基于粒子群算法的分布式计算系统能效优化方法

ENERGY EFFICIENCY OPTIMIZATION METHOD FOR DISTRIBUTED COMPUTING SYSTEMS BASED ON PARTICLE SWARM OPTIMIZATION

在分布式计算领域中,需要合理分配计算资源。传统方法通过建立精确的复杂模型并转换为近似线性整数规划模型进行求解,会造成较大误差。对此,提出一种粒子群优化算法,直接使用精确非线性整数规划模型,在基本的粒子群优化算法中加入问题的约束。同时对粒子进行离散化,使用真实的目标函数值来评估粒子的适应度,能够在较短时间内收敛且离最优解的误差相对较小。仿真实验表明,该算法在单位利润、完工时间、总能量消耗方面均取得了不错的效果。

In the field of distributed computing,it is necessary to allocate computing resources rationally to improve energy efficiency and maximize profits.Traditional methods solve by establishing accurate complex models and optimizing them to convert to approximate linear integer programming models,but it could causes large errors.In this paper,a particle swarm optimization algorithm is proposed.It used an accurate nonlinear integer programming model directly.The problem constraint was added to the basic particle swarm optimization.The particles were discretized,and the real objective function value was used instead of the approximation to evaluate the particle.The fitness of the algorithm can converge in a short time and the error from the optimal solution is relatively small.The simulation experiments show that the algorithm has achieved good results in terms of unit profit,completion time and total energy consumption.