Power transformer is one of the key electrical apparatus in power system and its reliability is directly correlated with the safety of power system. However

there are some controversies in existing standards

for example

gas species is incomplete

transformer classification is rough

etc. Therefore

we proposed a method to calculate the differentiated attention and warning values of gas concentration and gas production rate. This method was based on large amounts of DGA data in field and it was classified according to intrinsic properties of transformers firstly. Secondly

the distribution model of gas concentration and gas production rate under each classification was established. Then the ratio of defect and fault was associated with cumulative probability. At last

the attention and warning value of gas concentration and gas production rate under different classification were calculated by using inverse cumulative distribution function. The attention and warning values of transformers with different voltage properties were calculated by the proposed method. Examples in field are used to verify the method

and results show that the differentiated values calculated by this method can reduce the incidence of false negatives and the accuracy rate of warning is 100% for this example in field.