传统智能算法在求解复杂的带有多峰特点的优化问题时,由于其计算量和变异方式的限制很容易陷入局部最优,并且不具备跳出局部最优进行二次搜索等能力。针对这一问题,本文提出了混合差分的化学反应算法,在利用化学反应算法(CRO)良好的全局搜索能力的同时,使用差分变异策略来加强算法的计算精度。对于优秀分子可能在反应中被消耗掉的现象,有针对性地加入了精英保留机制来保持种群的优良。本文选取了CEC2005中的测试函数,特别是几个带有多峰特点的复杂测试函数来分析改进算法的各项性能,并与几个改进的智能算法进行了对比实验。最终验证改进算法在提高计算精度和全局搜索能力两方面具有良好的效果。
Classic intelligent algorithm has the ability of global search, but it is still easy to fall into local optimum when dealing with composition multi-modal problems, and hard to jump out of it. For this question, a hybrid DE chemical reaction optimization algorithm was proposed. In this approach, DE mutation was used to improve searching accuracy, and elitist reserve was utilized to retain the quality of the whole population. 8 benchmark functions are chosen from CEC 2005, and the accuracy of calculation and the performance of global search will be tested through solving these benchmarks. At last, simulations on these benchmarks are performed with HDECRO and compare the results with several modified intelligent algorithms to draw conclusion.