绝大多数现有演化动态优化研究假设问题是黑盒，但现实应用中许多动态优化问题是灰盒：虽然没有具体的数学表达式，但影响问题最优解的环境状态参数可以观测。合理利用这些参数，有助于提高算法寻优效率。针对这类灰盒问题，本项目研究学习增强的演化灰盒动态优化算法，重点解决两个问题：如何学习参数与最优解的关系，以及如何利用习得的关系增强算法性能。首先针对不同关系类型，分别研究基于函数模型的学习、基于条件概率模型的学习和基于代理模型的学习。然后研究应用模型，分别在种群初始化阶段和种群演化阶段增强演化算法的性能，以及多种增强策略的集成方法。最后，研究灰盒动态优化问题的建模和基准测试函数集的构造，以及学习增强的演化灰盒动态优化算法在动态电力潮流计算和动态旅行商问题中的应用。本项目有广泛的应用背景，研究内容有特色和创新，具有一定的理论和应用价值。

Most of the existing evolutionary dynamic optimization studies assume that problems are black boxes, but many dynamic optimization problems in real-world applications are gray boxes: although the specific mathematical expressions of the problem are not available, the environmental state parameters that affect the optimal solution can be observed. Using these parameters properly can help to improve the performance of optimization algorithms. For such gray box problems, leaning enhanced evolutionary gray-box dynamic optimization algorithms are studied in this proposal. We focus on two issues: how to learn the relationships between parameters and optimal solutions, and how to apply learned relationships to enhance algorithm performance. Firstly, for different types of relationships, functional model, conditional probability model, and surrogate model are studied, respectively. Secondly, the application strategies of the learned models are investigated to enhance the performance of evolutionary algorithms, at the population initiation and evolution stages, respectively, as well as the ensemble of multiple strategies. Finally, the modeling and benchmark function sets for gray box dynamic optimization problems are investigated, as well as the application of proposed algorithms to dynamic power flow and dynamic traveling salesman problems. This proposal has a broader application background. Its research content is somehow novel. Therefore, this proposal has theoretical and application value to some degree.