基于改进人工鱼群算法的神经网络优化  

Optimization of Neural Network Based on Improved Fish Algorithm

作　　者：魏立新[1,2] 张峻林[2] 刘青松[3] 

机构地区：[1]国家冷轧板带装备及工艺工程技术研究中心,河北秦皇岛066004 [2]燕山大学工业计算机控制河北省重点实验室,河北秦皇岛066004 [3]上海西门子工业自动化有限公司,上海200030

出　　处：《控制工程》2014年第1期84-87,93,共5页Control Engineering of China

基　　金：国家自然科学基金资助项目(61074099)

摘　　要：针对人工鱼群算法的寻优速度慢,后期收敛性差等缺陷提出了一种并行运行方式的改进人工鱼群算法(Improvement Artificial Fish Swarm Algorithm,IAFSA)。进而应用IAFSA算法对BP神经网络初始权值进行寻优,以解决BP网络初始权值选取困难且优化过程中容易陷入局部极值的问题。最后,将IAFSA-BP网络混合算法应用于PID参数的优化,从而克服了PID控制参数难以整定的难题。通过仿真实验,结果表明:改进的人工鱼群算法寻优速度更快,优化值更加合理。应用IAFSA-BP混合算法得到的PID控制参数使得系统响应更快,稳态误差更小,系统性能得到提升。First this article for slow speed of artificial fish algorithm optimization and some shortcomings of convergence bad puts for- ward a IAFSA of parallel operation mode. In order to solve the BP network initial weight value selecting difficult and optimization process easy to fall into local extremum problems, applying IAFSA algorithm to optimize BP neural network initial weights. Finally, the IAFSA-BP network hybrid algorithm is applied to the optimization of the PID parameters, so as to overcome the PID control parameters difficult to setting problem. The result of simulation experiment shows that the modified artificial fish algorithm optimization speed fas- ter, optimization value more reasonable. The PID control parameters of optimization makes the system response faster, steady state error smaller, the system performance improved.

关 键 词：人工鱼群算法 神经网络 局部极值 参数寻优 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]