Advanced analytics for transformer asset management
Case study on a 25 MVA transformer
It is very hard to manage the fleet of transformers and to estimate their reliability, probability of failure, and to plan activities accordingly without the aid of modern calculation methods and algorithms
The CIGRE working groups 12.05 and A 2.37 published the transformer failure survey ELT 088  first in 1983 and the latest version 642  in 2015, respectively. Based on these surveys, failure modes, failure causes (based on kV) and position of failures for power transformers were identified. Key findings included:
· Failure mode: based on 964 transformers, the dielectric failure mode was the predominant with 36.62 %, followed by the mechanical failure at 20.02 %.
· Failure causes: failure causes include ageing (12.34 %), external short circuit (11.62 %), improper repair (6.02 %), etc.
· Failure position: the three most common failure positions include winding (47.4 %), bushing (14.4 %) and tap changer (23.2 %), while other positions contributed to the failure as well, but the percentages were minor.
Based on the above, different models have been formulated to determine the ‘health index’ of power transformers [3, 4]. Some of the common parameters used in these models include dissolved gas analysis (DGA), oil quality parameters (dielectric strength, dissipation factor, acidity, moisture, colour, and interfacial tension of the oil), furans, transformer age, loading history, tap changer and bushing data, maintenance data, etc.
A new decoupled approach was presented in [5, 6] which forms the backbone of APM Edge software . The procedure developed selected those failure modes and brought to the ‘analysis matrix’ those operational parameters that play a role in that specific failure mode. It is very important to note that the those that do not contribute to a given failure mode, are not analysed together with those directly associated to a failure mode to enable decoupled analysis. This decoupled failure mode approach based on reliability centred maintenance philosophy is illustrated in Fig. 1.
For example, a bushing may fail due to several reasons, such as design and manufacturing issues, storage, maintenance and operations, external causes, etc. Each of these possible failures may require different data inputs in order to be properly assessed, such as:
· Bushing installation date
· Bushing power factor and capacitance
· Bushing reference power factor and capacitance as per manufacturer
· Bushing voltage class
· Bushing construction type
· Bushing inspection results – hot spots, cracked, oil, oil leakage, dirty bushings, etc.
· Bushing maintenance date.