In December 2005, Japan experienced a major outage in the Niigata Kaetsu area that lasted up to 30 hours and was caused by snow accretion on insulators. Porcelain long-rod insulators on several 154 kV and 66 kV lines had become covered by wet, packed snow of relatively high conductivity, attributed to salt transported from the sea by strong winds. The large amounts of wet snow, mixed with sea-salt, reduced insulation strength of the insulator strings and caused flashovers. While extensive research has been performed on the effect of insulator ice and snow accretion on flashover characteristics, knowledge related to the effect of salt-containing wet snow is still limited. In order to increase reliability of Japanese networks, a comprehensive project was undertaken to study the effects of ice and snow accretion on overhead lines. Part of this project was to clarify the snow accretion properties of long-rod and cap & pin insulators by field observation and artificial laboratory tests. A 154 kV class full-scale snow test procedure was also developed to evaluate various insulator designs and flashover voltage properties of snow-accreted insulators. This article, contributed by Hiroya Homma, Kohei Yaji, Teruo Aso, Masato Watanabe and Gaku Sakata of the Central Research Institute of Electric Power Industry (CRIEPI) in Japan, summarizes the procedures and results of field observation and artificial laboratory tests.
Blackout in Niigata Kaetsu Area
Details of the wet snowstorm in the Niigata case have been widely reported. Basically, a strong low-pressure system in the Pacific moved from south to north along the east coast of Japan’s main island while another low-pressure system in the Sea of Japan moved across the island on 22 Dec. 2005. Ambient temperature in the Niigata Kaetsu area, located in the northwest of the main island facing the Sea of Japan, stabilized in the range 0 to +2°C from 03:00 to 17:00 on 22 Dec. with heavy precipitation and wind. Total precipitation and the maximum 10 minutes average wind speed observed at Niigata Meteorological Local Agency was 26 mm from 03:00 to 17:00 and 14 m/s just before 09:00. The observation system of Tohoku Electric Power Co. recorded maximum wind speed of more than 25 m/s at 11:00. Cascading electrical failures on 154 kV and 66 kV transmission lines started just before 09:00 and resulted in numerous tripped lines. At about the same time, a couple of 275 kV transmission lines also tripped as a result of conductor galloping. In total, 30 transmission lines with 49 circuits tripped and induced a blackout covering a large area. Many porcelain long-rod insulator strings, used exclusively on 154 kV and 66 kV transmission lines, were packed with wet snow, the shape of which on some insulators was cylindrical while on others was eccentric pennant into the wind direction. These shapes of snow on insulator strings are different from those that generally result from ice accretion or snow cover without strong wind. Volume density of the packed snow ranged from 0.54 to 0.94 g/cm3 and maximum melted water conductivity was approximately 200 S/cm at 25°C.
The worst of the weather ended before the night of the 22 Dec. and restoration work started at midnight. More than 2500 workers climbed towers and removed packed snow from insulator surfaces by hand. This was difficult as the wet snow had by then nearly transformed into ice. Some 31 hours after start of the interruption, the system was restored to normal. During restoration, numerous traces of power arcs on insulator surfaces and hardware were observed on both horizontally and vertically mounted insulators, including V-strings. These were found over a wide area and many traces were also located some 30 to 40 km inland. Measured conductivities of the packed snow were highest in this area. Apparently, the sea-salt was carried inland by strong winds. After the outage, Tohoku Electric Power Co. replaced long-rod insulators on one transmission line circuit by cap & pin strings in the Niigata Kaetsu area and the same countermeasures are being conducted in other areas where similar events might occur.
Evaluation on Snow Accretion Properties of Various Insulators
Observation System & Method
The Niitsu observation site was set-up in an open field near the area where multiple ground faults were triggered during the snowstorm, 6 m above sea level and 17 km from the Sea of Japan shoreline (see Fig. 2). Several insulator samples were installed in a vertical position on the empty cross-arms at a height of 17.5 m, typical of 154 kV transmission towers. Three CCD cameras with internet terminals were employed and photographs of insulators have been taken every 10 min. Snow accretion on non-energized insulators was continuously observed under natural conditions. Table 1 summarizes the specifications of the 66 kV long-rod and cap & pin insulators tested. Although both types have been used by Japanese utilities, only long-rods were installed on the 154 and 66 kV lines in affected areas. After the outage, long-rod insulators on one of two-circuit transmission lines were replaced by anti-fog type cap & pin insulators as one of the selected counter-measures. For this reason, long-rod and the cap & pin insulators have both been selected as test insulators.Meteorological instruments have also been used to monitor the ambient climatic conditions in 10-min intervals. Relative humidity (RH) was measured with a capacitive sensor.
Effect of Insulator Geometry on Shed Bridging
Recently, field observations of glass and composite insulators under snow conditions have been conducted in Norway. The objective was to estimate ice or snow coverage and swing angle. However, a large amount of snow accretion could not be observed because temperatures might have been low, humidity low and wind velocities high. On the other hand, the authors carried out remote observations over the 5 sequential winters and recorded 26 events of snow accretion with shed bridging on either of the different insulators – although most events were caused by snowfall where 10-min average wind velocities (v10) were less than 10 m/s. Table 2 summarizes results on numbers of sheds bridging each year. The number of shed bridging events for long-rod insulators obviously exceeded that of cap & pin insulators and this was due to larger shed spacing.
Effect of Wind Speed on Shed bridging
The most interesting event was observed only once (Jan. 13, 2010), with v10 > 10 m/s. During this event, strong wind blew continuously with snowfall. The highest v10 was 16.2 m/s and cumulative precipitation (Pc) reached 9.5 mm. Figs. 3a and 3b show photos from of this high-v10 and typical low-v10 event. In the low-v10 case, Pc was about 3.5 mm. This value was smaller than that in the high-v10 case, however snow accretion volume was larger.
These observations correspond to the following snow accretion mechanisms, given certain assumptions:
• Under low-v10 conditions, snow falls relatively vertically and accumulates onto insulator sheds resulting in snow-covered insulators, not in packed snow accreted insulators. Thus, volume of snow accretion is larger while density is lower than under high-v10 conditions. Snow can easily bridge sheds.
• Under high-v10 conditions, snowflakes hit long-rod insulators horizontally by strong wind, resulting in packed snow conditions. In this case, volume of accreted snow is smaller but density is higher than under low-v10 conditions. Snow cannot easily grow enough to completely fill gaps between sheds.
• Massive precipitation is required under high-v10 conditions – compared with low-v10 conditions – in order to achieve shed bridging.
Test Facility & Procedure
High-v10 snow events are infrequent under natural conditions. In addition, ambient conditions related to snow accretion always change with time and complicate quantitative evaluation. In order to complement field observations and evaluate snow accretion properties under natural
conditions, a laboratory snow accretion test was carried out using a wind tunnel (see Fig. 4 for experiment set-up). In these tests, fresh natural snow was collected from the ground near the lab and stored in a freezer until testing. Snowflakes were made to fall over the outlet of the wind tunnel and blown at controlled wind speed. Test conditions are summarized in Table 3. Wind velocity (v) was varied from 3.0 m/s to 12.5 m/s and equivalent precipitation intensity (Peq) in mm/hour was measured using a customized cylinder set-up nearby the test insulators. The snow flux, expressed by the product of Peq and v per unit time, was maintained at a constant value of 350 to ensure repeatability. Ambient temperature was kept +1°C ± 0.5°C. Liquid water content (LWC) of snow was adjusted to almost 7% by pre-conditioning the snow, however was scattered from 5% to 9%. Anti-fog cap & pin discs, long-rods and also 254 mm standard cap & pin discs were selected as test insulators. The other part of the system consists of leakage resistance measurement instruments. This aims to detect shed bridging of snow-accreted insulators. A pair of thin electrode tapes attached to the body of test insulators and DC 100 V was applied to measure the leakage resistance (Rlab).
Reproduction of Cylindrical Snow Accretion from 2005 Snowstorm
In the snowstorm in 2005, cylinder-like shapes of snow accretion were observed. Fig. 5 shows typical photos of cylindrical snow accretion on a horizontal oriented long-rod insulator. In this case, wind velocity was 5.0 m/s, and the mechanisms explaining development of snow accretion on long-rod insulators were found to be similar to those confirmed for wires. Fig. 6 shows the assumed process of development of the cylindrical shape. Firstly, the snowfall was on the surface along the windward direction.
Then, accreted snow slides downward according to its weight and wind pressure. The snow was on the bare windward side surface again. This process was repeated a couple of times, finally resulting in a cylinder-like shape. In contrast, such cylindrical snow accretion was difficult to produce for vertically mounted insulators during the laboratory test. It is assumed that, under natural conditions, continuously varying wind direction helps snow accretion become cylindrical but constant wind direction in the test makes it pennant like in shape.
Comparative Simulation of Snow Accretion Characteristics of Various Insulators IUsing SAP
Fig. 7 shows an example of the transitions of the leakage resistances during the test. In this case, both long-rod and cap & pin insulators were mounted vertically and v was 5.0 m/s. Elapsed time (t) to the shed bridging was 25 min for the long-rod insulator and 76 min for the cap & pin insulator. Leakage resistances simultaneously showed rapid decreases with the shed bridging. Photographs of the snow accretion conditions are also shown for t = 25 min and 76 min in Fig. 7. These conditions replicate those observed in the field observations well, as shown in Fig. 3. In order to estimate the amount of snow flux on the insulators caused by a snow event for given climatic conditions, an index for snow accretion potential ability, SAP, is employed. SAP is expressed in kg/m2 as follows:
where P, v, t, and a (v) are the precipitation intensity in mm/s, wind speed in m/s, time in s, and coefficient of collision of snow particles, respectively. For the coefficient a (v), a constant value of 1.0 was used for all values of v, because a (v) around insulators has not never been investigated; the value was set to 1.0 presume the worst condition. In order to compare the snow accretion performance of various insulators, SAP until the shed bridging was measured. Fig. 8 shows SAP for each test insulator as a function of v. SAP for the long-rod insulator was approximately 1/9 times that for the anti-fog cap & pin insulator, and approximately 1/25 times that for the standard-disc cap & pin insulator. This result indicates that the cap & pin insulator exhibits superior performance when compared with the long-rod insulator. In addition, SAP increases with the wind velocity. This may be caused by the decreases of the coefficient of collision of snow particles, a (v) with the increase in v. Actually, in higher v test cases, snowflakes were frequently observed to strike and rebound from the accreted snow or insulator surfaces. These results agree with the conditions observed in nature, which were supposed to be evidence that SAP can be a feasible index for evaluation or/and prediction of snow accretion on transmission line insulators.
Evaluation on Flashover Voltage Property of Snow Accreted Insulators
Procedure of Flashover Voltage Test
To clarify the causes of the snow induced flashover and increase network reliability, a 154 kV class full-scale snow test procedure was developed for use in evaluating flashover voltage properties of different insulator designs. Various research has been carried out on the effect of insulator snow accretion on flashover characteristics and representative snow test procedures and evaluation methods have been established in IEEE standard 1783-2009. In the test procedure, natural snow gathered from the ground is mixed with salt with a snow blower. Then the snow is dumped into the snow pile jig mounted on the insulators to simulate accumulation of snow on horizontally oriented insulators. However, storing and handling natural snow presents difficulties. The proposed test required generation of snow with well defined conductivity, density, etc.. The target values of the snow parameters, such as snowflake size, snow density, liquid water content and snow conductivity, are shown in Table 4. The target snow conductivity was the same as observed after the Niigata Kaetsu blackout of 2005 and during continuous field observation in the same geographic region from 2007 to 2011.
The test procedure consisted of 4 steps: 1) generation of artificial snow with defined conductivity, 2) accretion of packed snow on the insulator, 3) increase of liquid water content in the accreted snow and 4) voltage application. A 154 kV class porcelain long-rod insulator and cap & pin suspension insulator string were utilized for the tests. The number of insulator sheds, connection length, creepage distance and dry arc distance of both insulators is shown in Table 5.
Table 4: Target values of Snow Parameters Table 5: Specification of Test Insulators
|Size of snowflakes||0.1-0.2 mm|
|Shape of snow||Cylindrical|
|Snow density||0.5 g/cm3 and higher|
|Liquid water content||20-30%|
|Snow conductivity, s25||200 and 700 mS/cm|
|Insulator type||Long-rod||Cap & pin|
|Shed number / profile||21||Anti-fog|
|Shed diameter [mm]||160||254|
|Connection length per unit [mm]||1,025||146|
|Creepage distance per unit [mm]||2,140||430|
|Number of units||2||13|
|Dry arc distance [mm]||1,774||1,960|
|Total creepage distance [mm]||4,280||5,590|
Artificial Snow Generation
During snow generation, water with conductivity of 200 or 700 mS/cm was sprayed inside a large climatic chamber of 18 m diameter and 23 m height at -9 to -10 °C, which generated fine ice particles in the form of artificial snow (Figure 9). Snowflake size was about 0.1 to 0.2 mm, and the visual appearance was very similar to natural snow (Fig. 10). The melted water conductivity of the artificial snow was 170 and 800 mS/cm as targeted, but the liquid water content was still zero. The collected artificial snow was kept in a storage freezer at -10 °C.
Accretion of Packed Snow
For the snow accretion and voltage tests, the artificial snow was brought back to the large climatic chamber where the temperature was adjusted to about +1°C. The snowflakes were blown onto the insulator by a small handheld vacuum cleaner operating in reversed mode until the snow accreted and filled the gaps between the sheds as observed in the Niigata case (Fig. 11). The distance between vacuum cleaner and insulator was about 500 mm and wind velocity at the insulator was approx. 15 m/s.
Since the amount of snow accreted onto the insulators influences on their performance in flashover tests, weight of the snow was recorded before each test. This was accomplished by hanging the complete suspension set using a load cell from the overhead crane. Fig. 12 shows photographs of insulators with well-packed, accreted snow at a density in the range of 0.5-0.6 g/cm3. Cylindrical snow accretion, as observed in the Niigata case, was achieved by rotating the insulator on a turntable during accretion. Thickness of the accreted snow at the shed surface was about 20 mm for both the long-rod and cap & pin insulators. For the long-rod insulator, all the shed-to-shed spaces were filled with high density packed snow, resulting in cylindrical shape (Fig. 12a). The cap & pin insulator strings maintained visible disk spacing after snow accretion (Fig 12b).
Increase of Liquid Water Content of Accreted Snow
Water with defined conductivity (the same as used for snow creation) was sprayed onto the accreted snow to increase the liquid water content to the range of 20-30%, which resulted in a snow density of 0.7-0.8 g/cm3 (Fig. 13). As the result, the target values for wet and packed snow with defined conductivity, density, etc. (Table 4) were attained and verified for the 154 kV insulators. Average weight of accreted snow on the long-rod and the cap & pin insulators was 21.2 kg and 23.8 kg, and the standard deviation was 1.6 and 2.1, respectively. Snow weight per dry arc distance of the tested insulators was derived as 120 g/cm and 121 g/cm for the long-rod and the cap & pin insulators. Upon completion of spraying, leakage resistance of the snow accreted insulator strings was measured using a mega-ohmmeter with results as shown in Table 6. The leakage resistance was very reproducible with repeated snow accretion for a given insulator type and snow conductivity.
Table 6: Leakage Resistance of Snow-Accreted Insulators
Measured Before Flashover Voltage Tests
Flashover Voltage Test
After spraying the conductivity-adjusted water onto the accreted snow surface, high voltage flashover tests were performed in the same test chamber used to create the samples, but at a temperature in the range of +1 to +2°C, using a 250 kVac, 500 kVA test transformer. Figs. 14 and 15 show the test set-up in the climate chamber and the test circuit. During the flashover voltage tests, leakage current was monitored through a shunt resistance of 150 W.
The applied voltage was increased to the desired value at a rate of 3-7 kV/s and thereafter kept constant until the insulator either flashed over or withstood, i.e., when the risk of flashover was considered negligible based on monitoring of leakage current levels. If flashover occurred at the applied test voltage, the test voltage was decreased about 7% for the next test.
Fig. 16 shows discharge activity at 134 kV during the voltage tests with the long-rod insulator. Fig. 17 shows the time variation of applied voltage and leakage current observed during the same test. Typically, three phases of flashover process were identified during the voltage test.
a. During the initial period of voltage application, small visible discharges appeared inside the accreted snow, and the leakage current increased with increased applied voltage.
b. A few tens of seconds after voltage application, intensive discharges developed and distributed along the insulator. A number of air gaps were created in the snow as a result of melting, and maximum currents in the range 100-600 mA were observed.
c. Thereafter, an arc grew along the surface and finally developed into a complete flashover after a few minutes.
Fig. 18 shows the typical sinusoidal waveform of the leakage currents observed during the initial period, which changes into a distorted waveform due to intense discharge activity.
Flashover voltage property of snow accreted insulators
Flashover Voltage Tests of 154 kV Insulators
Several flashover voltage tests of the snow accreted 154 kV class long-rod and cap & pin insulators were carried out with the snow conductivity of 170 and 800 S/cm. To understand the process of the snow-induced outage at Niigata Kaetsu and clarify usefulness of the countermeasure against snow-induced flashover, long-rod insulator and cap & pin insulators were compared. Table 7 shows conditions of flashover voltage tests.
In this research, withstand to the flashover test was defined as no flashover within one hour of voltage application with sufficiently low leakage current that future flashover seemed unlikely. For the cap & pin insulators, voltage tests were normally finished within ten minutes, because significant parts of snow often fell down rapidly. The minimum flashover voltage of an insulator was defined as being equal one voltage step higher than the maximum withstand voltage that the insulator withstood twice.
Table 7: Conditions of Flashover Voltage Test
|Temperature||0 to +2°C|
|Snow conductivity, s25||170 and 800 mS/cm|
|Liquid water content||About 20％
（water 4 L / snow 20 kg）
|Insulator type||Long-rod and cap & pin (anti-fog)|
|Sample set-up||Vertical position|
|Voltage application||Constant: Voltage step: 7％
(Ramp speed: 3 – 7 kV/s)
Relationship Between Snow Conductivity & Flashover Voltage
Fig. 19 shows the minimum flashover voltage of the snow accreted long-rod and cap and pin insulators at the two snow conductivities. The test results showed that the flashover voltage of both long-rod and cap and pin insulators decreased with the increase of snow conductivity. The lowest flashover voltage of the long-rod insulator at the higher snow conductivity was comparable to the maximum operating voltage 93 kV (normal operating voltage is 89 kV) of 154 kV transmission lines in the Niigata Kaetsu area. However, the difference between the lowest flashover voltage in the artificial snow tests and the operating voltage is presumably caused by some uncertainties of snow parameters measured at the sites during the Niigata case. In this research, snow accretion was performed without voltage, prior to the high voltage tests. However, if the insulators were energized during snow accretion as for operating transmission lines, the process leading to flashover might differ from that during laboratory tests. Leakage current through the accreted snow and partial discharge appearing on the snow surface may prevent snow accretion and increase liquid water content of snow. The effect of voltage application during snow accretion and changes in volume and conductivity of the melted water in accreted snow remain to be evaluated.
Comparing Long-rod Versus Cap & Pin Insulators
The cap & pin insulators showed significantly higher flashover voltage than the long-rod insulators for the same snow conductivity. The reason may relate to the configuration of the insulators, especially the difference in the shed spacing appears to be important. The small shed spacing of the long-rod insulator is easily bridged with packed snow, which results in a shorter creepage distance. Also, accreted snow tends to fall off the cap & pin insulator more easily during the voltage tests. Thus, substitution of the long-rod insulators with the cap and pin insulators seems to be reasonable as a countermeasure against snow induced flashovers.
This article described evaluation of snow accretion and flashover voltage properties of long-rod and cap & pin insulators focusing on the phenomena of wet snow accretion packed with sea-salt as occurred during the Japanese snowstorm of 2005. Results were as follows.
1. Field observation in natural environment and artificial laboratory tests under defined conditions were carried out in parallel. Wet snow accreted on the insulators in a temperature range particularly between 0°C and +1.0°C and the sheds of long-rod insulator were easily bridged compare to the cap & pin insulator because of the narrower shed spacing.
2. In the case of low wind velocity, volume of snow accretion was larger and snow easily bridged, but the density was lower than the high wind condition. It was assumed that massive precipitation was required under high wind condition to achieve shed bridging.
3. Cylindrical snow accretion seen during the snowstorm in 2005, was reproduced on the long-rod insulator in the laboratory tests. Mechanism of the development of snow accretion on the long-rod insulator was similar to those confirmed for wires.
4. Flashover voltage tests of snow-accreted insulators with controlled snow conductivity, liquid water content, density, etc. were carried out using the full-size 154 kV class insulators. The test procedure consisted of four steps, 1) generation of artificial snow with defined properties, 2) accretion and packing of snow on the insulator, 3) increasing the liquid water content of accreted snow in a controlled manner, and 4) AC voltage tests of the snow-accreted insulators.
5. The tests data indicate that the flashover voltage of both long-rod and cap & pin insulators decreased with increasing snow conductivity. The flashover voltage of the long-rod insulators with accreted snow of 800 S/cm conductivity was comparable to the service voltage in the Niigata Kaetsu area. Therefore, the test procedure appears suitable for evaluation of the flashover properties of insulators with wet snow accretion packed with sea-salt and is more practical than the existing snow test method, because it does not call for storing and handling natural snow.
6. Also, cap & pin insulators had significantly higher flashover voltage than long-rod insulators due to larger shed spacing. Thus, substitution of long-rod insulators with cap & pin insulators appears to be a reasonable countermeasure against snow induced flashovers.