基于代理模型的多目标启发式优化算法研究及应用
现存的代理模型大多采用粒子群算法优化的神经网络逼近真实模型，存在搜索效率低，收敛性差的问题。且在数据不均衡时会影响代理模型的精准度、可靠性。此外，基于蚁群算法求解代理模型存在寻优路径信息素增量难以把控，个体目标导向不强的问题，导致算法全局搜索能力弱、收敛性差，影响效率。针对上述缺点，本文提出一种基于代理模型的多目标启发式优化算法，针对数据不均衡的问题，本文采用中均值滤波和3-sigma 法则进行初次标记，接着采用改进的移动加权平均和动态阈值控制进行二次标记，保留体现真实数据分布的样本。针对基于代理模型的优化算法的优化效率问题，采用融入强化学习机制的蚁群算法进行优化调度，最终将本算法应用于万家寨引黄入晋经济运行研究项目之中。
[1] Metropolis N, Rosenbluth A W, Rosenbluth M N, et al. Equation of state calculations by fast computing machines [J]. The journal of chemical physics, 1953, 21(6): 1087-1092.
[2] Kirkpatrick S, Gelatt Jr C D, Vecchi M P. Optimization by simulated annealing [J]. science, 1983, 220(4598): 671-680.
[3] Colorni A, Dorigo M, Maniezzo V. Distributed optimization by ant colonies[C]//Proceedings of the first European conference on artificial life. Brussels, Belgium: Elsevier,1991, 142: 134-142.
[4] Noel M M, Muthiah-Nakarajan V, Amali G B, et al. A new biologically inspired global optimization algorithm based on firebug reproductive swarming behaviour [J]. Expert Systems with Applications, 2021, 183: 115-128.
[5] Singh G, Singh A. Extension of particle swarm optimization algorithm for solving transportation problem in fuzzy environment [J]. Applied Soft Computing, 2021, 110:607-619.
[6] F. Cuadros Bohorquez J, Plazas Tovar L, Wolf Maciel M R, et al. Surrogate-model-based, particle swarm optimization, and genetic algorithm techniques applied to the multiobjective operational problem of the fluid catalytic cracking process [J]. Chemical Engineering Communications, 2020, 207(5): 612-631.