The invention discloses an intelligent identification power distribution network fault prediction method fusing
relay protection data, and belongs to the technical field of a
power transmission and distribution network of a power
system, and the method comprises the following steps: S1, collecting and fusing multi-source heterogeneous data; s2, constructing
feature engineering and a dynamic
knowledge graph; s3, multi-
modal deep learning model training is carried out; s4, carrying out
incremental learning and
adaptation on a model based on transfer learning; and S5, fault early warning and decision support are carried out. The
relay protection action
event sequence and the fault recording data are fused with traditional
SCADA measurement data, standing book information and historical records, the limitation that only
SCADA or a single
data source is used in a traditional method is broken through, the protection action event reflects the most
direct response of the
system to abnormity, and fault symptoms can be captured earlier and more accurately.