基于多智能体的分销链预测优化模型研究

为提高分销链中多层预测的准确性,优化订货和库存,提出基于多智能体的预测模型.建立了具有约束条件的目标函数,并将惩罚函数引入遗传算法进行求解,使得分销链整体成本最小化.模拟四层分销链中各公司的预测订货量和订货成本,分析了单位成本参数对它们的影响.最后,比较了多智能体预测模型和传统预测方法.仿真结果表明,分销链中各公司的预测订货量与市场需求量趋于一致,在一定程度上改善了分销链中订货信息的多级放大现象(牛鞭效应);同时分销链的订货成本波动较小,证明多智能体预测模型优于传统的预测方法.

Forecasting optimization model in distribution chain based on multi-Agent

To improve accuracy of multi-echelon forecasting demand in distribution chain,and optimize ordering and storage management,a Multi-Agent-based Forecasting Model(MAFM) was proposed.Objective function with constrains was established,and Genetic Algorithm(GA) with penalty function was adopted to minimize total cost of distribution chain.Ordering variation and cost of four echelons in distribution chain were simulated,and influence of unit cost parameter on order variation and cost were analyzed.Finally,MAFM was compared to traditional forecasting methods.Simulation results showed that ordering variation of MAFM in distribution chain was approximate to the market's demand,and the enlargement of ordering variation(Bullwhip Effect) was mitigated.Furthermore,ordering cost also fluctuated smoothly.