摘要
为了更准确地描述有记忆效应的射频功放特性,提出了一种改进的简化粒子群优化(PSO)算法,并结合自适应模糊推理系统(ANFIS)建立模糊神经网络功放模型。改进的简化PSO算法仅保留粒子的位置项,加入了随机的个体最优候选解,由粒子的当前位置、个体最优解、全局最优解和随机的个体最优候选解共同决定其位置项;采用线性递减惯性权重,并利用异步变化的动态学习因子,且新颖地引入拉普拉斯系数,从而增加了种群多样性,加快了收敛速度,避免陷入局部最优。由模型仿真对比可知,该方法建立的功放模型结构简单、收敛快、误差小、精度高,从而验证了建模方法的有效性和可靠性。
In order to describe the memory effect of RF power amplifier model perfectly,this paper proposed an improved simplified particle swarm optimization（ PSO） algorithm,and built fuzzy neural network model for power amplifier combining with the adaptive neural fuzzy inference system（ ANFIS）. The improved simplified PSO algorithm had only position parameter,and added random local best candidate solution. The position parameter was directed by the particle＇s current position,local best solution,the global best solution and a random local best candidate solution. The algorithm employed linear decreased inertia weight,also used asynchronous changed dynamic learning factor,and novelly introduced Laplace coefficient. The new algorithm improved the swarm diversity,and speeded up the convergence,avoided being trapped in local optimal solution. The simulation comparison of the models shows that this modeling approach has the characteristics of simple structure,fast convergent,small error and high precision. Thus it verifies the validity and reliability of the modeling method.
出处
《计算机应用研究》
CSCD
北大核心
2015年第4期1000-1003,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(61372058)
辽宁省高等学校优秀科技人才支持计划资助项目(LR2013012)
辽宁工程技术大学研究生科研资助项目(5B2014032)
关键词
记忆功放模型
自适应模糊推理系统
简化粒子群算法
个体最优候选解
拉普拉斯系数
memory power amplifier model
adaptive neural fuzzy inference system
simplified particle swarm optimization
local best candidate solution
Laplace coefficient
