基于分解算法优化电力系统维修计划的研究

The consumption of electricity is growing in China with the rapid development. The reliability of the power plants and transmission lines in the power system is crucial for meeting this demand. Timely maintenance in the system is thus mandatory to reduce equipment breakdowns and avoid unexpected malfunctions. In this context, we propose a large-scale stochastic mixed-integer programming model for the power system maintenance scheduling problem. The optimal maintenance plan will be achieved from solving the model to determine over a given time horizon when to stop a generation unit or disconnect a transmission line for maintenance and when to re-start them again. Both strategic level and operational level decisions as well as their interactions are included in the model. Specifically, the decisions of whether to schedule maintenance for each equipment in the system at a given period are made at the strategic level, and the unit commitment decisions are made at the operational level to validate the feasibility and optimality of the maintenance plan. To the best of our knowledge, maintenance scheduling problems are often modeled with the assumption that the generation amount and customer demand are deterministic, though they are often subject to uncertainty in real-life system and the system would be highly exposed with the risk of not satisfying electric demand in the market. In this project we propose to apply conditional value-at-risk technique to measure and control the risk at a confidence level. Due to the complexity and large scale of the proposed model, we consider Benders decomposition as the solution method and propose to apply high computing technique, e.g., parallel computing, to further improve the computing efficiency. The success of the project would expect to deliver the solution assisting decision makers to make the most adaptive maintenance plan for power systems in terms of maintaining the system reliability, minimizing the cost, and managing the risk associated with uncertain parameters. The model and algorithm can also be extended to more applications in which the planning and operations can be integrated to formulate more effective model, such as power system capacity expansion, distributed energy system and facility location problem, etc.

为了帮助电力系统制定科学的维修保养计划，深入贯彻《电力发展“十三五”规划》中“统筹兼顾、优化布局、智能高效、保障民生”等基本原则，本项目研究为复杂的电力系统维修计划问题建立基于随机规划的大规模混合整数规划数学模型，全局考虑电力系统中各个设备的维修保养要求，利用条件风险价值方法控制设备维修所带来的风险，开发基于分解方法的优化算法并利用高性能并行计算实现该算法为数学模型求解。项目若成功实施，将能够帮助决策者制定最佳电力系统维修计划，在保障系统稳定运行的同时，将由于维修保养所造成的经济损失降到最低，并将由于设备维修与系统中不确定因素造成电力需求不能被满足的风险控制到可以接受的范围。而基于此开发的数学模型及优化算法还将能够拓展到其他类似的大规模复杂问题的求解，比如电力系统长期扩建计划的随机规划模型、分布式能源系统的优化和厂址选择问题等。

该项目为复杂的电力系统维修计划问题建立了基于随机规划的大规模混合整数规划数学模型，在全局考虑电力系统中各个设备维修保养要求的前提下，全面探索与分析维修计划阶段与电力系统运营阶段的关系，利用条件风险价值方法控制设备维修所带来的风险，并利用并行计算实现了该算法在高性能计算平台的求解过程。..研究成果预期能够帮助电力系统决策者制定最佳维修计划，在保障系统稳定运行的同时，将由于维修保养所造成的经济损失降到最低，并将由于设备维修与系统中不确定因素造成电力需求不能被满足的风险控制到可以接受的范围。在该项基金资助下，申请人为电力系统维修计划问题以及其它若干类似的大规模组合优化问题成功开发了一系列优化模型和算法，研究成果发表在管理科学类前沿期刊。