Abstract: Considering the complicated structure and the diversified information system of intelligent power network， a novel method for fault diagnosis was proposed. In the proposed method there were three parts including switch layer used for the simple fault diagnosis， feeder layer strived to resolve complex fault in the case of abnormal switch and protection information， and substation layer used to judge multitype fault in the complex system. Simultaneously， dynamic diagnosis strategy was adopted to adjust diagnostic entrance and structure longitudinally. And the improved depthfirst searching algorithm， Petri net reasoning and intuitionistic uncertaintyrough sets theory were applied to each layer respectively in the diagnosis. The simulation results showed that the adaptability of each layer diagnosis is enhanced， and the efficiency and accuracy of fault diagnosis are improved.In addition， kinds of complex fault can be accurately diagnosed with good practical application value.

Key words: fault diagnosis, multisource information, multilayer, Petri net, intuitionistic uncertaintyrough sets