The performance of overhead line insulation in DC differs from AC mostly because of the ionization of airborne particles resulting from the unidirectional electric field. Contamination levels in DC are therefore typically more severe than for AC, given the same service environment. Both CIGRE and IEC have published technical guides and mathematical models to handle this situation at the insulation design phase for a new DC line.
This edited 2017 contribution to INMR by J.M. George, E. Brocard, D. Lepley and F. Virlogeux of the Sediver Research Center in France compared insulation levels defined by these models with actual field observations as well as laboratory test results. The differences pointed out suggest a need to re-evaluate these models that, otherwise, might lead to string lengths being over-dimensioned under a ‘pessimistic’ approach.
Review of Key Parameters
Pollution deposits on the surface of an insulator are classified using Equivalent Salt Deposit Density (ESDD) and Non Soluble Deposit Density (NSDD). Both are important parameters. ESDD contributes to higher conductivity of water on the surface of insulators and subsequently increased leakage current. NSDD acts like a sponge, ‘sucking up’ moisture during morning dew or fog events and thereby providing water to the leakage current and dry band arcing mechanisms. In addition, contamination deposits in the field are usually non-homogeneous and the bottom of an insulator is typically more polluted than the top surface. CUR is the parameter that describes this property and is defined as the ratio of bottom to top ESDD. During laboratory testing, CUR is often set at 1, for practical reasons, thus providing more severe conditions than with a higher CUR. Other parameters such as shape and dimension of the insulator, altitude, dynamics of the contamination deposition process, polarity, etc. also matter. The focus here is on determination of suitable leakage distance for a string of insulators based on CUR, ESDD and NSDD.
The severity of contamination in any given environment is classified in standards and in Technical Brochures (i.e. CIGRE C4.303-TB518, Outdoor insulation in polluted conditions. Guidelines for selection and dimensioning. Part 2. The dc case IEC TS 60815-4 Ed. 1.0 2016-10 Selection and dimensioning of high voltage insulators intended for use in pollution conditions. Part: insulators for dc systems) from very light to very heavy. The specific creepage distance is then determined to avoid flashover of the insulator string under the given contamination conditions. An important element in this discussion is Reference Unified Specific Creepage (RUSC) distance, which is the minimum USCD calculated. Once this is defined, insulator design can be selected from a supplier’s catalogue and also the number of insulators required for each string.
String Design Based on Theoretical Approach
The following example shows a possible discrepency between results from the theoretical method and those from laboratory testing. A typical 300 kV DC string comprised of 20 toughened glass DC units, each having leakage distance of 550 mm, was tested with artificial pollution conditions established at ESDD = 0.047mg/cm² and NSDD = 0.1mg/cm² (see Fig. 2). This translates into a USCD = 36.6 mm/kV. According to the CIGRE & IEC documents referred to above, the theoretical RUSC for this environment would be 44.8 mm/kV. The performance of the string was excellent with leakage current peaks around 10 mA (see Fig. 3).
The difference of 20% is most likely highly conservative given the low currents measured during this test. It is therefore possible that the model is not adequate to accurately describe performance of a DC string under pollution. If such a line were to be designed based on results based from the model’s calculation method, string over design could generate unnecessary added cost.
Field Performance & Laboratory Testing
Several cases of actual DC lines have been evaluated, measuring real pollution levels by sampling on-site ESDD and NSDD levels. This information was then used to determine, using the theoretical model, the expected USCD. Laboratory tests were also performed on the actual string of insulators to determine flashover values. Among these cases, the pollution pattern of a 500 kV DC line, which did not experience specific pollution problems over 40 years’ service in a desert environment, showed non-uniform distribution along the strings, as described Fig. 4. This line is designed with a USCD = 25 mm/kV.
The theoretical model would recommend using at least 42 mm/kV given uniform contamination along the string and with ESDD = 0.04mg/cm², NSDD = 0.1mg/cm² and CUR=1. A laboratory artificial pollution test was performed on DC toughened glass insulators selected for a voltage upgrade of this line. The string was set with USCD = 23 mm/kV and tested under the same conditions as those used above for the theoretical estimate (i.e. ESDD = 0.04mg/cm², NSDD = 0.1mg/cm² and CUR=1). Maximum current during this test sequence was 60 mA. Going further in this investigation, another test was conducted with ESDD = 0.07mg/cm² and NSDD = 0.1mg/cm². In this case, maximum leakage current was 100 mA, but still without a flashover. Under these conditions, the theoretical model would recommend using a string with USCD = 51 mm/kV, i.e. twice what was tested. Even if allowing for the effect of possible non-linearity between short strings and full length strings (note: here the test was produced only on short strings), the difference with results of the model seems too high to accept.
Another example comes from Brazil and the 600 kV DC bipoles at Itaipu. Several strings were removed from Bipole 2 for evaluation. On this line, actual USCD is 28.5 mm/kV and pollution levels are classified as ‘heavy’ due to the agricultural and industrial pollution near São Paolo (see Fig. 5). It is interesting to note the CUR ratios in Fig. 6. While the utility’s Maintenance Department stated that there were no line interruptions related to pollution problems in the region, the theoretical model predicts intense flashovers unless strings are re-designed with at least USCD = 47mm/kV. In this case, it could be argued that the CUR level is outside the classical range considered in design of the model (i.e. the typical range of CUR in the model is CUR < 10), yet this again demonstrates the limits and inaccuracies of the mathematical approach.
Mitigation of Severe Pollution
Most DC lines are built in relatively clean environments, except for some cases located in urban/industrial areas, such as in China, or along coastlines, such as in Italy and New Zealand. Nevertheless, even in clean areas, the electrostatic effects of a DC line will attract airborne particles to form a higher pollution level than an AC line in the same location. The problem for higher contaminated areas is even more critical. Polymeric insulators are sometimes considered in these conditions, mostly in China. In the rest of the world, however, due to lack of consensus and standards describing maximum stress levels for silicone housings and seals for DC application, usage of these insulators is still limited. One solution for such applications is use of silicone coatings applied to the surface of traditional glass or porcelain insulators. For example, the new Ximeng-Taizhou-Shanghaimiao-Shandong 800 kV DC line uses several hundred thousand factory pre-coated toughened glass insulators. Similarly, Terna in Italy has been using such insulators on a 200 kV DC line in Tuscany and Sardinia (Figs. 7 and 8), thereby eliminating need for any washing for more than 5 years. Some areas, classified as ‘very heavy’, are shown in Fig. 7.
A comparative pollution test between pre-coated toughened glass insulators and polymeric insulators was performed in salt fog conditions for a coastal application. The results, summarized in Fig. 9, show the benefits of a classical glass string coated with RTV silicone material.
These results highlight the difficulty in arriving at some generalized performance statement without doing actual testing. Here, the two test objects were considered as hydrophobic materials and in both cases the respective leakage distances were equivalent during testing. But material and shape matter as well. It is also noteworthy that silicone-coatings so far have been done either with application of the silicone material on site or using an industrial factory pre-coated approach, e.g. as in the case of a new 800 kV DC line in China (Fig. 10) where nearly 400,000 pre-coated insulators have been used in all tension applications. The example in Fig. 9 suggests that the theoretical model needs more work and modification for hydrophobic materials (i.e. HTM) – especially if coated conventional insulators are being used instead of polymeric types. As in AC, shape matters and assumptions made for HTM in the theoretical model are made mostly for polymeric insulator shapes, not for cap & pin insulators.
Conclusions & Future Directions
The spread of results and inconsistencies between laboratory tests, field experience and theoretical models demonstrate the need to cary out actual design tests for any new project and not rely solely on equations as found in various documents used today. Fig. 10, for example, gives an idea of the possible margin of error in theoretical evaluation of RUSCD versus laboratory testing or field performance. The graph needs to be further adjusted by more test results to fine tune the reference to an RUSC. The types of errors possible with a mathematical model can lead to possible overdesign of 20% to 30% or more, with consequent huge added cost that can even make design of a new line unacceptable to a utility. Large rounds of pollution tests in DC are required to better document actual performance. Similarly, most laboratories carry out tests with a CUR = 1 while, in reality, CUR is often between 3 and 5 or even as high as 7. There is also a need to define new contaminant deposit methods so as to duplicate, with consistency, such variable CUR levels during testing. This will help better simulate actual field conditions without relying only on theoretical correction factors.