Ultra-short-term Prediction Of Photovoltaic Power Based On Phase Space Reconstruction And Meteorological Factors

With the introduction of various carbon emission reduction policies,photovoltaic power generation technology has been widely used,reducing carbon emissions to a certain extent and alleviating environmental crises.However,photovoltaic power generation systems are susceptible to various meteorological factors,resulting in intermittent and fluctuating characteristics of photovoltaic power generation output power,which will impact the main grid when connected to the grid and bring certain impact to the dispatch of the power system.difficult.Precise prediction of the output power of photovoltaic power plants can ensure the smooth operation of the grid when photovoltaics are connected to the grid.Therefore,in order to improve the accuracy of the ultra-short-term prediction of photovoltaic power,the paper uses the measured historical data of a photovoltaic power station to make the following two researches:One is to select the input data of the photovoltaic forecasting model.This paper uses the Pearson correlation coefficient to analyze the correlation between various environmental factors and the output of photovoltaic power plants,and obtain the correlation coefficients between various meteorological factors and the output power of photovoltaic power plants.In this paper,the mutual information method and Cao’s method are used to solve the delay time and embedding dimension of the phase space respectively,and the small data amount method is used to solve the maximum Lyapunov exponent to prove the chaotic characteristics of the photovoltaic time series.The original photovoltaic output time series are reconstructed in the phase space.It is constructed as a high-dimensional vector to mine the impact factors contained in the data itself.Based on the phase space reconstruction of the photovoltaic power time series,environmental factors(solar irradiance and temperature)with strong correlation coefficients with photovoltaic output are considered as additional input factors to further improve the prediction effect.The second is to construct a prediction model for ultra-short-term photovoltaic power prediction.Because the required meteorological data is difficult to obtain accurately,this paper proposes a CNN-LSTM prediction model based on chaotic phase space reconstruction and meteorological factors,and applies it to photovoltaic power prediction.In this hybrid prediction model,the high-dimensional vector sum and the corresponding meteorological factors(solar irradiance and temperature)after reconstructing the phase space of the historical power sequence are first used to extract the spatial characteristics of the data through the convolutional neural network model,and then through the long and short-term memory network model extracts the time characteristics of the data,and finally outputs the prediction results.In order to verify the good performance of the proposed model,the RBF neural network photovoltaic power prediction model based on meteorological conditions,the long short-term memory network photovoltaic power prediction model based on chaotic phase space reconstruction,and the CNN-LSTM based on chaotic phase space reconstruction The prediction model is compared and verified by calculation examples.By using error evaluation indicators to evaluate the analysis results of the calculation examples,The MAE of the ultra-short-term photovoltaic power prediction model proposed in this paper under different weather conditions has been reduced by 12.36%,24.68%,and 46.02%;The RMSE has been reduced by 14.19%,22.92%,and 44.12%;Hill’s unequal coefficient dropped by 14.06%,24.25%,43.01%.It shows that the proposed prediction model method has a good prediction effect in different weather types.