计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (3): 120-123.DOI: 10.3778/j.issn.1002-8331.1505-0057

• 网络、通信与安全 • 上一篇    下一篇

基于改进粒子群算法的模糊小波神经网络建模

南敬昌,田  娜   

  1. 辽宁工程技术大学 电子与信息工程学院,辽宁 葫芦岛 125105
  • 出版日期:2017-02-01 发布日期:2017-05-11

Fuzzy wavelet neural network for modeling based on improved particle swarm optimization algorithm

NAN Jingchang, TIAN Na   

  1. School of Electronic and Information Engineering, Liaoning Technical University, Huludao, Liaoning 125105, China
  • Online:2017-02-01 Published:2017-05-11

摘要: 随着射频功放非线性对射频前端的影响日益增大,使得功放建模变得越来越重要。提出了一种自适应模糊小波神经网络模型结构,并利用改进的粒子群优化算法,建立有记忆的功放模型。将小波函数融入到自适应模糊推理系统的模糊规则中,得到新的网络模型;在粒子群算法中引入最差位置影响因子,提高搜索效率,并进一步简化,忽略粒子的速度项,同时采用与适应度函数值相关的动态变化惯性权重,加快了收敛速度,避免出现“早熟”现象。仿真结果表明:该方法建立的功放模型误差小、精度高,能够有效地表征功放特性。

关键词: 模糊小波神经网络, 小波函数, 自适应模糊推理系统, 改进粒子群优化算法, 记忆效应, 功放模型

Abstract:  As the influence of Power Amplifier(PA) nonlinear increases for Radio Frequency(RF) front-end, PA modeling has been more and more important. An adaptive fuzzy wavelet neural network is proposed, using improved simplified particle swarm optimization algorithm to build PA model with memory effect. First, the wavelet function is combined with the rules of adaptive neural fuzzy inference system to build the new model. The improved particle swarm algorithm not only introduces the worst position influence factor but also simplifies for neglecting the velocity of particle. The inertia weight is dynamic with the change of fitness function value. The new algorithm improves the convergence rate, avoids being trapped in local optimal solution. The simulation results show that this modeling approach can characterize PA feature effectively with small error and high precision.

Key words: fuzzy wavelet neural network, wavelet function, adaptive neural fuzzy inference system, simplified particle swarm optimization algorithm, memory effect, power amplifier model