Data-driven photovoltaic power generation short-term prediction method

The invention discloses a data-driven photovoltaic power generation short-term prediction method, and the method comprises the steps: carrying out the standardization of the historical power generation data of photovoltaic power generation and factors affecting the photovoltaic power generation, calculating the correlation degree between the standardized factors and a historical power generation data curve of photovoltaic power generation based on a gray correlation analysis method, and carrying out the sorting of the correlation degree; selecting factors with association degrees greater than a preset value as input data; establishing an LSTM network model, and adjusting parameters of the LSTM network model according to the MSE prediction error; according to the method, photovoltaic power generation can be effectively predicted by adopting the grey correlation analysis method and the LSTM network model in combination with future meteorological data, and off-line training and on-line photovoltaic data predicti