基于深度强化学习的海上风电机组状态维护与备件库存联合优化  

JOINT OPTIMIZATION OF STATE MAINTENANCE AND SPARE PARTS INVENTORY OF OFFSHORE WIND TURBINES BASED ON DEEP REINFORCEMENT LEARNING

作　　者：董文康 吴雨芯 姚琦 郭俊 赵天阳 Dong Wenkang;Wu Yuxin;Yao Qi;Guo Jun;Zhao Tianyang(Energy and Electricity Research Center,Jinan University,Zhuhai 519070,China;College of Information Science and Technology,Jinan University,Guangzhou 510632,China;State Key Laboratory of Disaster Prevention and Reduction for Power Grid Transmission and Distribution Equipment(State Grid Hunan Electric Company Disaster Prevention and Reduction Center),Changsha 410007,China)

机构地区：[1]暨南大学能源电力研究中心,珠海519070 [2]暨南大学信息科学技术学院,广州510632 [3]电网输变电设备防灾减灾国家重点实验室(国网湖南省电力有限公司防灾减灾中心),长沙410007

出　　处：《太阳能学报》2023年第12期190-199,共10页Acta Energiae Solaris Sinica

基　　金：电网输变电设备防灾减灾国家重点实验室开放项目《国网湖南防灾减灾中心台风灾害下高比例海上风电并网系统弹性管理》。

摘　　要：维护与备件库存管理是海上风电运维的两个密不可分的关键环节。为提高海上风电机组设计寿命周期内的运维经济性,构建状态维护与备件库存联合优化策略。首先,将海上风电机组构建为由叶片、齿轮箱、电气、偏航、轮毂、制动、传动链、发电机8个子系统组成的系统,然后将各子系统的劣化过程构建为多状态马尔可夫随机过程,建立维护与备件库存的交互模型,其中包括被动维护时间与随机故障以及备件库存的关系、备件库存对维护活动的影响等。随后,设计子系统的劣化状态与备件库存状态、状态检修动作与备件订购的表征方法,并以此形成深度强化学习Dueling DQN的框架,通过对深度网络的迭代训练,求解海上风电机组的最优维护与备件订购决策序列。最后,以某海上风电场内的风电机组为例,验证所提联合优化方法的优越性,并讨论强化学习的探索率、风电场的可达率对运维成本的影响。Maintenance and spare parts inventory management are two fundamental sequential procedures in offshore wind farm operation and maintenance(OM).To improve the OM efficiency of offshore wind turbines(OWTs)in the design life cycle,a joint optimization strategy of condition-based maintenance and spare parts inventory control is proposed.Firstly,the OWT is constructed as a series system of subsystems including blades,gearbox,electrical,yaw,wheel hub,braking,transmission chain,generator,the deterioration process of each subsystem is modeled as a multi-state Markov process,and the coupling model of maintenance and spare parts inventory is established.Secondly,the representation and update method of subsystems deterioration,spare parts inventory are formulated as the framework of deep reinforcement learning.Through the iterative training of the deep networks,the optimal maintenance and spare parts ordering decision making are obtained.Finally,taking an OWT of a practical offshore wind farm as an example,the effectiveness of the proposed joint optimization method is verified.The impacts of the exploration rate and the accessibility of OWTs on the OM cost are further discussed.

关 键 词：海上风电机组 状态维护 库存控制 强化学习 联合优化 

分 类 号：TK83[动力工程及工程热物理—流体机械及工程]