Photovoltaic data clustering method, photovoltaic power prediction method, storage medium and device

A photovoltaic data clustering method provided by the present invention is characterized by setting the clustering categories of a neural network according to the characteristic index system of the irradiance containing the total irradiance time sequence difference sequence variance, using the samples to train a clustering model and then clustering the acquired light irradiance. The method is strong in data regularity for obtaining the clustering results, can achieve a high-precision short-term photovoltaic power prediction purpose.