约束动态环境优化免疫算法设计、理论分析及应用

Firstly, a few adaptive constraint handling approaches, based on the basic charactaristics of dynamic constrained optimization problems, is presented to deal with constrained optimization problems in dynamic environments. Secondly, an effective environment recognition module, relying on the function of clonal proliferation and immune response and immune memory in bio-inspired artificial immune systems, is designed to examine dynamically whether the environment changes or whether a change takes place in the optimization problems. Furthermore, combining immune mechanisms with differential evolution methods to form a hybrid constrained immune algorithm that consists of dynamic immune operators, we compare the advantages and disadvantages of algorithm obtained with other existing algorithms in terms of testing and analysis, and discuss the applications of algorithm proposed. Thirdly, a parallel processing immune algorithm with constraint handling methodology and adaptive immune operators is proposed to solving complex dynamic constrained optimization problems, the framework of algorithm is constructed involving of subpopulation parallel evolution each other, information change and transformation, network suppression and other principles of biological immune systems. Meanwhile, discussing the applications of algorith designed. Performance criteria of algorithm is established and the convergence is proved in time-varying environments. Lastly, we apply these algorithms on the real-word problems, so as to improve the disadvantages of algorithm proposed by comparison and testing. The software of algorithm is developed to apply real-word dynamic engineering constrained optimization problems, the results has profound significance for the development of information science.

研究约束优化动态环境问题的基本特征，提出适合于约束优化动态环境算法的自适应约束处理策略和方法。挖掘生物免疫系统的克隆繁殖、免疫应答及免疫记忆等功能，设计环境识别模块自适应识别环境变化，构建动态免疫算子，结合微进化策略，提出混杂约束优化动态环境免疫算法，测试分析算法与其他算法的优越性，讨论算法的实用价值。基于生物免疫系统自适应学习及网络抑制等原理，设计自适应免疫算子，结合约束处理策略，提出并行约束优化动态环境免疫算法，采用子群并行进化、信息传播交换、网络抑制等策略构建算法框架，讨论算法的实用价值。提出约束优化动态环境算法的评价准则，证明算法的收敛性问题。将新设计的算法用于大量测试问题及实际问题检测算法的实用性，并比较算法的优越性，改进和完善算法的不足，进而获得高级的免疫算法，实现算法软件的开发和利用，为工程和实践人员提供技术和方法，对信息科学技术的发展具有深远研究意义。

动态约束优化问题(DCOPs) 在控制科学与工程中有着广泛的应用, 传统的数学优化方法对优化问题的特征很敏感, 对复杂的 (非凸的、离散的、动态的等)DCOPs的处理具有极大的挑战。然而, 基于生物免疫系统的免疫算法具有高度并行性、自适应性和抗体多样性等特征, 其更适合求解DCOPs。本课题充分挖掘生物免疫系统的内部运行机理, 开发智能技术设计高级的约束优化算法解决复杂的静态约束优化问题(SCOPs) 和动态约束优化问题(DCOPs), 并探索其在控制科学与工程中的应用。具体工作如下 :.1) 针对多模态的、高维的、复杂的Pareto最优前沿类SCOPs, 基于生物免疫系统的多层响应模型, 提出一种多层响应免疫算法(CMIGA) 求解SCOPs。2) 为了克服已有约束处理技术对非可行解直接淘汰而仅保留可行解, 致使算法易于陷入局部搜索或早熟的现象, 基于生物免疫系统的固有免疫和自适应免疫交互运行模式, 提出目标约束融合的并行约束多目标免疫算法(PCMIOA)。3) 针对DCOPs的测试函数目前研究很不成熟, 基于SCOPs 的基准测试函数探索性提出一系列DCOPs。结合CMIGA的设计框架提出了一种邻域搜索的免疫算法 (DCMOIA)。4) 提出简化的克隆选择算法(SCSA)应用于逆变器PWM开关序列优化。克服了已有的群智能算法所获的最优PWM开关序列产生的电流谐波失真大、计算复杂度高、不适合实时控制等缺点。5) 针对CSA求解高维背包问题(KPs)时可行抗体比率低且易于陷入局部搜索的缺点, 提出受体编辑机制, 并设计二次修补策略增强约束处理能力, 获得提高的克隆选择算法(CSA-ER)。6) 提出一种不重复的精英提高记忆和自适应环境响应的免疫算法(MERIA)解决动态优化问题(DOPs)。