Abstract: Transfer logistics is the key link of logistics cost control in the steel industry,and the decision of finished product transfer flow direction also plays a decisive role in the efficiency of transfer logistics.For the problem of finished product transfer flow direction between the end warehouse and finished product warehouse of steel enterprises,considering the supply and demand relationship between the end warehouse and the finished product warehouse,the Stackelberg game model was constructed with the objective function of maximum revenue and minimum cost of transfer warehouse,and the game model was validated and solved by the Q-Learning algorithm,which fit the feature constraints reflecting the priority of waterway transportation into the algorithm's coding.Based on the characteristics of different sizes of finished goods transshipment data,the algorithm was generated,and the experimental results showed that the Q-Learning algorithm was able to find the optimal transshipment strategy for different sizes of transshipment problems,and compared with the related solution algorithms,the Q-Learning algorithm had a stronger superiority.

Key words: steel industry, transfer of finished products to storage, supply and demand relationship, Stackelberg game, Q-Learning algorithm

摘要： 针对钢铁企业末端库与成品库之间的产成品转库问题,考虑末端库与成品库之间的供需关系,以最大化转库收益最小化转库成本为目标函数构建了Stackelberg博弈模型,结合模型特征,采用Q-Learning算法对博弈模型进行验证求解,将体现水路运输为优先考虑因素的特征约束拟合到算法编码中。最后基于不同规模的产成品转库数据特征生成算例,实验结果表明Q-Learning算法能够对大、中、小不同规模的转库问题求得最优转库策略,并具有更强的优越性。

关键词: 钢铁企业, 产成品转库, 供需关系, Stackelberg博弈, Q-Learning算法