

Recently, more and more power electronic equipments are widely used. And the harm form harmonic is getting much more seriously, which make the harmonic level rising quickly and the losing being larger. In order to develop the economics and power electronic, we have to pay more attention to the research of harmonic.When there is harmonic source in the system, we will analysis and calculate load flow of system. Concretely, when analysis harmonics of power supply system, not only analyse harmonic source type and harmonic source characteristics, but also analyse harmonic currents distribution and the number of harmonic voltage of each node produces. That process is harmonic load flow calculation. Know the characteristics of harmonic to be able to optimize the allocation of filters.There are lots of methods could be used in reducing harmonic. In this thesis, an effective method which filter harmonic by allocating filters at the source of harmonic. Now the common method is allocating active power filter or passive power filter alone, in this thesis, a rational optimal planning model of harmonic filters is presented. Take the total investment cost of allocating all the filters as the target function, optimize the allocation position and parameter of the filters. Compare to other optimizing algorithms, genetic algorithm have the distinctive superiority in irregular optimizing, it could not only find out the global optimal answer but also fall into part optimal answer easily. To optimize genetic algorithm, roulette and optimum keeping are adopted in this thesis. At the same time, the probability of selecting and crossing is improved, solve the optimizing problems with an improved genetic algorithm, then get the global optimal answer.In this thesis, according to the rational optimal planning mathematical model, develop program with MATLAB to solve a classical case of 12 points system and compare the answer with other answers of several different conditions. The result shows that the rational optimal planning model and arithmetic are feasible and effective. The improved genetic algorithm with higher stability could convergence quickly and won't fall into part optimal answer.