Abstract: To overcome the pre-distortion limitations of structure and precision for nonlinear power amplifiers with memory in wireless communication system, an adaptive pre-distortion method with a dual-loop learning structure based on fuzzy neural network model identification is proposed. This method is based on real-valued time-delay fuzzy neural network model, and uses simplified particle swarm optimization algorithm to ensure network parameters with indirect structure off-line training as the initial of the pre-distortion model. Then it uses least mean square algorithm to adjust the parameters of pre-distorter adaptively with direct structure on-line training, and fits the nonlinearity and memory effect of power amplifier. This method has simple structure, fast convergence speed and high precision, and avoids falling into the local optimum. The results show that this scheme makes adjacent channel power ratio improve about 7 dB than the method of classic dual-loop learning structure, and improve the linearity of the power amplifier obviously. Therefore the simulation results verify the feasibility of the method.

Key words: power amplifier, pre-distortion, fuzzy neural network, nonlinearity with memory, simplified particle swarm optimization algorithm

摘要： 针对无线通信系统中记忆非线性功率放大器预失真结构不足和精度不高等问题，提出了一种基于模糊神经网络模型识别的双环学习结构自适应预失真方法。该方法以实数延时模糊神经网络模型为基础，采用改进的简化粒子群优化（Simplified Particle Swarm Optimization，SPSO）算法进行间接学习结构离线训练模糊神经网络来确定模型参数，作为预失真器的初值，再利用最小均方（Least Mean Square，LMS）算法进行直接学习结构在线微调整预失真器参数，拟合功放的非线性和记忆效应。该方法结构简单，收敛速度快且精度高，避免了局部最优。实验结果表明，该方案邻信道功率比经典的双环结构预失真方法约改善7 dB，功放的线性化性能明显提高，由此验证了其可行性。

关键词: 功率放大器, 预失真, 模糊神经网络, 记忆非线性, 简化粒子群优化算法