Prediction Of Photovoltaic Panel Temperature And Photovoltaic Power Generation Based On Deep Learning

The reserves of non-renewable energy sources including oil,coal and natural gas are limited.At the same time,along with the large-scale exploitation and consumption of fossil fuels,serious environmental problems are arising.Countries are attaching increasing importance to the development of new energy sources.In recent years,the photovoltaic industry has been developing rapidly under the influence of various government policies supporting renewable energy.However,photovoltaic power generation is random and intermittent.If the electricity from photovoltaic power generation is directly connected to the power grid system,it will bring great challenges to the stability and security of the existing power grid.Accurate prediction of photovoltaic power generation can effectively mitigate the impact of photovoltaic power generation on power system operation,and photovoltaic panel temperature plays a decisive role in photovoltaic generation electric power prediction,so accurate prediction of photovoltaic panel temperature is a further step to improve electric power accuracy prediction.Based on the deep learning technology and the historical data of Ningxia Zhongwei Photovoltaic Power Station,in this paper the photovoltaic panel temperature and electric power prediction were stiudied.The research contents include the following parts:First,a single long and short term memory network(LSTM)model was used to predict the photovoltaic panel temperature and photovoltaic electric power.LSTM,as a variation model of cyclic neural network,solved the problems of gradient disappearance and gradient explosion in the process of gradient reverse transmit.Finally,a more accurate result was obtained.Second,on the basis of the single LSTM network model,CNN-LSTM network model was proposed.When using short-term and long-term memory networks for photovoltaic panel temperature and photovoltaic generation electric power prediction,given that the convolutional neural network had strong ability to extract the data,firstly convolutional neural networks was adopted for feature extraction of photovoltaic panel temperature,photovoltaic generation electric power and the related data,and then the characteristic vector data was taken as input data of LSTM,and LSTM used its internal memory unit to dynamically describe the sequence changes of the photovoltaic panel temperature or photovoltaic electric power.In order to further improve the accuracy of photovoltaic panel temperature and photovoltaic generation electric power prediction,adaptive moment estimation algorithm(Adam)was used to optimize the weights and thresholds of the long and short term memory network,to gain photovoltaic panel temperature and photovoltaic electric power.The empirical results showed that the proposed CNN-LSTM model was more accurate than the LSTM model.Finally,Boosting integration algorithm was proposed based on the improved CNN-LSTM network model.Boosting integration algorithm's training process was ladder-like,and its training set was converted,and converted into a strong classifier in the end,according to a certain method after the basis classifiers were trained as the order.Prediction accuracy of photovoltaic panel temperature and photovoltaic generation electric power based on Boosting integration algorithm was better than that of using single LSTM model and CNN-LSTM model...