基于粒子群和模拟退火算法的混合算法研究

Study on hybrid algorithm based on particle swarm optimization and simulated annealing algorithm

在标准粒子群优化算法的基础上给出了一种改进策略,利用混沌变量的随机性、遍历性、规律性对粒子群进行初始化选择。同时为了增加粒子多样性又不流失适值较好的粒子,在一定的周期内对所有粒子重新进行有选择的初始化,并对除了种群最优之外对应的所有个体最优变异。计算结果表明,改进的粒子群算法提高了收敛精度和速度,但是个别函数寻优失败。将改进的粒子群算法结合模拟退火算法再次计算了测试函数,结果表明,改进的混合算法可以达到目标函数的全局最优点。

An improved method is derived from the standard particle swarm optimization. Firstly, the particles are initialized by chaos technique which is ergodic, stochastic and regular. To enhance the diversity and to hold the better particles, all the particles are initialized selectively again for a period oftime. Finally, all the individual best positions except the index ofglobal best position are mutated. The results show that convergence speed and accuracy of algorithm are enhanced by the improved particle swarm optimization. But it is im- possible to find the best positions of some functions. Then test functions are calculated again by a hybrid algorithm based on improved particle swarm optimization and simulated annealing algorithm, and the results show that the global optimal solution is achieved by the hybrid algorithm.