Photovoltaic power ultra-short-term prediction method and system

The invention discloses a photovoltaic power ultra-short-term prediction method and system, and the method comprises the steps: obtaining a photovoltaic power sequence of a similar power station according to the distribution condition of a photovoltaic power station; determining a highly correlated power station of the target power station by combining the photovoltaic power sequence of the similar power station and the photovoltaic power sequence of the power station to be predicted with grey correlation analysis; dynamically capturing the correlation between each characteristic variable of the target photovoltaic power station at the to-be-predicted moment and the photovoltaic power by introducing an attention mechanism; and dynamically capturing the correlation between the to-be-predicted moment and each historical moment by using a multi-layer nested space-time network, and further optimizing the prediction result by considering the change relation of the photovoltaic power along with the time. The correl