With the change of energy consumption, the consumption proportion of renewable energy power generation rises gradually. Photovoltaic power is chosen as the research object to study the characteristics of power curve under different weather conditions and to analyze the correlation of the efforts of various meteorological factors and photovoltaic power generation. Then by combining Empirical Mode Decomposition (EMD) and Long-short Term Memory (LSTM), a load forecasting model suitable for photovoltaic power generation is proposed. Firstly, the historical sequence of preprocessed photovoltaic power generation is reconstructed. Then, the reconstructed historical time series is decomposed in EMD model, and the decomposed sub-sequences are respectively predicted in the LSTM network. Finally, the results of each sub-sequence are superimposed to obtain the predicted result of photovoltaic power generation. The actual output data of photovoltaic power generation in a certain region in China is used to test the model. Compared with the ARIMA model, SVM model, LSTM model and EMD-SVM model, the model proposed in this paper has lower prediction error and can effectively improve the prediction accuracy of photovoltaic power generation.