Application Of Gated Recurrent Unit Neural Network In Photovoltaic Maximum Power Tracking

Solar photovoltaic power generation is a technology that uses solar energy to convert into electrical energy and is being used more and more widely.However,it is prone to not working at the maximum power point during operation,resulting in wasted energy.Therefore,it is important to study the high-precision solar PV system maximum power point tracking algorithm.In this paper,the maximum power point tracking algorithm of PV system is studied based on gated recurrent unit(GRU)neural network,which mainly includes the following three parts.1)Simulation of the output characteristics of the solar cell.In Matlab/Simulink simulation environment,the solar cell simulation model is built,the electrical characteristics of the solar cell are analyzed,and the solar cell output characteristics under different conditions are obtained by modifying the environmental data characteristics,including uniform solar irradiance and solar cell temperature,and inhomogeneous solar irradiance and solar cell temperature.2)When the solar irradiance is uniform,a hybrid GA-GRU-MPPT algorithm is proposed.Genetic algorithm(GA)is used to optimize the GRU neural network to obtain the optimal GRU network parameters.The prediction comparison experiments with BP-MPPT and GA-BP-MPPT algorithms show that the GA-GRU-MPPT hybrid algorithm has some advantages in prediction accuracy.(3)When the solar irradiance is not uniform,the GAN-CNN-GRU-MPPT hybrid algorithm is proposed.Generative adversarial network(GAN)is used to expand the environmental data when the solar irradiance is not uniform;convolutional neural network(CNN)is used to extract the data features of solar irradiance and solar cell temperature received by three solar cells,and the GRU neural network is used to extract the time-series features of the environmental data to predict the voltage value at the maximum power point.Compared with the prediction results of BP,RNN,LSTM and GRU algorithms,the hybrid GAN-CNN-GRU-MPPT algorithm predicts less error and the prediction results are closer to the theoretical voltage values when the solar irradiance is not uniform.The environmental data in this thesis include solar irradiance and solar cell temperature,and the original data are obtained from the PV cloud platform of Xiaoxing Group.Four simulation scenarios are considered in the paper,and the simulation results show that the maximum power point tracking accuracy of the GA-GRU-MPPT hybrid algorithm always remains above 98% when the solar irradiance is uniform;when the solar irradiance is not uniform,the maximum power point tracking accuracy of the GAN-CNN-GRU-MPPT hybrid algorithm can reach 99%.