基于PSO-ELM的射频功率放大器行为建模  

Behavior Modeling of RF Power Amplifier Based on PSO-ELM Algorithm

作　　者：邢光宇 安文星 Xing Guangyu;An Wenxing(School of Microelectronics,Tianjin University,Tianjin 300072,China)

机构地区：[1]天津大学微电子学院,天津300072

出　　处：《南开大学学报（自然科学版）》2022年第2期31-35,共5页Acta Scientiarum Naturalium Universitatis Nankaiensis

基　　金：国家自然科学基金(61504092);国家973项目(61331901)。

摘　　要：为了有效地解决原始极限学习机算法中由于网络输入参数选择的随机性而引起的在射频功放行为建模应用中的建模精度不理想以及不稳定的问题,粒子群优化的极限学习机算法首次被引入到射频功率放大器的行为建模当中.利用粒子群优化方法来选择原始极限学习机算法中单隐藏层前馈神经网络的输入参数(包括输入权重和偏置).对E类射频功放的行为建模实验结果表明,粒子群优化的极限学习机可以有效改进原始极限学习机对射频功率放大器的外部行为的建模和预测能力.The extreme learning machine algorithm optimized by the particle swarm optimization is introduced into the behavior modeling of RF power amplifier for the first time, which can effectively solve the problem that the original algorithm is not ideal and unstable in the application of RF power amplifier behavior modeling due to the randomness of network input parameter selection. In the improved algorithm,the particle swarm optimization method is used to select the input parameters(including input weights and bias) of the single hidden layer neural network in the original extreme learning machine algorithm. The behavioral modeling experiments of a class-E RF power amplifier show that the particle swarm optimization extreme learning machine can effectively improve the modeling and prediction ability of the traditional overlimit learning machine for the external behavior of modeling the RF power amplifier.

关 键 词：粒子群优化 极限学习机 射频功率放大器 行为模型 

分 类 号：TN722.7[电子电信—电路与系统]