Dielectric and Thermal Conductivity Characteristics of Epoxy Resin-Impregnated H-BN/CNF-Modified Insulating Paper

High-voltage direct-current (HVDC) dry bushing capacitor-core insulation is composed of epoxy resin-impregnated insulating paper (RIP). To improve the thermal conductivity, breakdown strength, and space charge characteristics of RIP, 0.1 wt % nano-cellulose fiber (CNF)-modified RIP (CNF/RIP), 2.5–30 wt % hexagonal boron nitride (h-BN)-modified RIP (h-BN/RIP), and 2.5–30 wt % h-BN + 0.1 wt % CNF-modified RIP (h-BN + 0.1 wt % CNF/RIP) were prepared. Scanning electron microscopy (SEM) was implemented; the thermal conductivity, DC conductivity, DC breakdown strength, and space charge characteristics were tested. The maximum thermal conductivity of h-BN + 0.1 wt % CNF/RIP was 0.376 W/m.K with a h-BN content of 30 wt %. The thermal conductivity was 85.2% higher than that of unmodified RIP. The breakdown strength and charge suppression were the best in the case of 10 wt % h-BN + 0.1 wt % CNF/RIP. The maximum breakdown strength was 11.2% higher than that of unmodified RIP. These results can play a significant role in the research and development of insulation materials for HVDC dry bushing.


Introduction
To meet the demand for electricity, attempts are being made to realize high-voltage power transmission with large capacity. Direct-current (DC) transmission is better than alternating current (AC) transmission for asynchronous grid connection; furthermore, power loss is less in the former case. Specifically, for long-distance and under-sea transport, DC transmission is more advantageous than alternating AC transmission [1]. High-voltage direct-current (HVDC) bushing plays an important role in HVDC transmission. To meet the requirements of an oil-free valve hall, environmental friendliness, convenient transportation, a flexible installation angle, and avoidance of secondary injury caused by equipment failure, HVDC dry bushing is gradually replacing the traditional oil-impregnated paper insulation bushing in the HVDC transmission system [2]. The development of HVDC dry bushing is rapid. The capacitor core is the key component of HVDC dry bushing; it features a multi-electrode concentric capacitor structure composed of epoxy resin-impregnated paper (RIP) and aluminum foil. The RIP (for insulation) and aluminum foil play the role of homogenizing the electric field distribution [3].

Sample Preparation
Unbleached coniferous Kraft pulp was used as raw material for the insulating paper. Unmodified insulating paper was produced through the processes of pulping, dissociating, shaping, compressing, and drying. Nano-modified insulating paper was prepared by the Mechanical Co. ultrasonic mixing method. Epoxy resin, curing agent, and accelerator were included in the weight ratio of 100:85:0.3. Then, they were stirred under vacuum at 333.15 K for 15 min. This mixture was used to impregnate the unmodified and modified insulating paper under vacuum at 333.15 K. Lastly, the modified and unmodified RIP were obtained by staged curing of the impregnated insulating paper. The curing conditions are 353.15 K for 2 h, 373.15 K for 2 h, and 393.15 K for 4 h. The process of producing nano-h-BN + CNFs modified RIP is detailed in Figure 1.

Sample Preparation
Unbleached coniferous Kraft pulp was used as raw material for the insulating paper. Unmodified insulating paper was produced through the processes of pulping, dissociating, shaping, compressing, and drying. Nano-modified insulating paper was prepared by the Mechanical Co. ultrasonic mixing method. Epoxy resin, curing agent, and accelerator were included in the weight ratio of 100:85:0.3. Then, they were stirred under vacuum at 333.15 K for 15 min. This mixture was used to impregnate the unmodified and modified insulating paper under vacuum at 333.15 K. Lastly, the modified and unmodified RIP were obtained by staged curing of the impregnated insulating paper. The curing conditions are 353.15 K for 2 h, 373.15 K for 2 h, and 393.15 K for 4 h. The process of producing nanoh-BN + CNFs modified RIP is detailed in Figure 1.  Silane coupling agent KH-550 was used as a surface modifier of nano-h-BN, and polyethylene glycol (PEG) with a long chain structure was used as a dispersant [16]. The material parameters and equipment models used in Figure 1 are listed in Table 1. Silane coupling agent KH-550 was used as a surface modifier of nano-h-BN, and polyethylene glycol (PEG) with a long chain structure was used as a dispersant [16]. The material parameters and equipment models used in Figure 1 are listed in Table 1.

Measurement System
According to ASMT-E1530, a thermal conductivity test was performed using the thermal conductivity tester (DTC-300, TA Instruments, Newcastle, PA, USA) at 298.15 K [17]. The diameter of the test sample was 50 mm. The thickness of the sample was measured and a thermal paste was applied on both sides of the sample before testing.
A three-terminal electrode system was employed during the DC conductivity measurement. The sample surfaces were aluminized as electrodes. The electric field was 10 kV/mm, and the temperature was maintained at 298.15 K. The electrometer (6517A, Keithley Instruments, Inc., Cleveland, OH, USA) was used to record the stable current (I). The average conductivities samples and the standard deviation of 4 samples were used to evaluate the conductivity characteristics.
A DC breakdown strength test was conducted using a cylindrical electrode, in accordance with ASTM-D149. The temperature was maintained at 298.15 K. The entire test system was placed in transformer oil to avoid surface breakdown [18]. The average breakdown strength and standard deviation of 15 samples were used to evaluate the breakdown performance.
The pulsed electro-acoustic (PEA) technique was applied during the space charge test. The test field strength was 10 and 20 kV/mm, respectively, and a pulse field strength of 2 kV/mm was applied simultaneously. The test temperature was maintained at 298.15 K. The structure of the PEA test system is illustrated in Figure 2.

Measurement System
According to ASMT-E1530, a thermal conductivity test was performed using the thermal conductivity tester (DTC-300, TA Instruments, Newcastle, PA, USA) at 298.15 K [17]. The diameter of the test sample was 50 mm. The thickness of the sample was measured and a thermal paste was applied on both sides of the sample before testing.
A three-terminal electrode system was employed during the DC conductivity measurement. The sample surfaces were aluminized as electrodes. The electric field was 10 kV/mm, and the temperature was maintained at 298.15 K. The electrometer (6517A, Keithley Instruments, Inc., Cleveland, OH, USA) was used to record the stable current (I). The average conductivities samples and the standard deviation of 4 samples were used to evaluate the conductivity characteristics.
A DC breakdown strength test was conducted using a cylindrical electrode, in accordance with ASTM-D149. The temperature was maintained at 298.15 K. The entire test system was placed in transformer oil to avoid surface breakdown [18]. The average breakdown strength and standard deviation of 15 samples were used to evaluate the breakdown performance.
The pulsed electro-acoustic (PEA) technique was applied during the space charge test. The test field strength was 10 and 20 kV/mm, respectively, and a pulse field strength of 2 kV/mm was applied simultaneously. The test temperature was maintained at 298.15 K. The structure of the PEA test system is illustrated in Figure 2.  Figure 3a shows that nano-h-BN is stacked in layers. Figure 3b illustrates nano-h-BN sheets covered with CNFs. CNFs can prevent the agglomeration of h-BN to some extent. Figure 3d shows that there are many nanofibers between pulp cellulose and presents a good combination of CNFs and pulp cellulose. Figure 3e,f shows that CNF-modified h-BN dispersed more evenly in pulp cellulose. It is due to CNFs being able to serve as a "bridge" for strengthening the connection between pulp cellulose and nano-h-BN sheets. Moreover, CNF can separate h-BN from each other.  Figure 3a shows that nano-h-BN is stacked in layers. Figure 3b illustrates nano-h-BN sheets covered with CNFs. CNFs can prevent the agglomeration of h-BN to some extent. Figure 3d shows that there are many nanofibers between pulp cellulose and presents a good combination of CNFs and pulp cellulose. Figure 3e,f shows that CNF-modified h-BN dispersed more evenly in pulp cellulose. It is due to CNFs being able to serve as a "bridge" for strengthening the connection between pulp cellulose and nano-h-BN sheets. Moreover, CNF can separate h-BN from each other.

Thermal Conductivity Test Results of the RIP
The thermal conductivity and rate of rise in thermal conductivity of 0-30 wt % h-BN + 0.1 wt % CNF/RIP and 0-30 wt % h-BN/RIP are presented in Figure 4. When the h-BN content is the same, the thermal conductivity of 0.1 wt % CNFs/RIP is slightly higher than that of h-BN/RIP. The thermal conductivity of 0-30 wt % h-BN + 0.1 wt % CNF/RIP and 0-30 wt % h-BN/RIP increased from 0.203 W/m.K to 0.376 W/m.K and 0.362 W/m.K, respectively. The higher the content of h-BN, the higher the thermal conductivity. The rate of rise in thermal conductivity of 30 wt % h-BN + 0.1 wt % CNFs/RIP and 30 wt % h-BN/RIP reached 85.2% and 73.4% respectively.
The rate of rise in thermal conductivity, η, is given by where λm and λ0 are the thermal conductivities of m wt % h-BN + 0.1 wt % CNF/RIP (or m wt % h-

Thermal Conductivity Test Results of the RIP
The thermal conductivity and rate of rise in thermal conductivity of 0-30 wt % h-BN + 0.1 wt % CNF/RIP and 0-30 wt % h-BN/RIP are presented in Figure 4. When the h-BN content is the same, the thermal conductivity of 0.1 wt % CNFs/RIP is slightly higher than that of h-BN/RIP. The thermal conductivity of 0-30 wt % h-BN + 0.1 wt % CNF/RIP and 0-30 wt % h-BN/RIP increased from 0.203 W/m.K to 0.376 W/m.K and 0.362 W/m.K, respectively. The higher the content of h-BN, the higher the thermal conductivity. The rate of rise in thermal conductivity of 30 wt % h-BN + 0.1 wt % CNFs/RIP and 30 wt % h-BN/RIP reached 85.2% and 73.4% respectively.
The rate of rise in thermal conductivity, η, is given by where λ m and λ 0 are the thermal conductivities of m wt % h-BN + 0.1 wt % CNF/RIP (or m wt % h-BN/RIP) and unmodified RIP, respectively.
Polymers 2020, 12, x FOR PEER REVIEW 7 of 12 BN/RIP) and unmodified RIP, respectively.               when the contents of h-BN were the same. The breakdown strength of 10 wt % BN + 0.1 wt % CNF/RIP was the highest-11.2% higher than that of unmodified RIP.

Space Charge Characteristics of the RIP
The results of the space charge tests conducted on unmodified RIP, CNF/RIP, 10 wt % h-BN/RIP, and h-BN + 0.1 wt % CNF/RIP are presented in Figure 7. Figure 7 shows that with the increase of applied voltage time, the amount of space charge injection increases gradually, the depth deepens, and finally and gradually stabilizes. The figure indicates that more space charges are injected and the injection depth is deeper when the electric field strength is higher. Figure 7a,b indicate considerable space charge injection in the unmodified RIP and CNF/RIP. It can be concluded that only the addition of CNFs cannot inhibit space charge injection. Figure 7a,c-f, and g indicate that space charge injection in h-BN + 0.1 wt % CNF-modified RIP is less than that in the unmodified RIP. However, the amount of charge at the cathode has increased. Furthermore, the space charge injection in 10 wt % h-BN + 0.1 wt % CNF-modified RIP is very little. According to the results in our previous research, only adding h-BN can also suppress space charge injection [1]. Figure 7h shows that 10 wt % h-BN/RIP can suppress space charge injection to some extent. This proves that h-BN plays an important role in suppressing the space charge. However, it was observed that changes in specific mass fraction of h-BN + 0.1 wt % CNF had a more significant effect on space charge suppression than the addition of h-BN alone.

Space Charge Characteristics of the RIP
The results of the space charge tests conducted on unmodified RIP, CNF/RIP, 10 wt % h-BN/RIP, and h-BN + 0.1 wt % CNF/RIP are presented in Figure 7. Figure 7 shows that with the increase of applied voltage time, the amount of space charge injection increases gradually, the depth deepens, and finally and gradually stabilizes. The figure indicates that more space charges are injected and the injection depth is deeper when the electric field strength is higher. Figure 7a,b indicate considerable space charge injection in the unmodified RIP and CNF/RIP. It can be concluded that only the addition of CNFs cannot inhibit space charge injection. Figure 7a,c-g indicate that space charge injection in h-BN + 0.1 wt % CNF-modified RIP is less than that in the unmodified RIP. However, the amount of charge at the cathode has increased. Furthermore, the space charge injection in 10 wt % h-BN + 0.1 wt % CNF-modified RIP is very little. According to the results in our previous research, only adding h-BN can also suppress space charge injection [1]. Figure 7h shows that 10 wt % h-BN/RIP can suppress space charge injection to some extent. This proves that h-BN plays an important role in suppressing the space charge. However, it was observed that changes in specific mass fraction of h-BN + 0.1 wt % CNF had a more significant effect on space charge suppression than the addition of h-BN alone.
Polymers 2020, 12, x FOR PEER REVIEW 8 of 12 when the contents of h-BN were the same. The breakdown strength of 10 wt % BN + 0.1 wt % CNF/RIP was the highest-11.2% higher than that of unmodified RIP.

Space Charge Characteristics of the RIP
The results of the space charge tests conducted on unmodified RIP, CNF/RIP, 10 wt % h-BN/RIP, and h-BN + 0.1 wt % CNF/RIP are presented in Figure 7. Figure 7 shows that with the increase of applied voltage time, the amount of space charge injection increases gradually, the depth deepens, and finally and gradually stabilizes. The figure indicates that more space charges are injected and the injection depth is deeper when the electric field strength is higher. Figure 7a,b indicate considerable space charge injection in the unmodified RIP and CNF/RIP. It can be concluded that only the addition of CNFs cannot inhibit space charge injection. Figure 7a,c-f, and g indicate that space charge injection in h-BN + 0.1 wt % CNF-modified RIP is less than that in the unmodified RIP. However, the amount of charge at the cathode has increased. Furthermore, the space charge injection in 10 wt % h-BN + 0.1 wt % CNF-modified RIP is very little. According to the results in our previous research, only adding h-BN can also suppress space charge injection [1]. Figure 7h shows that 10 wt % h-BN/RIP can suppress space charge injection to some extent. This proves that h-BN plays an important role in suppressing the space charge. However, it was observed that changes in specific mass fraction of h-BN + 0.1 wt % CNF had a more significant effect on space charge suppression than the addition of h-BN alone.

Discussion
In this section, we shall discuss the thermal conductivities of unmodified RIP and CNF/RIP. The structures of CNFs and insulating paper cellulose are similar. Besides, the CNF content in CNF/RIP is very low. Therefore, adding CNFs alone did not lead to changes in the thermal conductivity of RIP; the thermal conductivity of h-BN + 0.1 wt % CNF/RIP was higher than that of the unmodified RIP. The thermal conductivity of h-BN is very high, and the h-BN in RIP can build a thermal network. Moreover, the higher the h-BN content, the more complete the established thermal network [1]. Therefore, thermal conductivity increases with an increase in the h-BN content. The thermal conductivity of 0.1 wt % CNFs+h-BN/RIP is higher than that of h-BN/RIP, when the h-BN content is the same. It may be due to CNFs improving the interface of h-BN and pulp cellulose, which reduced the interface thermal resistance.
The DC conductivities of unmodified RIP and CNF/RIP are almost the same because their structures are similar and CNF contents are low; the DC conductivity of h-BN/RIP is higher than that of unmodified RIP and decreases with an increase in the filler loading of h-BN. The significant difference between the sizes of h-BN and paper cellulose is a reason why the interface interaction is weak. h-BN affects the conductivity of h-BN/RIP by "scattering" [19]. The scattering effect is weak because the h-BN content is low; furthermore, the adsorption of h-BN on the pulp cellulose will reduce cross-linking points to affect cellulose cross-linking, which can make the movement of the cellulose molecular chain easier. The carrier is easy to transport, so its conductivity increases. As the h-BN content increases, the scattering increases, and the h-BN fills some voids between the pulp cellulose. The carrier transport is inhibited, the conductivity of h-BN/RIP decreases. Therefore, the conductivity of h-BN/RIP increases and then decreases as the filler loading of h-BN increases; the specific surface area and surface energy of CNFs are large, so that they can interact with h-BN. The structures of CNFs and paper cellulose are similar, which makes the CNFs easily combine with insulating paper cellulose. The CNFs can be used as a "bridge" between h-BN and insulating paper cellulose. They can help in combining h-BN and paper cellulose more effectively. In addition, coating h-BN with CNFs can play a supporting role and prevent the agglomeration of h-BN. More interfaces are introduced, and carrier migration is limited [20]. In addition, Figure 3e shows that there are many CNFs linking between the pulp cellulose and pulp cellulose. These CNFs will make the movement of the cellulose molecular chain difficult and fill the voids between the pulp cellulose. At the same time, the presence of CNFs enables h-BN not only to be adsorbed on the pulp cellulose, but also to be fixed between the pulp cellulose by CNFs. The decrease in free volume reduces the free travel of carriers and makes carrier transport difficult. Therefore, the conductivity of h-BN + 0.1 wt % CNF/RIP is lower than those of unmodified RIP and h-BN/RIP. When the filler loading of h-BN is up to 15 wt %, its conductivity increases owing to the weakening of the interface. When the filler loading of h-BN reaches 30 wt %, its conductivity decreases owing to the enhanced "scattering" of h-BN.
For the DC breakdown strength variations in unmodified, CNF/RIP, h-BN/RIP, and h-BN + 0.1 wt % CNF/RIP, the free path of the carriers affects the progress of breakdown. The longer the free path of the carrier, the higher the energy obtained in the process of carrier transmission, and the lower the breakdown strength [9]. The carrier-transport process can be reflected by its DC conductivity. The DC conductivity of h-BN/RIP is higher than those of unmodified RIP and h-BN + 0.1 wt % CNF/RIP, and the carrier transport is easier. As a result, the breakdown strength of h-BN/RIP is lower than those of unmodified RIP and h-BN + 0.1 wt % CNF/RIP. However, although the conductivity of 30 wt % h-BN/RIP is lower than that of 15 wt % h-BN/RIP, its breakdown field strength is lower than that of 15 wt % h-BN/RIP. Because the content of h-BN is very high, some defects are introduced [21]; because the conductivities of unmodified RIP and CNF/RIP are almost the same, the breakdown strengths of RIP and CNF/RIP are almost the same; due to the strong interface effect, the breakdown strength of 2.5-10 wt % h-BN + 0.1 wt % CNF/RIP increases as the h-BN content increases. When h-BN content ≥15 wt %, the interface effect becomes weaker. Therefore, the breakdown strength of 15-30 wt % h-BN + 0.1 wt % CNF/RIP decreases.
The space charge characteristics of unmodified RIP, CNF-modified RIP, and h-BN + 0.1 wt % CNF/RIP are affected by the interface and h-BN "scattering". According to our previous study, appropriate use of h-BN can achieve space charge suppression by "scattering" [19]. In addition, the interface effect also helps in suppressing the space charge. According to the results of DC conductivity and DC breakdown strength, the interface effect is the largest in the case of 10 wt % h-BN + 0.1 wt % CNF/RIP. Therefore, the effect of space charge suppression is the best in the case of 10 wt % h-BN 10 wt % h-BN + 0.1 wt % CNF/RIP. Furthermore, as h-BN continues to increase, the "scattering" becomes stronger. Therefore, 15-30 wt % h-BN + 0.1 wt % CNF/RIP can still achieve space charge suppression.
The 0.1 wt % CNFs+10wt % h-BN/RIP with the thermal conductivity of 0.29 W/m.K, the high breakdown strength of 374.75 kV/mm, the low conductivity of 6.47318ˆ− 16 S/m, and the good space charge suppression effect is a good choice for HVDC dry bushing insulation.

Conclusions
This study investigated the dielectric properties of h-BN/RIP, CNF/RIP, and h-BN + 0.1 wt % CNF/RIP. The following conclusions were drawn: (1) CNF/RIP and h-BN/RIP did not increase the breakdown strength of RIP; however, they did increase the breakdown strength of 2.5-10 wt % h-BN + 0.1 wt % CNF/RIP. h-BN and CNFs were synergetic. (2) The breakdown and space charge suppression performances were the best in the case of 10 wt % h-BN + 0.1 wt % CNF/RIP. It can be a potential insulating material in the manufacturing of HVDC dry bushing. (3) In comparison with unmodified RIP, the h-BN/RIP and h-BN + 0.1 wt % CNF/RIP can suppress the space charge but CNF/RIP cannot. h-BN plays a significant role in suppressing the space charge, and CNFs can enhance its role.