Distributed photovoltaic power data restoration method based on multivariate information

The invention discloses a distributed photovoltaic power data restoration method based on multivariate information, and belongs to the field of data processing. On the basis of realizing missing value reconstruction by utilizing satellite remote sensing information and an interpolation function, a time dependency relationship between a power missing sequence and continuous complete power sequences on two sides is fully considered, and a power restoration model based on a bidirectional time sequence feature modeling unit and a cross correlation focusing algorithm is established; in the power data restoration process, multivariate information such as double-side time sequence correlation information and satellite remote sensing information is effectively integrated. The method is suitable for a distributed photovoltaic field station which is not equipped with special meteorological measurement equipment and is high in output fluctuation and small in number of high-correlation space-time similar samples, and the