The invention provides a photovoltaic power station generating capacity prediction system and method. The method comprises: collecting environmental influence factors, internal influence factors, power generation historical data and to-be-measured day data; performing denoising processing on the collected data; calculating the actual photoelectric conversion efficiency of the assembly according tothe internal influence factors; using historical data of the environmental influence factors and the actual photoelectric conversion efficiency of the assembly as input vectors; training the elasticadaptive BP neural network model; performing model optimization by repeatedly utilizing the solar radiation intensity, the atmospheric temperature, the actual photoelectric conversion efficiency of the assembly and the generating capacity of each day in a historical time period; using the environmental influence factors of the to-be-measured day and the actual photoelectric conversion efficiency of the assembly as model input. According to the method, the real-time photoelectric conversion efficiency value of the photovoltaic module is obtained through fitting analysis, so that the influence caused by the internal change of the photovoltaic module is reduced, the prediction accuracy of the generating capacity of the photovoltaic power station is greatly improved, and an accurate data reference is provided for power dispatching.