基于深度强化学习的多充电器部分充电策略  

Deep Reinforcement Learning-based Multi-charger Partial Charging Strategy

作　　者：李立志 冯勇[1] LI Li-zhi;FENG Yong(Faculty of Information Engineering and Automation,Kunming University of Science and Technology)

机构地区：[1]昆明理工大学信息工程与自动化学院

出　　处：《化工自动化及仪表》2024年第6期1053-1060,共8页Control and Instruments in Chemical Industry

基　　金：国家自然科学基金(批准号:62062047)资助的课题。

摘　　要：在大规模无线可充电传感器网络(WRSN)中,为了解决可分离充电模式下移动充电车(MCV)充电效率低下的问题,提出一种多充电器部分充电策略(PCSMC),以避免由于等待传感器节点进行完全充电而导致剩余能量低的节点失效。首先将充电时长这种连续动作空间转换为离散的动作空间,然后使用指针网络规划MCV的路径并动态调整传感器节点的充电时长,最后使用深度强化学习算法(DDQN)训练指针网络,从而生成近似最优解。仿真结果表明,与其他充电策略相比,PCSMC在失效节点数和平均充电延迟方面分别降低了23.56%和27.71%。In large-scale wireless rechargeable sensor network(WRSN),considering low charging efficiency of the mobile charging vehicle(MCV) in separable charging mode,a multi-charger's partial charging strategy(PCSMC) was proposed to avoid any failure of the nodes with low residual energy due to waiting for nodes to be fully recharged.In which,having the continuous action space in charging time converted into a discrete action space;then,having the pointer network used to plan the path of MCV and dynamically adjust the charging time of sensor nodes and finally,having a deep reinforcement learning algorithm DDQN(Double Deep Q-Network) adopted to train pointer network to generate a near-optimal solution.Simulation results show that,as compared to other charging strategies,the PCSMC can reduce the number of dead nodes and average charging delay by 23.56% and 27.71% respectively.

关 键 词：无线可充电传感器网络 可分离充电模式 部分充电 指针网络 深度强化学习 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]