Transformers are expensive and difficult to regularly maintain and check. In order to achieve fault detection of transformers and to ensure that they can be in normal operation for a long time, a simple on-line monitoring algorithm is proposed. First, a temperature model of a transformer hot spot is studied, and the accuracy of the model is verified by simulation. Secondly, the regression algorithm of transformer fault detection is theoretically analyzed, and the simulation is verified by the real data program algorithm on the transformer. The simulation results show that the parameters from the proposed algorithm would change significantly during a transformer fault, while the parameters remain basically unchanged when the transformer load changes slowly or suddenly. This verifies that the proposed algorithm could effectively identify a transformer fault. This work is supported by Science and Technology Project of State Grid Corporation of China (No. 522804170002).