The present disclosure relates to a distributed photovoltaic short-term power generation power prediction method,
system, device and medium, belonging to the technical field of photovoltaic power prediction, the distributed photovoltaic short-term power generation power prediction method, by acquiring continuous
multiple frame satellite cloud picture sequence, identifying and matching independent cloud layer target, calculating its moving
direction vector angle, generating the moving trajectory line of each cloud layer based on the moving
direction vector angle, by from the generated cloud layer obscuration
time sequence, the thickness parameter value of each cloud layer is parsed out, and the corresponding solar
radiation attenuation coefficient is further obtained, while accurately analyzing the coverage area percentage and coverage duration of the cloud layer above the
power station, by accurately depicting the dynamic characteristics of the cloud layer and its influence on the solar
radiation, the prediction accuracy of the distributed photovoltaic short-term power generation power is significantly improved, which can provide reliable data support for
power station operation scheduling,
power grid power balance and energy optimization configuration.