Basic principles of DGA – Part II Learning, using, and creating your own view on DGA

Basic principles of DGA – Part II Learning, using, and creating your own view on DGA

Vol. 8 issue 1

Limits values for DGA tests and results

The exact specification for DGA is something that all transformer users, managers, specific DGA software manufactures, online devices producers, and all the industry are looking for. For other oil test types, it is a usual behaviour to refer the measured value to the relevant limit from the corresponding guide or standard. Having limits like for dissipation factor or oil acidity does not work for DGA. It is probably the key obstacle for DGA manual or software-based diagnosis. The hard challenge here becomes selecting values that represent the specific equipment in the light of new developments and knowledge, and also considering the maintenance policy of the transformer owner. Adopting the published and already processed value is an excellent stage for the initial learning process for DGA interpretation. Using such general limits or normal or cut-off values may lead to substantial pricey consequences. In this essential aspect, a new version of IEEE C57.107 [9] published the 90 % approach of limits value, shown in Table 1. This approach was previously presented in CIGRÉ WG on DGA [11] and is a significant step forward to improve the uniqueness of the DGA limits values.


A statistical approach for the calculation of DGA limits may seem simple, however, it is not straightforward and easy to calculate a 90 % statistical limit, for example


The readers will be able to understand and exercise this statistical approach for DGA limits. Even if it may seem simple to calculate a 90 % statistical limit, it is not a straightforward statistical calculation. The ones who need to compute are demanded to select different options from the primary and extensive database to the specific 90 % limits. Of course, the challenge is inversely proportional to the size of the database. The constrains of data securities imply minimum size databases, and therefore more extensive efforts to develop such kind of limits.

The countenance of these difficulties also comes into view from the new CIGRÉ WG approach, as shown in Table 2. The selected values for two options of database size emphasise the importance of these selections and the meaning of the relation between the database size and the 90 % computed value.


Log in or subscribe to read the whole article.