PHOTOVOLTAIC POWER COMBINATION FORECASTING BASED ON CLIMATE SIMILARITY AND SSA-CNN-LSTM

Aiming at the problem that the forecasting accuracy of photovoltaic power may be affected by the lack of high-resolution meteorological data，a high-resolution photovoltaic power combination forecasting model is proposed，which combines climate similarity with singular spectrum analysis （SSA）， convolutional neural networks （CNN） and long short-term memory （LSTM）. SSA is employed to decompose the photovoltaic sequence into different subsequences，and CNN-LSTM based on day ahead prediction model is established to capture the continuous characteristics of photovoltaic output. Moreover，the climate similarity is used to select similar days from low-resolution meteorological data to achieve high-resolution photovoltaic output prediction. Finally，the grey correlation analysis is utilized to obtain the combination weights to get the final prediction results. The simulation results show that the combined prediction model can effectively improve the prediction results of high-resolution photovoltaic power，and obtain high prediction accuracy. © 2023 Science Press. All rights reserved.