New experimental study on the DC flashover voltage of polymer insulators: combined effect of surface charges and air humidity

: This research provides conclusions that can partially cover the lack of knowledge related to the effects of ambient atmospheric conditions on DC electric power transportation. Two polymer insulating materials are used for completing this research study. Inside a climate chamber, the relative humidity is controlled and adjusted from 20% to 80% RH. Then, the samples are charged with positive or negative charges by applying ±20kV to the corona-generating ring of needles. The surface potential is measured using an electrostatic voltmeter and is converted into surface charge density later by applying the probe response matrix method. The pre-charged samples are then stressed with high-voltage negative or positive DC values inside the climate chamber over a range of controlled values of the surrounding air humidity until flashover takes place. The space charges, which can drift in the air gap to reach the solid surface, are highly affected by the level of the relative humidity of the surrounding air. Also, increasing humidity results in a reduction of the DC flashover inception voltage and a shorter time to flashover regardless of the voltage polarity. Moreover, the positive or negative flashover inception voltage of both materials increases with pre-deposited negative charges while it decreases with pre-deposited positive charges.


Introduction
High voltage (HV) transmission lines transfer electric energy to domestic areas by cutting through hundreds of kilometers of arid, coastal and sometimes high-altitude areas. In fact, the insulation system carrying those transmission lines is highly affected by various environmental conditions such as humidity, air pressure and pollution, which may cause insulation breakdown and hence transmission outages. Keeping continuity of electric power transportation in such big systems is one of the main targets of electric power utilities for maintaining transmission security and avoiding revenue losses. Given this, many investigations have been completed on different scales, in recent decades, to provide a thorough understanding of the behaviour of HV insulators under different environmental operating conditions. However, there are some knowledge gaps that need to be filled by conducting further research into the nature of HVDC technology, which is different from HVAC, and about the corona-generated space charges that accumulate on the surface of insulators. Charge build-up on the insulation surface due to corona discharge from HVDC energised lines can further change the components of the electric field and influence the insulation performance under steady-state operating conditions. Therefore, priority has been given to study the flashover characteristics along the surface of pre-charged insulators under humid conditions. The surface charge distribution, due to DC corona discharges, along the surface of cylindrical samples of polymer insulators was previously studied by the authors in [1]. The authors investigated the effect of changing both the corona-charging voltage polarity and magnitude as well as the charging time on the profile, polarity and magnitude of space charges deposited on the insulation surface under room temperature conditions. The total amount of deposited charges increases as the magnitude of the charging voltage increases. Moreover, the amount of negative charges deposited on the surface during a negative charging process is found to be more than that of the positive charges deposited during a positivecharging process. Similar results have been reported in recent studies [2,3]. Moreover, the negatively charged area is found to be larger than the positively charged one which is exhibited due to the polarity effect.
The effect of pre-deposited charges on the negative DC flashover voltage of short samples of polymer insulators has been recently studied in [4]. The results show the dependency of the negative flashover voltage on the polarity of the charges. Moreover, the effect of the magnitude and polarity of the predeposited charges on both negative and positive flashover voltages has been introduced in different studies [5][6][7]. It has been found that the positive charges reduce the flashover voltage while the negative ones increase it regardless of the applied voltage polarity. However, the dc flashover voltage of fluorinated epoxy-based insulators decreases when hetero-charges are accumulated on the sample's surface while remaining unchanged with homo-charges [8]. In a recent study [9], the authors realised that whether under DC or AC voltage, the surface flashover voltage of the insulator decreases as the average charge accumulated on the surface increases. Also, the previous studies done by Manitoba Hydro [10,11] showed that the negative DC flashover voltage of a full-length fibreglass-reinforced plastic (FRP) hot-stick was reduced to a voltage level lower than the system's operating voltage.
Most of the previous studies in the literature have been carried out at room temperature and pressure conditions without any concern about the surrounding air humidity. Earlier researches have considered the effect of humidity only and studied the flashover test in the presence of dielectrics under AC and impulse voltages [12], as well as the effect of the absorption of contaminants by the insulator surface on the flashover voltage [13]. Some researchers have included the surrounding air humidity effect on the impulse flashover of pre-charged insulation samples [14,15] as well as the surface charge-decay characteristics of insulating materials [16,17]. Still, there is a need to determine the behaviour of insulating materials when stressed with HVDC under different ambient conditions. This paper is unique in providing a thorough understanding of the effects of surrounding air humidity on the performance of insulating materials. It focuses on determining and analysing the physics and characteristics of space charges deposited on the surface of insulating materials under humid conditions. Thereafter, the flashover performance of such precharged insulation samples at different values of surrounding air humidity is investigated and thoroughly analysed.

Surface charge measurement in humid air
The circuit diagram of the experimental set-up used for measuring the surface potential distribution on insulating materials in humid air is shown in Fig. 1. The experimental facilities, similar to that used in [18], were set up inside a climate chamber that is made of acrylonitrile butadiene styrene pipes and transparent plastic sheets. This allows us to conduct the experiments under a controlled level of relative humidity in a range of 20% up to 80%. This range covers the typical variation of the atmospheric relative humidity through most of the year, which may affect the on-site performance of HV insulators and/or the live-line maintenance work procedures. The dimensions of the humidity chamber are 210 cm (length), 130 cm (width) and 210 cm (height). Therefore, the inside space is sufficient for installing the XY table, the wooden frame supporting the sample and the humidifier, as shown in the circuit diagram of Fig. 1.
Two insulating materials are used to investigate the relation between the characteristics of deposited surface charges and the relative humidity of the surrounding air. One of those materials is FRP material, which is used to manufacture hot sticks used in liveline maintenance procedures. The other insulating material is fibreglass-reinforced epoxy covered with a sheath of hightemperature vulcanised silicone rubber (HTV-SiR), which is used for outdoor HV insulation purposes. FRP samples of two different lengths are used: 130 mm short and 250 mm long, while HTV-SiR samples of only short lengths are used, as shown in Fig. 2a. All the samples are cylindrical in shape and have standard diameters of 32 and 30 mm with relative permittivity of 4 and 3 for FRP and HTV-SiR, respectively. The samples are installed between two metallic electrodes mounted on a wooden frame. A non-contact corona discharge source is used for creating space charges in the surrounding air which is referred to as the corona-generating ring of needles (CGRN). It consists of a three-dimensional printed polyethylene frame on which 27 connected needles are mounted. The needles are distributed around the frame periphery with a 10 mm spacing between two adjacent needles (see Fig. 2b). When energised, the CGRN generates an electric field in the vicinity of the needles. A sample of electric field distribution on the surface of the sample when the CGRN is at −20 kV can be determined using numerical simulations (see Fig. 3).
The humidity-and temperature-controlling circuit is designed and implemented in-house using a commercial Arduino Uno microcontroller, 16A relay and a humidity and temperature sensor (DHT22). This system allows a range of 0-100% RH with an accuracy of ±2% for the relative humidity and −40 to 80°C with an accuracy of ±0.5°C for the temperature.
Before starting the test procedure, the humidity level is first raised to the required level, and the residual charges, which are deposited on the sample's surface from the previous test, are neutralised. After that, the surface is cleaned using isopropyl alcohol. The charging procedure starts when an equilibrium condition of the required relative humidity percentage is reached (i.e. the reference value ±0.5%). The CGRN is mounted symmetrically around the sample at the centre point and the gap between each needle tip and the surface underneath is maintained at 5 mm during the charging process. The CGRN is energised at +20 kV DC for 2 min. Afterwards, the voltage source is turned off and grounded and the CGRN is carefully removed while the electrostatic voltage probe is brought to the starting measuring point in about 30-45 s. The scanning span of the short samples contains 19 measuring points and 43 for long samples with a 5 mm separation between adjacent points. Every measurement is repeated at least five times under the same operating conditions (i.e. charging voltage and humidity) to achieve the measurement consistency. Finally, the measured values of surface potential distribution are converted into surface charge density distribution by applying the probe-response matrix method [18,19]. To find the elements of this matrix, a commercial finite-element method-based software (Comsol Multiphysics) is used. A detailed discussion of the modelling and numerical calculations can be found in [18]. The  Fig. 4 shows the circuit diagram of the experimental set-up used for studying the effect of relative humidity on the flashover performance of uncharged as well as pre-charged insulating materials. The detailed description of the DC source and circuit components can be found in [6]. A four-channel oscilloscope (Tektronix TDS3054) and a voltmeter (Tektronix DM255) are both connected to a mixed resistive-capacitive voltage divider through a switching box for double monitoring of the waveforms and values of the flashover voltage. Moreover, the leakage current is continuously monitored through a multimeter of minimum resolution of 0.1 μA, Fluke 116, which is connected in series with the ground return line. The short samples of FRP and HTV-SiR are used for the investigation. The two ends of the tested sample are equipped with sharp-edged copper rings of 7 mm width (the two samples at the top of Fig. 2a), which are similar to the end fitting used with live-line tools or for supporting HV insulators.

Flashover voltage measurement in humid air
The effect of the relative humidity of air on the flashover performance of uncharged as well as pre-charged insulating materials was experimentally investigated. To do so, the charging procedure, described in Section 2.1, has to be followed before each individual test of the pre-charged insulating materials. However, for testing uncharged samples, the same procedure is followed without the need of charging the sample's surface. During the study, the relative humidity of air is varied from 40% up to 80%. Before starting the test, the residual charges should be neutralised and the surface of insulator must be cleaned using isopropyl alcohol. Once the humidity inside the chamber reaches the desired level, the CGRN is mounted symmetrically around the sample and energised by ±20 kV for 2 min. After completing the charging process, the DC source is turned off and grounded before removing the CGRN. Then, the DC source is connected to one of the electrodes of the set-up while the other electrode is kept grounded. The DC source is turned on and the voltage is increased by a rate of 2 or 2.5 kV/s until flashover occurs for the FRP or HTV-SiR samples, respectively. This rate corresponds to ∼2% of the expected flashover voltage of the sample, which is chosen based on the recommendations given by IEC 60060-1 [20]. The test procedure is repeated n times, n ≥ 10, to find the average flashover voltage and its standard deviation. Ideally, for each trial, a new sample should be used. However, due to large number of flashover tests, it is difficult to use a new sample for every single test. As such, two sets of ten samples each are prepared for the flashover tests. One set is employed for the test and the other set is left for the recovery of its surface conditions. This procedure is known as the progressive stress test, according to IEEE Std. 4-2013 [21]. During all experiments, the temperature inside the climate chamber is maintained between 22.5 and 24.5°C.

Effect of humidity on surface charge characteristics
It is important to begin discussing some features of humidity and their impact on corona discharge in air before moving on to the result analysis. The relative humidity of air ranges between 0% and 100%, which reflects the percentage of water molecules in the surrounding air relative to saturation. Therefore, the low values are an indication of air dryness while the high values reflect the air dampness. The nature of water vapour in the air is well recognised as an electronegative gas which greatly affects the corona discharges as well as the ionisation process in humid air [22,23]. At higher humidity values, the number of free electrons initiating and sustaining partial discharge activity is reduced. Also, the formation of a stable ion cluster, which happens when water molecules capture electrons, reduces the effectiveness of the electronic avalanche process. Moreover, electrons cannot detach these clusters during molecular collisions [24]. As such, the mobility of these clusters of ions is reduced and in turn changes the local electric field distribution. Also, water molecules absorb the irradiated photons and weaken photoionisation in the air [25]. The presence of humidity enhances the conductivity on insulation surfaces, which can cause a faster local charge decay of electrons  that are diffusing over a large area [22]. All these humidity effects will cause a general decrease in the corona discharge activity in air and will affect the amount as well as the distribution of deposited charges over the nearby surfaces [23]. The effect of the relative humidity of the surrounding air on the characteristics of deposited charges along the surface of insulating materials is investigated and analysed. The importance of this research comes from providing findings that can fill in the lack of knowledge related to the effect of the operating atmospheric conditions on electric power transportation. The experiments are carried out on short samples of FRP and HTV-SiR as well as long samples of FRP. Only, the results of short samples of FRP will be shown and discussed. All the measurements are performed at each level of relative humidity after charging the sample's surface by corona discharges generated by applying −20 or +20 kV to the CGRN for 2 min. Seven values of the relative humidity of the surrounding air, ranging from 20 to 80% RH, are used to reflect possible, realistic operating conditions. Fig. 5 shows the effect of changing the relative humidity of the surrounding air on the potential and charge density distributions over the surface of the negatively charged FRP sample. The relative humidity values are grouped into three categories for simplicity of result analysis. These categories are low humidity (20-40%), medium humidity (50-60%) and high humidity (70-80%). At low humidity, the surface potential distribution is bellshaped with decreasing magnitude in the middle as the humidity increases, as shown in Fig. 5a. For medium and high humidity, the bell shape of the surface potential distribution no longer exists. However, the distribution uniformly covers the whole surface with a small hump at the middle. The measured values of the potential distribution dramatically decrease as the relative humidity increases from medium to high values.
On the other hand, Fig. 5b clearly shows a high consistency between the calculated surface charge density distributions and their corresponding surface potential distributions. At low humidity, there is an accumulation of small amounts of positive charge in the regions close to the electrodes, compared to negative charges in the middle. This can be related to the induction of image charges on the electrodes to compensate for those present in the middle of the insulator to keep the potential equal to zero. In contrast, high values of negative charges can be noticed at the regions close to the electrodes, for medium and high humidity levels, to compensate for the high values of surface potential measured beside the grounded electrodes. However, except for low humidity levels, the surface charge distribution along the entire surface of the sample tends to be uniform with decreasing values as the humidity increases from medium to high levels.
The effect of relative humidity on the surface potential and charge distributions of a positively charged sample is shown in Fig. 6. Similar to the negatively charged case, the bell shape of both surface potential and charge distributions is dominant at low humidity levels. However, the width of the bell shape seems to be narrower than the negatively charged case. The reason can be correlated to the mobility and diffusibility characteristics of both negative and positive ions. Although the mobility of a charged species decreases nonlinearly as the relative humidity increases, negative ions continue to have higher mobility than positive ions. Therefore, the drift time taken for both species to migrate in the air gap non-linearly increases with humidity yet negative ions consume less time than positive ions [26]. Thus, for the same charging time, negative ions can spread over a larger area on the surface as compared to the positive ions. For medium humidity levels, the surface potential measurement starts with high values, then gradually decreases after passing through the middle of the measurement span until reaching the other end side with values reduced by 65% of the starting point measurement. This is also reflected on the surface charge density distribution as the high values of charge density near the left side electrode compensate for the high values of the measured potentials on that side. Moreover, the surface potential values drop quickly as the humidity levels become higher than 60% and the surface charge distribution seems uniform covering the whole surface with a smaller concave up at the middle. Once the effect of relative humidity on surface potential and charge density distributions is determined, it becomes very important to quantitatively calculate how much charge is deposited on the surface during the charging process and their relation to the relative humidity level. The total amount of homo-charges, which have the same polarity as the charging voltage, is calculated using the surface integral of the obtained charge densities. Fig. 7a shows the relationship between the relative humidity and the total amount of homo-charges deposited on the surface of short FRP samples. In general, the two curves have similar characteristics that follow the same pattern. At low humidity levels, the total charge magnitude decreases with a slow rate as the relative humidity increases from 20 to 40%. Afterwards, it increases with a high increasing rate until it reaches 50%. At medium humidity levels, the total charge magnitude continuously rises until it reaches its maximum at 60%. However, it decreases again at a high rate as the humidity rises from medium to high levels and records its minimum value at 80%.
The effect of humidity is to decrease the mobility and the total number of charged species in the air gap [22,26]. Therefore, the authors could expect a linear decreasing relationship between the relative humidity of the surrounding air and the total number of deposited charges on the surface of the FRP sample. In Fig. 7a, the relationship is no longer linear and an unexpected increase of the deposited charges can be noticed at medium humidity levels for both positively and negatively charged case studies. The exact physical explanation for this is still unclear. However, the reason for that can be related to the induced charges which compensate for the high surface potentials measured near the grounded electrodes. These induced charges are the contributing factor in the increase of the total amount of deposited charges during the charging process. On the other hand, when the effect of the induced charges near the electrodes is eliminated by subtracting those values from the total amount of charges, the relationship becomes linear again. Therefore, longer FRP samples of double length, 250 mm, are used for the same test procedure to eliminate the effect of the grounded electrodes. Fig. 7b clearly indicates the linear relationship between the relative humidity of the surrounding air and the total amount of homo-charges deposited on the surface of the long FRP sample during the charging process. It is obvious that the amount of deposited homo-charges decreases as the relative humidity increases regardless of charge polarity. This relationship can be related to the electronegativity effect of humid air as explained in the first paragraph of Section 3.1. Moreover, a comparison between Figs. 7a and b reveals that the total amount of homo-charges deposited on the surface of a long FRP sample at different values of relative humidity is much lesser than that of the short FRP sample. For longer samples, the surface charge density distribution along the surface for both negatively and positively charged samples is shown in Fig. 8. The homo-charges are accumulated directly underneath the needles at the midpoint of the sample surface, covering a length of ∼1.5 cm along the sample surface. While hetero-charges, which have opposite polarity than that of the charging voltage, are widely spread, covering >90% of the surface, for shorter samples, homo-charges cover the whole surface of the sample with an incomparable amount of hetero-charges that only exist close to the grounded electrodes at low levels of relative humidity.
During the charging process, a HV is applied to the needles while the two electrodes are kept grounded. As such, the electric field strength in the air gap along the critical line between the two electrodes, which is a fictitious line parallel to the sample surface and 2 mm away from the surface, is affected by the sample length. The length of the sample alters the gap between the HV and the ground electrodes, which in turn changes the strength and distribution of the electric field. In general, the electric field strength is inversely proportional to the length of the gap separating the two electrodes. Also, the force exerted on a charged particle whether repulsive or attractive, due to the existing electric field, is directly proportional to the strength of that electric field. Moreover, the drift of the charged species in the inter-electrode region is field dependent and is characterised by the mobility of charged species, which is the ratio between the drift velocity and  the electric field. Therefore, the migration of the charged species in the inter-electrode region is highly affected by the direction and strength of the electric field. To calculate the electric field strength along the critical line between the two electrodes for short as well as long FRP samples, the Laplace equation ∇ 2 V = 0 was solved using a numerical finiteelement solver. Fig. 9a shows the electric field distribution for short and long FRP samples. One can see that the electric field strength is enhanced by ∼35% at the sample surface, directly underneath the tip of the needle, for the shorter sample. This reflects the effect of the separation gap length between the HV and ground electrodes on the electric field distribution. Near the grounded electrodes the electric field strength is reduced by ∼40% when the sample length is doubled. The distribution of charge density on the surface of short and long samples, which is shown in Fig. 9b, follows the corresponding pattern of the electric field distribution along the critical line between the two grounded electrodes.
For the HTV-SiR material, the effect of relative humidity on the total amount of homo-charges deposited on the sample surface, due to corona discharges of positively and negatively energised CGRN, is shown in Fig. 10. There are insignificant differences between the total homo-charge magnitudes deposited on the sample surface at the different relative humidity values. The magnitude of deposited charges is calculated corresponding to already distorted distributions of surface charges, which is a result of back discharges from the needles. These discharges neutralise partially the charges on the surface and results in a reduction of the charge density underneath the needles, which, in turn, forms the saddleshaped profile of the potential and charge distributions [2]. However, the relationship generally shows a slight decrease of charge magnitudes as the relative humidity of surrounding air increases. This is true except for values at 50 and 60% which shows a slight increase in the charge magnitudes over the values corresponding to low levels of relative humidity. Also, the magnitude of the total amount of homo-charges ranges between 15 and 20 nC for both negative and positive polarities. Moreover, the total amount of homo-charges deposited on the surface of the HTV-SiR material are higher than that deposited on the surface of the FRP material for both negatively and positively charged case studies.

Effect of humidity on flashover characteristics
The materials used to investigate the effect of humidity on the characteristics of DC insulation flashover are short samples of FRP and HTV-SiR. The flashover performance of uncharged samples is studied for both polarities of DC voltage at different humidity levels of the surrounding air. Moreover, the samples are then positively or negatively charged before conducting the flashover test to experimentally investigate the effect of pre-deposited charges on the DC insulation flashover performance under humid conditions. All the tests are carried out by following the standard progressive stress test method [21]. The atmospheric pressure during all the experiments is 101 kPa. Thus, no correction factors for air density are needed while studying the effect of humidity. The average temperature inside the climate chamber during the experimental period was 24.5°C. The absolute humidity varies between 9 and 18 g/m 3 , which corresponds to 40-80% of relative humidity at 24.5°C.
In general, the flashover voltage of an air gap increases as the humidity of the surrounding air increases, while it decreases when the air gap is bridged by insulating materials [14,15,[27][28][29]. Higher water vapour content (i.e. higher absolute humidity) affects the photon mean free path making it very short. As a result, the number of charged particles generated in the electron avalanche by photoelectron emission will be reduced and the streamer head part electric field strength will be weakened. In other words, a higher applied voltage is needed in humid air to reach a sufficiently high excitation energy state than in dry air. In addition, the electronegative water molecules in humid air adsorb more electrons from the streamer head to form negative ions, which prevent the development of the discharge near the streamer head in the air gap [30]. On the other hand, when the insulating material is bridging the air gap, the density of water molecules increases with increasing humidity in the surrounding air and forms a conductive wet layer on the surface of the insulator which in turn reduces its withstanding capabilities [22].
In the past two decades, there have been many studies concerning the effect of humidity and temperature on air breakdown under impulse as well as AC voltage stresses [14,15,28]. Some of these studies included insulating materials to detect the behaviour of solid insulation at higher humidity. However, there is a tendency to investigate the effect of air humidity on the characteristics of insulation flashover caused by overstressing with negative or positive DC voltages. Also, there are minimal research studies concerned with insulation flashovers, which have streamers that are propagating along the insulation surface not in the air gap, under different atmospheric conditions. Consequently, this study covers surface flashover of short insulation samples in humid air, in which the streamer is forced to propagate along the sample's surface and not in the air gap. For this purpose, two copper rings are installed at both ends of the sample while keeping in contact with the metallic electrodes. Fig. 11 shows the effect of the relative humidity of the surrounding air on the flashover performance of the FRP material. Increasing humidity reduces the flashover voltage and results in a shorter time to flashover regardless of the voltage polarity. Similar results are validated again by using different insulating material, HTV-SiR, as can be seen in Fig. 12. Indeed, the HTV-SiR material has improved flashover performance than the FRP material at different humidity levels. In general, the flashover voltage is inversely proportional to the intrinsic capacitance of the sample [31]. The relative permittivity of the HTV-SiR sample is lower than that of the FRP sample. Therefore, it is considered that the flashover voltage of the HTV-SiR sample becomes higher than that of the FRP sample. The silicone rubber material is characterised by its excellent insulating prosperities over other insulating materials. Moreover, it can recover its degraded surface hydrophobicity, caused by discharges, with time which enhances its withstanding characteristics. This property may predispose silicone rubber material to become the first choice over the other promising polymer-insulating materials for the purpose of HV insulators. However, it cannot replace the durable FRP material for live-line maintenance work, as silicone rubber materials are too vulnerable to mechanical damage if used in a hot-stick application.
The negative breakdown voltage of uncharged samples for both materials is slightly higher than the positive breakdown voltage. For negative polarity, after the applied voltage reaches its corona inception value, electron avalanches are generated close to the cathode, which forms the streamer's head and leaving positive space charges near the cathode. With a further increase in the applied voltage, more electron avalanches are produced at the same time, resulting in a dispersed plasma layer which in turn disperses the head of the streamer channel. As a result, the electric field is weakly enhanced at the head of the streamer channel and the development path becomes much more difficult under negative voltage. This weakness can be overcome by increasing the applied voltage to increase the applied external electric field. For positive polarity, the situation is different as the head of the streamer contains positive charges which strengthen the electric field of the streamer's head and maintains the development of the streamer channel without any dispersion. Therefore, the breakdown voltage of the negative polarity is higher than that of the positive polarity [30].
It is also noticed that the withstand characteristics of both materials increase when negative charges are deposited on their surfaces while decrease with positive charges. After the charging process, if positive charges accumulate on the sample surface, the electric field near the anode will be reduced while the external electric field will be slightly enhanced at the same time. In other words, positive charges reduce the electric field in the plasma channel and strengthen its head electric field. This strong electric field is continuously moving forward towards the cathode because of the positive charges at the front of the streamer's head. The increased electric field strength due to positive charge accumulation leads to insulation breakdown at voltage levels which are below the withstand characteristics of the uncharged sample. The effect of accumulated positive charges on insulation breakdown is analogous to a virtual reduction in the airgap between two electrodes which may break down at lower voltage levels. Therefore, positive charges reduce the breakdown voltage of the uncharged sample [30].
Moreover, the percentage of increase or decrease in the withstanding characteristics of insulating materials is directly proportional to the magnitude of the total amount of deposited charges on the sample surface. For example, the negative flashover voltage of FRP material is increased by 7.23% when a total of ∼20 nC negative charges are deposited on its surface at 50% relative humidity of the surrounding air, while at 80% of relative humidity, the FRP surface is only carrying a total of ∼6 nC negative charges which increases the sample's withstanding characteristics by 2.32%. Given that the higher the magnitude of the total amount of homo-charges deposited on the sample surface, the higher the

Conclusions
This research presents an experimental study that investigates the characteristics of space charge accumulation along insulation surfaces, due to corona discharges from energised objects, over a wide range of the surrounding air humidity values inside a climate chamber. In this research, space charges and the relative humidity of the surrounding air are considered together for studying their combined effect on the DC flashover characteristics of short samples of insulating materials. Short samples of FRP hot-sticks and HTV-SiR, which are used for live-line maintenance and HV insulation purposes, are employed in the study. After the charging process, the surface potential distribution is measured and then converted into surface charge density by using the ϕ-matrix method.
There is a high consistency between the calculated surface charge density distributions and their corresponding surface potential distributions at each humidity level, which reflects the accurate values of the numerically calculated probe-response matrix. Moreover, the polarity and magnitude of surface charges accumulated close to the grounded electrodes are humiditydependent, regardless of the polarity of the charging voltage. The bell shape characterises the surface charge density distribution at low-humidity values; however, at medium-and high-humidity values, the distribution uniformly covers the whole surface.
As negative ions have higher drift velocity than positive ions in humid air, their drift time to migrate in the air gap is less than that of positive ions. Therefore, for the same charging time, the total number of negative ions reaching the surface is more than positive ions and widely spread over the surface. Also, as the humidity increases, the number of water molecules increase in the surrounding air and capture more electrons from the medium and form ion clusters which in turn decreases their chance of reaching the sample surface during the charging process. Therefore, the total number of homo-charges deposited on the surface linearly decreases with increasing humidity.
In general, the breakdown voltage of the air gap decreases as the humidity of the surrounding air increases if the air gap is bridged by solid insulation. So, increasing the humidity reduces the flashover voltage and results in shorter time to flashover regardless of the voltage polarity. The negative breakdown voltage of uncharged samples for both materials is slightly higher than the positive breakdown voltage. It is also noticed that the withstand characteristics of both materials increase when negative charge is deposited on their surfaces and decrease with positive charge.

Acknowledgments
The financial support from Manitoba Hydro and the Natural Sciences and Engineering Research Council of Canada (NSERC) are acknowledged.