一种DNA-NSGA-ⅡRBF网络非线性动态系统建模

DNA-NSGA-Ⅱ nonlinear dynamic system modeling approach using RBF neural networks

基于DNA计算操作算子,提出了一种多目标非支配排序遗传算法,用于实现径向基函数(RBF)网络的优化设计。以RBF网络结构最简、拟合精度最高为优化指标,得到一组Pareto最优解,并根据测试数据的误差绝对值之和最小准则,从Pareto最优解集中筛选出最佳RBF网络。连续搅拌反应釜和pH中和过程建模仿真研究表明,该算法是一种有效的"黑箱"动态建模方法。

Based on the operators of DNA computing, a multi-objective non-dominated sorted genetic algorithm （DNA-NSGA-Ⅱ ） was proposed to optimize the radial basis function （RI3F） network. 13oth the structure complexity and the approximation performance were optimized. Once a group of Pareto optimal solutions were derived, the appropriate RBF network could be chosen in terms of the sum of absolute values of the testing errors. Simulation results of a continuous stirred tank reactor （CSTR） and pH neutralization process showed that the proposed method is an efficient black box dynamic modeling approach.