Metering device fault tracing method based on deep belief network

The invention discloses a metering device fault tracing method based on a deep belief network. The method comprises the following steps: constructing a sample library through the integration of the ledger data of an intelligent electric meter and the power utilization characteristic data of a fault electric meter, and carrying out learning on an offline ledger data sample library through a deep belief network model; inputting attribute data contained in an online ledger data sample library into the trained deep belief network model to judge whether the operation state of a metering device is normal or not; and carrying out further fault tracing on an electric meter detected to a have a fault through a power utilization curve fault characteristic sample library, and finally obtaining an accurate metering device fault tracing result. The metering device fault tracing method based on a deep belief network provided by the invention provides scientific and reasonable guidance for a meteringdevice rotation period a