The invention provides a super-short-term power prediction method based on a clearance model for a photovoltaic power station. The method comprises the following steps: data preprocessing; model training; and super-short-term power prediction for a photovoltaic power station according to the model. According to the super-short-term power prediction method based on the clearance model, the clearance model of the photovoltaic power station by combination of the clearance model of solar irradiance with output characteristics of the photovoltaic power station, the theoretical maximum output of the photovoltaic power station at specific time can be calculated, and the super-short-term power prediction model is established by employing normalization data and combining with autoregression time sequences on the basis of the theoretical maximum output, accordingly, the prediction accuracy is effectively improved.