Diagnostics and Monitoring of Insulators for Power System

A. Pigini* (Fellow IEEE), A. Colombo** and M. de Nigris (Senior Member IEEE)** *CESI; Via Rubattino 54, Milan, I 20134 Italy ** CESI RICERCA; Via Rubattino 54, Milan, I 20134 Italy
Abstract - Tests under AC voltage and under overvoltages were carried out on cap and pin insulator strings and on composite insulator set in dry and clean condition, simulating defects of different size and location along the insulator length. On the basis of the results obtained the severity of the various defects is derived, allowing the assessment of the characteristics of the defects still compatible with live line maintenance procedure and with the required reliability of the line. As a consequence, indication about the minimum sensitivity required for diagnostics is given. The efficiency of the various proposed methods for the diagnosis of insulators were analysed and compared by systematic investigation in laboratory and in the field. Index Terms- Cap and pin, Composites, Diagnostics, Insulators

guidelines for field inspection were set up and applied in the field. II. DETERMINATION OF THE MINIMUM REQUIRED
SENSITIVITY OF DIAGNOSTIC METHODS

I. INTRODUCTION Insulators are one of the most numerous components in the electrical system and thus the assessment of their conditions is essential to ensure the reliability of the line and therefore the availability of the whole system. The assessment of their reliability is of particular interest to assure safety conditions in relation to live line maintenance. Insulators are characterised by a two-fold functions that are to be guaranteed during the entire service life: the mechanical and electrical function. This paper deals mainly with the assessment of the electrical function. The insulator dielectric performance may be impaired by failures or damages of part of the insulator set (e.g. failure of one or a few caps on cap and pin insulators, degradation of the insulation on composite insulators). The insulator performance may be also affected by environmental conditions, especially contamination. The verification of the adequacy of diagnostic methods in respect to the electrical performance requires: - the assessment of the critical defect size, that is the maximum defect size still compatible with the acceptable system reliability and with live line maintenance procedures - the verification of the sensitivity of the diagnostic methods with reference to the defects to be identified Results of the research activity are reported in the following. On the basis of the research carried out

The dielectric performance of insulator strings made of cap and pin units depends on the number of damaged units, on the degree of damage of the units, and on their position along the string. The extent of the defect may also be described by the length of the string with damaged insulators with respect to the total string length. The dielectric performances of composite insulators can be seriously affected when the external or internal surface of the housing is subjected to degradation leading to the formation of conductive or semiconductive paths that reduce the withstand voltage of the configuration. The maximum allowable defective length of the insulator string giving rise to a certain minimum acceptable dielectric performance of the system is defined as critical defect length. This type of defect represents the absolute minimum that any diagnostic method should be able to detect thus defining the requirements for the diagnostic sensitivity. In the present study the minimum acceptable dielectric performances considered were those relevant to live line maintenance operations: i.e. the critical defect length is considered as that giving rise to a 50% probability flashover voltage under positive switching impulse equal to that of the reference configuration with a gap length equal to the minimum approaching distance, as defined in the Standards [1] for the system voltage considered. Many data are available for cap and pin insulator , leading to the following relationship, interpolating the lowest values of dielectric strength experienced, as a function of the number of damaged insulators or length of the insulator set damaged [2] U50 = K * Cd * U50rp with: (1)

- U50, U50rp: 50 % flashover voltage of the examined configuration and of the reference rod plane one under switching impulse; K: gap factor of the configuration;

Cd = (1-0,8 Nd/Nt) = (1-0,8 ld/l )

(2) with

- Nd damaged units and Nt total units in the string; - ld length of the damaged insulator set and lt total
insulator length. An important observation is that the lowest strength is obtained with grouped damaged insulators near the line side, but not always at the line end. This fact underlines the importance to detect damaged insulators at line end, but also along the string (at floating potential). Less data are available for composite insulator sets. Systematic tests were carried out in CESI Laboratories to determine the dielectric performance of composite insulators (non ceramic line insulators NCLI in the following) in configurations simulating actual field conditions and different types of defects. Insulators for 145 and 420 kV lines were considered. External defects on the NCLI were simulated by means of thin (conductive or semi conductive) layers having 5 mm width and variable length applied along the insulator surface; internal artificial defects were made applying on the insulator rod, prior to the application of the housing, thin layers of varnish having the required conductivity and length. The results are given in Fig. 1 as a function of the length of the insulator set damaged. In the same Figure they are compared with data from (2):

Comparing the dielectric strength of typical 145 and 420 kV configurations evaluated with (1), (2) with that of the minimum approach distances from [1], the critical defect length may be evaluated and results of about 30% of the insulator length [3]. The detection of smaller defects (15-20 % of the insulator length) may be desirable, since if extended to a significant number of insulator sets, can significantly affect the reliability of the system, increasing the risk of failure in normal operating conditions [4]. III. DIAGNOSTIC METHODS EVALUATION IN LABORATORY The detection of defective insulators is quite easy for glass insulator strings with shattered insulators, for which an accurate visual inspection can be sufficient. It is much more complicated for porcelain insulators and especially for composite insulators [5]. Special attention in the following will be paid to the composite insulator case. Visual inspection is traditionally used also for these insulators. The availability of powerful visual aids and sophisticated image conditioning and interpretation tools have increased the potential use of such diagnostic techniques, especially for what concerns helicopter mounted inspections. However the environmental conditions (daylight, sun reverberation, background color, etc.) may introduce such high disturbances that incipient defects can hardly be detected effectively. The efficiency of three additional diagnostic methods based on electric field measurement, on the measurements of corona emission in the ultraviolet range and of temperature rise in the infrared range were analysed and compared by systematic investigation in laboratory. A. Electric field monitoring The Electric Field Distribution Measurement (EFDM) diagnostic system for line insulators is based on the assumption that any defect along the insulator is likely to cause a distortion in the electrical field distribution. Systematic laboratory tests were carried out with power frequency voltage, to evaluate the EFDM approach. The field probe was sled along the insulator under test by means of an insulating arm moved by precision crane; the field distribution was determined both on the onward and backward runs of the probe along the insulator. Tests carried out on the same type of insulators by different testing teams have proven that the repeatability and reproducibility of the method is well within 5%. To establish the method sensitivity similar dielectric configurations as those for the determination of the critical defect length were used. The determination of the presence of a defect in a NCLI tested according to this method was performed [3] by comparing the pattern obtained on the defective insulator with a reference fingerprint for the configuration

Fig. 1. Dielectric strength (Cd in p.u.) under positive switching impulse for composite insulators as a function of the length of the insulator set damaged. Comparison with (2).

For 145 kV insulators all types of external defects generated dielectric performance depending mainly on the defect length having little dependence on defect conductivity or position along the NCLI. For 420 kV insulators the influence of the damage type and position resulted more important. Only a few results obtained with fully conductive layers are slightly lower than those given by (2). Considering that the defects along the insulators are to be expected not fully conductive in service, (2) may be retained as a good interpolation of the most critical condition for composite insulators also.

considered, as shown in Figure 2. On the defective insulator a tracking on the housing close to the live terminal, affecting the first two sheds live side, was simulated. It is evident from the figure that, although the damage on the NCI is not very heavy (defect covering 7% of the insulator length), the change in the pattern is evident.

all critical defects and even defects much lower than the critical value may be identified. Beyond being very sensitive, the power of the method is its capability of giving indications about defect size an location and thus criticality.

Fig. 4. length. Fig.2. Comparison of measurements on insulator without and with defects (shed 1: ground side, shed 15 line side) .

Normalized Electric Field for ground terminal defect

To enhance the detection of defects the data were normalized dividing the electric field value measured for each shed by the corresponding value of the plot taken as reference (1 p.u), as shown in Fig.3. The result is obtained on 145 kV NCLI with a conductive external defect at live terminal: it can be noticed that a defect at live terminal produces high electric field distortions. Lower distortion values are produced by conductive defects at ground and floating potential (Fig. 4 and 5). Semi-conductive defects produced similar results. From analysis of the deviation pattern, as those in Fig. 3, 4 and 5, implemented by field calculations, indications on the location of the defect can be obtained.

Fig. 5. Normalized Electric Field for defects at floating potential.

Fig. 3.

Normalization of Fig. 2 data.

The maximum electric field deviation as a function of the defect length for conductive defects in various positions are reported in Figure 6. It can be noticed that

Fig. 6. Maximum Electric Field deviation (p.u.) as a function of defect length and position.

B- UVC method The possibility of localizing initial corona activity constitutes an interesting technical challenge, especially in daylight conditions. The diagnostic indicator considered is the emission generated by the defects in the UVC range (i.e. with wavelength in the range 240-280nm [6]) a bandwidth in which the solar light is filtered by the atmosphere. Corona emissions intensity is evaluated making reference to the number of pulses of light emission (named blobs). A counter gives a number proportional to the quantity of blobs received by the sensor. The first investigation aimed at checking the sensitivity of the camera with respect to the standardized procedures for corona inception evaluation: for this purpose a standard-type 420 kV porcelain insulator string was used.

Fig. 8.

UVC sensitivity as a function of the defect length.

This means that this method, for this kind of defects, is extremely sensitive, indicating the presence of defects well below the critical one. However the relationship between defect length and blobs/m is not always monotonous and the calibration of the method is not easy to be assessed. Thus the method can give only qualitative results and by itself can not furnish indications about the defect criticality. Furthermore the detection of the defect is possible only if the defect makes corona at service voltage, which is not the case for some defects at ground or floating potential. C. Infrared Thermography Measurement (ITM) Thermal emission is associated with pre discharge activity along the insulator and with significant current flowing along the insulator, as for instance in presence of tracking. With ITM the temperature distribution along the insulator axis is measured by means of an infrared camera, looking for hot spots, associated with possible local defects. Laboratory tests were carried out to evaluate the characteristics of the ITM diagnostic approach. The thermo-camera used had a wavelength of 8-12 m, a field of view of 5 to 20 degrees (depending on the lens used) and a resolution of 0.64 - 0.16 mrad respectively. A sensitivity analysis was carried out to check the effect on the accuracy of the emissivity assumption and of the ambient temperature. The investigation indicated that the measuring error can be limited if the measurements are performed establishing accurately the emissivity and avoiding atmospheric conditions critical for the measurements. Tests were carried out on insulators with simulated defects For the examined insulators, an emissivity coefficient of 0.95 was used .The results are reported in Fig. 9. The vertical axis shows the difference in K between the maximum temperature measured on the insulator (hot spot temperature) and the insulator temperature (Tamb).

Fig. 7.

Sensitivity of UVC vs. Visual Corona.

The standard laboratory procedure (Standard Test) for visual corona has evidenced the inception of corona activity around the ground electrode of the protective gap at 400 kV with extinction voltage at 385 kV. The curve in Fig.7 indicate that the sensitivity of UVC, with a gain of 160, is higher than that of the operator. An equivalence between the two methods can be obtained when reference is made to a blob rate of 500-800 blobs/min. The experience suggests a possible use of the methodology also for standard tests, thus removing the very critical human factor of such tests. Once established the sensitivity, measurements were performed on composite insulators with artificial defects. The corona emission level, in terms of blobs/min, is reported in figure 8 as a function of the defect length for defects live side. The data refer to tests at 100 kV. It is evident that even very small conductive defects, covering less than 4% of the total insulation, if at line potential, may produce noticeable emissions.

Fig. 9. Over temperature along the insulator as a function of the defect length. Fig. 10. Composite insulator with an age-related erosion close to the live electrode: the defect is not conductive.

As can be seen from Fig. 9, the deviation in temperature is quite limited and does not show a clear relation with the position or length of the defect. As a matter of fact thermography can detect strong corona, as from the tips of the defect, but also heating due to flowing of current along defective parts of the insulators (eg. along the simulated tracking, simulated by semiconductive paint). In this last case the heating can be related more to the resistivity of the path than to its length. Critical defects (30% of the insulator length) show deviations of few K. IV. COMPARISON OF THE METHODS AND DISCUSSION As an example, comparison of the different methods is made with reference to a silicon rubber line insulators with 27 sheds and 1150 mm length. The defect consisted of an ageing-related erosion on the core. The defect, in spite of its limited visible extension, is considered critical since exposing the fiber glass rod. In this particular case the damage was already clear from visual inspection. The defect was clearly identified by EFDM (maximum field deviation of 0,12 p.u.). Corona visualization of such a critical defect appeared only at the voltage of 100 kV. The use of infrared thermography revealed at 100 kV a localized over temperature of about 1.2 K over the average temperature of the insulator. A comment on the comparative sensitivity of the methods can be made starting from the example considered and from the analysis of the results reported in the previous paragraphs. The EFDM method demonstrates to be sensitive to the examined defect as to most of the defects, giving indications also when they do not cause thermal or UVC emissions.

Fig. 11. Corona visualisation at 100kV (AF mode, gain 80).

For the examined case, the UVC was at the limit of detection, since observable only at maximum service voltage. There is thus the risk of non detection of important defects. On the contrary even small defects, not significantly affecting the insulator performance, if of suitable shape and location, may give rise to UVC, thus affecting the selectivity of the method. Finally the thermal emission of the defect considered is quite low, confirming that the diagnosis may result critical and that an high accuracy is necessary when performing ITM measurements. The pre-discharge phenomena along the insulators may be significantly affected by atmospheric conditions, such as humidity, pressure presence of rain, and by the condition of the insulator surface (e.g. presence of a contamination layer).

V. CONCLUSION Live line maintenance requires the verification that critical conditions are not present along the line, and in particular that the dielectric withstand of the insulators along the line is higher than that applicable to the minimum approach distances. This implies that at least defects affecting more than 30% of the insulator length are to be identified. Visual inspection remains of primary importance to assess the insulator condition. EFDM results the most sensitive method for establishing the insulator condition. Unfortunately the method can be applied only within live line maintenance operations, and thus it finds an intrinsic limitation when critical insulator conditions are expected. UVC method is very sensible to defects causing corona, being sometimes too sensitive, since some of the defects detected can be far from critical. Some of the critical defects, e.g. long defects with high resistivity at floating potential, may not cause corona, but may produce heat. ITM is thus an important method to complete the information on the insulator condition in service. The measurement are to be made in dry condition and low humidity to optimize defects identification . The investigation performed and the field experience has allowed to set up guidelines for field inspection. Some of the critical defects unfortunately cannot be detected by any of the methods proposed (e.g. conditions close brittle fracture), pointing out the need of additional research. REFERENCES
[1] IEC 61472, Live working - Minimum approach distances for A.C. systems in the voltage range 72,5 kV to 800 kV - A method of calculation, July 2004. [2] CIGRE WG 33.07 Guidelines for insulation coordination in live working, CIGRE Brochure 151, 2000. [3] M.de Nigris, F. Tavano, F. Zagliani; R.Rendina, Diagnostic methods of non ceramic insulators for H.V. lines, CIGRE general Session 2000 paper22-207. [4] G. Marrone, E. Garbagnati; C.Valagussa, D. Perin, M. Ricca, R. Bonzano, Investigation on the dielectric strength of damaged insulator strings of HV overhead lines during repair operations by live working, CIGRE general session 1994, Paper 33-305. [5] CIGRE WG 22.03 Review of in service diagnostic testing of composite insulators, ELECTRA N 169, December 1996. [6] P. Lindner Inspection for corona and arcing with the Daycor camera, 2005 World Congress and Exhibition on insulators, arresters & bushings Hong Kong.

Fig. 12. ITM detection at high humidity level method on insulator without defects.

Fig. 13. UVC detection at high humidity level method on insulator without defects.

Examples of recordings on insulators without defects at humidity level higher than 95% are shown in Figs. 12 and 13. The UVC emissions was not stable in terms of location and time along the insulator: the number of blobs/min were also very variable. A similar behavior was observed with ITM, with maximum temperature deviation of about 5 K. The examples show that emissions of the same order of those caused by defects can be observed also on insulators without defects in conditions of high humidity, possibly in presence of contamination. Thus the campaign for defect detection are to be carried out avoiding conditions of high humidity. UVC and ITM may be used to have preliminary indication about the presence of pollution, but the analysis of the suitability of the methods to this purpose is outside the scope of the present paper.