The invention relates to the technical field of photovoltaic power prediction, and discloses a medium and long term photovoltaic power probabilistic prediction method and
system based on data fusion, and the method comprises the steps: constructing a
time sequence feature extraction module which is used for the
feature extraction of a three-dimensional all-
sky image, and extracting static space
feature data from the three-dimensional all-
sky image through a visual converter, static space
feature data are processed through a self-adaptive bidirectional gating circulation unit module, and change characteristics of the static space
feature data in the time dimension are captured; extracting space-time characteristics of the preprocessed meteorological data through a characteristic
pyramid fusion module; meteorological data
time sequence features, historical photovoltaic power data and three-dimensional all-
sky image
time sequence features are combined, and probabilistic prediction of photovoltaic power is achieved through an MQ-Informer module. According to the invention, high-precision photovoltaic power prediction can be realized, an
electric power company is helped to better manage an
electric power network and a
power transmission line, and safe and stable operation of the
system is ensured.