Reactive Power Optimization In Ship Power System Based On Improved Genetic Algorithm

The distribution of the reactive power directly influenced the stability and economy ofthe power system. By optimizing the distribution of reactive power, the total loss of thesystem can be reduced, and the voltage quality can be improved. The study of reactive poweroptimization means a lot.There have been a lot of research of land-based power system, but due to thecharacteristics of ship power system, there are not many research about it. Therefore, to studythe experiences of reactive power optimization of land-based power system and trying toapply it to the ship is the main focus of this paper.As we all know, the genetic algorithm is an artificial intelligence algorithms whichdeveloped in recent years. It is suitable for solving the reactive power optimization problemsfor its high-efficiency and the ability of global search in parallel, If combined with the idea ofother artificial intelligence algorithms, the performance will get more improved.First of all, a model which contains9buses of ship power system is proposed, and themodeling for components in the system is also completed. Based on the structuralcharacteristics of the ship power system, the topological analysis and flow calculation arecompleted. In order to analyze the power flow of the system, the method which is used tonumber the system based on graph theory is proposed. Sometimes the flow calculation forland-based power system might not converge. Therefore, in this paper, the flow calculationmethod of forward and backward substitution is used. The main idea is to use the busincidence matrix which export the bus layered matrix and the corresponding upper bus matrix.Based on this, the calculating of the flow by using the forward branch for current and backsubstitution for voltage is done. In order to prove that the method is effective, the9-bus shippower system is used to test. The final convergence means the method is feasible.When use the simple genetic algorithm, some problems such as convergence to localoptima, or the excessive time to find the best solution may be encountered. To solve theseproblems, many improved methods have been proposed. This paper proposes an improvedmethod which combines with the idea of simulated annealing algorithm, and improves thearrangement of the fitness function. The strategy of the generation of the initial population,combined with roulette and tournament populations are also improved. Fixed crossover andmutation probability may lead to the reduction of population diversity, therefore the adaptivecrossover and mutation probability is proposed. When use the function to test the performance of the algorithm, both the accuracy and speed of convergence is improved.Finally, the proposed method has been applied to the IEEE9buses power system and9buses ship system. The computation results show that this approach can find optimal solutionmore efficiently than the simple genetic algorithm.