基于粒子群算法的大型网络计划资源优化

Resource Optimization for a Large Network Plan

现有求解网络计划资源优化的方法中,解析法不能解决大型复杂网络优化问题,启发式方法过多依赖具体问题、求解效率低,遗传算法生成新一代优化解种群依据的三个算子的实现参数选择,大部分依靠经验并严重影响解的品质,粒子群算法存在大型网络计划资源优化计算量过大和缺少大型网络计划资源优化算例问题.借助设计网络计划时间参数的计算机算法、建立评价函数、设计进化方程等基础工作,选择与工作开始时间相关的变量作为粒子空间位置,用蒙特卡洛方法和限制条件优化初始粒子群,设置可行解范围,用二维动态数组解决大型网络计划资源优化运行image超限问题,通过粒子群算法进化,寻求大型网络计划资源优化解,算例表明基于粒子群算法的大型网络计划资源优化效果明显,粒子群算法参数分析表明:粒子群算法的参数会影响网络计划资源优化结果,而且初始粒子群限制条件和优化目标设置的影响程度较大.

Among the existing methods to solve resource optimization for network plan, analytical methods cannot solve the optimization problem of large and complex networks, whereas heuristic methods rely excessively on specific issues and have low solving efficiency. Genetic algorithm generates a new generation of population on the basis of optimal solution of the realization of the three operators parameter selection most rely on experience and seriously affect the quality of the solution. For particle swarm optimization some problems exist including excessive computing and the lack of large-scale network planning resource optimization example. Based on designing computer algorithm for parameters of the network planning, establishing the evaluation function, and designing evolution equations, etc., select the variables associated with work starting time as the spatial position of the particle, optimize the initial particle swarm optimization using Monte Carlo method and restrictions, set feasible solution range, solve the image overrun problem for large-scale network planning optimization using two-dimensional dynamic array, seek large-scale network planning resource optimization solution by particle swarm evolution. Examples show that the effect of a large network resource optimization based on particle swarm optimization is obvious. Parameter analysis of particle swarm optimization shows ： parameters of PSO algorithm affect the network planning resource optimization results, initial PSO restrictions and set optimization goal has greater degree of influence.