comparing pd results with visual inspection enel iris power

Iris Rotating Machine Conference Scottsdale, June 2005

Page 1 of 19

COMPARING PD RESULTS WITH VISUAL INSPECTION

Alessandro Rossi ENEL (Italy)

G.C. Stone

Iris Power Engineering (Canada)

Abstract: ENEL is the main generation utility in Italy, with more than 200 hydro units and 120 turbine generators, totaling 39,000 MW of capacity. To help identify the maintenance needs of the stator windings in these units, ENEL has been gradually equipping the generators with partial discharge (PD) sensors that facilitate an on-line measurement of the PD. The paper describes results from several machines that have PD. In all cases, the high PD was confirmed by visual inspections. Although PD usually found the units with severe insulation problems, it was not always possible to determine the causes of the deterioration from the PD pattern.

INTRODUCTION

ENEL is the largest generator of electricity in Italy. At present it has 39,000 MW of generation, of which about 1/3 is hydraulic, and 2/3 is from fossil sources. There is no nuclear generation in Italy. There are about 180 hydro generators (>than 20 MW), 72 steam turbine generators and 48 gas turbine generators. As the fleet of hydro and fossil units ages, ENEL has found it prudent to be proactive in on-line condition assessment of the generators in the system. One of the tools used to find stator winding insulation problems is partial discharge (PD) testing. In the first part of the 1980s, ENEL experimented with on-line PD testing on 10 hydro generators. In addition, ENEL used devices such as the TVA corona probe during major overhauls. By the end of the 1990s, ENEL started to permanently install sensors on large hydro generators that have suffered from PD in the past, and on turbogenerators installed in new and renewed plants. Although there are many technologies available for on-line PD monitoring that are effective, ENEL decided to employ the high frequency technology using pairs of 80 pF sensors for hydro generators and smaller turbine generators, and stator slot couplers (SSCs) for the large hydrogen-cooled turbine generators [1,2]. This technology allowed ENEL to install the sensors, perform the testing and do basic interpretation of the results with only moderate training. Since 1999, 72 generators (comprising 13,500 MVA of generation) have been equipped:

33 hydros are equipped with 80 pF capacitive PD sensors installed within the winding, in the PDA mode

16 hydros and 4 turbos have been equipped with 80 pF sensors on the machine terminals (in the Bus mode)

23 large hydrogen-cooled generators have been equipped with SSCs. ENEL staff measure the partial discharge activity using the TGA-SP instrument manufactured by Iris Power Engineering. The interval between tests is based on how high the past results were, as well as the importance of the plant. The test is usually done 6 to 18 months apart. The process of testing these 72 generators, as well as analyzing and documenting the results, requires from 100


Page 2 of 19

to 150 man-days per year. Most of the machines have relatively low PD activity, in comparison with a published database of results [3]. However, a number of machines were identified by the test as having high PD. All the machines undergo a periodical visual inspection, according to established maintenance criteria; in some cases there has been also the opportunity to do off-line tests. This paper summarizes the comparison of the PD data with the known condition of the stator winding insulation. Some difficulties encountered with interpretation are also pointed out.

PLANT A, UNIT 2 This pumped storage plant consists of 4 units, rated 165 MVA, 17 kV. Unit 2 first ran in 1990, has an epoxy-mica insulation system and has two PDA sensors per phase. Figure 1 shows recent PDA data from Unit 2. Figure 1 (as for all of the similar figures below) has six plots, each of which shows the magnitude of the PD pulses versus the 50 Hz AC phase position. The color shows the pulse repetition rate. The top two plots show the PD activity in the C1 and C2 sensors in A (or U) phase, the middle two plots are from B (or V) phase and the bottom two plots are from the PD sensors in C (or W) phase. Note that the plots from the three phases are phase shifted from each other, so that a vertical line through the plots for the three phases show PD activity that is occurring at the same time. Such plots sometimes enable one to determine if the PD is occurring in the slot (driven by phase to ground voltage) or in the endwinding (driven by the phase to phase voltage) [3]. The highest PD is occurring in Phase A, near coupler 2. The peak PD magnitude (Qm) is +732 and 679 mV. This is a significant PD magnitude, since it is higher than almost 90% of 16-18 kV machines in a published statistical distribution of PD magnitudes [4]. The plots seem to indicate that there is phase to phase PD between phases A and B, since the PD is occurring at the same time but with opposite polarity in these two phases. [3]. On this unit, sensors were installed following a first visual inspection that revealed both phase to phase and slot activity. When the tests showed this high PD, a wider visual inspection of the stator was planned and then conducted in 2003. A number of different problems were found. Photo 1 shows the expected phase-to-phase PD that is occurring in the endwinding. However, in addition to this problem, significant PD deterioration was found to be occurring at the interface between the slot semiconductive coating and the silicon carbide stress relief coating interface. When a boroscope was used to examine the sides of bars in the slot, signs of slot discharge were found (Photo 2). Thus instead of one deterioration process, three were found (although the latter two may be caused by the same manufacturing problem). The PD test indicated there were problems, but the PD from the endwinding must have overshadowed the PD due slot discharge. Repairs for this unit are planned for the next year or so.

Photo 1: Endwinding PD from Unit 2, Plant A.


Page 3 of 19

Photo 2: Slot discharge on the surface of a stator bar in the slot, as viewed via a boroscope down

a vent duct in the core. PLANT E, UNIT 2 This pump storage plant is similar to Plant A, except the units were commissioned in the late 1970s. It has 9 units, each rated 170 MVA, 17 kV. The insulation system of this unit is of the polyester-mica splitting type. The PD activity for all three phases is shown in Figure 2. B phase Coupler 1 is the part of the winding with the highest PD activity with a Qm of +300 mV and 709 mV. This is a similar level as found in Plant A Unit 2 and indicated that the PD levels were high. With the strong negative predominance of the PD, and the classic phase position of the PD pulses (between 0 and 90 degrees and 180 to 270 degrees of the AC cycle), this PD would typically be occurring in voids within the groundwall insulation, close to the copper conductors. A stator bar was identified as having high PD by the TVA probe. This bar was then removed for examination and it was found that the groundwall insulation was easily delaminated (Photo 3). Such delamination is typically caused by operation at high temperature and/or load cycling, as occurs in a pump storage hydrogenerator. Although the insulation was delaminated, the winding still seems to have significant life, since the unit passed a DC hipot test at 37 kV. Discharging within the groundwall insulation of a Roebel bar is a very slow deterioration process.

Photo 3: Delaminated insulation in a stator bar removed from Plant E, Unit 2.

PLANT E, UNIT 4


Page 4 of 19

This unit is similar to the one described above, but in this case the insulation is an epoxy- mica type. The PD activity for all three phases is shown in Figure 3. The highest PD is on Phase C, near coupler 2. The Qm is +1119 and 760 mV. This is exceptionally high PD, higher than about 95% of similar voltage machines [3]. The phase position of the PD in Figure 3 indicates that there is significant phase to phase PD, especially between A and C phases, near coupler 2 which is often caused by bars that are too close together in the endwinding. In order to prove if the readings were due to phase-to-phase activity, ENEL performed some tests when the machine was operated in the pump mode (which has reverse rotation). As expected, it was noted that the phase-to-phase activity changed position by 60 degrees. This kind of comparison between the pump and generate mode has been repeated on others pump storage units within ENEL and often showed the 60 degree difference. So this phase shift is a good indicator on if phase-to-phase PD is occurring in the endwinding due to insufficient spacing between stator bars. In addition, the phase pattern in Figure 3 for Unit 4 shows there is PD within the slot near coupler 2 in A phase although some experience is needed to distinguish this from the complicated pattern. In early 2005, the rotor was extracted for repair and there was an opportunity to visually inspect the complete winding along the slots. The areas of PD activity between bars in the endwinding were found. In addition, using a boroscope, 6 stator bars were found to be affected by slot PD, although only one of them was near the high voltage end of the winding. The PD probably developed in air spaces between the side of the bars and the core, because the wedges were found to be tight. Taking advantage of the unit shutdown due to other reasons, a stator rewind has been planned.

PLANT R, UNIT 4 Plant R is another pump storage plant, with 8 units rated 140 MVA, 17 kV. The stator insulation system is of the epoxy-mica paper type. In Unit 4, a PDA test in December 2003 revealed exceptionally high PD activity. As shown in Figure 4, the worst PD is in Phase B, near coupler 2, where the Qm is +1429 and 2485 mV. This is amongst the highest PD ever recorded. The PD patterns in Figure 4 are complex- indicating that two or more deterioration processes are occurring. Although we know the PD activity is very severe, it is not easy to determine from the patterns what the root causes are. As a result of the high PD, the winding was visually inspected in March 2004. This inspection showed that the semiconductive coating in the slot was extensively deteriorated (Photo 4), and the filler strips between bars were eroded. In addition, there were signs of endwinding PD. With this visual confirmation, a general overhaul and the substitution of the affected bottom bar is planned for the 2nd half of 2005.

PLANT R, UNIT 8 Unit 8 in the same plant also has some significant PD activity, although not as high as in Unit 4. Figure 5 shows that the Qm in the worst phase (Phase A, near coupler 1) has a Qm of +548 and 195 mV. The PD pattern only reveals classic PD in the slot. Due to the positive polarity predominance, one would expect PD on the bar surface. When PD results were taken at different loads, +Qm was found to increase from 350 mV to 550 mV as the load increased. As a result of these characteristics, it was concluded that loose bars within the stator slot caused the PD. A visual examination of the winding in October 2004 did indeed find evidence of loose bars within the slot, even if there arent yet strong signs of bars deterioration. Since loose windings can lead to relatively rapid failure in comparison to other processes, corrective actions are being planned.


Page 5 of 19

Photo 4: Boroscope image of the slot discharge damage on the side of a bar, as viewed down a vent duct in the stator core. This is from Plant R, Unit 4. The PD activity is shown in Figure 4.

PLANT P, UNIT 4 Plant P has 4 pump storage units rated 300 MVA, 12 kV. Unit 4 has an epoxy-mica paper insulation system. The PD activity is shown in Figure 6, where a band of high- level PD activity (Qm +590 mV and 622 mV) in Phase C near coupler 1. For a 12 kV winding, this PD is higher than almost 95% of similar windings. The band PD pattern is unusual, so it was decided to closely inspect the bars from that parallel in June 2002, when the stator was rewound for a mechanical damage suffered from a part of the rotor. Some stator bars belonging to Phase C and positioned near coupler 1 showed signs of slot discharge and erosion (Photo 5). It is not known why the PD pattern is so unusual.

Photo 5: Erosion of the groundwall insulation on a bar removed from Plant P, Unit 4.


Page 6 of 19

PLANT P, UNIT 2 Unit 2 is a sister unit of the same design and ratings as Unit 4, above. This unit was visually inspected both in 2001 and 2003. No significant stator winding insulation issues were found in these inspections, which included measurement of the contact resistance between the bar and the core and a boroscope inspection through ventilation ducts. The PD tests in July 2004 on the four parallels per phase equipped with PD sensors showed that the highest PD was in the C2 coupler in B phase, where the Qm was only about 140 mV. In comparison to the published PD database, this is not a high value for a 12 kV machine. Unfortunately this machine experienced a stator ground fault in May 2005. The visual examination revealed the failed bar, which was connected to the phase-end, had severe signs of abrasion in the slot, as well as slot discharge. In addition several other bars were found to be deteriorated. The unit was quickly returned to service by cutting out the failed bar. A repeat PDA test in May 2005 after the unit was returned to service did show high activity (Qm of about 900 mV) - at least in one parallel that was known to have visual signs of looseness and PD deterioration (Figure 7). What was surprising was that the PD test performed less than a year before did not show the deterioration. We can only conclude that the rate of deterioration must have been much faster than is normal since the 2003 visual inspection and the 2004 PDA test did not reveal any issues. It seems that some failure processes can progress more quickly than previously suspected, arguing for a shorter time between tests.

PLANT PG, UNIT 21 This is a combined cycle plant with 4 generators; two of them are rated 370 MVA, 20 kV. These units are hydrogen cooled, and usually operate near 45 psi. Unit 21 was commissioned in 1975 and has a polyester mica splitting insulation system. In 1991, visual inspections indicated that some bars were damaged by slot PD, and they were replaced. Similarly, a bar was replaced for the same reason in 2001. The PD sensors are SSCs installed in slots with bars operating at high voltage. Figures 8 and 9 show the PD activity in 6 of the SSCs in 2004. The highest PD activity was measured in Slot 27 in Phase A. The PD was all in the slot, and not in the endwinding (SSCs are able to explicitly distinguish between endwinding and slot discharges [2]). Slot 27 Qm was +683 and 710 mV. This is exceptionally high for a hydrogen-cooled winding measured with SSCs. In comparison to a published database of PD measured with SSCs, this is the highest reading ever recorded. The patterns are consistent with classic PD within the slot. Slots 27, 20 and 8 in phases A, B, and C respectively all have exceptionally high PD (Figure 8). However, in comparison, slots 28, 18 and 7 all have very low PD (Figure 9). The bars that were replaced all exhibit relatively low PD in comparison to the original bars installed in 1975. Thus it seems that problems are still occurring in the bars from the original winding. The PD patterns in Figure 9 have an odd polarity dependence compared to all the other measurements shown above. In particular, the positive PD is occurring in the positive portion of the AC cycle. This may because there is some cross coupling from the extremely high PD in the other slots, or perhaps this is an artifact of PD occurring immediately below the SSC.

CONCLUSIONS

On-line PD testing has identified a number of windings that have significant insulation deterioration problems, which has been confirmed by visual inspections. Thus PD testing is useful for identifying units that need further attention. However, in some of the ENEL generators, the PD pattern was complex and it was not easy to determine what the actual cause of the deterioration was, prior to a visual inspection. Also, in one case, there was less than one year warning of a high risk of failure, thus more frequent testing than every 6 months may sometimes


Page 7 of 19

be needed. PD testing is considered by ENEL a powerful tool and its used together with other diagnostic methods, mainly visual inspections. Maintenance decisions are taken by ENEL on the combined results of two or more diagnostic methods; in this way, it is possible to have a confirmation of the problems detected and to enlarge the area of what is detectable.

REFERENCES 1. J.F. Lyles et al, Using Diagnostic Technology for Identifying Generator Winding Maintenance Needs, Hydro Review, June 1993, pp 58-67.

2. S.R. Campbell, et al, Practical On-line PD Test for Turbine Generators and Motors, IEEE Trans EC, June 1994, pp 281-287.

3. V. Warren et al, Advancements in PD Analysis to Diagnose Stator Winding Problems, Proceedings of the IEEE International Symposium on Electrical Insulation, April 2002, pp 497-500.

4. V. Warren, Partial Discharge Testing: A Progress Report, Proceedings of the Iris

Rotating Machine Conference, June 2004, New Orleans. NOTA: FIGURAS DAS P`GS. 11 A 19 FORAM PROPOSITALMENTE RETIRADAS PARA ADEQUAO AO NMERO M`XIMO DE P`GINAS. ENTRETA NTO, NO HOUVE PERDA SUBSTANCIAL COMPREENSO DO PAPER.


Page 8 of 19

Figure 1. PDA test data for Plant A, Unit 2. Note that Phase A is measured on a less sensitive scale. Also note that B and C phases are shifted form A phase by 120 or 240 degrees. A vertical line through the 3 graphs implies that the PD is occurring at the same instant of time.


Page 9 of 19

Figure 2: PD in Plant E, Unit 2. The PD is typical voids within the groundwall.


Page 10 of 19

Figure 3: PDA result from Plant E, Unit 4

comparing pd results with visual inspection enel iris power

Documents

pd results

large hydro generators

high pd

smaller turbine generators

steam turbine generators

gas turbine generators

online pd testing

enel staff