Research On Nonlinear Distortion Suppression Algorithm Based On Optimal Cascade Forward Neural Network

With the vigorous development of society,various businesses are increasingly dependent on communications,and the spectrum resources used by wireless radio communications are becoming increasingly tight.In order to improve spectrum utilization,modern wireless communication systems increasingly use non-constant envelope modulation technology and multi-carrier transmission technology.However,peak-to-average signal obtained by the new technology will produce more serious nonlinear distortion after processing by the power amplifier,resulting in a significant reduction in communication quality.As an indispensable part of the communication system,the power amplifier has a decisive influence on the performance of the communication system.In order to further improve the communication quality,the research on the nonlinear distortion suppression of the power amplifier is particularly important.As mentioned above,this thesis studies the post-distortion scheme at the receiving side.The main work and innovations are as follows:(1)Using the ability of neural network to fit arbitrary nonlinear functions,two types of neural networks with different structures are used to perform post-distortion processing on the signal at the receiving side.After the neural network structure,activation function and training algorithm are determined,the convergence performance of the two neural networks is compared,and the neural network with better convergence performance is selected for further research.(2)Aiming at the problem that the convergence performance of the neural network is affected by the initial weights and biases,a cascade forward neural network based on the whale optimization algorithm is designed.The algorithm optimizes the initial weights and biases.After the neural network is trained with the optimal initial weights and biases,the convergence performance is effectively improved.It is then used for post-distortion processing at the receiving side,while using traditional adaptive filtering algorithms as a control.The simulation results show that under the action of the whale optimization algorithm,the convergence performance of the cascade forward neural network is improved,and the nonlinear suppression performance is improved.(3)A post-distortion scheme of cascade forward neural network based on differential evolution algorithm is designed.After global optimization,the algorithm obtains the optimal initial weights and biases,and then uses the optimized network for post-distortion processing at the receiver.After simulation,it is verified that the differential evolution algorithm can effectively improve the convergence performance and nonlinear suppression performance of the cascade forward neural network.