1. Overview The Shuikou Hydropower Plant in Fujian Province has seven 240 MVA SFP9-240 000/220 type 242±2×2.5%/13.8 kV main transformers, with connection group YN.d11, no-excitation voltage regulation, and forced oil circulation air cooling (ODAF). Among them, transformer No. 6 (formerly transformer No. 2) has experienced two outlet short circuits since its commissioning in February 1994: a ground fault on phase A of the high-voltage side outlet on March 3, 1994, and a three-phase short circuit on the high-voltage side outlet on May 14 of the same year. Within three months of operation, the total hydrocarbon content in the transformer oil exceeded the warning value of 150 ppm, and the rate of increase was rapid. Dissolved gas chromatography analysis data in the transformer oil are shown in Table 1. Table 1. Statistical data of electricity consumption per capita in Northeast China region in 1995. Sampling Date: H2 CO CO2 CH4 C2H6 C2H4 C2H2 C1＋C2. Operating Time: 94-02-04: 45 190 1.6 0.6 0 2.2 72 hours; 94-03-07: 14 140 530 6.6 26 1.4 0 34.0 1 month; 94-05-11: 73 290 1200 110.0 41 190.0 0 340.0 3 months. [b]2 Fault Analysis[/b] The transformer is connected as a unit, with a capacity of 222.2 MVA, therefore the main transformer is not operating at full load. The transformer oil temperature is generally around 50℃. On May 14, a short circuit occurred at the main transformer outlet. The short circuit current multiple was 2.9, and the duration was about 0.15 seconds, which was far less than the transformer's allowable short circuit current and duration. Analysis of the gas in the oil revealed a high total hydrocarbon content, with C2H2 < 5 ppm, indicating a general overheating fault. Analysis using the "three-ratio method" showed C2H2/C2H4 = 0, CH4/H2 = 1.5, and C2H4/C2H6 = 4.6, coded as (0, 2, 2), indicating a thermal fault exceeding 700℃. The CO2/CO ratio was 1200/290 = 4.1 (above 3 but below 11), generally ruling out an insulation fault. Subsequent furfural content testing also confirmed that the solid insulation material had not undergone overall aging or localized deterioration. Gas production rate: after one month of operation, the absolute gas production rate ra = 1.5 ml/h, and the relative gas production rate rr = 1350%/month; after three months of operation, ra = 7.4 ml/h, and rr = 450%/month. It should be said that the absolute gas production rate more directly reflects the condition of the fault location. Table 2 shows the oil chromatographic analysis data from three tests conducted after the main transformer outlet short circuit and before it resumed operation. Table 2. Oil Chromatographic Analysis Data (ppm) After Short Circuit at Main Transformer Outlet Tab.2 Regional ROVTUE Sampling Date H2 CO CO2 CH4 C2H6 C2H4 C2H2 C1＋C2 ra (ml/h) 94-05-16 77 310 1200 120 44 210 0.8 370 7.4 94-05-20 95 360 1200 130 55 220 1.2 400 10.3 94-05-22 76 340 1300 140 64 230 1.1 430 26.6 As can be seen from Table 2, the gas production rate before and after the main transformer accident remained basically unchanged, and the accident had no direct impact on the change in gas content in the oil; the gas diffusion in the oil was relatively slow, and the total hydrocarbon content was still increasing one week after the accident, indicating that the fault location should be a place with poor oil circulation. The above analysis indicates an overheating fault exists within the main transformer, but the location and severity of the fault remain unclear. Diagnosing overheating faults within a transformer is a complex task. The type and location of the fault are closely related; different fault points reflect different fault types. Factors influencing fault diagnosis include equipment structure, auxiliary equipment failures, and other factors. The transformer's cooling system cannot be ignored, especially since a submersible pump failure significantly impacts the gases in the oil. The primary countermeasure remains oil chromatography analysis, with a correspondingly shortened testing cycle, supplemented by necessary electrical tests, and timely checks of the cooler's submersible pump to avoid external fault interference and misdiagnosis. Electrical tests were first conducted to determine if the transformer coils and core were faulty. The test results showed that the transformer's insulation resistance and DC resistance were normal. To monitor changes in the internal fault of the main transformer, tests were performed on four items: dissolved gases, moisture, furfural, and metal content in the oil. The results showed that the water content in the oil was 7.0 ppm, with no significant change compared to before production; the furfural content in the oil was 0.002 mg/l both before and after operation; and the metal content in the oil was shown in Table 3. Table 3: Test results of metal content in oil (unit: μg/g) Sample Iron Copper Aluminum Test Date 5.30 0.46 1.27 0.43 94-06-01 6.14 0.57 1.40 0.60 94-06-16 During the two months after the main transformer resumed operation on May 23, 24 samples were taken, totaling 56 samples (including oil samples from the lower drain valve of the main transformer, gas relay, and submersible pump outlet). Typical data from each gas analysis are shown in Table 4. Table 4. Typical data of oil chromatography analysis (ppm) after the main transformer resumes operation. (Sampling date: H2 CO CO2 CH4 C2H6 C2H4 C2H2 C1＋C2 ra/(ml/h)) 94-05-26 89 363 1475 145 57 260 1.2 460 66.4 94-06-02 115 410 1600 180 60 340 1.9 580 95.9 94-06-06 120 400 1500 180 67 355 1.8 605 29.5 94-06-10 125 425 1700 200 Table 4 shows that the total hydrocarbon content in the oil exceeds the warning value, mainly CH4 and C2H4, but the CO and CO2 contents are not high. Based on the "characteristic gas method," "three-ratio method," and "triangular spectrum method," it is determined that there is a localized overheating fault in the bare metal. Given the very low C2H2 content in the oil, the estimated hot spot temperature is no higher than 700℃. Statistical analysis of the changes in CO, CO2, and C2H2 in the oil gas was conducted using the "three-ratio method." The results showed that the nature and energy of the fault remained unchanged. Subsequent monitoring also indicated that the internal fault of the main transformer continued to develop, but the rate gradually slowed down. Since the main transformer was put into operation, the load has not changed significantly, and the changes in the oil gas may be related to the operation mode of the cooler (i.e., whether the submersible pump is engaged or not). After nearly two months of operation, the main transformer produced gas at a rate of ra = 13 ml/h and rr = 65%/month. This main transformer (2B) is exactly the same model as the previously put into operation No. 1 main transformer (1B), and the load is also almost the same. The 24-hour oil temperature changes of No. 1 and No. 2 main transformers are shown in Table 5. Table 5. 24-hour oil temperature variation of main transformers No. 1 and No. 2. Statistical data of ROVTUE of 9 typical industries in Northeast China region in 1995. Time/h: 0 2 4 6 8 10 12 14 16 18 20 22. Average 1B/℃: 50 49 47 46 52 53 53 54 54 54 53 51.4. 2B/℃: 50 49 48 47 53 54 54 58 58 58 56 55 53.1. Load/MVA: 207 208 206 211 214 214 214 219 219 217 213 210 213. Date: 1994-06-07; Temperature: 35/27℃. As shown in Table 5, the average oil temperature of main transformer No. 2 is 1-2℃ higher than that of main transformer No. 1, with a larger difference reaching 4℃ when the air temperature is high. This indicates that the efficiency of the cooler for main transformer No. 2 is low. Observation suggests that its oil pump flow rate is not reaching the rated value (135 m³/h). Furthermore, the sound of the oil flow at the outlet of the submersible pump in cooler No. 2 is also abnormal. Measurement of the cooler's oil pump current revealed that the operating current of submersible pump No. 2 is too high, and the starting current decreases slowly, indicating a longer starting process. Pump No. 3 has a higher starting current. Based on the above observations, combined with electrical tests and chromatographic data analysis, it can be determined that the submersible pump in cooler No. 2 is highly likely to be faulty. [b]3 Submersible Pump Disassembly Inspection and Fault Analysis[/b] The submersible pump model is 4B2.135-4.5/3 V. It was scheduled to be shut down along with the main transformer, and submersible pumps No. 2 and No. 3 were replaced. Submersible pump disassembly inspection and fault analysis were then conducted. Disassembly and inspection revealed severe overheating at both ends of the No. 2 submersible pump motor shaft near the iron core, resulting in a blackened and bluish tinge; the inner ring of the bearing (E307) at the non-impeller end was cracked. Based on the degree of shaft overheating, the estimated temperature was 500-600℃; and considering the approximately 150℃ temperature difference between the shaft thermal expansion and the resulting bearing crack, the hot spot temperature was also estimated to be above 500℃. This is largely consistent with the temperatures obtained from gas chromatography analysis of the oil. The No. 3 submersible pump motor shaft also exhibited a slightly lesser degree of overheating. Causes of the fault: Visual inspection revealed poor quality of the cast aluminum rotor, with poor integrity between the guide bars and end rings; the rotor overheating was caused by a broken rotor cage bar. Due to the broken rotor bar, the motor's starting torque decreased, resulting in a longer start-up process under load; the motor speed decreased under full load, leading to a reduced oil pump flow rate; and motor vibration increased the pump's operating noise. These findings are consistent with operational observations and analysis. After replacing submersible oil pumps No. 2 and No. 3, the main transformer operated for another 3 months, and gas analysis was performed on 24 separate tests of 50 oil samples each. The total hydrocarbon content in the oil increased from 820 ppm to 1100 ppm, with a gas production rate ra = 4.3 ml/h and rr = 10%/month. The main hydrocarbons were CH4 and C2H4, while C2H2 remained relatively constant. Furfural content was low, iron and copper content increased slightly, and aluminum content showed no significant change. Clearly, a localized overheating fault still existed inside the main transformer. Given the relatively high gas content in the oil at the time, vacuum degassing was performed on the transformer oil in conjunction with a minor overhaul. After the main transformer was put back into operation, the total hydrocarbon content continued to increase. The transformer coil deformation test results showed that the frequency response characteristics of the three-phase coils were relatively consistent, with small differences between them. Conclusion: No significant coil deformation occurred. To narrow down the suspected fault range, the main transformer coolers were operated in groups of (No. 1, No. 4), (No. 2, No. 3), (No. 1, No. 3), and (No. 2, No. 4). Further investigation and testing revealed that submersible pumps No. 1 and No. 4 also had faults. In conjunction with the pre-flood season maintenance of the main transformer, submersible pumps No. 1 and No. 4 were replaced. Disassembly and inspection results showed that the motor shaft of submersible pump No. 1, near the iron core at the non-impeller end, was severely overheated, turning black and blue; while both ends of the motor shaft of pump No. 4 showed overheating near the iron core. In summary, it was concluded that all four submersible pumps of the main transformer had varying degrees of motor rotor overheating, with pump No. 2 being particularly affected. Considering the total amount of gas (C1 + C2) generated by the main transformer fault, the reduction from eliminating the submersible pump fault could not offset the increase from the other fault point, indicating that the fault point was still developing. [b]4 Transformer Cover Inspection and Handling[/b] After eliminating the submersible pump fault of the main transformer, it was noticed that the three-phase DC resistance of the low-voltage coil was unbalanced and exceeded the warning value. This may be due to poor contact of the A-phase low-voltage coil leads, a problem with the welding quality between the leads and the copper busbars, or a partial short circuit fault in the iron core. After the main transformer was reinstalled, inspection and testing revealed no signs of overheating in the A-phase low-voltage coil leads or welded joints, and the insulation of all parts of the core was good. However, it was found that the copper busbar of the A-phase low-voltage coil leads was only 70 mm from the oil tank (the design value should be 100 mm). Three oil tank magnetic shielding plates near the lower leads of the A-phase low-voltage coil showed signs of overheating and discoloration, with the insulating adhesive scorched, cracked, and delaminated. Three areas of black debris deposited on the low-voltage side of the lower section of the oil tank, in considerable quantity. Infrared spectroscopy analysis (ICP method) revealed that the main components of the deposits were saturated hydrocarbons, silicon dioxide, and other metal oxides. The test results are shown in Table 6. Table 6. Infrared Spectroscopic Analysis Results of Sediments Sample No. Copper Iron Aluminum Loss on Ignition / % A 0.182 22.020 0.144 33.15 B 0.742 25.920 0.156 19.50 Based on transformer leakage flux calculation, the maximum axial magnetic flux density within the shield is 0.43246 T (RMS), indicating that the shield has not reached saturation. On-site analysis suggests that the overheating fault was caused by a closed loop formed due to the insufficient spacing between the tank's magnetic shield and the current-carrying copper busbar, as well as poor insulation at the fixed position of the magnetic shield. Infrared imaging temperature measurement showed that the temperature at the bottom of the A-phase coil on the low-voltage side of the main transformer was about 10°C higher than the top. This was also noticeably felt by touching the tank wall. Moreover, the location with the higher temperature was precisely at the lower lead wire connector of the A-phase low-voltage coil, but no signs of overheating or weld breakage were found at the connector after testing and observation with the cover. Analysis revealed that the weld joint between the low-voltage side A-phase lead and the copper busbar was only 30 mm from the magnetic shield of the oil tank, 10 mm smaller than the original design. The phase current passing through this point is approximately 5800 A. Due to the small distance, the leakage magnetic flux through the magnetic shielding silicon steel sheet increases, potentially causing oversaturation and overheating. If the silicon steel sheet is grounded at multiple points, it will also overheat due to a large circulating current. The debris deposited on the lower section of the oil tank, some of which can be attracted by a magnet, indicates the presence of magnetic metal particles. This phenomenon would only occur if the paint on the surface of the magnetic shielding silicon steel sheet had overheated and peeled off. By mid-May, the main transformer had been operating for over a year, especially since March, operating at almost full load (in the initial stage of operation, it operated at a low load due to low reservoir water levels). Prolonged overheating of some magnetic shielding components caused the surface paint to carbonize and peel off, potentially leading to insulation breakdown between the silicon steel sheets and increased losses. To address this, the entire magnetic shielding of the oil tank was replaced with a plate-type structure, and the distance between the low-voltage current-carrying copper busbar of phase A and the oil tank was adjusted to meet design dimensions. Simultaneously, efforts were made to fundamentally resolve the leakage flux and overheating issues. The chromatographic analysis results after the main transformer was put into operation are shown in Table 7. The fault gas analysis indicates that the main transformer's overheating fault is still developing, but it is unrelated to the main magnetic flux (total hydrocarbons do not increase during no-load operation). Table 7 Oil Chromatography Analysis Data (ppm) after Magnetic Shielding Overheating Treatment Sampling Date H2 CO CO2 CH4 C2H6 C2H4 C2H2 C1＋C2 Operating Conditions 96-12-22 6.4 4.1 130 12 7.2 24 0 43 No Load 96-12-26 38.0 10.0 130 100 47.0 170 0 320 Load 150MW 96-12-30 480.0 16.0 200 490 540.0 1300 4.8 2440 Load 200MW [b]5 Low-Voltage Coil Fault Diagnosis and Handling[/b] In order to confirm the location of the thermal fault in the main transformer, it is necessary to focus on finding out the reason for the high resistance of the low-voltage coil in phase A. Table 8 shows the measured DC resistance values ​​of the low-voltage coil of the main transformer at each stage, and Table 9 shows the measured DC resistance values ​​between phases of the coil at each of the three lifting stages. Table 8. Measured DC Resistance of Low-Voltage Coil of Main Transformer (μΩ) Test Date AB BC CA Relative Error/% 92-07-30 1202 1192 1206 1.17 95-05-19 1182 1179 1200 1.77 95-06-05 1211 1206 1222 1.32 96-10-13 1225 1222 1252 2.43 96-10-20 1192 1173 1201 2.36 96-12-11 1139 1130 1161 2.71 Table 9. Measured DC Resistance Between Phases of Low-Voltage Coil of Main Transformer (μΩ) Test Date Ax By Cz Relative Error/% 95-06-03 The table shows that the three-phase imbalance coefficient of the main transformer's low-voltage coil gradually increases. During the third inspection, no hot spots were found in the magnetic shielding or exposed conductive parts. However, a direct measurement of the phase resistance of the A-phase low-voltage coil showed a deviation of 5.10%, indicating a fault in the internal conductive circuit of the A-phase low-voltage coil. Half (61) of the 122 parallel-wound wires of the A-phase low-voltage coil were unsoldered, and the DC resistance of each wire was measured. Five wires showed lower measured values ​​(see Table 10), and the parallel value of every two wires was greater than the resistance of a single wire (0.22Ω) (see Table 11). Table 10. Measurement Results of DC Resistance of Some Single Conductors (Ω) | No. | 19 | 20 | 26 | 27 | 35 | Resistance | 0.1776 | 0.1157 | 0.1537 | 0.1406 | 0.1216 | Table 11. Measurement Results of Parallel Resistance Between Strands of Faulty Conductors (Ω) | Parallel Conductors | 19-20 | 19-26 | 19-35 | 26-27 | 26-35 | 27-35 | Resistance | 0.2726 | 0.3192 | 0.2940 | 0.2406 | 0.2615 | 0.2497 | Among these, the resistance values ​​of 5 conductors are too low, indicating that after the insulation between these 5 conductors was damaged, the circulating current caused overheating at the fault point, resulting in the conductors overheating and melting, reducing their cross-sectional area, and even, in severe cases, burning out. There is a parallel short circuit between the strands inside the coil. Judging from the overall high resistance, there is also a strand breakage phenomenon inside the coil. Inspection of the replaced A-phase low-voltage coil revealed that 7 of the parallel conductors in the 18th turn from the bottom were burned. Some conductors had most of their cross-sections melted, while others broke immediately after being pulled out. Analysis of the broken conductors after return to the factory for testing and analysis of the residue after burning indicated a problem with the conductor material itself, specifically the presence of impurities. The fault was likely caused by poor paper insulation at the welded ends of the conductors or damage from a short-circuit impact. 6. Fault Cause Analysis During the winding, drying, and assembly processes of transformer coils, defects in conductor material, poor insulation wrapping, improper welding, inadequate control of the winding and pressing process, or improper assembly operations often lead to insulation damage and short-circuit faults in the transformer coils. In particular, the mechanical vibration generated when the transformer experiences an external short circuit can cause coil deformation or induce coil faults. The short circuit between the 5 conductors in the A-phase low-voltage coil is a "short circuit between in-line conductors." Because the axial leakage flux varies along the coil's radial direction, the magnitude of the leakage flux linked at different positions of the parallel conductors differs, resulting in a potential difference between the parallel conductors. When a short circuit occurs between conductors in the same position, a circulating current will flow through the short circuit point, causing heat generation. Furthermore, the magnitude of leakage flux is also related to the transformer load; the larger the transformer load, the more severe the overheating and the greater the increase in fault gases, which is consistent with the results of the main transformer's chromatographic analysis. The coil strand insulation fault also confirmed the preliminary analysis's view that "the fault location should be where oil circulation is poor." It should be noted that although the inherent defects of this main transformer were detected in the oil chromatographic analysis, they were overlooked because the deviation in the measured DC resistance of the coil was not significant, making it even more difficult to diagnose the gradual fault inside the coil. After nearly three years of operation, the fault gradually developed to the point where the conductors melted, eventually causing a significant deviation in the coil's DC resistance. [b]7 Conclusion[/b] Because the main transformer fault was not singular but multiple and developing, and the potential main fault points were relatively hidden, coupled with the special nature of the fault, comprehensive analysis and judgment were very difficult. After nearly three years of extensive work, including approximately 500 chromatographic tracking analyses and various electrical and non-electrical tests and comprehensive diagnoses, the rare fault of both short circuits and open circuits between strands of the A-phase low-voltage coil was gradually identified after eliminating faults such as overheating of the submersible pump rotor, poor insulation between iron core poles, and overheating of the oil tank magnetic shield. The low-voltage coil was replaced on-site, including on-site hot oil spray vacuum drying and comprehensive testing. The latent thermal fault inside the main transformer, which had plagued the system for three years, was finally eradicated. Practice has proven that using gas chromatography to analyze dissolved gases in oil to detect latent faults inside oil-filled electrical equipment is a highly effective method. When the analysis of characteristic gases in the oil indicates a possible internal fault, a comprehensive analysis should be conducted in conjunction with electrical tests and other projects. Comprehensive analysis is a discipline. It requires a thorough understanding of the equipment (including its auxiliary equipment), a comprehensive grasp of the installation, operation, and maintenance status, as well as relevant design and manufacturing data; and a comprehensive judgment based on the results of electrical and chemical tests—that is, a systemic analysis of the entire process. This not only helps in determining the type of fault but also in accurately estimating the location of the fault.