手术室是医院手术服务的重要平台科室，负责手术相关资源的日常运作调度。在手术需求不断攀升的今天，手术室逐渐成为医院手术服务流程的瓶颈环节，手术室资源调度水平的高低直接影响着医院手术服务质量和成本。手术室资源调度问题不确定性显著，涉及多类设施和人力资源的协同优化利用，是一类不确定环境下的复杂资源调度问题。因此，研究手术室资源优化调度问题具有重要的现实和理论价值。按照决策逻辑顺序，手术室资源调度问题可以分为手术护士排班问题以及择期手术和手术护士调度问题。本文分别针对这两个问题展开深入研究。与现有的确定性排班方法不同，本文考虑了需求不确定条件下的手术护士排班问题，提出一个基于随机规划的随机排班模型。该模型考虑了急诊手术对手术护士需求的不确定性，并对多类手术护士排班重要约束进行刻画。针对模型求解，本文提出一个基于蒙特卡洛仿真和整数规划的优化算法。通过理论证明和数值实验验证，该算法具有良好的收敛性，可以较好的近似求解随机排班模型。数值实验结果肯定了随机排班模型对需求不确定条件下手术护士排班问题的优化效果。与已有的两阶段调度方法不同，本文针对择期手术和手术护士调度问题之间的紧密联系，提出两类整体调度方法：最优化方法和启发式方法。最优化方法基于0-1整数规划建立择期手术和手术护士的整体调度模型，考虑了各类重要调度约束。数值实验结果表明该模型对于提高手术间和手术护士整体利用水平的效果。为有效求解大规模问题，本文提出一个启发式分解迭代算法。该算法采用降序排列最早开始时间的最佳适应算法对择期手术调度问题进行快速求解，并设计多类启发式方法帮助提高算法的迭代搜索效率。数值实验验证了该算法的性能。本文基于由三甲医院收集整理的实际数据对手术室资源优化调度方法的相关应用问题进行研究。重点针对实际中手术时长不确定性的影响因素进行多因素方差分析，找出手术时长的显著影响因素，为实际中手术时长的准确估计提供理论依据，从而提高了择期手术和手术护士整体调度方法的应用性。本文对实际中急诊手术的手术护士需求经验概率分布进行推导，验证了随机排班模型的应用性。

Operating rooms play decisive role in providing surgical services in hospitals by scheduling various surgical resources. With ever-rising surgery demand nowadays, operating rooms tend to be the bottleneck in the overall surgical process. Scheduling of operating room resources has significant impact on the quality and cost of surgical services. Operating room scheduling is complicated in practice due to the considerable uncertainty with respect to surgery durations and emergency surgeries, as well as the need to coordinate the use of multiple surgical resources. Therefore, operating room scheduling problem is of great importance, both theoretically and practically. Operating room scheduling can be divided into two specific problems that are solved sequentially in practice: surgical nurse rostering problem; elective surgery and surgical nurse scheduling problem. These two problems are studied in depth in this research.Different from deterministic rostering methods, this research considers the uncertainty involved in surgical nurse rostering problem, and accordingly proposes a stochastic rostering model that is in the form of stochastic programming. Specifically, the model takes into account the uncertain demand for surgical nurses and incorporates various practical constraints regarding surgical nurse rostering. An almost exact method is developed to solve the stochastic model by using Monte Carlo simulation and integer programming. The convergence property of the method is proved theoretically and demonstrated numerically. The results of the numerical experiments reveal that better use of surgical nurses can be achieved using the stochastic rostering model given uncertain demand for surgical nurses.In contrast to the two-stage method that schedules elective surgeries and surgical nurses separately, this research takes into consideration the strong interactions between the surgery and nurse scheduling processes and proposes an exact method and a heuristic method to integrate the scheduling processes. The exact method relies on a 0-1 integer programming model that schedules elective surgeries and surgical nurses simultaneously and considers multiple important constraints. For large problem instances, an efficient heuristic method is developed that generates surgery schedules using a fast heuristic algorithm and solves the corresponding nurse scheduling problem to global optimality. The heuristic method adopts an iterative structure, in which several heuristic search procedures are developed to facilitate the search for a good solution. The performance of the exact and heuristic methods is demonstrated through numerical experiments.The applicability of the proposed methods for operating room scheduling is validated using real-life data obtained from general hospitals. Specifically, in order to reduce variability in surgery durations, the sources of variability are studied using a multi-factor analysis of variance, knowledge of which may ultimately be used to improve elective surgery and surgical nurse scheduling. Besides, the empirical probability distribution of the demand for surgical nurses is deduced from real-life data that further validates the proposed stochastic rostering model.