| In order to achieve the "double carbon goal"(carbon peak in 2030 and carbon neutral goal by 2060),building an efficient and safe energy system,vigorously developing renewable energy,and promoting clean energy transformation is an effective measure.As a typical renewable energy,solar energy has high research value,but photovoltaic power generation is greatly affected by climate and external environment,and its output is highly random.With the increasing proportion of photovoltaic power in the power grid,in order to better integrate photovoltaic power generation into the large power grid,reduce its impact on the power grid,improve grid safety and accuracy,dispatch and power quality,and accurately predict photovoltaic power generation output Very important.The basic principles and system composition of photovoltaic power generation systems are described,including the photoelectric effect of photovoltaic panels,the conversion of electrical energy by photovoltaic inverters,the charging and discharging process of batteries,and the classification of loads.Using the historical data of existing photovoltaic power plants in 2017,the impact of four key weather factors,including radiation intensity,temperature,humidity and wind speed,on the output power of photovoltaic power generation systems was analyzed,laying the foundation for subsequent photovoltaic power predictions.From the perspective of improving the accuracy of photovoltaic power prediction,evaluate and compare the advantages and disadvantages of existing forecasting methods a neural network prediction model based on integrated learning is proposed for photovoltaic power prediction,and MAE(Mean Absolute Error)and RMSE(Root Mean Square Error)are established.,R^2(coefficient of determination)three evaluation indicators are used to evaluate the accuracy of the prediction results.Select ANN(Artificial Neural Networks,artificial neural network),DNN(Deep Neural Networks,deep neural network),CNN(Convolutional Neural Network,convolutional neural network)as the base learner of the Bagging method to construct the prediction model.Overcome the limitations of a single prediction method,and improve the prediction accuracy of photovoltaic power generation.Modeling is carried out based on the data collected by the photovoltaic power station of a solar energy design and research institute,the random forest fills in the missing values,the XGBOOST method calculates the importance of features and visualizes them,and then builds models respectively,ANN builds a 3-layer neural network model,and DNN builds an 8-layer neural network.Network model,CNN builds a convolutional neural network model,uses full connection to connect adjacent layers of neural network,dropout prevents overfitting,grid search method selects the optimal hyperparameter combination such as epoch,batch＿size,and draws the corresponding Neural network diagram.Combining the model evaluation indicators MAE,RMSE and R2,the comparative analysis of the integrated model and the single model shows that the regression prediction effect of neural network integrated learning is better than that of deep learning single model,which significantly reduces the prediction error and greatly improves the prediction accuracy of photovoltaic power. |
