基于GHSOM神经网络模型的Doherty功率放大器非线性特性研究

目前，氮化镓基Doherty功率放大器的热效应、电效应等非线性问题以及构建相应的网络模型是功率放大器在无线通信工程中的研究热点。本项目探索研究动态增长型分层自组织神经网络（GHSOM）方法用于辅助分析功率放大器的相关机理，通过剖析GHSOM模型的成长型单元结构，研究GHSOM网络的建模理论、实现方法及鲁棒性，从理论上建立GHSOM整体网络模型；深入研究氮化镓材料的温度效应、Doherty放大器的记忆效应等非线性特性，结合GHSOM成长型单元结构可分层可分裂的优势，构建可表征氮化镓基Doherty功率放大器非线性特性多值函数的GHSOM整体网络模型。本研究将为高效便捷的复杂非线性电路模型设计以及微波大功率放大器的整体建模提供良好的理论指导，对促进功率放大器在未来通信、相控阵雷达以及航空航天等领域的应用具有重要的实际应用价值。

Currently, nonlinear problems such as pyrometric effects and electrical effects of GaN Doherty power amplifiers and the construction of the corresponding network models are the research hotspots on the use of power amplifiers in wireless communication engineering. This project aims at exploring new ideas about analyzing the power amplifiers using growing hierarchical self-organizing neural network (GHSOM) method. We start by analyzing the growing cell structure of GHSOM model. Then, based on the modeling theory, implementation and robustness of GHSOM network, we establish an overall theoretical GHSOM network model. Through in-depth study of the temperature effects of the GaN material, memory effects and other nonlinear characteristics of the Doherty amplifiers, combined with the advantages of growth GHSOM unit structure that they can be layered and split, we construct an overall GHSOM network model that can represent the nonlinear characteristics of the GaN Doherty power amplifiers. The study will be a good theoretical guidance for high-performance and convenient complex nonlinear circuit model design and overall modeling of high-power microwave amplifiers. It also has important practical value on spreading the use of power amplifiers in the fields like future communication, phased array radar and aerospace.