Short-term Photovoltaic Power Prediction Research Based On Neural Networks

With the continuous consumption of traditional energy,more and more renewable energy,such as wind and solar energy,has become an important strategy for sustainable development in many countries.Solar energy is a typical intermittent energy,and photovoltaic power generation is greatly affected by the environment,weather and other factors.When a large range of output is connected to the grid,it will affect the safe and stable operation of the power system.Therefore,accurate PV power prediction can provide reference for real-time regulation and control of power grid,which has certain theoretical and practical significance.The deficiency of traditional forecasting methods in data and models leads to low prediction accuracy.Aiming at the problems in PV power prediction,this paper carries out related research from two aspects: data mining and model building.Firstly,this paper introduces the basic principle and system composition of photovoltaic power generation,analyzes several factors influencing photovoltaic output based on actual data,and studies the output characteristics under different weather types.To improve forecast efficiency and experimental data for the outlier detection and replace and normalized processing,through the K-means clustering data can be divided into three small sunny,cloudy and rainy sample data set.Under the three small sample data set,considering various environmental factors,uses the principal components analysis method to a variety of environmental factors sequence dimension.Determine the final input characteristics of the prediction model.In order to avoid the limitations of a single model and give play to the advantages of the combined model,the principle of convolutional neural network(CNN)and bidirectional Long short-term memory network(Bi LSTM)are introduced,the CNN-Bi LSTM combined photovoltaic power prediction model is established and model parameters are selected.In order to further improve the prediction accuracy and reduce the low accuracy of the output prediction of rainy days caused by the fluctuation of environmental factors,based on the CNN-Bi LSTM model,adopts variational mode decomposition(VMD)to decompose the environmental sequence,selects the sequence with the highest correlation with PV output from the obtained sub-sequence.The principal component analysis method was used to reduce the redundancy among the input features and determine the final input features under the three weather types.A VMD-CNN-Bi LSTM combined PV power prediction model is established,and the prediction verification is carried out under the small sample data of three weather types.Finally,the experimental is verified that the CNN-Bi LSTM combined PV power prediction model adopted in this paper is better than the LSTM and CNN-LSTM prediction models.The VMD-CNN-Bi LSTM combined PV power prediction model constructed on the basis of CNN-Bi LSTM model has better prediction accuracy than the CNN-Bi LSTM,VMD-LSTM and VMD-CNN-LSTM PV power prediction models,which verifies the effectiveness of the proposed method in this paper.Figure [55] Table [12] Reference [82]...