Method for training photovoltaic power prediction model

The invention relates to a training method of a photovoltaic power prediction model, belongs to the technical field of photovoltaic power generation, and is used for solving the problems that an existing photovoltaic power prediction model is low in prediction precision and an existing deep learning network model is low in calculation efficiency when being applied to photovoltaic power prediction. The method comprises the following steps: classifying daily weather of a photovoltaic electric field to be predicted in a historical period in a region, and dividing the weather into stable weather and turning weather; based on the generalized weather type, classifying the weather of each historical time period in the steady weather day and the turning weather day, and dividing the weather into a plurality of sub-weather types; constructing a training data set of each sub-weather type; and training a pre-established photovoltaic power prediction model through the training data set of each sub-weather type to obtain