Photovoltaic power station super short term power prediction method based on least square method

The invention relates to a photovoltaic power station super short term power prediction method based on a least square method. The method includes the following steps: determining sunrise time and sunset time of the position of a photovoltaic power station in a prediction day, inputting prediction data into a prediction system in a rolling mode, screening abnormal data in actual power data input in the rolling mode, adopting the least square method to conduct parameter fitting on the historical short term predicted power and the historical actual power of corresponding time of the predicted time in the coming four hours, determining super short term predicted power values in the coming four hours and setting the predicted power values of time periods contained in the super short term predicted power values in the coming four hours and smaller than the sunrise time or larger than the sunset time to be zero. By means of the method, reliability of short term predicted data is effectively utilized, and a good prediction effect is obtained under the condition that output fluctuation is not large. The method is applied to a part of domestic province electric power company scheduling mechanisms. A calculation result meets the actual engineering requirements according to a verification result.