Abstract: Aiming at the problem that the scheduling plan was difficult to be implemented due to the influence of dynamic factors in the scheduling process of distributed photovoltaic operation and maintenance resources,a dynamic scheduling method of distributed photovoltaic operation and maintenance resources based on reinforcement learning was proposed.In this method,the priority of operation and maintenance task was adjusted synchronously by constructing dynamic scheduling rules.A dynamic scheduling model was established to minimize the completion cost and time of the new plan.Q-learning was used to solve the model.Through the experimental comparison,Q-Learning algorithm had a fast solving speed and good algorithm stability,which was more suitable for solving dynamic scheduling problems.The proposed dynamic scheduling method of distributed photovoltaic operation and maintenance resources could cope with the influence of dynamic factors in the process of distributed photovoltaic operation and maintenance,and improve the service quality.

Key words: distributed photovoltaic, maintenance service, dynamic scheduling, reinforcement learning, Q-Learning algorithm

摘要： 针对分布式光伏运维资源调度过程中因动态因素影响导致调度计划难以实施的问题,提出基于强化学习的分布式光伏运维资源动态调度方法。该方法通过构建动态调度规则同步调整运维任务的优先级,并以新计划完成成本最低和完成时间最短为优化目标构建动态调度模型。采用Q-Learning求解模型,通过实验对比,Q-Learning算法的求解速度快、算法稳定性好,更适合求解动态调度问题,所提资源动态调度方法可以应对分布式光伏运维过程中的动态因素影响,提升服务质量。

关键词: 分布式光伏, 维修服务, 动态调度, 强化学习, Q-Learning算法