Research And Application On Adaptive Parameter Adjusting Quantum-behaved Particle Swarm Optimization

Quantum-behaved Particle Swarm Optimization shows a huge advantage in solving optimizational problems. It has been widely used in the national economy, national defense and all aspects of social development and becomes a focus of the domestic and overseas research during recent years. However, the population diversity of the algorithm reduces in the later stages and the algorithm is easy to fall into the local optimization, which leads to the premature convergence. To further improve performance of Particle Swarm Optimization, an improved algorithm based on analyzing the influence of parameters on the performance of the algorithm is proposed in this paper--Adaptive Parameter Adjusting Quantum Particle Swarm Optimization algorithm, which is used for the study of the parameters optimization.In the improved algorithm, the mutation of the particle position is performed by the quantum Hadamard-gate to increase the diversity of the population. The inertia factors, self-factors and global-factors are adaptively determined according to the current fitness to improve the global search capability. The improved algorithm and traditional algorithm are tested by the function extremum optimization question. The result indicates that the improved algorithm has better the convergence rate and the success rate, which is superior to others in the global search capability and the optimization efficiency.In this paper, the improved algorithm is proposed to optimize the parameters of the RBF neural network and the pendulum Normalized Fuzzy Neural Networks controller. In this method, a multi-dimensional vector which is composed of the optimization parameters as the particles is calculated to acquire the global optimization parameters during the solution space. The superiority of the RBF neural network based on the improved algorithm is verified by the tests of the nonlinear function approximation and the chaotic time series prediction. The pendulum control test in different initial states demonstrates that the pendulum NFNN controller based on the improved algorithm has higher control accuracy and faster convergence and the improved algorithm has a high value of the engineering applications.