Simultaneous Determination of Dissolved Gases and Moisture in

Jun 19, 2001 - Institut de recherche d'Hydro-Québec (IREQ), 1800, boulevard Lionel-Boulet, Varennes, Québec, Canada J3X 1S1. This paper presents the...
0 downloads 0 Views 143KB Size
Download PDF
Anal. Chem. 2001, 73, 3382-3391

Simultaneous Determination of Dissolved Gases and Moisture in Mineral Insulating Oils by Static Headspace Gas Chromatography with Helium Photoionization Pulsed Discharge Detection J. Jalbert, R. Gilbert,* and P. Te´treault

Institut de recherche d’Hydro-Que´ bec (IREQ), 1800, boulevard Lionel-Boulet, Varennes, Que´ bec, Canada J3X 1S1

This paper presents the development of a static headspace capillary gas chromatographic method (HS-GC) for simultaneously determining dissolved gases (H2, O2, N2, CO, CO2, CH4, C2H6, C2H4, C2H2, C3H8) and moisture from a unique 15-mL mineral oil sample. A headspace sampler device is used to equilibrate the sample species in a twophase system under controlled temperature and agitation conditions. A portion of the equilibrated species is then automatically split-injected into two chromatographic channels mounted on the same GC for their separation. The hydrocarbons and the lighter gases are separated on the first channel by a GS-Q column coupled with a MolSieve 5-Å column via a bypass valve, while the moisture is separated on the second channel using a Stabilwax column. The analytes are detected by using two universal pulsed-discharge helium ionization detectors (PDHID). The performance of the method was established using equilibrated vials containing known amounts of gas mixture, water, and blank oil. The signal is linear over the concentration ranges normally found for samples collected from open-breathing power transformers. Determination sensitivity varies with the nature of the species considered with values as high as 21 500 A × 10-9 s (µg/ g)-1 for H2O, 46-216 A × 10-9 s (µL/L)-1 for the hydrocarbons and carbon oxides, and as low as 8-21 A × 10-9 s (µL/L)-1 for the O2 and N2 permanent gases. The detection limit of the method is between 0.08 and 6 µL/L for the dissolved gases, except for O2, N2, and CO2, where higher values are observed due to air intrusion during sampler operations, and 0.1 µg/g for the dissolved water. Ten consecutive measurements in the low and high levels of the calibration curves have shown a precision better than 12% and 6%, respectively, in all cases. A comparison study between the HS-GC method and the ASTM standard procedures on 31 field samples showed a very good agreement of the results. The advantages of configuring the arrangement with two PDHID over the conventional flame ionization and thermal conductivity detectors were clearly demonstrated. High-voltage (HV) power transformers are susceptible to malfunctions such as arcing, overheating, and partial discharges, * Corresponding author: (e-mail) [email protected].

3382 Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

which always result in the chemical decomposition of the mineral oil and cellulose insulation. Several gases are produced that are totally or partially dissolved in the oil, including H2, CH4, CO, CO2, C2H2, C2H4, C2H6, and C3H8. Because the relationship of dissolved gases with specific transformer problems is known, it is possible to detect the presence of faults at an early stage of development and take preventive action before failures occur. As a result of this, the analysis of dissolved gases (DGA) in mineral oils became over time one of the most important assays for electric utilities worldwide. Another critical HV power transformer condition is that the insulation should be kept at very dry levels in order to maximize the life of cellulose insulation and achieve adequate electrical strength and low dielectric loss. To monitor the dryness condition of the insulation, the same public utilities rely on periodic removal of an oil sample for analysis of the moisture. Dissolved gases and moisture are usually determined by applying two different standard procedures: the dissolved gases are assessed by gas chromatography using the basic vacuum extraction method to partially remove the species from oil samples (method A of ASTM D 3612 or IEC Publication 567) while Karl Fischer (KF) coulometric titration is used for the moisture (method B of ASTM D 1533 or IEC Publication 814).1,2 Since the first papers were published in 1993 on the use of a static headspace sampler coupled with gas chromatography (HSGC) for performing DGA in mineral insulating liquids,3-5 the technique has been successfully implemented by a large number of laboratories around the world. The interest in the technique was mostly related to the procedure’s capability to perform with one instrument over 40 consecutive analyses unattended. Briefly, the technique relies on establishing equilibrium with the species (1) Annual Book of ASTM Standards; American Society for Testing and Materials: Philadelphia, PA, 1998; Vol. 10.03 (ASTM D 3612: Standard Test Method for Analysis of Gases Dissolved in Electrical Insulating Oil by Gas Chromatography. ASTM D 1533-96: Standard Test Methods for Water in Insulating Liquids (Karl Fischer Reaction Method)). (2) Guide for the Sampling of Gases and of Oil from Oil-Filled Electrical Equipment and for the Analysis of Free and Dissolved Gases; IEC Standard Publication 567; International Electrotechnical Commission: Geneva, Switzerland, 1992. Determination of Water in Insulating Liquids by Automatic Coulometric Karl Fischer Titration; IEC Standard Publication 814; International Electrotechnical Commission: Geneva, Switzerland, 1985. (3) Leblanc, Y.; Gilbert, R.; Duval, M.; Hubert, J. J. Chromatogr., A 1993, 633, 185-193. (4) Leblanc, Y.; Gilbert, R.; Jalbert, J.; Duval, M.; Hubert, J. J. Chromatogr., A 1993, 657, 111-118. 10.1021/ac010063f CCC: $20.00

© 2001 American Chemical Society Published on Web 06/19/2001

Figure 1. Schematic diagram of the equipment for simultaneous determination of dissolved gases and moisture by the HS-GC technique. Representation of the standby condition.

between an oil sample and a gas phase in a closed vial. The vials are prepurged with argon, which also allows the atmospheric gases present in the mineral oil samples to be assessed. The equilibrium is achieved at 70 °C with mechanical shaking to ensure quick desorption of the volatile components into the headspace. After equilibration, a portion of the headspace gas is automatically injected into a chromatograph for capillary column separation and thermal conductivity (TCD) and flame ionization (FID) detection of the species. A step-by-step procedure based on the technique was submitted in 1997 to the ASTM D-27 Committee for the inclusion of a headspace procedure into the ASTM D 3612 Standard (Accepted in October 2000 for an inclusion in ASTM D 3612). Similar action was undertaken by the CIGRE TF 15-01-07 Working Group on behalf of the International Electrotechnical Commission (IEC). During the same period, another method using a headspace sampler mounted on a chromatograph equipped with a capillary column and a TCD was developed for measuring moisture in mineral oil at low parts-permillion levels.6 The analytical performance was shown to be comparable to the KF coulometric titration method with measurements unaffected by the oil matrix, contrary to what is experienced with direct sample titration.7-9 Considering the substantial field sampling and analysis cost reduction that could be achieved by using a unique HS-GC technique for determining in a single run both dissolved gases and moisture in mineral oil samples, considerable effort has been made in recent years to find the proper instrumental conditions for such application. To realize the merge, some technical difficulties had to be resolved, e.g., separation of a polar species such as moisture under DGA nonpolar conditions and the lack of detection sensitivity for some species when argon is used as a carrier gas. The first concern could be tackled by splitting the (5) Jalbert, J.; Gilbert, R. IEEE Trans. Power Delivery 1997, 12 (2) 754-760. (6) Jalbert, J.; Charbonneau, S.; Gilbert, R. A New Analytical Method for the Determination of Moisture in Transformer Oil Samples. Proc. 63rd Annu. Int. Conf. Doble Clients, Boston, MA, March 1996. (7) Jalbert, J.; Gilbert, R.; Te´treault, P. Anal. Chem. 1999, 71 (15), 3283-3291. (8) Gilbert, R.; Jalbert, J.; Te´treault, P. Anal. Chem. 2001, 73 (3), 520-526. (9) Cedergren A.; Nordmark, U. Anal. Chem. 2000, 72, 3392-3395.

sample volume injected between a polar column for the moisture separation and a nonpolar column for separating the other species. With this type of split injection, some moisture will inevitably reach the nonpolar column and conditions would have to be found to avoid peak interference with the hydrocarbons and permanent gases. For the second concern, one application needs argon as a carrier gas for a TCD quantification of H2 near the low part-permillion level, while for an acceptable H2O sensitivity, the carrier gas should be helium if the same detector is kept in the system. To overcome this, the conventional TCD and FID detectors could be substituted for two universal pulsed-discharge helium ionization detectors (PDHID). These detectors use a nonradioactive pulsed high-voltage discharge source for generation and pulsed collection of electrons.10,11 In a conventional DGA arrangement, a Ni catalyzer is fitted at the inlet of the FID to improve the sensitivity of CO and CO2 by converting these gases to CH4. An arrangement including highly sensitive detectors for atmospheric gases such as PDHID may eliminate the need for such a conversion. On the other hand, the use of these detectors may cause O2 and N2 peak saturation, especially when samples are collected from openbreathing power transformers where mineral oil is air saturated. The work performed to combine the two individual HS-GC techniques into a single method is described in this paper. The method performance will be presented and discussed as well as the results of a validation through a comparison with standard procedures on samples collected from power transformers. EXPERIMENTAL SECTION Apparatus. A schematic diagram of the equipment used for the simultaneous determination of dissolved gases and moisture by HS-GC is shown in Figure 1. Headspace sampling is done using a Hewlett-Packard device, model 7694, equipped with a 250-µL injection loop and a 44-sample outer tray feeding an 8-sample inner tray. The device allows for both temperature control and mechan(10) Wenworth, W. E.; Vasnin, S. V.; Stearns, S. D.; Meyer, C. J. Chromatographia 1992, 34 (5-8), 219-225. (11) Mendonca, S.; Wentworth, W. E.; Chen, E. C. M.; Stearns, S. D. J. Chromatogr., A 1996, 749, 131-148.

Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

3383

Table 1. Instrumental Conditions Headspace Sampler Parameters temperatures sample 70 °C transfer line 150 °C gas sampling valve and 150 °C injection loop pressure vial overpressure 70 kPa (0.7 bar) times equilibration at 70 °C 30 min with shaking pressurization 0.2 min loop fill 0.2 min loop equilibration 0.2 min injection 1.0 min shaking power maximum level Gas Chromatograph Parameters carrier He gas flow 8.2 mL‚min-1 oven 30 °C for 5.5 min 30-110 °C at 10 °C‚min-1 110 °C for 1.5 min 110-150 °C at 24 °C.min-1 150 °C for 10 min PDHID temperature 300 °C PDHID auxiliary gas He at 30 mL‚min-1 bypass valve 0.0 min, GS-Q and MolSieve in series 4.7 min, MolSieve bypassed 9.0 min, GS-Q and MolSieve in series 13.6 min, MolSieve bypassed

ical agitation of the samples in the inner tray. The transfer line of the sampler is connected directly to the inlet of a zero dead volume T-union (Valco Instruments Co. Inc.) located in the oven chromatograph where the capillary columns are fixed at the outlets (channels 1 and 2). The chromatograph is a Hewlett-Packard, model 6890, mounted with two PDHIDs, model D-3-I-HP (Valco Instruments Inc.). To optimize detector sensitivity, air intrusion into the system had to be minimized. One of the major sources of air was from the headspace sampler’s mechanical flow controller and pressure regulator. The GC carrier gas and vial pressurization flows were then brought to the sampler through separate stainless steel lines from the chromatograph’s auxiliary electronic pressurecontrolled port (EPC). With these modifications, the PDHID baseline signal was between 1 and 2 × 10-6 A. The hydrocarbons and the lighter gases are separated on channel 1 by a 40 m × 0.32 mm GS-Q column (J&W Scientific) coupled with a 20 m × 0.32 mm MolSieve 5-Å column with a film thickness of 30 µm (Chrompack) via a bypass valve. The bypass valve is used to temporarily isolate the MolSieve column by routing the effluent of the GS-Q column directly into the PDHID. The water is separated on channel 2 by a 60 m × 0.25 mm Stabilwax column (Restek) with a film thickness of 0.25 µm. The use of a 0.25-mmi.d. column allows the helium flow rate to be nearly equilibrated on both channels (4.3 mL‚min-1 in channel 1 and 3.9 mL‚min-1 in channel 2). The static headspace and chromatographic conditions are summarized in Table 1. To validate the moisture determinations, the results of the HS-GC method were compared with those obtained from a Mitsubishi Moisturemeter, model CA06 (Mitsubishi Chemical Inc.), equipped with a membrane titration cell, model CAMCL2 for coulometric measurements. The anodic compartment is filled with 70 mL of Hydranal-Coulomat AG-H and 3384

Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

30 mL of chloroform and the cathodic compartment with 2.5 mL of Hydranal-Coulomat CG for all the tests. Injection of the moisture contained in the oil samples is done indirectly by performing an azeotropic distillation with an Aquastar oil evaporator, model EV6L (EM Science). A transfer line is used to connect the evaporator chamber to the KF vessel. The full setup is completely protected from intrusion of atmospheric moisture. The instrumental conditions set for these KF determinations are identical to those described in a previous publication.8 Chemicals. The performance of the HS-GC method was established using equilibrated headspace vials containing known amounts of gas mixture, moisture, and Voltesso 35 mineral oil (transformer oil from Imperial Oil, Canada). A primary gas mixture standard in helium of 5000 µL/L for H2, O2, N2, CH4, CO, CO2, C2H2, C2H4, C2H6, and C3H8 (Scott Speciality) was used for these preparations. In the case of CO2, CO, O2, and N2, pure gases (minimum purity of 99.997%, Matheson Gas Products) were also used to spike the vials at a higher concentration range. The moisture was added into the vials by using a water-saturated octanol solution of 47.3 (1 mg of water/g of solution (WSO).12 The analytical performance was also established by using two certified gas mixture standards in helium (also from Matheson), one containing 10 µL/L H2, C2H4, C2H2, C2H6, CH4, and C3H8, 200 µL/L CO, 1510 µL/L CO2, 15 000 µL/L O2, and 65 100 µL/L N2, and the other, 100 µL/L H2, C2H4, C2H2, C2H6, CH4, and C3H8, 3240 µL/L CO, 28 000 µL/L CO2, 45 400 µL/L O2, and 66 700 µL/L N2. The KF reagents were purchased from Riedel-de Hae¨n. The toluene (Certified ACS) used for azeotropic distillation, methanol, and chloroform (Optima grade) were from Fischer Scientific. The helium gas (99.999% pure) supplied to the GC and the glovebox was from Union Carbide. The dry nitrogen used for carrying the azeotropic vapors into the KF vessel was from Matheson Gas Products, and the calcium carbide (Technical grade) was from Sigma-Aldrich Canada Ltd.. Preparation of Field Samples and Calibration Standards. The oil samples are collected from transformers using 30-mL Perfektum glass syringes (Popper & Sons Inc.) fitted with threeway stopcocks (Pharmaseal). Upon receipt of these syringes, a Precision Glide 18G1 needle (Becton Dickinson & Co.) is attached to the three-way stopcock. A 15-mL aliquot of the sample is then introduced into a 20-mL headspace vial sealed with a perforated aluminum cap fitted with a rubber PTFE-lined septum (Wheaton). The vials are purged with helium prior to the introduction of the oil sample. The positive pressure buildup during the introduction of the oil is released through a 26G1/2 needle mounted on a syringe to avoid direct contact of the content with air. In the case of the preparation of calibration standards in the presence of oil, a patent pending dehydrating process13 is first applied to the oil to remove the moisture. As explained in the application, the process consists of introducing 10 g of CaC2 into 50-mL polyethylene bottles (Nalgene Centrifuge Ware), which are then filled to the top with an oil and closed. The bottles are then agitated at 70 °C for 1 h and centrifuged at 14 000 rpm for 10 min to separate the liquid phase consisting of dried oil and the solid phase consisting of a mixture of Ca(OH)2 and unreacted CaC2. A first (12) Margolis, S. A.; Levenson, M. Fresenius’ J. Anal. Chem. 2000, 367 (1), 1-7. (13) Jalbert, J.; Gilbert, R. Process for Dehydrating a Mineral Oil or Other Solvents for the Preparation of Moisture-In-Oil or Moisture-In-Solvent Standards. Hydro-Que´bec, Que´bec, Canada, Patent Pending, 1996.

Figure 2. Typical HS-GC chromatograms obtained for the equilibration of a vial containing a blank oil sample under the conditions given in Table 1.

step of dried oil recovery is then carried out by introducing the bottles into a glovebox (under 0.7-1.7% RH) so that their contents can be decanted into a second series of polyethylene bottles. The new series of bottles is then again centrifuged under the same conditions as above. This centrifugation-decantation cycle is applied four times before pouring the oil of the bottles into a sparger located in the glovebox (Ultra ware mobile-phase reservoir from Supelco/Sigma-Aldrich). The helium bubbles are used to drag the dissolved gases from the oil to the reservoir headspace where a vacuum is applied. After ∼4 h of bubbling, 15 mL of oil is taken by connecting a 30-mL Perfektum glass syringe to a valve attached to the reservoir cap. A tube attached at the inlet port of the valve allows samples to be drawn from the bottom of the reservoir. The syringe containing the oil is then used to fill a 20mL open headspace vial. The vials are then sealed and brought out of the glovebox, where they are spiked with known amounts of WSO, the 5000 µL/L primary gas mixture, and pure gases using 10-µL/1-mL SGE syringes (Supelco). Finally, the samples are homogenized by applying 20-s vortex mixing with a Maxi-Mix II (Barnstead/Thermolyne). In the case of preparation of calibration standards in the absence of oil, sealed-headspace vials are purged at a rate of ∼2 L‚min-1 for at least 30 s with pure helium or one of the two certified gas mixture compositions used for this study. Safety Considerations. The Hydranal-Coulomat AG-H contains methanol, 1-pentanol, N,N-dimethyldodecylamine, imidazol, sulfur dioxide, and some bromide and iodide while the HydranalCoulomat CG contains diethanolamine hydrochloride, methanol, diethanolamine sulfite, diethanolamine, and hydriodic acid. These compounds are all toxic to varying degrees with respect to inhalation, ingestion, and eye irritation. Several solvents (methanol, chloroform, 1-pentanol) are flammable. In addition, imidazole is an in vitro mutagen and chloroform is an animal carcinogen. Calcium carbide reacts violently with water, liberating extremely flammable gases. Inhalation may result in spasm, inflammation, and edema of the larynx and bronchi, chemical pneumonitis, and pulmonary edema. Proper precautions should be taken in the laboratory to minimize exposure and to comply with the disposal regulations.

RESULTS AND DISCUSSION Figure 2 shows typical chromatograms of a dehydrated and degassed oil sample (blank sample) recorded simultaneously under the optimized conditions given in Table 1. The channel 1 signal shows three peaks at retention times of 4.9, 9.4, and 10.4 min corresponding to CO2, O2, and N2, respectively. The presence of these species in the chromatogram could be associated with a residual of atmospheric gases in the blank sample or air intrusion into the system during one of the sampler sequential operations, e.g., vial pressurization, loop fill, loop equilibration, and gassampling valve actuation. The signal was estimated at 12.8 µL/L for CO2, 1450 µL/L for O2, and 2290 µL/L for N2 by comparison with peak areas recorded from the analysis of a vial containing 15 mL of blank oil in which 35 µL of pure CO2, 500 µL of pure O2, and 500 µL of pure N2 were added (Concentrations expressed in µL/L (or ppm (v/v)) for the dissolved gases and in µg/g (or ppm (w/w)) for dissolved water according with the normal practice in ASTM and CEI standards). The headspace gas concentrations resulting from the equilibration of such gas volume additions are those that would have been obtained from an oil sample containing 2600, 37 120, and 37 120 µL/L CO2, O2, and N2, respectively. On the other hand, the channel 2 signal shows three peaks at retention times of 2.2, 6.1, and 9.0 min. The first peak located at the gas hold-up volume is attributed to the unretained atmospheric species (O2, N2, CO2). The second peak is attributed to a volatile contaminant that was present in the glovebox when the vial was prepared. The third peak is attributed to moisture associated with a residual of H2O in a blank sample and more likely with atmospheric moisture intrusion in the system during sampler operations. The headspace unit used in this study was definitively showing an abnormal air intrusion during operations when compared to previous systems used for the development of individual techniques. The amount of moisture present in this chromatogram is estimated at 1.3 µg/g by comparison with the signal obtained from the analysis of a vial containing 15 mL of blank oil in which 8 µL of WSO was added. The headspace moisture concentration resulting from equilibration of such a WSO Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

3385

Figure 3. Typical HS-GC chromatograms obtained for the equilibration of a vial containing a blank oil sample with a known amount of gas mixture and moisture under the conditions given in Table 1.

addition is the one that would have been obtained from an oil sample containing 24.4 µg/g H2O. Figure 3 shows typical chromatograms recorded from the analysis of a vial containing 15 mL of blank oil in which 200 µL of 5000 µL/L gas mixture, 500 µL of pure O2, 1 mL of pure N2, and 8 µL of WSO were added. The headspace gas concentrations resulting from the equilibration of such additions are those that would have been obtained from an oil sample containing 74, 1856, 78, 74, 74, 39 302, 74 316, 77, 742, and 74 µL/L H2, CO2, C2H4, C2H2, C2H6, O2, N2, CH4, CO, and C3H8, respectively, and 24 µg/g of H2O. As seen in this figure, all the peaks are well resolved under the column configuration, programmed oven temperature, and column isolation conditions retained for this application (Table 1). While developing the method, a Carboxen-1006 column (30 m × 0.53 mm i.d.) previously qualified for DGA5 was tested for the CO2, C2H4, C2H2, C2H6, and C3H8 species separation. Unfortunately, a broad peak was interfering in the elution region of the C2 hydrocarbons that was attributed to moisture introduced onto the column by splitting the injected volume over the two channels. By configuring the system with a GS-Q column, the moisture peak is delayed in the elution region of the unsaturated C3 hydrocarbons (peak maximum at 15.8 min) leaving the light hydrocarbons free of interference. The peak resulting from the volatile contaminant in the glovebox is also seen in the channel 2 chromatogram. As shown in this figure, it takes ∼14 min for the complete analysis of the species of interest; furthermore, an additional 10-min column purge is required, as well as 6 min to return to initial conditions. The headspace sample equilibration time was therefore set at 30 min, under which conditions the analysis takes 54 min for the first sample and 24 min for subsequent samples. The performance of the method was first evaluated by calibrating the system with equilibrated known amounts of species in blank oil. The air-saturated samples collected from openbreathing transformers normally contain more than 35 000 µL/L O2 and 70 000 µL/L N2. The linearity of the signal was therefore checked over the following concentration ranges in the oil: 0-50 µg/g for H2O, 0-100 µL/L for H2, CH4, C2H2, C2H4, C2H6, and C3H8, 0-800 µL/L for CO, 370-7800 µL/L for CO2, 8400-43 000 3386

Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

µL/L for O2, and 21 000-85 000 µL/L for N2. To cover these concentration ranges, vials containing 15 mL of blank oil were equilibrated with different known amounts of the 5000 µL/L gas mixture, pure O2, N2, CO and CO2, and WSO. Eight calibration points were used in the 0-50 µg/g range (0, 0.8, 1.6, 3.1, 6.1, 12.2, 24.4, and 48.8 µg/g), seven in the 0-100 µL/L range (0, 3.7, 7.4, 13, 26, 46, and 93 µL/L), nine in the 0-800 µL/L range (0, 3.7, 7.4, 13, 26, 46, 93, 371, and 785 µL/L) and four in the 370-7800 µL/L range (371, 1961, 3923, and 7845 µL/L), in the 8400-43 000 µL/L range (8491, 21 227, 33 963, and 42 453 µL/ L), and in the 21 000-85 000 µL/L range (21 227, 42 453, 59 435, and 84 907 µL/L). Each concentration level was prepared in triplicate, giving a total of 42 vials to be processed by the system. Each individual point of the calibration graphs was obtained by averaging the results over the triplicate analyses performed at a given concentration. The parameters of the regression lines are given in Table 2. As seen in this table, the signal is linear over the concentration ranges of interest with r-values almost greater than 0.99. Some examples of the straight-line relationship obtained with the calibration plots are illustrated in Figure 4, where the error bars show the variability of the signal over the triplicate runs. The data indicate that the sensitivity varies with the nature of the species of interest. It could be as high as 21 500 A × 10-9 s (µg/g)-1 for H2O, between 46 and 216 A × 10-9 s (µL/L)-1 for the hydrocarbons and carbon oxides, and as low as 8-21 A × 10-9 s (µL/L)-1 for the O2 and N2 atmospheric gases. To understand the rationale behind these figures, one needs to take into account the relative concentration of each species in the headspace gas at injection, which is controlled by the solubility of the species in the oil at 70 °C, and the ionization efficiency of the species by the photons produced from the helium discharge.14 It is evident that a higher sensitivity could be achieved by using a larger injected sample volume (loop volume >250 µL). However, the use of a larger gas sample size with such equilibrated airsaturated oil samples will directly affect the linearity of the O2 and N2 signals by causing peak saturation. Another interesting (14) Madabushi, J.; Cai, H.; Stearns, S.; Wentworth, W. Am. Lab. 1995, (Oct), 21-30.

Table 2. Regression Line Parameters for Calibration with Equilibrated Vials Containing Known Amounts of Gas Mixture, Moisture and Blank Oil

gas

Na

corr coeff r

slope m (A × 10-9 s (µL/L)-1)

H2O H2 CO2 C2H4 C2H2 C2H6 O2 N2 CH4 CO C3H8

8 7 4 7 7 7 4 4 7 9 7

0.9980 0.9995 0.9862 0.9991 0.9992 0.9995 0.9934 0.9882 0.9987 0.9980 0.9992

21574.5b 63.5 46.0 132.8 108.7 128.8 21.4 7.8 216.3 152.1 82.8

a

error in the slope Sm (A × 10-9 s (µL/L)-1)

intercept b (A × 10-9 s)

error in the intercept Sb (A × 10-9 s)

570.4b 0.9 5.5 2.5 2.0 1.9 1.7 0.9 4.9 3.7 1.5

62139.8 114.8 43207.5 236.9 324.8 202.0 128578.9 408227.1 636.7 336.9 183.0

11364.8 38.2 23228.0 105.9 81.6 76.1 47678.0 42729.1 204.7 1072.0 60.6

Number including a blank reading point except for O2 and N2. b Expressed in A × 10-9 s (µg/g)-1.

Table 3. Detection Limits of the Method (3SB/m)a this work 250-µL injection loop work5

gas

prev 3-mL injectn loop with TCD and FID detectn

calcd with slope of calibrtn curves obtained with vials containing equilibrated known amts of species in oil

H2 CO2 C2H4 C2H2 C2H6 O2 N2 CH4 CO C3H8 H2O

0.6 0.1 0.04 0.05 0.04 11 11 0.06 0.09 0.2 0.3b

0.08 25 0.04 0.3 0.1 11 45 0.5 1 6 0.1

calcd with slope of calibrtn curves obtained with vials purged with diff gas mixtures 0.09 17 0.02 0.1 0.03 1070c 4490c 0.06 0.1 6

a Expressed in µL/L for the dissolved gases and in µg/g for dissolved water. b From injection of equilibrated samples at 150 °C using a sampler coupled with a chromatograph equipped with a TCD.6 c Two-point calibration.

observation from Table 2 is the nonnegligible intercept value for O2, N2, CO2, and H2O as a consequence of air intrusion during sampler operations. As will be seen in the next section, this will affect the lowest amount of atmospheric species that could be detected by the system. The detection limit of each species, 3SB/m, was estimated by using the standard deviation of the noise (SB) in the vicinity of each peak of interest on a chromatogram recorded from a headspace vial containing 15 mL of blank oil and the corresponding slope of the calibration curves (m) given in Table 2. The calculated values are presented in Table 3 together with those previously reported for the individual HS-GC techniques. The detection limits are between 0.08 and 6 µL/L for all the dissolved gases, except for O2, N2, and CO2, where higher values are observed due to a contamination during sampler operation. When comparing the detection limits with those of the individual techniques,5,6 one needs to take into account that the headspace volume injected in the PDHID-configured arrangement is reduced from 3 mL to 250 µL and also that the reduced volume is split over two channels. A sufficiently low detection limit is achieved for CO and CO2, which allows the system to be configured without the need of a Ni catalyzer. A very low detection limit is also achieved for H2O (0.1 µg/g), even if samples are equilibrated at

70 °C (150 °C for the individual HS-GC method; see ref 6). This opens the possibility of substituting the PDHID for a TCD on channel 2. However, the use of a PDHID on this channel allows the detection of other volatile species that could eventually be used to improve the current fault diagnosis methods or to extend diagnosis to other specific transformer problems. The precision of the method was estimated by performing 10 consecutive measurements in the low and high levels of the calibration curves. Two series of 10 vials containing 15 mL of blank oil were equilibrated in the presence of known amounts of the 5000 µL/L gas mixture, pure O2, N2, CO, and CO2, and WSO. In the case of the low-level vial series, the headspace gas concentrations at equilibration are those that would have been obtained from an oil sample that contains 9 µL/L H2, C2H4, C2H2, C2H6, CH4, and C3H8, 1856 µL/L CO2, 19 620 µL/L O2, 39 240 µL/L N2, 185 µL/L CO, and 9 µg/g H2O. For the high-level series, it corresponds to an oil sample containing 74 µL/L H2, C2H4, C2H2, C2H6, CH4, and C3H8, 3710 µL/L CO2, 39 302 µL/L O2, 74 316 µL/L N2, 742 µL/L CO, and 24 µg/g H2O. The relative standard deviations obtained from the analysis of these two sets of samples are given in Table 4. In all cases, the data indicate a precision better than 12% and 6% at low- and high-concentration levels, respectively. Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

3387

Figure 4. Examples of a typical straight-line relationship obtained with the calibration plots. Table 4. Precision of the Method (n ) 10) % RSD gas

low levels

high levels

H2 CO2 C2H4 C2H2 C2H6 O2 N2 CH4 CO C3H8 H2O

9.7 2.3 8.0 7.9 9.3 2.9 1.5 8.1 9.4 11.2 7.7

3.1 2.0 2.7 2.1 3.7 2.3 1.6 2.0 3.7 6.0 1.1

Analysis of duplicate field samples with the HS-GC technique and the ASTM standard procedures was done in order to assess the accuracy of the measurements. Two 30-mL syringe samples were collected from each of the 31 open-breathing transformers selected for this study. Upon receipt of the syringes, one set was 3388 Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

Figure 5. Typical HS-GC chromatograms obtained for the equilibration of a vial containing a sample collected from an open-breathing transformer under the conditions given in Table 1.

routed to an independent laboratory where the dissolved gases were determined using method A of ASTM D 3612. From the remaining set, two 15-mL oil aliquots were collected from each syringe, one for assessing the moisture by method B of ASTM D 1533 and the other for simultaneously determining dissolved gases and moisture using the headspace technique. In applying method B of ASTM D 1533, the moisture contained in the oil samples was introduced into the KF titrator by performing an azeotropic distillation with toluene. A previous investigation showed that this type of sample injection provides KF results that are free of any systematic errors.7,8 Typical chromatograms obtained with one of these field samples are shown in Figure 5, where the advantage of the technique using two PDHID detectors is well illustrated. For example, the peaks associated with the products resulting from the decomposition of insulating papers (2-furaldehyde, furfuryl alcohol, 2-acethylfuran, 5-methylfurfural) and the oil antioxidant (2,6-di-tert-butyl-p-cresol (DBPC)) could be seen in the 12-20-min region of the channel 2 chromatogram. The sensitivity of the PDHID may be sufficiently high for determining furanrelated compounds under the actual system configuration at the

Figure 6. Comparison of the HS-GC method with the gas extraction method (method A of ASTM D 3612) and KF coulometric titration (method B of ASTM D 1533). System calibrated with equilibrated vials containing known amounts of gas mixture, moisture, and blank oil.

sub-ppm level. Many peaks are also well resolved in the region of 2.5-7.5 min that could be tentatively associated with polar lowmolecular-weight volatile species. The results reported at 273 and 101.325 kPa are compared in Figure 6 with the y-axis of the graphs for the HS-GC data and the other axis for the data obtained by applying the standard ASTM procedures. The number of data points for each graph could be less than 31 because some of the species were present in the samples at an undetectable level and, in few cases, at a level well above the calibration concentration ranges chosen for HS-GC. As seen in the graphs where the

diagonal corresponds to a perfect correlation of the data, a very good fit is observed for most of the species of interest. In the case of the HS-GC technique, the total gas volume in percent is obtained by the sum of the individual gas concentrations expressed in parts-per-million. The results show a very good correlation with the ASTM values obtained by measuring in a buret the volume of the gases extracted from the oil. As discussed in a previous presentation,15 it is not necessary to use equilibrated known amounts of gases in vials containing 15 mL of blank oil for calibrating the chromatographic signal. At Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

3389

Figure 7. Comparison of the HS-GC method with the gas extraction method (method A of ASTM D 3612). System calibrated with equilibrated vials containing known amounts of gas mixture.

equilibrium, the following relationship exists between the remaining concentration of a gas in the oil, CL, its concentration in the headspace, CG, and its initial concentration in the oil, C0L: C0L ) CG(K + VG/VL), where VG and VL are the volume of the gas phase and liquid phase in the headspace vial, respectively, and K is the gas partition coefficient at 70 °C. The C0L of each gas can be calculated from the chromatographic determination of CG and then the only response to calibrate is CG through headspace sampler injections of different standard gas mixtures. To further evaluate the method’s performance, this less tedious system calibration was used to compare the results obtained on the 31 field samples. In using this approach, however, the gas partition coefficients need to be known precisely as well as the phase ratio VG/VL because they could affect the accuracy of the determinations. In the present case, the averaged volume over 20 vials taken from the same batch as the ones used for field sample analysis was measured and the value utilized to calculate the VG/VL ratio (VG/VL ) 0.363). Meanwhile, the following set of gas partition coefficients for the (15) Gilbert, R.; Jalbert, J. Dissolved Gas Analysis in Insulating Oils by Controlled Headspace Sampling Coupled With Capillary Gas Chromatography. Proc. 8th BEAMA Int. Electr. Insul. Conf. 1998, 444-451.

3390

Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

70 °C ) 0.094, Voltesso 35 was applied in the above relation: KH 2 70 °C 70 °C 70 °C 70 °C KO2 ) 0.220, KN2 ) 0.121, KCO2 ) 0.901, KC2H2 ) 0.968, °C 70 °C KC702H°C4 ) 1.417, KC702H°C6 ) 1.985, KCH ) 0.428, and K70 CO ) 0.136. 4 These coefficients, which are slightly different from those previously used for such calculations,5 were obtained using the HSGC vapor-phase calibration method proposed by Kolb.16 This method was found in our laboratory to be very accurate for determining coefficients in the range of values covered by the species of interest (0.07-10). However, the solubility of these species may be slightly affected by different mineral oil compositions, and consequently, the use of this set of coefficients for other oil types may slightly affect the accuracy of the determinations. The system was then calibrated by analyzing the following samples: one vial purged with pure helium (zero concentration point), one vial purged with the certified gas mixture of lower concentration, and one vial purged with the certified gas mixture of higher concentration (see Experimental Section for the certified gas mixture compositions). Three-point calibration curves were constructed for all the species except for O2 and N2, where the

(16) Kolb, B.; Welter, C.; Bichler, C. Chromatographia 1992, 34 (5-8), 235240.

zero concentration point was omitted due to presence of nonnegligible peaks in the chromatogram resulting from the air intrusion problem discussed above. The slope of the calibration curves (m) and the standard deviation of the noise (SB) as estimated from the zero concentration point chromatogram in the vicinity of each peak of interest were used to get a second estimate of the detection limits (3SB/m). As seen in the last column of Table 3, it is possible to reach lower limits for some species when compared with a system calibrated with vials containing equilibrated known amounts of species in oil. However, higher values are observed for O2 and N2, and in these cases, a lower concentration standard would have been needed to obtain a better estimate of the slope of the calibration curve used in the calculation (only two-point calibration; see Table 3). The interpolated values from such calibration curves are compared in Figure 7 with the data obtained by applying the standard ASTM procedure. These results show that the type of system calibration may affect the degree of correlation achieved with the standard procedure. For instance, a slightly better correlation is noted in Figure 7 for CH4 and CO when compared to the Figure 6 data. In conclusion, it was shown that the use of this new technique for performing simultaneously dissolved gases and moisture in insulating oil brings many advantages over the conventional approaches. Only 15 mL of sample is required for performing both assays under a single chromatographic run. It is no longer required to use a Ni catalyzer to convert CO and CO2 to CH4, which is recognized as a critical piece of equipment when the

reliability of the conventional systems is considered. Besides, helium is the unique gas required for driving all the system: headspace sampler, chromatographic separation, and detection. In addition, it allows for the possibility of extending our fault diagnostic capabilities by giving access to additional molecules that could be key factors in determining the state of the equipment insulation. However, the long-term reliability of the PDHID still remains to be established before an implementation of the technique for routine analyses. Finally, when considering the use of the technique in such an application, the availability of commercial gas-moisture standards or moisture-in-oil standards for calibrating the chromatographic signal will definitively make the approach more attractive. ACKNOWLEDGMENT The authors thank P. Gervais from Hydro-Que´bec’s TransEÄ nergie division for the substantial financial support attributed to the project. They are particularly grateful to Dr. G. Be´langer from IREQ for assistance at different levels. Special thanks go to G. L. Gauthier and J. McCurry from Agilent Technologies and Dr. J. Madabushi from Valco Instruments Co. for their support in the acquisition of the instrumentation.

Received for review January 16, 2001. Accepted May 3, 2001. AC010063F

Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

3391