不确定性资源调度分配问题的适应性离散差分进化算法研究

Differential evolution(DE), as a new type of intelligent optimization method, Due to advantages in high reliability, strong robustness and excellent optimization performance, DE has become one of the hottest topics in the research field of evolutionary computation. DE has shown a good potential optimization capability. Uncertain resource scheduling and allocation problem belongs to the typical combinatorial optimization problem. The uncertainty has not been considered comprehensively in the existing research. The suitable method is lacked for handling uncertainty and constraints. With regard to the problems existing in the uncertain resource scheduling and allocation, this project aims to establish the multi-goals, multi-constraints in accordance with the uncertain characteristics. In addition, the differential mutation and crossover operators are designed in the new discrete differential evolution algorithm. And a memetic discrete differential evolution algorithm is built, combining the discrete differential evolution with the algorithm which has powerful local search ability. The improvement can balance the exploration and the exploitation abilities and realize reasonable scheduling and allocation of uncertainty resources. Furthermore, the coupling degree function and coupling coordination degree function are established to evaluate the system utility and coupling state of utility metrics so as to form the evaluation system. Research of the project not only expands the application field of differential evolution algorithm, but also realizes the performance and efficiency of the uncertain resource scheduling and allocation. At the same time, the research provides a solid basis for the theory and methodology to other similar problems. As a result, the study of the project has important theoretical value and practical significance.

差分进化算法是一种新型的智能优化算法，具有高可靠性、强鲁棒性以及良好的优化性能，属于进化计算领域的热点课题，该算法在组合优化领域表现出良好的优化潜能。不确定性资源调度分配属于典型的组合优化问题，现有研究对资源在调度分配中的不确定性特征考虑不全面，并缺乏适合处理不确定性和多约束的调度分配方法。本项目针对不确定性资源调度分配存在的问题，建立符合不确定性特点的多目标、多约束数学模型；设计离散差分进化算法的差分变异和交叉算子，将其与具有较强局部搜索能力的算法相结合，形成memetic算法，平衡算法的探索与开发能力，实现不确定性资源的合理调度分配；建立耦合度函数和耦合协调度函数，对效用和效用指标的耦合状态进行评价。该项目的研究不仅拓宽了差分进化算法的研究领域，实现了不确定环境下资源调度分配的性能与效率。同时为其它领域中相似调度分配问题的处理提供了坚实的理论和方法指导，具有重要的理论价值和现实意义。

不确定性资源的调度分配问题是复杂系统决策优化中的一类重要问题，该类问题的核心是将具有不同价值和任务完成概率的资源进行调度分配，使系统的整体期望效能最大化。该项目全面考虑调度分配过程的不确定性特征，并充分考虑资源约束和可行性约束，建立多目标、多约束的数学模型；采用离散差分进化算法求解不确定性资源调度分配问题，设计解相量编码方式，确定解相量与决策变量之间的关系，实现不确定性资源调度分配的任务-资源分配对，同时，设计离散DE 算法的差分变异算子和交叉算子，提出不可行解的修正方案；针对算法的探索与开发能力进行平衡，确定学习机制的协同作用策略，对任务与资源的调度分配过程进行动态优化，增强优化算法的全局寻优能力。该项目的完成有效解决了针对包含不确定性特征问题的数学建模、智能优化算法探索与开发能力平衡的科学问题，为其它领域中相似调度分配问题的处理提供坚实的理论和方法指导。