The invention relates to a PCNN electric power fault area detection method based on local features. According to the method, a pulse coupled neural network (PCNN) synchronous ignition mechanism acts as the basis, the internal parameters are properly simplified, and the PCNN model iteration end rules are set through combination of the neighborhood boundary local features of the fault area and the non-fault area under parameter optimization configuration so that the model is enabled to acquire an infrared image fault area through adaptive iteration. The experiment result indicates that the electrical equipment fault area can be automatically and efficiently identified by the method for aiming at the actual infrared detection image so as to have great fault area detection performance.