Permutation entropy is a kind of model which calculating the turbidity of data adjacent size relation. The construction method of spatial permutation entropy is employed to establish the state diagnosis model for generator.The original vector is constructed through multiple measurement points of stator coil temperature

and the ordering vector space is reconstructed through the vector dimension and spatial delay parameter. The permutation entropy of original vector is calculated to indicate the generator state. It is found that the permutation entropy value calculated from historical monitoring data of stator coil temperature can indicate different load conditions of different classes of generators. With the increase of power generation load

the mean value of permutation entropy increases and the standard deviation of ranking entropy decreases. The spatial permutation entropy formed by appropriate vector dimension and spatial delay parameter can diagnose the five abnormal conditions of generator stator coil temperature. It can provide accurate signal for state identification and fault alarm of a large number of temperature measuring point groups of generators.