三级领导式的快速自适应狼群优化算法

为提高狼群算法的收敛速度,在此提出了一种称为三级领导式和微粒进化方程的自适应狼群算法,人为地把灰狼分成两类,领导层三只灰狼:如αβ和δ,剩下的为猛狼w。在游走搜索阶段随机设定一个猎物位置,利用狼群与猎物之间的距离来指导游走搜索猎物;在召唤阶段,利用三个领导层灰狼作为头狼来引导猛狼向猎物靠近,避免了传统狼群算法只有一只头狼引导整个狼群就容易陷入局部最优的情况;在围攻猎物阶段利用惯性因子来表示以往奔袭的经验、学习因子与随机数之间的乘积来表示猛狼自身经验的认识与总结、迭代影响因子来表示整体狼群经验的认识与调整,综合起来狼群粒子奔袭速度加快收敛速度和跳出局部最优,从而找到真实的整体最优值。本次选取的8个测试函数对应的对比性实验结果表明:该方法较为精确地实现寻找到了测试函数的最优值且较早地快速收敛到最优解,在后期也平稳收敛到真实的最优值,该算法适用于多维多波峰函数求极值问题。

Fast Adaptive Wolf Swarm Optimization Based on Three-Level Leadership

In order to improve the convergence speed of wolf swarm algorithm, an adaptive wolf swarm algorithm called three-level leadership and particle evolution equation is proposed, man-made grey wolves are divided into two categories：the leadership grey wolves, such as alpha, beta and delta, and the rest are the fierce wolf. In the wandering search stage, a prey position is set, and the distance between wolves and prey is used to guide the wandering search. In the calling stage, three best gray wolf particles are used to guide the wolves to approach the prey, avoiding the traditional situation that only one wolf is easy to fall into local optimum. In the siege stage, inertia factor is used to express the past experience of running. The product of learning factor and random number represents the knowledge and summary of the experience of the wolves themselves, and the iteration influence factor represents the knowledge and adjustment of the experience of the wolves as a whole. The speed of attack of wolf swarm particles accelerate the convergence speed and jump out of the local optimum, so as to find the real global optimum. The comparative experimental results of the eight test functions selected in this paper show that the method finds the optimal value of the test function more accurately, and converges to the optimal value earlier and faster, and to the real optimal value smoothly in the later period. The algorithm is applicable to multi-dimension and multi-peak function extremum problem with unknown search space.