Experimental Study on Glaze Icing Detection of 110 kV Composite Insulators Using Fiber Bragg Gratings.

Icing detection of composite insulators is essential for the security and stability of power grids. As conventional methods have met difficulties in harsh weather, a 110 kV composite insulator with embedded Fiber Bragg Gratings (FBGs) was proposed for detecting glaze icing in this paper. FBG temperature compensation sensors in ceramic tubes were adopted for simultaneous measurement of icicle loads and temperature. Then, temperature calibration experiments and simulated icicle load experiments were carried out to obtain temperature and icicle load characteristics of FBGs. The results showed that temperature sensitivities of FBG strain sensors and FBG temperature compensation sensors were 18.16 pm/°C, and 13.18 pm/°C, respectively. Besides, wavelength shifts were linearly related to icicle loads within the polar angle range of −60° to 60°, and the load coefficient of FBG facing the icicle was -34.6 pm/N. In addition, the wavelength shift generated by several icicles was equal to the sum of wavelength shifts generated by each icicle within the polar angle range of −15° to 15°. Finally, icicles can cause wavelength shifts of FBGs within a big shed spacing. The paper provides a novel icing detection technology for composite insulators in transmission lines.


Introduction
Since the 21st century, the global climate has changed dramatically, and the shortage of conventional energy has become increasingly severe. It is significant to construct smart grids for the electric power industry [1]. Currently, the sensing technologies of electrical equipment are an essential foundation to ensure safe operation, promote controllability, and achieve intelligence of smart grid [2].
In March 2019, the Internet Department, State Grid Corporation of China, published an outline for the construction of the Ubiquitous Power Internet of Things (UPIoT). UPIoT includes four layers: sensing layer, network layer, platform layer, and application layer. In particular, the sensing layer is comprised of various sensors, edge computing equipment, and local communication networks to achieve acquisition, aggregation, and processing of data from conditions of power equipment. In summary, intelligent sensing technologies are one of the core foundations of the sensing layer in UPIoT [3].
Insulators in transmission lines are easily covered by ice in cold and high humidity regions, resulting in flashover accidents and even severe disasters of power grids. To date, there have been several methods (e.g., ultraviolet imaging, unmanned aerial vehicle (UAV), and video camera) to detect icing conditions on insulators. Wu and Liao [4,5] processed aerial insulators images from UAV

FBG Sensing Principle
An FBG is a spatial phase grating inscribing into the photosensitive fiber core. The FBG has a significant property of serving as a narrow band filter or reflector. More specifically, it can reflect a narrow band of light with a specific wavelength from an incident broad source, and the rest of the light with other wavelengths passes through the FBG. The FBG reflection wavelength is defined as λ B [29]: where n eff is the effective refractive index of the optical fiber, and Λ denotes the periodicity of the grating. Temperature, stain, or other environmental parameters would change the refractive index and periodicity of the FBG resulting in a wavelength shift. In this case, the FBG wavelength shift caused by temperature change (∆T) and axial strain (ε) can be expressed as [30]: where α f denotes the thermal expansion coefficient, ξ is the thermo-optic coefficient, P e is the elastic-optic coefficient, and K B,T and K B,ε are both constant which denote the temperature sensitivity (pm/ • C) and the axial strain sensitivity (pm/µε), respectively.

Simultaneous Measurement of FBG Temperature and Strain
Glaze icing can change the temperature and strain of insulator sheds, which both cause the wavelength shifts of FBG at the same time. Therefore, it is necessary to propose a method for measuring FBG temperature and strain simultaneously. Specifically, two FBGs (i.e., FBG-1 and FBG-2) are inscribed in an optical fiber. FBG-1 is affected by temperature and strain, whose wavelength shift can be described as Equation (3). In contrast, FBG-2 is packaged by a designed housing that is not affected by strain. Thus, the wavelength shift of FBG-2 can be described as Equation (4).
Therefore, the wavelength shift of FBG-1 caused by strain can be described as:

Fiber Bragg Grating
In this paper, FBGs with different reflection wavelengths have been inscribed in single mode optical fibers with polyimide coating using phase mask technology. The diameter, effective length and reflectivity of each FBG are about 250 µm, 10 mm and 90%, respectively. The initial wavelengths of FBGs in room temperature are within a range of 1510 to 1590 nm. The strain measurement of FBG ranges from −2500 to 2500 µε, and the accuracy rate is 1% F.S. Moreover, the operating temperature ranges from −40 to 120 • C. More specifically, three optical fibers were embedded into a composite insulator and labeled as 1#, 2#, and 3#, respectively.
In order to measure icicle loads and temperature simultaneously, the FBG temperature compensation sensor was packed in a ceramic tube, which is cylindrical, with a length of 30 mm and a diameter of about 2 mm ( Figure 1). Besides, epoxy resin adhesive was injected into the ceramic tube Sensors 2020, 20, 1834 4 of 13 for insulation. The FBG temperature compensation sensors are not affected by icicle loads due to the high elasticity modulus of the ceramic tube. ranges from −2500 to 2500 με, and the accuracy rate is 1‰ F.S. Moreover, the operating temperature ranges from −40 to 120 °C. More specifically, three optical fibers were embedded into a composite insulator and labeled as 1#, 2#, and 3#, respectively.
In order to measure icicle loads and temperature simultaneously, the FBG temperature compensation sensor was packed in a ceramic tube, which is cylindrical, with a length of 30 mm and a diameter of about 2 mm ( Figure 1). Besides, epoxy resin adhesive was injected into the ceramic tube for insulation. The FBG temperature compensation sensors are not affected by icicle loads due to the high elasticity modulus of the ceramic tube.

A 110 kV Composite Insulator with Embedded FBGs
The type of 110 kV composite insulator is FXBW-110/100, whose parameters are listed in Table 1. The insulator has 13 big sheds, which are projected from the insulator trunk, and is intended to increase the creepage distance [31], as shown in Figure 2a.
Three optical fibers, 1#, 2#, and 3#, were embedded vertically into the composite insulator during silicone rubber injection process. Besides, the optical fibers were at a 120° interval around the central axis of the composite insulator ( Figure 2). Fourteen FBGs were inscribed in 1# and 2# optical fiber respectively, and thirteen FBGs were inscribed in 3# optical fiber. The FBGs were labeled as FBG-mn, where m was the label of the optical fibers (i.e., 1, 2, and 3), and n was the label of the FBGs (i.e., 1, 2… and 14). For example, the fourteen FBGs in the 1# optical fiber were separately labeled as FBG101 -FBG114. Specifically, FBG101 -FBG113 were all FBG strain sensors, which were designed above each big shed root in silicone rubber housing (Figure 2b), and FBG114 was an FBG temperature compensation sensor. Similarly, FBG201 -FBG213 in 2# optical fiber and FBG301 -FBG313 in 3# optical fiber were all FBG strain sensors. In addition, FBG214 was also an FBG temperature compensation sensor.
After proper preparation of the core rod surface, three optical fibers with FBGs were put on the core rod surface according to the arrangement in Figure 2c. Then, two ends of optical fibers were fixed, and midpoints between two FBGs were adhered on the core rod using room temperature vulcanized silicone rubber (RTV) until curing. The core rod and end fittings were joined by a pressing process. Finally, the core rod and optical fibers were placed in a vacuum injection machine, and the pressure remained constant during the process of injecting silicone rubber and curing.

A 110 kV Composite Insulator with Embedded FBGs
The type of 110 kV composite insulator is FXBW-110/100, whose parameters are listed in Table 1. The insulator has 13 big sheds, which are projected from the insulator trunk, and is intended to increase the creepage distance [31], as shown in Figure 2a. Three optical fibers, 1#, 2#, and 3#, were embedded vertically into the composite insulator during silicone rubber injection process. Besides, the optical fibers were at a 120 • interval around the central axis of the composite insulator ( Figure 2). Fourteen FBGs were inscribed in 1# and 2# optical fiber respectively, and thirteen FBGs were inscribed in 3# optical fiber. The FBGs were labeled as FBG-mn, where m was the label of the optical fibers (i.e., 1, 2, and 3), and n was the label of the FBGs (i.e., 1, 2 . . . and 14). For example, the fourteen FBGs in the 1# optical fiber were separately labeled as FBG101 -FBG114. Specifically, FBG101 -FBG113 were all FBG strain sensors, which were designed above each big shed root in silicone rubber housing (Figure 2b), and FBG114 was an FBG temperature compensation sensor. Similarly, FBG201 -FBG213 in 2# optical fiber and FBG301 -FBG313 in 3# optical fiber were all FBG strain sensors. In addition, FBG214 was also an FBG temperature compensation sensor.
After proper preparation of the core rod surface, three optical fibers with FBGs were put on the core rod surface according to the arrangement in Figure 2c. Then, two ends of optical fibers were fixed, and midpoints between two FBGs were adhered on the core rod using room temperature vulcanized silicone rubber (RTV) until curing. The core rod and end fittings were joined by a pressing process. Finally, the core rod and optical fibers were placed in a vacuum injection machine, and the pressure remained constant during the process of injecting silicone rubber and curing.

The Sensing System
The sensing system was made up of three parts, i.e., the 110 kV composite insulator with embedded FBGs, an FBG interrogation system, and an artificial climate chamber, as depicted in Figure 3. The type of interrogation system is JEME-iFBG produced by Shenzhen JEMETECH Company in China. Specifically, the interrogation system can acquire signals of 6 channels simultaneously based on a high-power wavelength scanning laser, and each channel can detect 48 FBGs at most. With a wavelength range of 1510 to 1590 nm and a wavelength accuracy of 1 pm, the interrogation system can satisfy the requirement for detecting glaze icing on composite insulators.
The artificial climate chamber was used to simulate the low-temperature environment during glaze ice accretion on composite insulators. Specifically, the artificial climate chamber offered a temperature change from −15 °C to room temperature (20 °C) with an accuracy of 1 °C. In experiments, three optical fibers were pulled out from the chamber, and the interrogation system was linked to the computer using the USB port for data recording, analysis, and transfer.

The Sensing System
The sensing system was made up of three parts, i.e., the 110 kV composite insulator with embedded FBGs, an FBG interrogation system, and an artificial climate chamber, as depicted in Figure 3. The type of interrogation system is JEME-iFBG produced by Shenzhen JEMETECH Company in China. Specifically, the interrogation system can acquire signals of 6 channels simultaneously based on a high-power wavelength scanning laser, and each channel can detect 48 FBGs at most. With a wavelength range of 1510 to 1590 nm and a wavelength accuracy of 1 pm, the interrogation system can satisfy the requirement for detecting glaze icing on composite insulators.
The artificial climate chamber was used to simulate the low-temperature environment during glaze ice accretion on composite insulators. Specifically, the artificial climate chamber offered a temperature change from −15 • C to room temperature (20 • C) with an accuracy of 1 • C. In experiments, three optical fibers were pulled out from the chamber, and the interrogation system was linked to the computer using the USB port for data recording, analysis, and transfer.

Temperature Calibration Experiments
Previous studies have demonstrated that glaze ice is easy to accrete on insulator sheds when the temperature is between −8 and 0 °C in the field [32]. Besides, according to IEEE Standard 1783-2009, the air temperature should remain between −15 and −5 °C for depositing artificial ice on insulator sheds in a climate chamber [33]. Thus, in the temperature calibration experiments, the air temperature in the climate chamber was set between −15 and +15 °C. The process was as follows: (1) The 110 kV composite insulator with embedded FBGs was put into the artificial climate chamber. Three optical fibers were pulled out from the chamber, and connected with the demodulator and computer through outlet fibers. The wavelengths of all FBGs were recorded as initial wavelengths when the temperature was 0 °C.
(2) The temperature range of the artificial climate chamber was set from −15 to +15 °C, and the wavelength shifts were recorded at intervals of 3 °C. A previous research has indicated that the temperature of FBG in the composite insulator was equal to the temperature in the chamber after 8 hours [18]. Therefore, after being placed into the chamber for 8 h, the temperature of the composite insulator could be considered as a constant and recorded if a wavelength rate was less than 3 pm/min [18]. After averaging 10 sets of data, the results of temperature calibration experiments were created.

Simulated Icicle Load Experiments
Many studies have indicated that the common glaze ice accretion on insulators causes the highest probability of ice flashover occurrence [34,35]. Therefore, in order to find the relationship between glaze ice and wavelength shifts, weights were hung on the edge of a big shed for simulating glaze ice loads generated by icicles [36] (Figure 4).
In order to describe icing conditions of composite insulators quantitatively, icicle length and icicle bridged degree were proposed, which can be described as [37][38][39]: where η denotes the bridged degree of a big shed spacing, S is the big shed spacing, d is the air gap distance, and L is the icicle length.
In order to describe positions of simulated loads and FBGs on insulator sheds quantitatively, a polar coordinate system was proposed, where the midpoint of the core rod was the original point

Temperature Calibration Experiments
Previous studies have demonstrated that glaze ice is easy to accrete on insulator sheds when the temperature is between −8 and 0 • C in the field [32]. Besides, according to IEEE Standard 1783-2009, the air temperature should remain between −15 and −5 • C for depositing artificial ice on insulator sheds in a climate chamber [33]. Thus, in the temperature calibration experiments, the air temperature in the climate chamber was set between −15 and +15 • C. The process was as follows: (1) The 110 kV composite insulator with embedded FBGs was put into the artificial climate chamber.
Three optical fibers were pulled out from the chamber, and connected with the demodulator and computer through outlet fibers. The wavelengths of all FBGs were recorded as initial wavelengths when the temperature was 0 • C. (2) The temperature range of the artificial climate chamber was set from −15 to +15 • C, and the wavelength shifts were recorded at intervals of 3 • C. A previous research has indicated that the temperature of FBG in the composite insulator was equal to the temperature in the chamber after 8 hours [18]. Therefore, after being placed into the chamber for 8 h, the temperature of the composite insulator could be considered as a constant and recorded if a wavelength rate was less than 3 pm/min [18]. After averaging 10 sets of data, the results of temperature calibration experiments were created.

Simulated Icicle Load Experiments
Many studies have indicated that the common glaze ice accretion on insulators causes the highest probability of ice flashover occurrence [34,35]. Therefore, in order to find the relationship between glaze ice and wavelength shifts, weights were hung on the edge of a big shed for simulating glaze ice loads generated by icicles [36] (Figure 4).
In order to describe icing conditions of composite insulators quantitatively, icicle length and icicle bridged degree were proposed, which can be described as [37][38][39]: where η denotes the bridged degree of a big shed spacing, S is the big shed spacing, d is the air gap distance, and L is the icicle length.
In order to describe positions of simulated loads and FBGs on insulator sheds quantitatively, a polar coordinate system was proposed, where the midpoint of the core rod was the original point ( Figure 5). The position was considered as (r, θ, n), where r denoted the radius (mm), θdenoted the polar angle ( • ), and n denoted the number of big sheds in the insulator. For example, the coordinates of FBG101, FBG201, and FBG301 were (9, 0 • , 1), (9, 120 • , 1), and (9, −120 • , 1), respectively. When the simulated load was on the edge of the 1st big shed facing the 1# optical fiber, its coordinate was (61, 0 • , 1). In this paper, the r of simulated loads were all 61 mm, so the coordinates of simulated loads could be simplified to P(θ, n).
The process was as follows: (1) The 110 kV composite insulator with embedded FBG was placed in the artificial climate chamber, and the air temperature was set as -10 • C during the experiments. The optical fibers were pulled out from the chamber, and connected with the demodulator and computer. (2) Geometric characteristics of an icicle on insulators were proposed in [37]. In this paper, the weights were set as 0.5, 1.0, 1.5, 2.0, and 2.5 N, corresponding to icicle lengths and bridged degrees, as shown in Table 2. The icicle lengths and bridged degrees are common parameters in icicle growth experiments [38,39]. (3) Weights were hung on insulator big sheds. The wavelength shifts were continuously recorded when wavelength rates were stable (< 3 pm/min) [18]. After averaging 10 sets of data, the results were created.
The process was as follows: (1) The 110 kV composite insulator with embedded FBG was placed in the artificial climate chamber, and the air temperature was set as -10 °C during the experiments. The optical fibers were pulled out from the chamber, and connected with the demodulator and computer.
(2) Geometric characteristics of an icicle on insulators were proposed in [37]. In this paper, the weights were set as 0.5, 1.0, 1.5, 2.0, and 2.5 N, corresponding to icicle lengths and bridged degrees, as shown in Table 2. The icicle lengths and bridged degrees are common parameters in icicle growth experiments [38,39].
(3) Weights were hung on insulator big sheds. The wavelength shifts were continuously recorded when wavelength rates were stable (< 3 pm/min) [18]. After averaging 10 sets of data, the results were created.   ( Figure 5). The position was considered as (r, θ, n), where r denoted the radius (mm), θ denoted the polar angle (°), and n denoted the number of big sheds in the insulator. For example, the coordinates of FBG101, FBG201, and FBG301 were (9, 0°, 1), (9, 120°, 1), and (9, −120°, 1), respectively. When the simulated load was on the edge of the 1st big shed facing the 1# optical fiber, its coordinate was (61, 0°, 1). In this paper, the r of simulated loads were all 61 mm, so the coordinates of simulated loads could be simplified to P(θ, n).
The process was as follows: (1) The 110 kV composite insulator with embedded FBG was placed in the artificial climate chamber, and the air temperature was set as -10 °C during the experiments. The optical fibers were pulled out from the chamber, and connected with the demodulator and computer.
(2) Geometric characteristics of an icicle on insulators were proposed in [37]. In this paper, the weights were set as 0.5, 1.0, 1.5, 2.0, and 2.5 N, corresponding to icicle lengths and bridged degrees, as shown in Table 2. The icicle lengths and bridged degrees are common parameters in icicle growth experiments [38,39].
(3) Weights were hung on insulator big sheds. The wavelength shifts were continuously recorded when wavelength rates were stable (< 3 pm/min) [18]. After averaging 10 sets of data, the results were created.    Table 2. Icicle load corresponds to the icicle length and icicle bridged degree with density of 0.9 g/cm 3 [36,37].

Weights
Corresponding

The Relationship between Wavelength Shifts of FBGs and Temperature
The wavelength shift mean deviation (WSMD) was proposed to discuss the temperature characteristics of FBG strain sensors in an optical fiber. For example, the WSMD of 13 FBG strain sensors in 1# optical fiber, i.e., FBG101 -FBG113, was expressed as f 1 , which can be defined as: where n denotes the label of FBG strain sensor in 1# optical fiber, λ n denotes the wavelength shift of FBG1n at a temperature, and λ denotes the average value of wavelength shifts of 13 FBG strain sensors in 1# optical fiber at the temperature. Similarly, f 2 and f 3 represent the WSMD of FBG strain sensors in 2# and 3# optical fiber, respectively. As shown in the right Y axis of Figure 6, the mean deviations were all within 8 pm, which showed that wavelength shifts of FBG strain sensors in an optical fiber were seen as the same at a temperature.

A Simulated Load of An Icicle on A Big Shed
The wavelength shift of FBG102 changed with the simulated load of one weight at P1 (0°, 2), as the red line in Figure 7. When a simulated load on the edge of the 2nd big shed moved at a 30° interval, the wavelength shifts of the FBG102 were shown in Figure 7.
It was found that the wavelength shifts of FBG102 were linearly related to the simulated load of one weight at P1(0°, 2), P2(30°, 2), P3(60°, 2), P10(−60°, 2), and P11(−30°, 2), respectively, corresponding to a polar angle range of -60° to 60°. The linear fitting formulas were described as: Therefore, wavelength shifts dependent on temperatures of FBG101 -FBG113, FBG201 -FBG213, and FBG301 -FBG313 were shown in red, black, and yellow solid lines, respectively ( Figure 6). The solid lines showed that wavelength shifts of all FBG strain sensors were linear with temperature, and goodness of fit R 2 were all greater than 0.997. Therefore, the temperature sensitivities of all FBG strain sensors were approximately 18.16 pm/ • C. The green and blue lines showed that wavelength shifts of FBG114 and FBG214 were linear with temperature, and goodness of fit R 2 were approximately 0.997. The temperature sensitivities of FBG temperature compensation sensors were 13.18 pm/ • C. Therefore, according to Equation (5), the wavelength shift of an FBG strain sensor caused by strain was as follows: where ∆λ S denoted the wavelength shift of an FBG strain sensor, and ∆λ T denoted the wavelength shift of an FBG temperature compensation sensor.

A Simulated Load of an Icicle on a Big Shed
The wavelength shift of FBG102 changed with the simulated load of one weight at P 1 (0 • , 2), as the red line in Figure 7. When a simulated load on the edge of the 2nd big shed moved at a 30 • interval, the wavelength shifts of the FBG102 were shown in Figure 7.

Several Simulated Loads of Several Icicles on A Big Shed
In order to simulate several icicles accretion on different positions of an insulator shed, three simulated loads were hung on a big shed in the insulator. It is clear from Section 4.2 that wavelength shifts were easily affected by the icicle load with polar angle ranges of −60° to 60°. Therefore, a simulated load Fa of 1 N was at Pa (0°, 3), and the other two simulated loads Fb and Fc of 0.5 N were at (θ°, 3) where θ changed in a polar angle range of -60° to 60°. The relationship between the wavelength shifts and the changing polar angles (θb, θc) were shown in Figure 8.
It was found that the absolute value of wavelength shift decreases with the distance from Pa(0°, 3) to load point (e.g., Fb and Fc). In addition, when the polar angle of Fb and Fc were between -15° and 15°, the wavelength shifts of FBG103 were between 67 and 79 pm. In contrast, the wavelength shift generated by the sum of Fa, Fb, and Fc at 0° was 79 pm. The deviations of the wavelength shifts were within 12 pm. In view of the experimental error, it can be found that when the polar angle of Fb and Fc were between −15° to 15°, the wavelength shifts were similar to that generated by the sum of Fa, Fb, and Fc in 0°.
Comparing the Equation (10) with Equation (13), Equation (11) with Equation (12), the load coefficients of FBG102 were approximate when the loads were symmetrical about the polar angle of 0 • . It can be concluded that the wavelength shifts of FBG were equal when two loads were symmetrical about the extension line from the original point to the FBG.
In contrast, when the polar angles of icicle load were from 90 • to −90 • , with the increase of the icicle load, the wavelength shifts appeared to remain stable. The reason for this was that FBG102 was located 86.27 mm away from P 4 and P 9 , even 140 mm away from P 6 . The strain transferring to FBG102 was so small that the wavelength shifts were all less than 12 pm.

Several Simulated Loads of Several Icicles on A Big Shed
In order to simulate several icicles accretion on different positions of an insulator shed, three simulated loads were hung on a big shed in the insulator. It is clear from Section 4.2 that wavelength shifts were easily affected by the icicle load with polar angle ranges of −60 • to 60 • . Therefore, a simulated load F a of 1 N was at P a (0 • , 3), and the other two simulated loads F b and F c of 0.5 N were at (θ • , 3) where θ changed in a polar angle range of -60 • to 60 • . The relationship between the wavelength shifts and the changing polar angles (θ b , θ c ) were shown in Figure 8.

Wavelength Shifts of All FBGs in An Optical Fiber
To analyze wavelength shifts of all FBGs in an optical fiber when an icicle was generated on a big shed, a simulated load of 2 N was generated at (0°, 1), (120°, 1), and (−120°, 1), respectively.
It was found from Figure 9 that the wavelength shifts of the FBGm01 were -80 pm, those of the FBGm02 were less than −10 pm, while those of FBGmn (3 ≤ n ≤ 11) were almost 0 pm. The reason for this is that the silicone rubber is soft material, whose elasticity modulus is 2.6 MPa. The deformation of the 1st big shed was large under high loads, meanwhile, the strain transferring to FBGm02 was large enough to detect the wavelength shifts. As for FBGmn (3 ≤ n ≤ 11), the transferred strain was too small to change the wavelength shifts. In conclusion, icicle loads can cause wavelength shifts of the FBGs within one big shed spacing. It was found that the absolute value of wavelength shift decreases with the distance from P a (0 • , 3) to load point (e.g., F b and F c ). In addition, when the polar angle of F b and F c were between -15 • and 15 • , the wavelength shifts of FBG103 were between 67 and 79 pm. In contrast, the wavelength shift generated by the sum of F a , F b , and F c at 0 • was 79 pm. The deviations of the wavelength shifts were within 12 pm. In view of the experimental error, it can be found that when the polar angle of F b and F c were between −15 • to 15 • , the wavelength shifts were similar to that generated by the sum of F a , F b , and F c in 0 • .
In conclusion, when the polar angles are between −15 • and 15 • , the wavelength shift generated by several icicles was equal to the sum of wavelength shifts generated by each icicle. In contrast, when the polar angles are between ±15 • and ±60 • , the absolute values of wavelength shifts would decrease with the distance from icicle positions to P a .

Wavelength Shifts of All FBGs in an Optical Fiber
To analyze wavelength shifts of all FBGs in an optical fiber when an icicle was generated on a big shed, a simulated load of 2 N was generated at (0 • , 1), (120 • , 1), and (−120 • , 1), respectively.
It was found from Figure 9 that the wavelength shifts of the FBGm01 were -80 pm, those of the FBGm02 were less than −10 pm, while those of FBGmn (3 ≤ n ≤ 11) were almost 0 pm. The reason for this is that the silicone rubber is soft material, whose elasticity modulus is 2.6 MPa. The deformation of the 1st big shed was large under high loads, meanwhile, the strain transferring to FBGm02 was large enough to detect the wavelength shifts. As for FBGmn (3 ≤ n ≤ 11), the transferred strain was too small to change the wavelength shifts. In conclusion, icicle loads can cause wavelength shifts of the FBGs within one big shed spacing.

Wavelength Shifts of All FBGs in An Optical Fiber
To analyze wavelength shifts of all FBGs in an optical fiber when an icicle was generated on a big shed, a simulated load of 2 N was generated at (0°, 1), (120°, 1), and (−120°, 1), respectively.
It was found from Figure 9 that the wavelength shifts of the FBGm01 were -80 pm, those of the FBGm02 were less than −10 pm, while those of FBGmn (3 ≤ n ≤ 11) were almost 0 pm. The reason for this is that the silicone rubber is soft material, whose elasticity modulus is 2.6 MPa. The deformation of the 1st big shed was large under high loads, meanwhile, the strain transferring to FBGm02 was large enough to detect the wavelength shifts. As for FBGmn (3 ≤ n ≤ 11), the transferred strain was too small to change the wavelength shifts. In conclusion, icicle loads can cause wavelength shifts of the FBGs within one big shed spacing.

Conclusions
In this paper, a detection method for glaze icing of 110 kV composite insulators has been proposed. The 110 kV composite insulator with embedded FBGs is helpful to sense glaze icing conditions in all weather, offering much better versatility than conventional approaches. Three optical fibers with FBGs were embedded symmetrically into a 110 kV composite insulator. Besides, the FBG temperature compensation sensors were packaged in ceramic tubes so as to solve the problem of icicle loads and temperature cross-sensitivity. Then, temperature calibration experiments and simulated icicle load experiments were performed.
The results showed that temperature sensitivities of FBG strain sensors and FBG temperature compensation sensors were 18.16 pm/ • C, and 13.18 pm/ • C, respectively. FBG wavelength shifts were linearly related to the icicle loads within the polar angle range of −60 • to 60 • . In particular, the load coefficient of FBG facing the icicle was −34.6 pm/N. Moreover, within the polar angle range of −15 • to 15 • , the wavelength shift generated by several icicles was equal to the sum of wavelength shifts generated by each icicle. Finally, icicle loads can cause wavelength shifts of the FBGs within one big-shed spacing. Therefore, the icicle loads of composite insulators can be detected using FBGs in the laboratory.
In further research, a mathematical model and a location algorithm will be proposed for monitoring multiple icicle loads and positions simultaneously on the composite insulator. In addition, uncontrollable outdoor factors (e.g., sunlight, wind, rain, etc.) should be taken into considerations as well.