The invention provides a photovoltaic power station power prediction method and system based on grid-connected inverter operation data, and the method comprises the steps: building a photovoltaic module model according to the parameters of a photovoltaic module in a photovoltaic power station; constructing a power prediction model based on an artificial neural network algorithm; collecting outputdata of the photovoltaic array under the shielding of static shadows with different thicknesses and different shielding sizes, constructing a training set, training the power prediction model, and obtaining a trained power prediction model; the output power of the real-time operation data of the inverter of the photovoltaic array under the sunny sky working condition is acquired; classification and normalization are carried out, the output power of the whole photovoltaic power station is predicted through the trained power prediction model, and power prediction comprises rolling prediction ofthe output power of the photovoltaic power station under the sunny sky working condition and minute-level power prediction of the photovoltaic power station under dynamic cloud cluster shielding. According to the method, the equipment cost is reduced, and the defect that cloud clusters with different thicknesses influence the photovoltaic array power prediction precision is overcome.