DISSOLVED GAS ANALYSIS CONDITION MONITORING OF VACUUM TYPE TAP-CHANGERS FOR POWER TRANSFORMERS

DISSOLVED GAS ANALYSIS CONDITION MONITORING OF VACUUM TYPE TAP-CHANGERS FOR POWER TRANSFORMERS Rainer Frotscher, Dr. Dazhong Shen - Maschinenfabrik Reinhausen GmbH Falkensteinstraße. 8, 93059 Regensburg, Germany, r.frotscher@reinhausen.com Tim Farrell - Reinhausen Asia-Pacific Sdn. B ABSTRACT Vacuum type on-load tap-changers have been designed to perform up to 300 000 or 600 000 operations without maintenance. For the majority of network applications, this enfolds time spans of thirty years and more which can represent the whole transformer life. Adequate monitoring techniques, such as Dissolved Gas Analysis (DGA), are an appropriate tool to ensure maximum operational availability. While DGA is an established method to be used on transformers, the application of DGA to on-load tap-changers (OLTCs) turns out to be more complex, due to the big variety of different tap-changer models. While DGA on oil-switching type OLTCs does not seem to be an applicable diagnostic method, it seems to be appropriate for vacuum type OLTCs. If OLTC DGA shall be successful, a deeper understanding of gas sources and their typical gas patterns is necessary. Superposition of the gas patterns lead to model-specific DGA fingerprints which strongly vary with the operational data. As the absolute values of dissolved gases are very low, they can also be masked by stray-gassing effects. Established interpretation methods like quotient methods or the Duval tap-changer triangle may not always be applicable. Results of a field study and from power switching tests on various VACUTAP models enabled the definition of typical gas increasing rates for network service. As high amounts of acetylene (C 2 H 2 ) usually indicate unusual behaviour of vacuum type OLTCs, C 2 H 2 is the most important gas to be monitored. Based on actual field data, typical gas increasing rates for C 2 H 2 (in ppm per 1000 operations) have been defined for the different VACUTAP models. In the same way, a general caution limit for acetylene could be set. Online sensors for two to four gases can be used to perform a meaningful trend analysis. Trend data must be correlated with operational data and evaluated by statistical methods to enable an intelligent DGA condition monitoring. These methods can be implemented as software in modern monitoring systems like TAPGUARD. KEYWORDS: Vacuum type on-load tap changer, OLTC, DGA, trend analysis, condition monitoring 1. INTRODUCTION Regulated power transformers have a regulating winding with multiple taps which are connected to a tap-changer to adjust the transformer ratio. On-load tap-changers (OLTCs) are used if the ratio shall be changed while the transformer is under load. These tap-changers consist of a tap selector which selects two adjacent taps of the regulating winding and a diverter switch which performs the load switching operation between the two selected taps without interrupting the load current. If both functions are combined in a single functional unit, the tap-changer is called selector switch. Diverter switches respectively selector switches feature an own oil compartment which separates the tap-changer oil from the transformer oil. In all modern tap-changers, these oil compartments are pressure tight and gas tight. OLTCs in traditional oil switching technology use contact systems which produce switching arcs during their switching operation and which deteriorate the oil in the diverter switch compartment. Regular maintenance is necessary to replace the oil, clean the oil compartment and replace worn contacts. Modern OLTCs use vacuum interrupters for making and breaking the load current. These vacuum interrupters fully encapsulate the switching arcs so that the diverter switch oil is no longer carbonized. This enables prolonged maintenance intervals up to 300 000 or 600 000 operations, depending on the respective OLTC model. For standard applications in power supply networks with 5 000 to 10 000 switching operations per year, this means that the tap-changer mechanism is free of maintenance for the whole transformer life of 30 to 60 years. It is a common understanding that, over such long periods, the reliability of any technical device will decrease. To compensate for that, monitoring techniques can be used which, if intelligent enough, allow the 1

sensible evaluation of operational parameters online. By doing so, incipient faults can be detected before they can cause a failure. 2. METHODS 2.1. Tap-Changer Condition Monitoring Many parameters can be measured on an OLTC. The most important parameters refer to the oil quality, more precisely: to dielectric strength (U d ) and moisture content (H 2 O), which must be maintained in a sufficiently good state. Appropriate limit values for U d and H 2 O for units in service are defined in IEC60422 respectively in the operating manual of the tap-changer. As moisture content has a high impact on dielectric strength, moisture eventually is the determining parameter for a good oil quality. Even if the tap-changer is free of maintenance, the oil quality of the tap-changer oil must be regularly controlled in the same way as it is commonly done for the transformer. It is important to check the dehumidifying granulate of the oil conservator regularly or to use an automatic system like MTraB with self-drying cartridge. Ensuring a sufficiently good oil quality is essential to achieve high operational reliability of the tap-changer. Aside from oil quality, other parameters can be controlled either to protect the tap-changer from running into inadmissible states which could cause a fault (excessive heat, mechanical malfunction) or to gain additional information on the actual stress (switching count, load current). Advanced diagnostic techniques like innovative vibro-acoustic analysis [1] or Dissolved Gas Analysis (DGA) use intelligent algorithms for evaluation. Table 1 gives an overview on parameters which can be sensibly measured. Table 1: Parameters for tap-changer condition monitoring Parameter Method Value Tap-changer position Switching supervisory control Torque Vacuum interrupter monitoring system Oil temperature Switching count Load current through OLTC Dielectric strength (U d ) Moisture content (H 2 O) Switching sequence (vibro-acoustic) Switching sequence (DRM) DGA online offline offline / online offline / online offline offline / online Supervision of limit values, inadmissible operating conditions, information on actual stress Oil quality (obligatory) Advanced diagnostics, evaluation by intelligent algorithms DGA is a powerful diagnostic tool which has been established for many years to be applied on transformers, bushings and instrument transformers. DGA for tap-changers is far more complex, due to the high diversity of tap-changer models and additional influence factors. In the following, the usability of DGA for tap-changers will be discussed. 2.2. DGA for oil switching OLTCs In oil switching OLTCs, the switching arcs deteriorate the oil in the tap-changer compartment in manifold ways and produce huge amounts of gases, which can be dissolved in the oil or occur as free gas. Depending on the gas saturation level of the oil, the gas bubbles can fully or partially dissolve in the oil or escape via the breather. Switching arcs produce a mixture of arcing and heating gases, their amount and composition are determined by the arc intensity and arc duration as well as by the oil brand and grade of oil carbonization. It has been found that, depending on the grade of oil carbonization, the typical arcing gas pattern (mainly H 2, C 2 H 2, some C 2 H 4, CH 4 and other heating gases), often changes along the tap-changer s working life to produce mainly H 2 and C 2 H 4, but much less C 2 H 2. The mechanism behind this phenomenon is not fully understood yet, but there is evidence that the switching arcs are more expanded in carbonized oil, possibly due to the high amount of carbon particles which lead to a fostered inflammability of the oil. Additionally, heating gases are produced by the transition resistors, which can reach peak surface temperatures between 20 and 450 C during normal operation, depending on their design (which is specific to the respective application) and load conditions. As a result, the gas compositions in an oil switching OLTC turn out to be highly individual. DGA values and patterns may differ in almost any order and all patterns are principally normal. So, general rules for DGA interpretation of oil switching OLTCs are not 2

existent. Even a trend analysis will be only of limited use, because the tap-changer oil is usually exchanged with every scheduled maintenance. Therefore MR does not consider DGA to be an applicable diagnostic method for oil switching OLTCs (OILTAP series). 2.3. DGA for vacuum type OLTCs By encapsulating the switching arcs inside vacuum interrupters, the conditions for the tap-changer oil are comparable to the conditions for the transformer oil. Under normal operating conditions (including overload), no free gases are produced, and the ppm values of hydrocarbon gases dissolved in the oil are in many cases in the 1- or low 2-digit range. Over the working life of the tap-changer, they turn out to be fairly constant, which enables a proper evaluation and trend analysis. Periodic changes, e.g. caused by daily temperature cycles, can be compensated by integrating the measured oil temperature in the evaluation algorithms. 2.3.1. Gas Sources For a full understanding of the development of gases inside a vacuum type OLTC, it is advisable to first define the gas generating components and the determining parameters. a) The vacuum interrupters are sealed systems, which do not release or produce any gases in the surrounding oil. In case of an unlikely leakage, huge amounts of arcing gases will occur which indicate the fault. b) In some OLTC types, the main contact (MC) respectively by-pass contact commutates the load current from the continuous current path onto the main path, causing it to flow through the vacuum interrupter; see Fig. 1, positions 1 and 2 as an example. Because inductance and resistance of the main path are higher than in the continuous current path, some sparks or low-energy arcs are generated at the MC. c) The transition resistors are heated when the load current flows through the transition path; see Fig. 1, positions 3-5 as an example. For operation under nominal load, the design value has been set not to exceed a temperature rise of 190K per operation. With the oil temperature added, the maximum surface temperature will be below 300 C, which causes a certain heat gas pattern. Operations under overload, or multiple consecutive operations as they can occur in service, will cause higher temperatures, but because they are rare, they will not change the typical gas pattern significantly. On the other hand, when switching operations under no load or partial load are performed, the resistors heat up very moderately and so will produce none or only negligible amounts of hydrocarbon gases. At surface temperatures of less than 100 C, only CO and CO 2 are produced due to incipient oil ageing which can be used as indicator for the thermal stress of the oil, caused by the transition resistors [2]. The use of CO and CO 2 as thermal markers is possible here because tap-changers only contain negligible amounts of cellulose, whose degradation usually causes CO and CO 2. d) Transition reactances, as they are used in combination with reactor type OLTCs, show very low losses; therefore, they don t heat up significantly and so produce only negligible amounts of gases. Furthermore, they are located in the transformer tank and so cannot influence tap-changer DGA. 3

Fig. 1: switching sequence of VACUTAP VR e) The tap selector can be equipped with a change-over selector which doubles the regulating range of the transformer, either by reversing the regulating winding or by adding a coarse tap winding. When the change-over selector is operated, the potential of the regulating winding is only determined by the capacitances to the neighboured winding(s) and/or the transformer core or tank; see Fig. 2. This condition can cause significant potential drift of the regulating winding. Even if the change-over selector does not have to switch the load current, it must be able to break capacitive currents up to 500mA and withstand recovery voltages on the open contacts up to 50kV. It is obvious that such stress will cause switching sparks and arcs in the transformer oil, which can be visible in the transformer DGA in case the number of change-over selector operations is extraordinarily high. In the majority of cases, switching of the change-over selector will not disturb transformer DGA. It also does not influence tap-changer DGA. The only exception is the compartment type OLTC VACUTAP RMV-II, which houses the change-over selector inside the tap-changer Fig. 2: Capacitive coupling of compartment. regulating winding during operation of change-over selector (reversing switch) 2.3.2. Gas Patterns The typical gas patterns for each gas source are illustrated in Table 2. The patterns originate from various laboratory and field test data and have been aligned with literature [3], [4]. As the intensity of commutation sparks, arcing of change-over selector and resistor heating strongly depend on the individual application, the absolute ppm amount of each gas and the composition of gases with them will differ significantly. 4

Table 2: Typical gas patterns of components of vacuum type OLTCs Source Gas Pattern Determining Factors commutation contacts, by-pass contacts sparking / low energy arcing Inductance and resistance of main current path load current H 2 C 2 H 2 CH 4 C 2 H 6 C 2 H 4 CO transition resistors heating <300 C design values of step voltage and through-current (application-specific) actual load, load profile oil temperature H 2 C 2 H 2 CH 4 C 2 H 6 C 2 H 4 CO change-over selector low energy arcing / sparking H 2 C 2 H 2 CH 4 C 2 H 6 C 2 H 4 CO capacitances of regulating winding to neighboured windings resp. core/tank voltage of regulating winding design (type-specific) operating speed 2.3.3. Additional influence on gas patterns Because the ppm gas values are very low during normal service of any VACUTAP OLTC, the gas patterns may be masked by stray-gassing processes, which is a self-gassing activity of strongly hydro-treated oils, preferably when they are in contact with certain metals [5]. It has been found by tests that the material which is used for the transition resistors stimulates the generation of significant amounts of H 2, but also produces CH 4, C 2 H 6 and C 3 H 8 in low amounts. Steel and copper also promote stray-gassing activity. With this, the gas pattern caused by the commutation processes (sparking) may be tampered. Depending on the applied oil brand, stray-gassing can suggest more commutation activity than actually present. 3. RESULTS 3.1. Gas patterns of VACUTAP OLTCs Inside the tap-changer oil compartment, the gas patterns superimpose, depending on the gas sources present. Concerning VACUTAP VV and VR, this will lead to a mixture of sparking and heating gases. For VACUTAP VM, only heating gases will be present, because there is no commutation contact. In contrast, the compartment type OLTC VACUTAP RMV-II will only show sparking/arcing gases, due to current commutation and operation of the change-over selector in the OLTC tank. Because this OLTC type has a much higher oil volume than VACUTAP VR, VV or VM, the absolute values will be lower. Additional H 2 and CH 4 may be introduced by stray-gassing effects. As a result, the typical gas patterns of oil-filled VACUTAP OLTCs can look like this (Figure 3): H 2 C 2 H 2 CH 4 C 2 H 6 C 2 H 4 CO H 2 C 2 H 2 CH 4 C 2 H 6 C 2 H 4 CO H 2 C 2 H 2 CH 4 C 2 H 6 C 2 H 4 CO H 2 C 2 H 2 CH 4 C 2 H 6 C 2 H 4 CO VACUTAP VR VACUTAP VV VACUTAP VM VACUTAP RMV Fig. 3: Typical gas patterns of VACUTAP OLTCs (qualitative) It can be seen that the usual ranges of gas levels will lead to significant variations of the gas patterns observed in practice. This is mainly due to the low absolute levels, which are influenced by the specific OLTC configuration 5

(single-phase/three-phase, U m, design of transition resistors) the operating mode (transformer load, total number of OLTC operations per day, switching profile), the breathing conditions (equilibrium between gas generation and gas loss via the breather) and also the oil brand. Principally, each application shows its individual fingerprint. 3.2. Gas generation rates of VACUTAP OLTCs Power switching tests with all oil-filled OLTC types of the VACUTAP series (VV, VM, VR, RMV-II) have been performed to determine the gassing behaviour under different load conditions (partial load, nominal load, overload). 30 000 to 60 000 switching operations have been performed within a few weeks. The development of gases has been recorded with different online-dga monitoring systems (for hydrogen, composite gas and four gases) and was supported by regular laboratory analyses. The three online sensors have been mounted on a special test stand which provided the same amount of oil to stream along each sensor. In a closed circuit, a special pump with minimized turbulences transported the oil out of the OLTC oil compartment to the sensors and back. A sealed expansion tank with rubber bag was used to compensate for the thermal expansion of the oil without gas losses. Due to the extreme time-squeezing the amount of dissolved gases developed on a level which could reliably be measured by the online sensors. Free gases did not occur. From the recorded data, gas Table 3: Typical gas generation rates [ppm/1000 operations] for VACUTAP OLTCs under nominal load VV VM VR RMV-II H 2 0.4 2.1 3 0.8 CO 1.2 2.1 0.3 0.3 C 2 H 2 <0.1 n.d. 0.3 <0.1 C 2 H 4 <0.1 0.2 0.3 <0.1 generation rates (ppm per 1000 operations) could be derived; see Table 3. H 2, C 2 H 2 and C 2 H 4 can be attributed to the commutation activity, but H 2 could (partially) also have been generated by stray-gassing processes. As described in section 2.3.1., CO can be used as indicator for incipient thermal oil ageing, caused by the heat of the transition resistors. Unlisted gases in Table 3 were not reliably detectable. During normal network operation, the actual gas generation rates will be lower, because a) gas loss will occur via the breathing system (if not sealed) and b) under partial load, commutation processes will be weaker and resistor heating will be lower than in the tests under nominal load. 3.3. Field study with VACUTAP VM The VACUTAP VM is the only OLTC type which does not generate C 2 H 2 amounts which can be reliably measured (>10x detection limit), because of the missing main contact. In this OLTC type, the load current flows continuously through the vacuum interrupter in the main path. In a two years field study, the free-breathing OLTC performed approximately 4000 operations. The average transformer oil temperature was 27 C, indicating a very moderate average load. After two years in service, low amounts of H 2, CH 4 and CO could be measured, with H 2 and CH 4 near the limit of quantification. The CO content was significantly higher which gives an indication on light thermal oil ageing, caused by the transition resistor heating. Following Duval Triangle No. 4, H 2 and CH 4 can be attributed to stray-gassing [6]; see Fig. 4. Not determined Stray-Gassing of mineral oil Overheating (T<250 C) Corona / Partial Discharge OLTC DGA Hot spots with carbonization of paper (T>300 C) Fig. 4: Duval Triangle No. 4 for stray-gassing and low-temperature faults 4. CONCLUSIONS / RECOMMENDATION 4.1. Practical experience with DGA More than 40 000 VACUTAP OLTCs are operating in service worldwide, many of them trouble-free since more than 15 years. Nevertheless, the manufacturer s data base of VACUTAP DGA data is very sparse, due to the very limited access to field data. Usually, the tap-changer manufacturer is contacted only in isolated cases if DGA values look abnormal and then is asked to give a statement. In the majority of cases, the values don t reveal any irregular behavior but in some cases, incipient faults or single exceptional events have been detected and measures have been initiated in time to prevent the equipment running into a failure. Besides that, there are 6

also cases with exceptional gassing behavior without allocation to a definite root cause. The reasons for such behavior can be manifold: a) data may be inconsistent or erroneous due to sampling or analyzing errors b) only a single DGA without any reference to the history is available c) important operational parameters are unknown, such as temperature of oil sample / oil brand name total no. of OLTC operations; average number of operations per day actual load of the transformer; information on recent load changes application data (network service, industrial application, ) date of last OLTC maintenance OLTC serial no. d) relevant information on specific features of the application is missing, like information on additional equipment like oil filter units, coolers or heaters which can influence the gassing behavior breathing situation (sealed / free-breathing) e) unusual wave forms of load current or voltage f) exceptional combination of part tolerances g) pollution of the oil with gases from industrial environment In such cases, the DGA values must be accepted as they are and suggest a more intensive observation by monitoring. 4.2. Typical C 2 H 2 values for VACUTAP OLTCs Abnormally high C 2 H 2 values inside a vacuum type OLTC represent distinct arcing activity. More indicative is the increasing rate of C 2 H 2 since the last sample and the number of operations under load between two samples. The gas may have developed slowly due to a beginning malfunction or may also be the result of a single event, e.g. a non-destroying flashover with low energy. With this, C 2 H 2 is the most significant gas to be monitored. Field experience has shown that, in case the absolute value of dissolved C 2 H 2 exceeds 50 ppm, an inspection of the OLTC is recommendable. As shown in section 3.2, the typical gas generation rates per 1000 operations for network service are very low. Nevertheless, higher values may occur, due to exceptional load conditions, an increased number of consecutive operations within a short period or for other reasons. Anyhow, the gas generation in case of a malfunction will be much higher. With this, it is possible to define limit values for normal service conditions which are clearly different from fault cases. Covering exceptional cases of gassing without indication of a fault, the following maximum gas generation rates for C 2 H 2 are defined for normal operation in network service: Table 4: Maximum C 2 H 2 gas generation rates for VACUTAP OLTCs in network service ppm C 2 H 2 / 1000 operations VACUTAP VV <2 VACUTAP VR <4 VACUTAP VM 0 VACUTAP RMV-II <3 These values cover also extreme cases observed in service without indication of a fault. They should be used as guiding values which may occur between two consecutive DGAs. They may not be used to define limit values for C 2 H 2 for a defined number of switching operations. 4.3. Trend analysis While a single DGA is only meaningful when showing extreme values, trend analysis also allows the evaluation of low ppm values or low gas increasing rates. Intelligent algorithms can suppress stray-gassing effects and can detect periodic variations due to daily load changes or other influences. To be successful, DGA data must be correlated with corresponding operational parameters, such as actual load, oil temperature and time stamps of OLTC operations. Determination of relevant correlation factors or equations (linear or non-linear) may be complicated, but for identifying correlations, statistical methods (data-mining techniques) are available which can be applied on reference data bases containing faulty and non-non-faulty equipment. The correlation parameters can be tuned then on the specific application by learning algorithms. These functions are the elemental part of intelligent interpretation algorithms which are integrated as software in suitable online monitoring systems, such as TAPGUARD. As a result of such intelligent evaluation, caution and warning limits are optimized for the specific application and so can be individually set. More general limit values which 7

are valid for a specific OLTC population or model family are also possible by combining data from several installations. The greater the data base, the more stable are correlations and the more reliable are the results. With increasing experience, limit values and interpretation results can be adjusted and refined. In this context, it is not necessary to monitor and evaluate all 11 gases which are commonly used for DGA (9 without C 3 gases). It is absolutely sufficient to trace the development of 2 to 4 selected gases over time. For example, the MSENSE x2.5 online gas monitor can trace H 2 and CO with high sensitivity and accuracy. While sudden changes of H 2 can be related to unusual sparking or arcing activity, a dynamic increase in CO can represent thermal irregularities in an early state, with or without cellulose involved. If an irregular behavior is detected, it is advisable to draw an oil sample and perform a laboratory analysis for 11 (9) gases, also including determination of the oil quality (U d, H 2 O). Depending on the result, the following measures might be adequate: immediate reaction, such as shutdown of the unit scheduled maintenance gathering additional information by applying other diagnostic methods, such as Dynamic Resistance Measurement (DRM) or vibro-acoustic measurement increased focus on the application, repeated laboratory analysis in shorter intervals If there is a fault, then it will worsen, which means that specific gas values will increase long-term, or trend analysis will show a gas increasing rate which is unacceptably high. In the latter case, immediate reaction will be adequate as the fault will develop quickly and cause a failure. By applying such methods, the high reliability of maintenance-free vacuum type tap-changers can be maintained over the whole lifetime. 5. REFERENCES [1] M. Foata, R. Beauchemin, K. Viereck, A. Saveliev, H. Hochmuth, MR: New Vibro-Acoustic Tap-Changer Diagnostic Method: First results and Practical Experience, paper A2-104, CIGRE Session Paris, 2016 [2] I. Höhlein-Atanasova, SIEMENS, R. Frotscher, MR: Carbon Oxides in the Interpretation of Dissolved Gas Analysis in Transformers and Tap-Changers, IEEE Electrical Insulation Magazine, Nov/Dec 2010, Vol. 26, No. 6, pp. 22-26 [3] D. Dohnal, MR, Classification of on-load tap-changers (LTCs) with regard to dissolved gas analysis (DGA), Tutorial Session, IEEE Transformers Committee Meeting, Vancouver, Spring 2002. [4] M. Shirai, S. Shimoji, T. Ishii, Mitsubishi Electric Corp., Thermodynamic Study on the Thermal Decomposition of Insulating Oil, IEEE Transactions on Electrical Insulation, Vol EI-12 No 4, Aug 1977. [5] I. Höhlein-Atanasova, SIEMENS: Unusual Cases of Gassing in Transformers in Service, IEEE Insulation Magazine, Vol. 22, No. 1, January/February 2006 [6] M. Duval: The Duval Triangle for Load Tap Changers, Non-Mineral Oils and Low Temperature Faults in Transformers, IEEE Electrical Insulation Magazine, Nov/Dec 2008, Vol. 24, No. 6, pp. 22-29 8