Usage of nanotechnology based gas sensor for health assessment and maintenance of transformers by DGA method

Electrical Power and Energy Systems 45 (2013) 137–141

Contents lists available at SciVerse ScienceDirect

Electrical Power and Energy Systems journal homepage: www.elsevier.com/locate/ijepes

Usage of nanotechnology based gas sensor for health assessment and maintenance of transformers by DGA method Anjali Chatterjee a,⇑, Partha Bhattacharjee a, N.K. Roy b, P. Kumbhakar b a b

Central Mechanical Engineering Research Institute, Mahatma Gandhi Avenue, Durgapur 713 209, India National Institute of Technology, Durgapur 713 209, India

a r t i c l e

i n f o

Article history: Received 20 January 2012 Received in revised form 14 August 2012 Accepted 19 August 2012 Available online 6 October 2012 Keywords: Analyzer Dissolved gas analysis Fault diagnosis Gas sensor Nanotechnology Optimum temperature

a b s t r a c t Present day power system is essentially a complex mesh of various important components with power transformer as one of the key elements. For the reliability of power supply, a robust maintenance tool for power-transformer is highly essential. To cater for this demand a portable, online diagnostic device is developed which can record the temperature and quantify the concentration of some of the dissolved gases in transformer oil with the help of a non-invasive sensor fabricated by nanotechnology. After conditioning the signals, the data are transmitted to the nearest substation for storage in computer. For the purpose of analysis and health assessment of the transformers from a remote place, the computer is made accessible through network. From the data of five gases, different faults, if they are occurring inside the transformer, can be predicted. The fault diagnosis is performed by dissolved gas analysis (DGA), which is one of the proven methods widely used during the last two decades. Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction In the present day industrial world no compromise can be made on the quality and reliability of uninterrupted power supply. Power transformers are one of the major capital investments in a power system network. Appropriate maintenance, including insulation reconditioning and timely filtration, can extend the life of a transformer to more than even 60 years. In the existing system, maintenance personnel periodically take transformers and circuit breakers offline, in order to check the operating conditions of the equipments. With this method, still there are catastrophic failures, not to mention about the fruitless maintenance. Researchers like Bernadic and Leonowicz have introduced practical approach to power system fault location in power network [20], which could be helpful to locate a faulty transformer. A growing need for lower cost and more accurate diagnostic tools, have led to an introduction of online monitoring system based on artificial intelligence and various analytical techniques in maintenance of electrical power substation. One of the most informative methods for the detection of fault gases is the dissolved gas analysis (DGA) technique [1,2]. Souahla et al. have suggested a new multilayer perception neutral network for decision making in DGA [21]. In this method a sample of the oil is taken from the unit and the dissolved gases are extracted. Then the extracted gases are separated, identi⇑ Corresponding author. E-mail address: [email protected] (A. Chatterjee). 0142-0615/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijepes.2012.08.044

fied, and quantitatively determined. At present this entire technique is best done in the laboratory since it requires precision operations. Since this method uses an oil sample it is applicable to all type units and like the gas blanket method it detects all the individual components. The main advantage of the DGA technique is that it detects the gases in the oil phase giving the earliest possible detection of an incipient fault. This advantage alone outweighs any disadvantages of this technique. Traditionally, DGA has been carried out by taking a sample of oil from the transformer, sending it to a laboratory and waiting for the results from the gas chromatograph [3]. Unless there was any suspicion of a problem in the transformer, samples might be taken at intervals of up to one year, depending on the maintenance regime of the operator. This means that a fault that develops over a shorter period of time than the sampling interval can be missed, leading to possible catastrophic failure of the transformer. Manual sampling can also lead to errors in the analysis. For example, bad sampling techniques can introduce contaminants into the oil, and inappropriate storage can mean loss of gas from the oil between the time interval taking the sample and analysis in the laboratory. Results can vary from laboratory to laboratory, and even between users of the same equipment. A Round Robin test co-ordinated by Michel Duval for Cigré showed spreads of several hundreds of percent between identical samples tested in different laboratories using various techniques. This can lead to uncertainty in diagnosing the results, especially when looking at trending of gas concentrations. It has been observed that if service provided to the transformer is scheduled

138

A. Chatterjee et al. / Electrical Power and Energy Systems 45 (2013) 137–141

properly, it extends the life of the equipment [22]. Till now experiences with laboratory analysis of dissolved gases (DGA) has been considered as a baseline. However the dynamic behavior of dissolved gases requires continuity and trending, which are unlikely to be captured through periodic manual sampling. To overcome these short comings a gas analyzer is developed to sample the oil continuously in 247 modes in the present work. The main component of the gas analyzer is the gas sensor developed with the help of nanotechnology. Semiconductor oxides like SnO2 [4], Ga203 [5], WO3 [6], TiO2 [7] were reported for detection of different hydrocarbon gases but at a quite high sensing temperature (150– 450 °C). However nanoparticles help to sense the gases at low temperature and with higher sensitivity. In this research work effort has been given to design and develop a gas sensor from the thin film of doped zinc oxide (ZnO). It is inexpensive, relatively abundant, chemically stable, easy to prepare and non-toxic. For these properties it finds its usage in a large number of areas. Nunes et al. [8], published a report on ZnO sensor for methane detection at low temperature (100–200 °C) but no detail study on dynamic response is reported. In work we have synthesized ZnO by chemical precipitation method [9]. To increase the sensor sensitivity pure ZnO was doped by various concentration of manganese (0.5%, 1.5% and 2.5% by weight). There are a few commercially available techniques currently used for online detection of dissolved gases in the transformer oil. Different technology based methods are applied in management of power system [23]. Most of the techniques are based on the measurement of a particular gas concentration in a sample of gas extracted from the oil, mainly by using gas chromatograph. These devices are costly and require offline operations (calibration, maintenance, etc.), and the need of sample preparing. In the case of other transducers which are based on infrared technology, there are others inconveniences, such as the limited range of measurement in the low ppm region. Broadly, the three main families of gas sensors are spectroscopic, optical and solid state. Spectroscopic systems are those based on the direct analysis of fundamental gas properties, such as molecular mass or vibration spectrum; optical sensor systems are based on the measurement of the absorption spectra after light stimulation; and solid-state ones are based on the change of physical and/or chemical properties of a sensing material after gas exposure. These changes depend on the gas sensing material and usually involve changes in its electrical properties. In the past Zylka and Mazurek [10] has developed a portable analyzer, fitted with more than one electrochemical gas sensors, for online monitoring of the transformer. In his prototype he used four sensors to detect four types of gases dissolved in the oil. Benounis et al. [11] developed an optical fiber sensor to follow the aging of the transformer oil. An effort has been made by the authors to develop a prototype for online monitoring of the transformer in their research work [24], this work further deal with the fault diagnosing techniques by detecting more number of gases. Numerous researchers have done fault prediction of transformers with the help of DGA and majority of them deals with the fault classification based on the results of gas chromatograph [12]. Novelty of this work lies in the development of a prototype gas analyzer for monitoring the health of the transformer. The gas analyzer uses thin film sensor to detected five dissolved gases in the oil and thus helped in diagnosing major faults occurring inside the transformer. The characteristic of the sensor is that it can sense five gases at different temperature.

2. Methodology of dga The volume of the system has to be considered when talking about rates of gas evolution. The gases are reported in terms of

concentration in parts per million (ppm) and the total gas generated by a fault will be dependent on the total volume of oil the system. To determine rates of gas generation it is necessary to collect samples at different times. Normal aging of the insulating oil will give rise to a slow accumulation of gases over a semiannual sampling period. A moderate accumulation of gases over a monthly interval can indicate an incipient fault, while a rapid accumulation (i.e. over 10% per month) of gases is indication of an active fault. The transformer oil contains number of dissolved hydrocarbon gases; their concentration gives us an indication of the health of the equipment [17]. The fundamental chemical reactions are involved in the breaking of carbon–hydrogen and carbon–carbon bonds. Analysis of the quantity of each of the fault gases present in the fluid allows identification of fault processes such as corona, sparking, overheating and arcing. Table 1 indicates the upper threshold limit of concentration for the dissolved gases in the transformer oil developed at California State University Sacramento in co-operation with Pacific Gas and Electric Company [19]. Table 2 gives the average gas composition present in the transformer oil in volumetric proportion. From Tables 1 and 2 we can infer that H2, CH4, C2H6, C2H2 and CO gas constitute nearly 97% of the total dissolved combustible gases (ppm content of six gases namely hydrogen, methane, ethane, ethylene, acetylene and carbon monoxide constitute TDCG). Increase in the content of these gases gives an indication of a fault occurring inside the transformer. According to the properties of ZnO, it is highly selective towards hydrogen, methane, ethane, acetylene and carbon monoxide at different optimum temperature but they also demonstrate fairly significant cross-sensitivities. Though transformer oil has number of dissolve gases which can be sensed by ZnO, we have studied five main combustible gases namely hydrogen [7], methane [13], ethane, acetylene and carbon monoxide [14,15]. 3. Experimental setup Fig. 1 shows the schematic representation of the placement of the gas analyzer near the transformer. The gas extraction unit along with the sensor is mounted inside a pipe, which is fitted to the oil drain valve at the bottom of the transformer oil tank. The arrangement is such that the sensor is protected against the environmental hazards. Two solenoid valves that are fitted at two ends of a pipe, controls the sampling time of the oil. This is done by adjusting the opening and closing timings of the valves. For some time the oil is trapped inside the pipe, during that time period all the five gases are detected and quantified. With this one cycle of operation is over, after that the outlet valve opens, the tested oil is pumped back into the transformer tank. The sampling time can be adjusted by controlling the timings of the valve operation. Fig. 2 shows a prototype setup for testing the fabricated sensor, which is an important part of the gas analyzer. Initially test runs were done with high purity (99.9%) hydrogen, methane and carbon monoxide of known concentration. Thin film of manganese doped zinc oxide material, synthesized by chemical precipitation method acts as a gas sensing material. This thin film is coated on a glass

Table 1 Shows the guidelines developed at California State University Sacramento to indicate the level of the gases. Gas

Normal (ppm)

Abnormal (ppm)

Hydrogen Methane Ethane Ethylene Acetylene Carbon monoxide

150 25 10 20 15 500

1000 80 35 150 70 1000

A. Chatterjee et al. / Electrical Power and Energy Systems 45 (2013) 137–141

139

Table 2 An average gas composition in volumetric proportion. H2 (%)

CH4 (%)

C2H2 (%)

C2H4 and others (%)

70

10

15

5

Fig. 3. Variation in sensitivity of the sensor, with respect to temperatures for C2H6, CO, H2, CH4 and C2H2.

Fig. 1. Schematic diagram for gas analyzer and arrangement of the sensor placement.

substrate; it is an electrical insulator but good thermal conductor. A heater coil is placed below the glass substrate, as it is required to raise the temperature of the thin film so that the target gases are sensed at their optimum temperature and the control of the coil is by PWM [16]. In case of glass substrate, heat transfer from heater to sensing material through substrate is usually high. The ohmic contact with the nanoparticle of the thin film was established by using silver conductive paste and the connections were made by wires having very low resistance. During the initial experiments high resolution multimeter which works on 2  4 wire method is used to measure the sensor resistance at different temperature. Experimentations were performed to detect the temperature where the sensor exhibited highest sensitivity to individual gases and study their cross sensitivity effect. Numbers of readings were taken for sensor resistance response at different temperature in air and in the gaseous ambience keeping the gas concentration

fixed. The concentration of gas was controlled, by adjusting the gas flow through the flow meter for a fixed duration of time. This procedure helped to detect the optimum temperature where ZnO exhibited maximum sensitivity to individual gases. Sensitivity is the ratio of reduction of sensor resistance in the gas by its resistance in air. Inference can be drawn that thin film of ZnO shows maximum sensitivity towards ethane at 100 °C, carbon monoxide at 172 °C, hydrogen at 190 °C, methane at 209 °C and acetylene at 220 °C (shown in Fig. 3). With the sensor data one can now quantify the dissolved gases in the transformer oil. In the test chamber about 200 ml of transformer oil is taken inside a flask, a test tube on which thin film of doped ZnO is coated on the outer surface, is inserted from the top. For the extraction of gas from the oil, thermal degassing method is adopted. The flask is placed on a special heating platform, which simultaneously performs heating and stirring action on the transformer oil. The electrical parameters of the sensor material are measured continuously but recorded in the microcontroller only at the optimum temperature of the sensor material for a particular target gas. Besides the sensor, the internal circuitry of the analyzer as shown in Fig. 1, consists of a microcontroller for processing the

Fig. 2. Experimental setup for thermal gas extraction.

140

A. Chatterjee et al. / Electrical Power and Energy Systems 45 (2013) 137–141

Fig. 4. Shows variation of sensitivity with concentration CO.

Table 3 Cross sensitivity exhibited by different gases at their optimum temperature. Gas

°C

Individual response (mV)

Combined response (mV)

Cross-sensitive (%)

C2H6 CO H2 CH4 C2H2

100 172 190 209 220

110 164 168 371 396

132 212 313 437 450

18 29 86 18 15

sensor data and displaying it on LCD display. An analog to digital convertor is used to convert the analog signal from the thermocouple and resistance drop across the thin film to digital form, which is feed to the microcontroller. The microcontroller does the job of a linearizer and regulates the heater temperature by PWM control [11]. The resistance drop at optimum temperatures for all the five gases are recorded in the microcontroller. These data are compared with the ppm values mentioned in Table 1 for the normal condition for individual gases. If the concentration of the gases goes beyond the threshold value, indication is given through red LED and a buzzer. The different gas concentration values are stored and processed in the microcontroller. Later they are periodically transferred to the control room using an RF transmitter working at 434 MHz frequency. Modulation is based in short range wireless communication transmission. The receiving antenna is placed in the control room of the switchyard. After number of experimentations with known concentrations of hydrogen, methane, ethane, acetylene and carbon monoxide gases, the equations of sensitivity corresponding to different concentration of the gases, at its corresponding optimum temperature is derived from the curve. The curve in Fig. 4 shows the sensitivity of carbon monoxide for its corresponding concentration at 172 °C (optimum temperature). With the help of these equations for any value of sensor resistance, corresponding gas concentration present in parts per million (ppm) can be found out at any point of time. As the transformer oil contains the mixture of all the three target gases, hence their cross-sensitivity parameter plays an important role. To detect the cross sensitivity of the sensor, voltage response to individual gases of known quantity were recorded. Keeping the volume of the container and quantity of individual gases constant, the sensor response of the mixture is noted, shown in Table 3. While measuring the gases, its percentage of cross sensitivity is multiplied in the software to read the actual quantity of the gas. For calibrating the sensor two samples of transformer oil one from a faulty transformer and another from a new one were used. Output voltage recorded by the sensor for the two samples were compared with the gas chromatograph results obtained from an oil testing laboratory for the same samples. Later on investigation tests were run on the oil samples taken from the existing transformers which were also subjected to gas chromatography DGA examinations. In one test cycle as the tem-

Fig. 5. Voltage response of the sensor to individual gases in transformer oil at different temperature.

perature of the sensor increases, C2H6, CO, H2, CH4 and C2H2 are detected in a sequence one after the other as depicted in Fig. 5. After quantifying the acetylene gas, which is sensed at the highest temperature the heater voltage is deactivated and the cycle is repeated after a time interval. Auto start arrangement is provided by a timer linked relay circuit, which triggers at preset time intervals. The whole procedure is programmed in the microcontroller. The concentration of individual gas in parts per million (ppm) is displayed on a LCD panel in a cyclic fashion. The above mentioned analyzer has a capacity to store data of 2 months. 4. Theory and discussion 4.1. Sensor working Semiconductor metal oxide (SMO) gas sensors have become a prime technology in several domestic, commercial, and industrial gas sensing systems. Nanostructure of metal oxide films/pellets has a direct influence on the sensitivity, selectivity, and recovery and response time for a particular gas. In addition to the conductivity change of gas sensing material, the detection of the gases can be performed by measuring the change of capacitance, work function, mass, optical characteristics or reaction energy released by the gas/ solid interaction. The existing semiconductor gas sensors have response time around 20 s while the sensors manufactured from nano of metal oxide have response time in microseconds. Nanomaterials improve the 3R’s – reliability, reproducibility and robustness of the sensor due to improve surface area, increased functionality and amenability integration with existing sensor platforms. In our work we have used the resistance modulation function of the sensor of sense the gases. At low temperature there is no change in resistance, even if the concentration of the gas is increased. This shows the available energy is not enough to carry out the reactions on the surface. The temperature of the senor is increased from room temperature to 280 °C, but only a part within 100–270 °C is considered in operation because in this range reduction in resistance of the sensor is noticeable. The sensitivity of the sensors S was calculated using the following mathematical expression:

S ¼ Rg n Ra

ð1Þ

where Ra is the resistance of the sensor in the presence of air and Rg is the resistance in the presence of the test gas. The sensitivity of the sensor is influenced by many factors, including internal and external causes, such as natural properties of base materials, surface areas and microstructure of sensing layers, surface additives, temperature, humidity. Upto a certain limit the increase in the operation temperature leads to an enhancement of the films sensitivity till its peak value is reached, after that the sensitivity starts reducing (as shown in Fig. 3). At peak sensitivity, the temperature value

A. Chatterjee et al. / Electrical Power and Energy Systems 45 (2013) 137–141

Fig. 6. Drop in resistance of the sensor with rise of temperature for Mn doped and undoped ZnO thin film samples in air.

Table 4 Diagnosis of some of the faults by key gas method. Corona A. Oil B. Cellulose Pyrolysis oil (low temp) Arching

H2 CO, H2 CH4, H2 C2H2, H2

corresponds to the optimum temperature. Initially samples of doped and undoped ZnO were tested for its electrical characteristics; manganese doped thin film performed much better in gas sensing as shown in Fig. 6. Later doping concentration was varied from 0.5%, 1.5% and 2.5% by weight percentage. It was observed that with the increase of doping the target gases were sensed at much lower temperature. Hence we have selected thin film (area 19.97 cm2) of zinc oxide doped by 2.5% of manganese as our sensor. 4.2. Fault diagnosis By sensing the five major dissolved gases present in the transformer oil one can predict some of the faults. Even at the earliest stages of fault development, both thermal and electrical stresses on transformer oil will generate dissolved hydrogen. In addition, the highly mobile hydrogen molecules are quickly diffused throughout the oil, providing an early and reliable fault signature. Carbon monoxide is a by-product of cellulose degradation. Although generated in all transformers under normal operating conditions, anomalous amounts of carbon monoxide may be generated when localized cellulose overheating occurs. Presence of methane in large quantity in the oil is a sure indication for a fault occurring inside the transformer. When its increases along with hydrogen and ethane, even at low temperature indicate the decomposition of the oil as mentioned in Table 4. As the oil temperature rises due to fault, the formation of the degradation gases change from methane to ethane to ethylene. Here key gas method [18] is used for fault diagnosis. Increase in small quantity of acetylene points towards the possibility of arcing taking place inside the oil or inter-turns of the windings. Thus we can classify the type of fault by knowing the concentration of the individual gases in the oil and its generation rate. 5. Conclusion From the study performed it is possible to conclude that the most suitable ZnO thin film for application on gas sensors should be the ones with a low thickness and a high resistivity, in order to enhance the values of sensitivity, since the absorption/desorption phenomena occurs in the films surface. The operation temperature is an important parameter to control the sensitivity of the

141

sensor. In the case of acetylene, methane and hydrogen, the sensitivity of the sensor increase with the temperature and the highest values of sensitivity were obtained at 220 °C, 209 °C and 190 °C. For the test with ethane and carbon monoxide the highest value of sensitivity occurs with an operation temperature of 100 °C and 172 °C, which means that the operation temperature of the sensor should be chosen depending on the gas used. The thin film presents a linear variation of sensitivity with the gas concentration. However it has a low selectivity since it detects the presence of several gases. With the application of nanotechnology it was possible to miniaturize the gas sensor, which makes the DGA analyzer highly portable and cost effective. It should be noted that the aforementioned types of sensors and monitoring systems cost several thousand dollars in the commercial market; hence an effort is given in the present work to develop a reliable device at a much cheaper cost. References [1] Rogers RR. IEEE and IEC codes to interpret faults in transformers using gas in oil analysis. IEEE Trans Electr Insul 1978;13:349–54. [2] Fredi Jakob, Dissolved gas analysis – past, present and future, Weidmann – ACTI Inc. [3] Kelly Joseph J. Transformer fault diagnosis by dissolved gas an analysis. IEEE Trans Ind Appl 1980;IA-16(6). [4] Bose S, Chakraborty S, Ghosh BK, Das D, Sen A, Maiti HS. Methane sensitivity of Fe-doped SnO2 thick films. Sens Actuators B 2005;105:346–50. [5] Fleischer M, Meixner H. A selective CH4 sensor using semiconducting Ga2O3 thin films based on temperature switching of multigas reactions. Sens Actuators B 1995;24/25:544–7. [6] Kim JC, Jun HK, Huh JS, Lee DD. Tin oxide-based methane gas sensor promoted by alumina-supported Pd catalyst. Sens Actuators B 1997;45:271–7. [7] Dwivedi D, Dwivedi R, Srivastava SK. The effect of hydrogen-induced interface traps on a titanium dioxide-based palladium gate MOS capacitor (Pd-MOSC): a conductance study. Microelectron J 1998;29(7):445–50. [8] Nunes P, Fortunato E, Lopes A, Martins R. Influence of the deposition conditions on the gas sensitivity of zinc oxide thin films deposited by spray pyrolysis. Int J Inorg Mater 2001;3:1129–31. [9] Santhosh kumar D, Chandra Mouli K. Studies on Mn(1x)ZnxFe2O4 nanoparticles synthesized by co-precipitation method. Int J Nanotechnol Appl 2010;4(1):51–9. [10] Zylka P, Mazurek B. Rapid dissolved gas Analysis by means of electrochemical gas sensors. In: 14th International conference on dielectric liquids, July 2002. p. 325–8. [11] Benounis M, Aka-Ngnui T, Jaffrezic N, Dutasta JP. NIR and optical fiber sensor for gases detection produced by transformation oil degradation. Sens Actuators A 2008;141:76–83. [12] Muhammad NA. Comparative study and analysis of DGA methods for transformer mineral oil. IEEE Power Tech 2007:45–50. [13] Galstyan VE, Aroutiounian VM, Arakelyan VM, Shahnazaryan GE. Investigation of hydrogen sensor made of ZnO thin film. Am J Phys 2008;1(4):242–6. [14] Gruber D, Kraus F, Müller J. A novel gas sensor design based on CH4/H2/H2O plasma etched ZnO thin films. Sens Actuators B 2003;92:81–9. [15] Gong H, Hu JQ, Wang JH, Ong CH, Zhu FR. Nano-crystalline Cu-doped ZnO thin film gas sensor for CO. Sens Actuators B: Chem 2006;115(1):247–51. [16] Nagy Michael J, Romen Steven J. The effect of pulse width modulation (PWM) frequency on the reliability of thermoelectric modules. T E Technology, Inc. [17] Atanasova-Höhlein I, Rehorek C, Hammer T. Gassing and oxidation behaviour of insulating fluids under thermal stress. CIGRE 2009;C107:1–5. [18] Duval Micheal. Calculation of DGA limit values and sampling intervals in transformers in service. IEEE Trans 2008;24(5):7–13. [19] Digiorgio Joseph B. Dissolved gas analysis of mineral oil insulating fluids. Northern Technology and Testing. In: Proceedings of IEEE 14th International conference on dielectric liquids, July 2002. p. 274–8. [20] Bernadic A, Leonowicz Z. Fault location in power network with mixed feeder using complex space phasor and Hilbert Huang transform. IJEPES 2012;42(1):208–19. [21] Souahla Seifeddine, Bacha Khmais, Chaari Abdel Kader. MLP neutral network based decision for power transformer fault diagnosis using an improved combination of Roger’s and Doerrenburg ratios DGA. IJEPES 2012;43(1):1346–53. [22] Chatterjee Anjali, Bhattacharjee Partha, Roy Nirmal Kumar. Mathematical model for predicting the state of health of the transformers and service methodology for enchanting their life. IJEPES 2012;43(1):1487–94. [23] da Rosa Mauro A, Leite da Silva Armando M, Miranda Vladimiro. Multi agent systems applied to reliability assessment of power system. IJEPES 2012;42(1):367–74. [24] Chatterjee A, Bhattacharjee P, Roy NK. Online monitoring of power transformer by using gas sensors manufactured from nano-particles. In: UPEC 2011. p. 1–4.