Pollution Flashover Voltage of Transmission Line Insulators: Systematic Review of Experimental Works

Over the past decades, extensive experimental-based research works have been carried out to investigate the flashover phenomenon on the performance of polluted transmission line insulators. The critical focus has been on developing methods that can determine the safety, reliability, and sustainability of the overall power transmission network based on experimental results obtained from polluted insulators’ flashover voltage tests. In this paper, a systematic review of available scientific works, published as early as the 1990s, for the analysis of pollution flashover voltage, is undertaken. The review mainly focuses on factors influencing the efficiency of transmission line insulators under polluted conditions. Specifically, publication databases utilizing various synonyms and keywords associated with the terms “contaminated insulators” and “flashover voltage test” have been scrutinized. The search has resulted in 1364 articles, from which 97 articles have satisfied the review requirements and have been subsequently analyzed to determine the parameters associated with polluted insulators. Major factors that affect the performance of insulators, including electrical and environmental impacts, are discussed. Variations in factors affecting flashover test development and insulator efficiency are also considered. Overall, the current analysis provides an important insight toward successful evaluations of the health of transmission line insulators and research advancements of electric power transmission line insulators.


I. INTRODUCTION
With the rising demand for modern-day electricity, the topic of overhead power transmission lines has become important and prominent to ensure minimally interrupted electricity supply. A transmission line comprises several types of components with different functions, and one of the most important components is the insulator, as illustrated in Figure  1. Due to pollution and unstable weather in the outdoor environment, insulators may get damaged from time to time [1]. The damage or failure of insulators may arise from the flashover voltage, or in general, due to a combination of multiple damaged components, including fitting faults, which may result in power outage [2]. Once the flashover occurs in a power transmission line, it may lead to super regional blackouts and may even cause catastrophic accidents as suggested by [2]. Often, site engineers can provide actual insights into factors that influence the performance of insulators, as well as yielding guidance into the fitting characteristics of the transmission line insulators. It is therefore crucial to understand the flashover mechanisms of insulators based on the views and perspectives of engineers and researchers worldwide [3][4][5]. VOLUME XX, 2017 9 To date, various pollution flashover models, including static and dynamic models, have been pursued by many researchers in predicting the flashover voltage characteristics of polluted insulators [6][7][8][9][10][11][12][13]. For this, the collection of experimental data in identifying factors affecting the performance of polluted insulators are key elements to understanding the failure of polluted insulators. Moreover, in-depth exploration of experimental parameters in relation to the performance of polluted insulators is vital to gain an insight into problems faced by high voltage insulators.
Nevertheless, no systematic reviews have thus far been carried out with regard to factors affecting pollution flashover voltage tests. This paper therefore aims to systematically review experimental studies dealing with pollution flashover voltage tests conducted on high voltage insulators under polluted environmental conditions. Specifically, pollution flashover tests carried out by researchers between 1990 and 2021 have been reviewed and factors, limitations, and criteria relevant to pollution flashover characteristics have been critically discussed.

A. BACKGROUND OF INSULATOR POLLUTION FLASHOVER
Failure of insulators occurs due to flashovers or discharges. The main cause of insulator flashovers or discharges is the environmental conditions [12]. These include contamination, aging, and moisture. These conditions are subjected to insulators during the insulators' in-service lifetime, thus resulting in insulator failures and subsequent electrical grid outages [14]. To date, a significant amount of both theoretical and experimental works have been devoted to the study of flashover that occurs on polluted insulators. This huge amount of works resulted in the development of numerous models capable of predicting various characteristics related to the pollution flashover phenomenon. For example, the insulator shape, the pollution distribution layer and its resistivity, the heat exchange, and the presence of moisture have been correlated with pollution flashover to determine key factors affecting the pollution flashover phenomenon.
Significantly, the contamination of an insulator associated with the presence of moisture degrades the dielectric performance of the insulator [2]. The presence of moisture in addition to the pollution of insulator results in the dissolution of salt contaminants, leading to the formation of a conductive layer on the insulator [6]. The conductive layer, which is subjected to different values of voltages across the insulator, becomes an easy path for the leakage current to flow. This causes the conductive layer to heat up through the Joule effect, thus resulting in the formation of dry bands on the insulator. The potential difference, which initially appears between the line and the ground electrode of the insulator, will be at the limits with the presence of the dry band regions [12]. Due to high electric fields, a spark will occur, originating from the wet area where the voltage is high, above the dry bands, ionizing the surrounding air [8]. The spark will grow at high intensity and may propagate over the whole length of the polluted layer. This causes the passage of the current in the AC state and leads to the line short-circuit, changing the insulator to a conductor [15]. Of note, a flashover will take place whenever the electric field exceeds a threshold value, commonly known as the threshold voltage [15]. Figure 2 demonstrates the flashover process of a polluted insulator.

B. RESEARCH QUESTIONS
Although a huge number of insulator flashover tests have been carried out globally in the last two decades, a systematic review that summarizes the development of flashover tests on polluted insulators has yet to be available in the literature. Therefore, the current work intends to review available literature with regard to pollution flashover tests on polluted insulators. For this, the following research questions have been addressed.
i. What have been done in the studies of contaminated insulator flashover?
ii. What are the techniques, parameters, methods, conditions, and insulator samples used in pollution flashover tests? iii. What are the crucial factors affecting the flashover voltage due to the pollution of high voltage insulators?

II. RESEARCH METHODOLOGY
The research methodology of this study consisted of four main stages as demonstrated in Figure 3. The first stage was meant for incorporating the right strategy for searching the literature and setting the criteria for the literature to be included in the current review. This was needed to complement extensive theoretical questions implemented on any existing research, which were proposed to include adequate search questions and answers from the related scientific literature. The second stage was intended for the selection of the literature, and it comprised data classification and data extrapolation. This information processing operation constituted of data collation accompanied by data definition. Then the extraction and evaluation of the data was carried out in the third stage by applying accurate estimation criteria. Finally, in the fourth stage, synthesis of data was carried out, in which a step-by-step analysis of data was performed to deliver a satisfactory conclusion of the selected study characteristics.

1) INCLUSION AND EXCLUSION CRITERIA
A systematic review of all peer-reviewed and published papers relevant to the study of pollution flashover of polluted insulators were finalized for exploration. The articles that were written in the English language on the tested pollution flashover during the last four decades from 1980 to 2020 were reviewed. Accurate criteria for the addition were established to include relevant articles and exclude articles that were not related to the study of pollution flashover. Therefore, only peer-reviewed papers that concentrate essentially on pollution flashover studies were taken into consideration. For this, articles that presented the results from laboratory tests, field tests, static models, dynamic models, analytical studies, numerical models, mathematical models, statistical analysis, and prediction models on flashover of polluted insulators were included. Papers selected from references of relevant papers were also included. Finally, all papers were compiled. The inclusion and exclusion criteria are summarized and tabulated as in Table 1. 2) SEARCH TERMS analytical) and (pollution* OR polluted OR contamination* OR contaminated) AND (insulator* OR dielectric*). Figure  4 shows the bibliometric analysis of 97 articles issued on the pollution flashover of insulators according to the Scopus database using particular keywords such as "pollution flashover" AND ''insulators".

3) RESOURCES
Six research databases were used as sources for this systematic review, namely, Web of Science (WoS), Scopus, Science Direct, IEEE Xplore, Springer, and Google Scholar. Various existing papers were also searched for their title, abstract, and keywords and were employed to operate more search terms to locate published journals or papers, conference proceedings, workshops, symposiums, books chapters, and IEEE bulletins.

1) LITERATURE SEARCH PROCESS AND SCRUTINY
The search process steps advocated by Khan et al. [16] were adopted in this work to choose the relevant papers. Figure 5 shows the steps used to classify papers that are relevant to the pollution flashover topic. First, comprehensive searches of six research databases or publishers (as mentioned above), as well as a database that relates to systematic reviews, were administered to select relevant articles.
Second, the journals that include papers relevant to studies on the flashover models and tests of polluted insulators were selected. Then, the reference lists of all the relevant papers that matched the conditions of inclusion were checked. Each list of references was searched for any additional citations that could point to new articles and collected. Finally, when the search process entered the saturation stage when searches did not add any extra studies, the filtering process began. The primary selected list of articles was filtered and examined to ensure relevance. The steps involved in this filtering have been summarized as follows: First, the titles were evaluated for pertinence, and the article's contents were scanned to ensure relevance to the subjects under examination. They were subjected to further assessment according to the following conditions: typed in English, journal rank, and conference indexing. Papers that discussed the same topic and the most recent were included.

C. EXTRACTION (ASSESSMENT OF DATA QUALITY)
The quality assessment was aimed to boost the reliability of the selected articles in this work and in ascertaining the eligibility and comprehensiveness of the results [17]. The papers selected were assessed based on quality by a scoring method to decide the reliability, significance, and relevance of the articles. The studies were judged using a set of 10 criteria as exhibited in Table 2. The obtained papers were of different types of studies. To evaluate their quality, the included articles were classified into four categories as suggested by [18]. The categories were: 1) Evaluation Research (ER) articles: These articles comprised the techniques that were developed and implemented, and a stringent evaluation was carried out on them. 2) Validation Research (VR) articles: These articles comprised studies that discussed the assessment of a technique or method by using some experiments, i.e., work conducted in laboratories. To our knowledge these studies were novel and were carried out in practice. 3) Solution Proposal (SP) articles: These papers provided the solution to new issues or an extension of studies concerning the techniques underused. The suitability and efficiency of the solution were illustrated by a descriptive scenario or case study or discourse. 4) Opinion Articles (OA): These articles reflected the points of view of the authors on the application of the case definition methodology, emphasizing their advantages and disadvantages. These forms of research were not dependent on the relevant studies and the research methodology. The quality of evaluated questions was used for each category as given in Table 2 and is proposed by [19], [20], [21], as well as [17]. Each question was assessed based on three potential responses: "Yes" (score = 1), "Partially" (score = 0.5) or "No" (score = 0). This was done by the help of StArt (State of the Art through Systematic Reviews) software [20]. Subsequently, the sum of answers scores would determine the quality of the relevant study. × × × -3 Is there a satisfactory report of the context (equipment function descripts, experiment setting, products used, and so on) in which the work was carried out? [21] × × - 4 Is the sample representative of the group to which the findings can be generalized? [19] × × - 5 Was the analysis of the results adequately accurate? [18] × × × - 6 Is there a debate about the research outcomes? [21] × - 7 Are the limitations of this work explicitly addressed? [21] × × - 8 Are the concepts learned exciting? [19] × -Is there enough analysis of relevant studies? [20] × × × -

D. EXECUTION OF DATA SYNTHESIS
The information gathered from the answers to the research problems were tabulated using Microsoft Excel spreadsheets. The information was extracted according to the data described in Table 3 from each of the total number of 88 articles involved in this review. In the choice method too, the StArt tool [22] was used to aid in data extraction.
Through the synthesis stage, the terms which described the study subject was normalized and the most common keywords was employed. Three taxonomies were built utilizing those terms: (1) the approaches applied on the flashover studies of polluted insulators (model, analysis and predict); (2) the contamination methods applied in pollution flashover studies; (3) the relationships between pollution flashover-related information.

A. STUDIES SELECTION
The selection process of studies was conducted as in Figure 6. The articles were collected from electronics database with the help of the keywords that were mentioned in Section II (search term). These papers were carried out strictly according to the requirements for inclusion and exclusion. 1364 articles were obtained, and their abstracts were reviewed. As shown in Figure 7, the obtained papers consisted of 567 articles from Scopus, 423 articles from IEEE Explore, 147 articles from Science Direct, 136 articles from Web of Science, 51 articles from Springer, 16 articles from Institute of Physics, 7 articles from references lists, and 19 articles from other sources. Out of 423 IEEE articles, there were 159 journals papers, 247 conferences papers and 11 magazines. The collected studies (1364 articles) were inserted and arranged into the StArt software to start the selection process. Moreover, 5 papers were included manually. Out of 1364 articles, 376 appeared to be duplicated studies and were excluded. After reviewing the abstract for 988 articles, 747 papers were rejected based on the lack of relevance to the noted study. Therefore, the remaining 244 papers were eligible for full text review and 97 met all eligibility criteria and were included in this systematic review process. Figure 8 depicts the processing of selection and extraction for relevant studies using Start software. Out of 241 full text articles, 144 articles were excluded because the flashover on polluted insulators was not the main concern of the papers or included only secondarily study or as a case study in a larger study. Figure 8 illustrates the selection process using StArt tool software. Based on the eligibility criteria, rejected papers were classified as low and very low levels while accepted papers were classified as high and very high levels. Accepted papers that achieved three or four of inclusion criteria were classified as very high level while accepted papers with less than three of inclusion criteria were VOLUME XX, 2017 9 classification as high level. Meanwhile, rejected papers that met all the exclusion criteria were classified as very low level; rejected papers that met less than four exclusion criteria were classified as low level. Figure 9 demonstrates the number of papers that met inclusion and exclusion criteria. As can be seen from Figure 9, most of the included papers were concerned with experimental studies for the pollution flashover of insulators. In addition, most papers were rejected because the objectives of those articles were improved in new articles under the same authorship members.

B. ELIGIBLE STUDIES CHARACTERISTICS
Ninety-seven eligible studies were published between 1990 and 2021, comprising 103 journal articles and 18 conference papers. The keywords of the included studies were analysed and plotted in word-cloud software as shown in Figure 10. Table 4 demonstrates the number of conference papers and journal publications by year of publication. The number of publications has increased dramatically since 2010. Of note, the number of review papers about pollution flashover of outdoor insulators appeared little throughout the last decade. Although there have been many studies on pollution flashover, comprehensive reviews of insulator pollution flashover have not been done.

TABLE 4. Temporal view of studies included in the review by the year of publication
Published year Journals papers Conferences papers Figure 11 shows the share of journals publishing papers on pollution flashover of high voltage insulators. From Figure 11, majority of the paper were published in IEEE Transactions on Dielectrics and Electrical Insulation (TDEI) (40 articles, 31.7%), Electric Power Systems Research (15 papers, 12%), International Journal of Electrical Power and Energy Systems (9 articles, 10.7%), and IEEE Transactions on Power Delivery (8 articles, 7%). The rest of the journals published less than 5 articles per journal.

C. Description and Key Findings
In this section, the fundamental information and the key findings of pollution flashover voltage of high voltage insulators are discussed. The pollution flashover of contaminated insulators was investigated using several parameters. According to data in Table 6, majority of the selected studies investigated pollution flashover based on the factor of equivalent salt deposit density (SDD), followed by non-soluble salt deposit density (NSDD), pollution distribution, humidity, dry band, coating, pressure, aging, polarity, wetting rate, and insulator shape, respectively. The impact and results of these parameters are discussed in this section. Flashover voltage, as one of the most important indicators of polluted insulator tests, has been studied frequently in 94 papers. Most researchers performed their experimental studies only on the flashover voltage and flashover voltage gradient of the polluted insulator. Meanwhile, some researchers used experimental tests to validate their proposed models. Accordingly, a good model would result in less error in predicting the flashover voltage [24], [33], [47], [56], [88], [121], [124], [129], [130]. The flashover voltage has therefore been the main parameter used to determine factors affecting insulation flashover. Figure 13 shows factors affecting the flashover voltage of insulators and the number of published studies for each factor.
Similarly, [31] studied the effect of SDD and fog-water conductivity on the contamination flashover voltage of three different types of insulators under non-uniform pollution. From the experiment data, the relationship between the flashover voltage, SDD, and fog water γ was established as [31]: where b is a factor implying the rate of fog water on the flashover voltage and e is a constant equals to 2.718. In [66], an experiment was carried out on three separate insulators (porcelain, glass, and polymer) to investigate the effect of contamination variations under salt fog (additional salt deposit density ASDD) on the flashover voltage. From [66], the relationship between ASDD in fog water and the pollution level SDD was represented as: where k denotes the coefficient describing the effect of SDD and fog water conductivity on ASDD, determined experimentally as 0.179, 0.191, and 0.230 for porcelain, glass, and composite insulators, respectively. Then, the flashover voltage was expressed as [66]: where the negative of characteristic exponent (-a) indicates the flashover voltage stress decreases with an increase in both ASDD and SDD. The authors in [58] tested the characteristics of the flashover voltage of polluted polymeric insulators under three different polluting methods, namely, Brushing Method (BM), Dipping Method (DM), and Spraying Method (SM). The study reported some variations in contamination flashover parameters, and that their effects on the flashover voltage were different. The relationship between the flashover voltage and pollution severity under these three methods of polluting was calculated using equation (1). The different pollution methods was noticeable with changes in coefficient A, but negligible in the characteristic exponent a. Meanwhile, [79] extracted the relationship between the flashover voltage and SDD with the changing numbers N of insulator units in an insulator string, as seen in Figure 14.
The authors of [63] also extracted the relationship between the flashover voltage and SDD under ring-shaped nonuniform pollution for porcelain insulator as follows [63]: where O and I are the outer and inner of the insulator surface area, respectively. Meanwhile, r:R represents the ratio of pollution diameter on the insulator surface to the whole insulator diameter.
became higher [49], [141], [144]. When comparing fan-shaped non-uniform pollution with fanshaped uniform pollution, the flashover voltage dropped with fan-shaped pollution. The increase of the degree of fan-shaped non-uniformity pollution (L/W) resulted in a decrease in the flashover voltage, as shown in Figure 16 (c). For ring-shaped non-uniform pollution, the flashover voltage increased References [45], [49], [143] and [154] reported that, compared to uniform contamination, the flashover voltage increased with longitudinal non-uniform contamination. Furthermore, the flashover voltage U50 increased as the longitudinal nonuniformity level (H/M or L/M) increased for all types of investigated insulators, as shown in Figure 16 (a). As shown in Figure 16 (b), with smaller mean value of the electrical conductivity of contamination on the insulator surface and higher surface of the insulator in (Top/bottom)-shaped, the flashover voltage as the level of I/O increased, and under some situations, it increased by approximately 36% compared to a uniform pollution condition, as shown in Figure 16 (d).
Meanwhile, the flashover voltage would initially increase and subsequently decrease as the radius (r) of extremely contaminated areas increased [63], [142].

3) AIR PRESSURE DISTRIBUTION IMPACT ON FLASHOVER VOLTAGE
The effect of air pressure on the flashover voltage were investigated by the references [34], [72], [73], and [120]. The results in [34] showed that the AC flashover voltage on polluted insulators decreased with the reduction of air pressure under a specific value of SDD. For example, for the cap-andpin type glass insulator with SDD of 0.03 mg/cm 2 , the flashover voltage decreased from 238.5 kV to 191.0 kV due to a decrease in air pressure from 98.6 kPa to 70.1 kPa. In addition, the results showed that the distinctive exponent n in equation (7) describing the effect of air pressure on contamination flashover voltage U f , which has an impact on the flashover voltage of contaminated insulators, is variable, and n value is related to the contamination degree and geometric structure of the insulator.
where U 0 is the flashover voltage at the normal air pressure P 0 , P is the experimental air pressure, and n is the exponent describing the effect of air pressure on U f . In the study of [72], the data of a polluted porcelain insulator string composed of 7 units of insulators indicated that the flashover voltage decreased remarkably reduced air pressure. As arc radius dropped from 89.9 kPa to 61.7 kPa, the flashover voltage decreased by 13.1 %. In addition, the arc radius of flashover under the effect of air pressure was considered. The results indicated that the arc radius was 1.5 mm and 3.5 mm corresponding to air pressure of 89.9 kPa and 61.6 kPa, respectively [72]. The electron density ne of the flashover arc channel at low pressure was also investigated experimentally by [73]. The results reported that, with increasing pressure, the electron density of the arc channel increased based on equation (8): where α is the index determined using the least-square method to be 0.58 and n o is electron density in atmospheric pressure. The effect of the flashover phenomenon of polluted porcelain insulators under low air pressure on arc levitation was also studied in [120]. According to the findings, higher contamination and lower air pressure resulted in more serious arc levitation.

4) DRY BAND IMPACT ON FLASHOVER VOLTAGE
The flashover voltage under the influence of dry band was discussed in references [36], [62], [70], [71], [74], [112], [132]. The flashover voltage of glass insulators under dry bands with five scenarios of pollution distribution were tested in [36]. The results showed that dry bands had a significant effect on the flashover voltage, in which the flashover voltage increased with a decrease in the dry band area [36]. Authors in [62] investigated the effect of the number of dry bands formed on the contaminated plate insulator surface on voltage distribution. It was discovered that increasing the number of dry bands resulted in less discontinuity in the voltage distribution along the insulator, which improved the voltage grading and had a significant impact on the insulator's withstand capability.
The propagation of flashover on polluted porcelain post insulators under dry bands was investigated in [70], where the findings indicated that the discharge characteristics was affected by the formation of dry bands. In addition, the study compared the flashover voltage values between rain and fog factors. It was reported that the flashover voltage in rain was obviously higher than that in fog. An increase in the number of dry bands on the insulator surface would lead to a reduction in flashover activities and, as a result, an increase in the flashover voltage [71]. As suggested by [74], the effectiveness of insulation of polluted insulators increased with increased dry band length. Meanwhile, the effect of the dry band position on the performance of plate insulators was investigated by the authors in [112] (Figure 17). From the results, the flashover voltage with the dry band placed in the middle of the insulator was the highest; the flashover voltage with the dry band placed near electrodes of the insulator was lower than that with the dry band placed in the middle of the insulator.

5) NSDD IMPACT ON FLASHOVER VOLTAGE
The influence of NSDD on the flashover voltage of polluted insulators was investigated in [59], [60], [64], [91]. The critical voltage (before flashover) of porcelain insulators that contained heavy NSDD had significant distortions as shown in Figure 18 [59]. Furthermore, the greater the NSDD and/or contamination width, the lower the flashover voltage, and the more extreme and long-lasting the discharges are [60]. The influence of NSDD on the flashover voltage gradient of three different insulators was studied in [64]. The flashover voltage gradient E L recorded from the experiment can be fitted to is fitted to equation (9) to determine the characteristic exponent c that characterizes the effect of NSDD on the flashover voltage [64] [91].
where L represents the insulator length in cm. According to the fitting results, the characteristic exponent c value for NSDD between 0.078 and 0.103 mg/cm 2 is within 0.12-0.14 for glass and porcelain insulators, and within 0.13 -0.16 for composite insulators [64]. According to the results in [91], the flashover voltage subsided as NSDD and ESDD (also known as SDD) increased. The effects of NSDD and ESDD on the flashover voltage are independent of one another, as seen in Figure 19, which depicts the combined effect of NSDD and ESDD on glass and porcelain insulators.

6) HUMIDITY IMPACT ON FLASHOVER VOLTAGE
Humidity has a significant influence on the flashover voltage. The influence of humidity on the flashover voltage of glass insulators under different contamination profiles were tested and reported in [36], [150], [153]. The results showed that increased humidity led to a decrease in the flashover voltage. For a heavy pollution case, for example, under 0.25 mg/cm 2 ESDD, increasing the humidity from 75% to 95% decreased the flashover voltage by 10 kV (32 % reduction) ( Figure 20). [27], [84] tested the flashover voltages of contaminated composite silicone rubber insulators (conventional vs. textured) and glass insulator string (3 units) under variable wetting rates. The findings revealed that the flashover voltage of the conventional insulator decreased from 24.9 kV to 22.8 kV when the wetting rate increased from 3 l/h to 8 l/h, whereas the flashover voltage of the textured type insulator decreased 6.9 kV with the same increase in wetting rate [84]. According to [27], increasing the wetting rate of contaminated insulators caused a decrease in the flashover voltage gradient, jeopardizing the insulator's dielectric properties. For the glass insulator string, under medium contamination (0.12 mg/cm 2 ), the flashover voltage gradient Uniform a b Flashover VOLUME XX, 2017 9 reduced by 31.5 % and 47.69 %, respectively, due to increased wetting rate from 2.5 l/h to 5 l/h and 7.5 l/h. This implies that greater wetting rates has a major impact on electrical insulator efficacy and flashover risk [27]. In [153], the flashover voltage under DC and AC fields were compared under the effect of humidity. The results indicated that the flashover was more obvious under AC field especially if the humidity exceeded 80%. Figure 21 shows the surface flashover under DC and AC fields.

7) INSULATOR SHAPE IMPACT ON FLASHOVER VOLTAGE
According to [37], [81] and [155], an important factor influencing the flashover voltage is the insulator shape.
Boudissa et al. [37] provided the results of the test that enabled the influence of porcelain insulators' geometry on the flashover voltage to be quantified. Non-uniform contamination methods were used to quantify the effect of insulator shape on the flashover voltage, where non-uniform insulator surface contamination was reported to reduce the flashover voltage; greater non-uniformity resulted in lower flashover voltages. This voltage decrease was demonstrated by a change in the length of the insulator.
Li et al. [81] tested the flashover voltage of four different insulators with different structures, i.e., Π-type glass insulator, plate-type glass insulator, CA-590EZ porcelain insulator, and CA-878EY porcelain insulator. Based on the findings, a double-arc method for calculating contamination flashover was developed. The correlation analysis showed that calculating the flashover voltage based on different insulator types can be an effective method in determining insulation flashover.

8) POLARITY IMPACT ON FLASHOVER VOLTAGE
Some studies considered the effect of voltage polarities on the flashover voltage [39], [88], [98]. From these studies, it was concluded that the flashover voltage of plate insulator was different under positive and negative polarities. The average flashover voltage discrepancies for the single arc were obtained to be 21% and 28% for positive and negative polarities, respectively [39]. According to [88], the critical voltage was larger when the supply was in the positive polarity rather than the negative polarity.

9) WETTING RATE IMPACT ON FLASHOVER VOLTAGE
The effect of wetting rate on the flashover voltage was studied in [27], [36], [135]. In the work of [17], the authors determined the relationship between the flashover voltage gradient and the leakage current index R hi extracted from experimental work. The results showed that the flashover voltage gradient decreased by 46.92% with increasing R hi by 2.5 under a wetting rate of 2.5 ± 0.1 l/h for a glass insulator string with 3 units of glass insulators. For a porcelain insulator string with 3 units of porcelain insulators, the flashover voltage gradient decreased by 48.32% under the same change in R hi . Meanwhile, the flashover voltage of the polymeric insulator under a high wetting rate is higher than that obtained under a low wetting rate [36]. The authors in [135] concluded that the relationship between the flashover voltage and wetting rate can be determined from equation (10): where W r is the wetting rate, C p is a constant of pollution, and β accounts for the effect of the wetting rate on the flashover voltage at a constant SDD. Figure 22 depicts the results from [135] in correlating the flashover voltage with the wetting rate of different insulators. For each insulator, the flashover voltage continued to decrease as the wetting rate increased, to the point where the flashover VOLUME XX, 2017 9 voltage was three times lower in cases of high wetting rates compared to low wetting rates [145]. Regarding insulators type, the flashover voltage was the lowest for the porcelain insulator.

10) COATING IMPACT ON FLASHOVER VOLTAGE
The flashover voltage performance of insulators under different coatings was investigated in [50], [67] [146 -150]. In [50], the critical flashover voltage was measured to assess the reliability of room temperature vulcanizing (RTV)-coated porcelain insulators under different contamination levels. Figure 23 depicts the achieved flashover voltages for the porcelain specimens, without and with RTV coating, at various pollution levels. Compared to the uncoated specimen, the critical flashover voltage of the RTV-coated insulator increased noticeably. Moreover, the reduction in voltages was clearly proportional to the severity of contamination. It can be noted that the RTV-coated insulator had a better performance compared to the uncoated insulator at all levels of pollution, particularly under medium and light pollution levels [50]. Literatures [146], [147] investigated the flashover voltage of a porcelain insulator with varied coating damages. As shown in Figure 24, the flashover voltage was the lowest for insulators with fan-shaped damage. When the damaged region gets bigger, the critical leakage current increased whereas the flashover voltage reduced.
Reference [150] also tested the influence of the dimension and location of the coating damage on the flashover voltage of RTV-coated insulators. Different shapes of coating damage on a plate insulator were investigated. Generally, the flashover voltage reduced as the area of damage increased, either longitudinally or laterally, as illustrated in the Figure 25. When there are many damages, the contamination flashover voltage can calculated by the minimum "effective path" distance, defined as [150]: where L ep123 and L ep14 is the "effective path" distances along l1-l2-l3 and l1-l4 respectively while L mep is the minimal "effective path" distance. According to the findings, the flashover voltage and the "effective path" distance had a linear relationship.

11) AGING IMPACT ON FLASHOVER VOLTAGE
The flashover voltage for four various types of unaged and aged polymeric insulators under AC voltage was discussed and studied in [43], [48] and [152]. Flashover voltage tests were performed in 0, 2, 4, and 6 weeks of aging duration. As shown in Figure 26, increasing the aging duration of the composite insulators reduced the flashover voltage and weakened the insulator's performance. According to Figure  26(c), with SDD equal to 0.05 mg/cm 2 , the flashover voltage of the insulators aged for 4 and 6 weeks decreased by around 3.56 kV (28%) and 12.15 kV (95%), respectively, as compared to the unaged insulator. This implies that the hydrophobicity of the polymer surface decreases with increasing aging duration. Aging therefore has a substantial negative impact on the resistance and flashover voltage of insulators.

IV. Remarks and recommendations
From the review, SDD is the main parameter used to determine the flashover voltage of insulators. The advantage of determining the flashover voltage based on SDD is that it enables the pollution severity to be mapped with appropriate insulator selection criteria. The use of SDD, when coupled with the measurement of conductivity, allows the quantification of the amount of pollution deposited on insulators, thus representing both contamination accumulation and wetness.
However, the determination of the flashover voltage from SDD measurements suffers from the drawback that the flashover voltage values do not always reflect precisely the insulation pollution condition. This is because the composition of soluble salts in natural pollution is complicated. For example, soluble salts in most polluted locations are dominated by calcium sulfate (CaSO4), a particularly difficult compound to dissolve in water. Since a huge amount of water is used in the SDD measurement approach, CaSO 4 dissolves and adds to the solution's conductivity. In reality, however, the amount of water on the surface of in-service insulators is little, typically equaling just around 1% of the water utilized during an SDD test. Very little CaSO 4 can therefore be dissolved in such a little amount of water. The presence of CaSO 4 does not, in practice, contribute to solution conductivity and hence the pollution flashover voltage.
Since artificial contamination employs NaCl to imitate soluble salts, the pollution flashover voltage of a naturally contaminated insulator is frequently substantially greater than that of an artificially polluted test sample, even under the same SDD. Furthermore, because soluble salt contents vary greatly by area, the pollution flashover voltage testing of natural samples under the same SDD value might result in a wide range of findings. Of note, the most crucial criterion for determining the suitability of a parameter in indicating pollution severity is the ability of the parameter to correlate well with the pollution flashover voltage, and this is sadly lacking from the SDD correlation of the pollution flashover voltage.
To address some of the shortcomings in the SDD measurement approach, researchers examined the chemical composition of soluble salts found in natural pollution deposits. The SDD value obtained using the conventional method was then normalized based on the percentage of monovalent and bivalent salts in the soluble layer. These adjusted SDD values, also known as 'effective SDD,' have a better correlation with the flashover voltage. Nevertheless, chemical examination of soluble salt compounds of inservice insulators is expensive, time consuming, and energy intensive. Therefore, conducting chemical analysis for each SDD measurement is impractical. From the review, the examination of pollution distribution can also provide a preliminary view of the behavior of insulators with regard to pollution deposition owing to wind and rain effects. This would assist in finding the distribution of pollution layers that further explains the natural state while analyzing the distribution of pollution experimentally. Of note, the humidity can be monitored during testing using sensors, and the humidity can be controlled to a desired level. Also, under the same test conditions, the influence of an insulator shape on flashover voltage may be determined by employing different insulators of the same material with varying diameters. Although the humidity and insulator shape parameters can also have an influence on the flashover voltage, their interpretation of the flashover voltage becomes meaningless in the absence of SDD data. Therefore, the use of the humidity and insulator shape parameters in addition to SDD would be preferable in obtaining better correlation of the parameters with the pollution flashover voltage.
While NSDD can have impact on the flashover voltage of insulators, most research employs kaolin as insoluble materials in determining NSSD. This may be one of the limitations in correlation the NSDD parameter with the flashover voltage, since majority of the non-soluble natural materials are sands and free carbon particles (soot). Therefore, the use of different compounds in NSDD formulation could lead to better correlations of NSDD with the flashover voltage. It should be noted that the rate of NSDD deposition is affected by various factors, including desert sandstorms.
Meanwhile, the investigation of the impact of coatings on insulators regardless of pollution level is one of the greatest studies that reflect the beneficial results of coatings in increasing insulator performance. This is due to the usage of the same coating material even in experimental research. However, these studies are not viable if the insulator's coating has been in place for more than 10 years, where exposure to sunlight and artificial lighting can have adverse effects on the useful life of insulator coatings.

V. Positive and negative factors influencing flashover voltage
Based on Figure 13, eleven parameters obtained from 94 papers were discussed in this review study. It was clear that five parameters (SDD, NSDD, aging, humidity, and insulator shape) had negative impacts on pollution flashover voltage while only two parameters (coating and pressure) have positive effects. Four other parameters (pollution distribution, dry band, polarity and wetting rate) can have positive or negative effects on the pollution flashover voltage depending on the condition of the insulator and the change of the parameter value. These parameters were extracted from eighty-one papers. The fact that some of the parameters have a negative impact on the flashover voltage indicates that increasing the value of the variables leads to a decrease in the flashover voltage on contaminated insulators [151]. On the contrary, with the value of parameters having positive impacts increases, the flashover voltage also increases. Figure 27 illustrates the type of influence for factors flashover voltage.

VI. Future research directions related to flashover voltage of polluted insulators
Although the systematic review method as recommended by [152] has been used in the current work, there is a need for additional comprehensive research on the flashover voltage characteristics of insulators to enable researchers to further explore parameters affecting the flashover voltage issues of insulators. Future research directions linked to the flashover voltage might include, but are not limited to, the following work: 1. Research studies of flashover voltage and other characteristics related to polluted insulators. 2. Critical review of studies pertaining to materials characteristics that can aid in the improvement of the performance of insulators. 3. Critical review of artificial intelligence-based optimization techniques in predicting the flashover voltage. 4. Investigation of the degradation of insulator surfaces on the flashover voltage. VOLUME XX, 2017 9

VII. CONCLUSION
The current work has examined peer-reviewed publications focusing on pollution flashover voltage tests on high voltage insulators using a systematic methodology. This has offered value-added knowledge about pollution flashover voltage studies carried out by different researchers around the world. Critical parameters that affect the pollution flashover voltage have been discussed by reviewing the literature published between 1990 and 2021. The impact of eleven parameters, i.e., SDD, NSDD, aging, humidity, and insulator shape, coating, pressure, pollution distribution, dry band, polarity, and wetting rate, on the flashover voltage has been mainly discussed. The emphasis has been placed on the importance of knowing the pollution flashover voltage of insulators as a critical component for the evaluation and detection of the condition of the insulator on transmission lines. Nevertheless, research challenges remain especially on how one can properly monitor insulator pollution and propose techniques for analyzing insulator conditions prior to catastrophic failures. While several prototype devices have been proposed to monitor the condition of insulators using various ways, including those with infrared thermal imaging technology, much work is needed to ensure reliable monitoring and analysis of insulator conditions.