IEC 60599 Guide to the interpretation of dissolved and free gases analysis
IEC 60599 Mineral oil-impregnated electrical equipment in service - Guide to the interpretation of dissolved and free gases analysis
8 Gas concentration levels in service
8.1 Probability of failure in service
8.1.1 General
The probability or risk of having an incident or a failure in service is related to gas concentration levels.

Below certain concentration levels (quoted as typical values or normal values), the probability of having a failure is low. The equipment is considered healthy, although a failure cannot be totally ruled out, even at these low levels, but it is improbable. A first rough screening between healthy and suspect analyses can therefore be obtained by calculating typical values for the equipment.

The probability of having a failure may increase significantly at values much above typical concentration levels. The situation is then considered critical, for even though a failure may never occur at these high levels, the risk of having one is high. Such failures may be divided into two categories:
- failures that develop within a very short time (which are therefore impossible to detect by oil sampling/laboratory analysis, but only by on-line detectors);
- failures developing over an extended time span. Only this second category may be detected by DGA laboratory analysis.

8.1.2 Calculation methods
Utilities with large DGA and equipment maintenance databases are able to calculate the probability of failure in service for a given type of equipment and at a given concentration level of a gas. This can be obtained by calculating the number of DGA analyses which have led to an actual failure or incident in service (gas alarm, failure, repair, outage, etc.), and comparing it to the total number of DGA analyses on this type of equipment and at this gas concentration level.

A large number of analyses is necessary to get reliable values of failure probability. Knowledge of these values, however, is useful when choosing the normality percentage most appropriate for a given network and type of equipment (see 8.2.3).

8.2 Typical concentration values
8.2.1 General
Typical concentration values are the acceptable gas quantities below which field experience shows no detectable or possible incipient fault, and which are overpassed by only an arbitrarily low percentage of higher gas contents, for example 10 %. Typical concentration values will be referred to in such an example as the 90 % typical values.

However, typical concentration values are preferably to be considered as initial guidelines for decision making, when no other experience is available. They shall not be used to ascertain whether or not a fault exists within an equipment. They should be viewed as values above which the rate of gas formation may permit the detection of a probable fault.

Typical concentration values are affected by a number of factors, chiefly the operating time since commissioning, the type of equipment and the nature of the fault (electrical or thermal). For power transformers, the type of oil protection, load factor and operation mode are other influencing factors.

Typical concentration values may be calculated as follows and should be obtained by the equipment users on the specific types of equipment.

8.2.2 Calculation methods
The simplest method of calculation consists in gathering all the DGA results concerning a specific type of equipment. For each characteristic gas considered, the cumulative number of DGA analyses where the gas concentration is below a given value is calculated, then plotted as a function of gas concentration. Using the plotted curve, the gas concentration corresponding to a given percentage of the total cumulative number of analyses (for instance 90 %) is the 90 % typical concentration value for that gas and type of equipment.

8.2.3 Choice of normality percentages
If the normality percentage chosen (e.g. 90 %, 95 % or other) is too low, suspicion will be placed on too many pieces of equipment, with a loss of credibility in the diagnosis and recommendations, and an increase in maintenance costs. If the normality percentage is chosen too high, failure may occur without advance warning, also involving considerable costs.

The choice of a normality percentage is often an educated guess, left to the experience of the user of similar equipment. A certain amount of leeway in the choice of a normality percentage is also provided by considering the probability of failure and the actual failure rate of the equipment in service. In the absence of such information or experience, users may choose conservative normality percentages such as 90 % as a rough screening value. If adequate databases are not available to calculate typical values, users may also adopt as a first step, values observed on other networks and indicated in the application notes. It is apparent from these values that, in general, typical values are lower in instrument transformers and bushings than in power transformers, especially those with a communicating OLTC.

8.3 Alarm concentration values
Alarm concentration values are those values of concentration above which the probability of an incident is sufficiently high to require urgent competent decisions and/or actions.

Alarm concentration values should be set by users, manufacturers or independent experts, based on previous experience with equipment with similar characteristics (voltage, type, manufacturer, loading practices, age, etc.).

8.4 Rates of gas increase
If there is no increase in gas concentration over the last analysis, chances are that the fault has disappeared (or that a very small one is still there, in the case of air-breathing equipment, compensated by diffusion losses through the conservator, see 6.1).

What should be considered as a typical or alarm rate of gas increase depends very much on equipment type and age, type of identified faults, air-breathing and load patterns, and volume of the insulation involved. Ultimately the rate of increase should be decided by the user of the equipment, the manufacturer and/or other experts.

An increase in gas concentrations of more than 10 % per month above typical concentration values is generally considered a prerequisite for pronouncing the fault as active, provided it is clear that the precision on DGA values is better than 10 % after one month. Much higher rates of gas increase, such as 50 % per week, and/or evolving towards faults of higher energy (e.g. D2 or T3), are generally considered very serious, especially if they exceed alarm concentration values. In the case of power transformers, typical rates of gas production in millilitres per day are also reported (see table A.3). Special attention should be given to cases where there is an acceleration in the rate of gas increase.

NOTE In the case of carbon oxides, rates of gas increase are dependent on the oil-to-paper ratio, which may be quite different depending on the equipment considered.

On-line gas detectors, including those already available for hydrogen, may be particularly well-suited for detecting non-typical rates of gas increase occurring within minutes, hours or weeks, which is generally not possible with routine oil samplings done at monthly or yearly intervals. However, when such increases are detected by a gas detector, complete DGA analyses shall be made at the laboratory to confirm the detector readings, evaluate the rates of increase of the various gases and identify the fault.