基于异维变异的差分混合粒子群算法

针对粒子群(Particle Swarm Optimization,PSO)算法和差分进化(Differential Evolution,DE)算法存在容易陷入局部极值、进化后期收敛速度慢和收敛精度低的局限性,提出了一种基于异维变异的差分混合粒子群(UDEPSO)算法。首先,为了提高群体多样性,使用熵度量初始化粒子；其次,在粒子迭代的过程中,根据粒子的分布特点,引入异维变异学习策略和维度因子以引导粒子及时跳出局部极值达到最优解；最后,将所提算法在10个典型的测试函数上进行了仿真,其在9个测试函数的收敛精度和标准差上取得了显著的效果,远优于PSO算法、DEPSO算法以及CDEPSO算法。实验结果表明,UDEPSO算法在优化收敛精度和效率上具有较强的优势。

Differential Hybrid Particle Swarm Optimization Algorithm Based on Different Dimensional Variation

Considering the limitations that particle swarm optimization(PSO) algorithm and differential evolution(DE) algorithm are difficult to control the initial distribution of the particles and easily fall into local optimum to reduce the convergence accuracy at later process,a differential hybrid particle swarm optimization algorithm based on the different dimensional variation was proposed. Firstly,in order to improve the diversity of particle swarm,the entropy measure me-thod was introduced to initialize particles.Secondly,during the process of particle iteration, learning strategy of different dimensional variation and dimension factors were adopted to guide the immersed particles to jump out of the local optimum to reach the best solution in a timely manner according to the particle distribution.Finally,this algorithm was simulated on ten typical test functions.The convergence precision and standard deviation show the superior performance of the proposed method on nine testing functions comparing with PSO,DEPSO and CDEPSO.These experiments prove that the algorithm has a strong advantage in convergence accuracy and optimization efficiency.