动态惯性权重粒子群算法研究与实现

Research and Implementation of Dynamically Changing Weighted Particle Swarm Optimization

为提高基本粒子群优化算法(PSO)的收敛性能,并达到一定收敛率和精度,提出了动态惯性权重粒子群算法(DCWPSO),给出了改进算法在采用固定控制参数取值时,参数的最佳取值范围,并应用测试函数进行仿真研究。和以往文献中多种改进的PSO算法进行比较,仿真分析表明:动态惯性权重粒子群算法(DCWPSO)加快了收敛速度,提高了解的精度,这种改进的粒子群优化算法在解决实际优化问题中很有潜力,并通过神经网络仿真研究,深入研究分析了原始算法PSO和DCWPSO算法在优化不同结构网络中控制的作用。

To improve the convergence performance of Particle Swarm Optimization(PSO) and achieve certain convergence rate and accuracy, dynamically changing weighted particle swarm optimization(DCWPSO) was proposed, and the best value range of parameters was given while fixed control parameters were adopted in the improved algorithm. The simulation was researched by using the test function and compared with a variety of improved PSO algorithms in previous literature, which indicated DCWPSO quickened the tempo of convergence, improved the accuracy of the solution and had great potential in solving practical optimization problems. Meanwhile, through the research on neural network simulation, the original PSO and how DCWPSO algorithms play an important role in optimizing the control of different structured networks were studied and analyzed deeply.