Performance indices of power system fault diagnosis methods are summarized， followed by presentation of the method for construction of a test sample set to ensure an objective and precise evaluation. In order to enhance the fault tolerance ability of the fault diagnosis model， information correction is thus adopted based on the maximum likelihood decoding theory. Then the power network fault diagnosis process is carried out by the encoded elementary net（EEN） system， of which a new method for protection reclassification is introduced to simplify the EEN model as well as to enhance its translatability. Based on the test results of all the signals in the test sample set（two examples are given to show the diagnosis process）， it is shown that the power network fault diagnosis model is simple， fast and is tolerant for information aberrance， which can meet the requirements of application.