The knowledge base of black box fault diagnosis of electric power metering device contains a large number of abnormal features, which leads to unsatisfactory fault diagnosis results. Therefore, a black box fault diagnosis method of electric power metering device based on digital filtering and EMD method is proposed. Filtering the black box information of the power metering device under the digital filter; According to EMD method, a knowledge base is established for the black box fault diagnosis of power metering devices, and the local information increment matrix based on local sampling data information is used to judge whether there is a fault, thus completing the black box fault diagnosis of power metering device. According to the experimental findings, the proposed fault diagnosis method exhibits a negligible distinction and the calculated theoretical catastrophe value, the diagnosis result is the most accurate and consistent with the actual value, and the abnormal value of the diagnosis is 0, which has a good diagnosis effect.