Research On Distribution Voltage And Reactive Power Multi-Objective Optimization Based On Modified Genetic Algorithm

With the development of society, electric customers pay more and more attention to the quality of distribution systems operation. Voltage and reactive power optimal control is an effective means to secure the security, economy, reliability and quality of its operation. In our country, on-load tap-changing transformer and shunt capacitor are the main equipments of voltage and reactive power control in high and medium voltage distribution system. This thesis mainly focuses on the study of optimal controlling strategy of the on-load tap-changing transformers and shunt capacitors in high and medium voltage distribution system. First, the background, existing problems and fundamental theories on the voltage and reactive power control in distribution system have been studied, and the need and feasibility of study on this subject have been explained. Second, the mathematic modeling and its solutions have been discussed in detail. Therefore, a multi-objective optimization model of distribution voltage and reactive power is established, and its new solution, modified genetic algorithm is put forward.According to the forecasted load curve of near future, the whole day's load data is grouped under 24 load levels. On the ground of the typical load data in each level, a time-interval model of voltage and reactive power optimization is presented: in the objective function, the whole network's losses, voltage quality, load disequilibrium, load importance and operation limits are taken into account; the power flow equation, the range of controls parameter, load voltage, load disequilibrium and the operation limits are taken as constraints. Genetic algorithm is an effective optimization algorithm to deal with the mixed integer programming, and it could solve the un-differential objective function effectively. Because of the traditional genetic algorithm's slow convergence and other shortcomings, to solve the problem of voltage and reactive power optimization in distribution system, it is modified in this thesis as follows: the neural networks is used to estimate the individual fitness, avoiding to solve the time-exhausting power flow equation frequently; the pattern search is used to improve the optimal individual fitness of several generations, therefore improving the partial optimization of the algorithm; the method of ranking selection with worst individuals replaced and variable parameters is used to improve convergence of genetic algorithm and the quality of result. At last, the examples demonstrate that, the modeling of voltage and reactive power multi-objective optimization in distribution system is rational and indispensable, and the modified genetic algorithm is feasible and valid to solve the above problems.