一类具有精英学习能力的增强型人工免疫网络优化算法研究

An Enhanced Artificial Immune Network with Elitist-Learning Functionality for Optimization Problems

中文关键词  人工免疫系统  精英学习  亲和力学习  微粒群优化  PID控制

英文关键词  artificial Immune network  elitist-learning  affinity-learning  particle swarm optimization  PID controller

中文摘要

      本文提出了一种用于求解优化问题的具有精英学习能力的增强型人工免疫网络(EaiNet-EL)算法. 该算法集成了亲和力学习和精英学习, 并采用两个控制因子分别调节其学习速度. 本文对控制因子的灵敏度作了探讨, 并在两个经典函数的优化实验中将提出的EaiNet-EL算法同传统aiNet算法作了比较研究. 实验结果表明, EaiNet-EL在最优解质量和收敛速度上都要优于传统aiNet算法. 作为应用实例, 工业PID控制器被用于测试EaiNet-EL算法的性能. 实验所得的阶跃响应曲线说明, 使用EaiNet-EL得到的系统性能要优于使用其它三种方法得到的系统性能.

英文摘要

      This paper proposed an new enhanced artificial immune network with elitist-learning (EaiNet-EL) for optimization problems. The proposed new algorithm introduces two types of learning mechanisms, affinity-learning and elitist-learning, which are respectively adjusted by two weighted control factors. The sensitivity of factor values is investigated and the comparative experiments are carried out between the proposed EaiNet-EL optimization and the canonical aiNet optimization on two classical benchmarks. The simulation results indicate that the proposed EaiNet-EL optimization outperforms the canonical aiNet optimization in both the final solution and convergence speed. Furthermore, a typical application — an industrial PID control system, is considered to test the performances of multiple optimization approaches including the proposed EaiNet-EL optimization. The step response of the control system shows better performance occurs under the proposed EaiNet-EL optimization than that under other three approaches.