Research On Transmitter Rf Distortion Of Massive MIMO System And Design Of Compensation Scheme

In recent years,the fifth generation of communication technology(5G)has developed rapidly,bringing great convenience to people's lives.Massive Multiple-Input Multiple-Output(MIMO)technology is one of the key technologies of 5G.Massive MIMO systems have the advantages of high spectrum efficiency and large system capacity.However,due to the technical limitations of radio frequency components at the transmitter,massive MIMO systems inevitably have multiple radio frequency distortion problems,which become an important bottleneck restricting system performance.In order to improve the system performance of massive MIMO,this thesis studies the compensation technology of radio frequency distortions at the massive MIMO transmitter.The main work is as follows:Firstly,this thesis focuses on the power amplifier(PA)nonlinearity,in-phase/quadrature(I/Q)branch imbalance,RF link crosstalk and other radio frequency distortion problems in massive MIMO systems.A scheme of RF distortion compensation for massive MIMO system transmitters based on Real-Valued Timed-Delay Neural Network(RVTDNN)is proposed.This solution uses multiple RVTDNN pre-distortion networks to pre-process the transmitted signal,compensate for the undesirable characteristics of the transmitter's radio frequency,and improve system performance.In addition,in order to determine the value of each hyperparameter in RVTDNN,this thesis also proposes a hyperparameter optimization scheme based on quantum genetic algorithm for the pre-distortion network.Compared with the traditional genetic algorithm-based hyperparameter optimization scheme,this scheme can obtain the optimal solution with small population number and low time complexity.Secondly,according to the hybrid beamforming structure and the time-varying characteristics of the transmitter,a radio frequency distortion compensation scheme based on online deep neural network is proposed for massive MIMO transmitter.This scheme is based on deep neural network to preprocess the transmitted signal and adopts online learning method so that the neural network can adapt to changes in transmitter characteristics and improve the performance of the system.