**3.4.1 Introduction**

Electric power facilities, such as substations, rely on instrument transformers for their functionality and protection. They are divided into voltage transformers (VT) and current transformers (CT) for measuring and controlling voltage and current, respectively. The role of the instrument transformer is to provide accurate signals for protection, control and metering systems, including revenue metering. These requirements place stringent demands on the accuracy and reliability of the instrument transformer to guarantee the correct functionality for protection systems and precise measurement for metering purposes.

Created over a century ago, they are reliable for over-voltage and over-current protection; allow 0.2% revenue metering accuracy and their behavior is well known under both normal and abnormal conditions. Nevertheless, these pieces of equipment are made entirely of copper, ceramic and iron with all empty spaces filled with oil, which are weighty materials, producing bulky, heavy and clumsy equipment. On top of that, they tend to explode without prior warning, resulting in the potential destruction of nearby equipment by pieces of sharp ceramics and furthermore putting the substation personnel at risk.

Optical voltage transducers offer many improvements on traditional inductive and capacitive voltage transformers such as linear performance and wider dynamic range, lighter weight, smaller size and improved safety.

The optical-fiber sensors industry has grown in recent years, and most of the efforts involving the sensors industry focused the use of Fiber Bragg Grating (FBG) as a sensor element. Among the parameters of interest most of the works found in the literature focus on temperature, strain, pressure, displacement, acceleration, vibration, voltage and current.

The behavior of optical current transformer (OCT) and optical voltage transformer (OVT) applied on electric power transmission system has been widely discussed in the literature because they present advantages when compared with conventional transformers. The innovations coming from the optical transformers circumvent problems such as the risk of explosion, high weight, electric safety, insulation oil, difficulty of installation, etc [Sawa et

Optical Fiber Sensors 23

For improving isolation for high voltage the entire assembly was immersed in a bath of insulating oil. The FBG with central wavelength of 1532.9 nm was stretched to 1535.18 nm as shown in Fig. 3.4.2, before bonded to the aluminum structure to allow measurements in

(a) (b) Fig. 3.4.2. (a) The FBG reflection spectrum before bonded to the aluminum base and (b) after

Notice that by bonding the FBG on the PZT stack as we described, we would have the strain on the FBG equals to the strain on the PZT. This is because, although the total displacement

Since the fiber is bonded to the ends of the stack, the displacement previewed by (8) will be

Now combining (5), (8) e (9) and considering ΔT=0 (constant temperature environment), we

ΔLPZT= ∆LFBG� (1)

is bigger, so is the length of the fiber, yielding therefore the same strain.

Fig. 3.4.1. Schematic diagram of the FBG-Piezostack.

bonded.

achieve:

transmitted to the fiber, so that

both directions, that is augmenting and retreating PZT thickness.

al, 1990, Cease et al, 1991, Werneck and Abrantes, 2004, and De Nazaré and Werneck, 2010].

However, the main drawback is still the high cost of this new technology, not only for acquisition but also maintenance, demanding specialty skills uncommonly available among company personnel. With this motivation, this case relates the development of a high voltage measuring system to be used as the core of a 13.8-kV-class OVT for the electric power industry application using a PZT(Lead Titanate Zirconate) crystal as voltage transducer and FBG as strain measuring sensor. This new technology can be developed at a cost fully compatible with conventional CTs and VTs.

FBG technology is one of the most popular choices for optical-fiber sensor for strain or temperature measurements due to their simple manufacturing, besides it is relatively easy to deal with and reliable. The use of piezoelectric ceramics in the last decade due to piezoelectric characteristics and transducer properties has attracted interest to electric power systems measurements because of their properties to convert electrical energy to mechanical energy [Niewczas et al, 2005, Yao and Yi, 2006 and Allil and Werneck, 2011].

This study relates to the development of a high voltage sensor system using a PZT piezoelectric crystal as transducer and an FBG as a sensor for an optical voltage transformer for 13.8-kV-class. In the present contribution, a voltage was applied in a combined PZT and FBG sensor by using a high voltage source. This voltage acts on the PZT ceramic causing a mechanical deformation and by using a FBG as interrogation system, the spectrum of the reflected light from the FBG is captured and demodulated to obtain a sinusoidal signal proportional to the applied voltage.

The results showed a linear relationship between the applied voltages to the PZT-FBG sensor with the reflected Bragg wavelength shift. The easy implementation and the low cost of the equipment used prove the viability of this project for applications in the electric power industries.

From previous experimental studies it has been proven that the exposure to ultraviolet radiation during the FBG inscription process decreases the silica yield strength, furthermore, when stretching the FBG to bond it to the stress element, it is necessary to remove the optical fiber coating, and this process can degrade the fiber strength **[**Miyajima, 1982, Olshansky, Maurer, 1976 and Kurkjian et al, 1989].

To study the mechanical strength and the fiber resistance to strain, in a previous paper [Ribeiro and Werneck, 2010] we measured the tensile strength of silica optical fiber. By providing information about mechanical strength it is possible to obtain a maximum life span for these devices.

#### **3.4.2 Experimental setup**

As mentioned above, we used a PZT crystal as voltage transducer and a FBG as strain measuring sensor. The experimental setup of the FBG-PZT sensor system is shown in Fig. 3.4.1. The ceramic stack was built using ten 4-mm-thick PZT rings, with d33 = 300 pm/V separated by 0.2-mm thick copper electrodes where the contacts were fixed. The electrodes were arranged on both sides of the ceramic discs and were connected in a parallel line. The ceramic disks were glued together separated by the cooper plates using EPO-TEK 302-3M resin and kept in the oven for 3 hours at a temperature of 65°C for the cure. A double aluminum structure was used to accommodate the ceramic stack and the 82-mm-length sensor was glued on the top of it.

al, 1990, Cease et al, 1991, Werneck and Abrantes, 2004, and De Nazaré and Werneck,

However, the main drawback is still the high cost of this new technology, not only for acquisition but also maintenance, demanding specialty skills uncommonly available among company personnel. With this motivation, this case relates the development of a high voltage measuring system to be used as the core of a 13.8-kV-class OVT for the electric power industry application using a PZT(Lead Titanate Zirconate) crystal as voltage transducer and FBG as strain measuring sensor. This new technology can be developed at a

FBG technology is one of the most popular choices for optical-fiber sensor for strain or temperature measurements due to their simple manufacturing, besides it is relatively easy to deal with and reliable. The use of piezoelectric ceramics in the last decade due to piezoelectric characteristics and transducer properties has attracted interest to electric power systems measurements because of their properties to convert electrical energy to mechanical

This study relates to the development of a high voltage sensor system using a PZT piezoelectric crystal as transducer and an FBG as a sensor for an optical voltage transformer for 13.8-kV-class. In the present contribution, a voltage was applied in a combined PZT and FBG sensor by using a high voltage source. This voltage acts on the PZT ceramic causing a mechanical deformation and by using a FBG as interrogation system, the spectrum of the reflected light from the FBG is captured and demodulated to obtain a sinusoidal signal

The results showed a linear relationship between the applied voltages to the PZT-FBG sensor with the reflected Bragg wavelength shift. The easy implementation and the low cost of the equipment used prove the viability of this project for applications in the electric

From previous experimental studies it has been proven that the exposure to ultraviolet radiation during the FBG inscription process decreases the silica yield strength, furthermore, when stretching the FBG to bond it to the stress element, it is necessary to remove the optical fiber coating, and this process can degrade the fiber strength **[**Miyajima, 1982,

To study the mechanical strength and the fiber resistance to strain, in a previous paper [Ribeiro and Werneck, 2010] we measured the tensile strength of silica optical fiber. By providing information about mechanical strength it is possible to obtain a maximum life

As mentioned above, we used a PZT crystal as voltage transducer and a FBG as strain measuring sensor. The experimental setup of the FBG-PZT sensor system is shown in Fig. 3.4.1. The ceramic stack was built using ten 4-mm-thick PZT rings, with d33 = 300 pm/V separated by 0.2-mm thick copper electrodes where the contacts were fixed. The electrodes were arranged on both sides of the ceramic discs and were connected in a parallel line. The ceramic disks were glued together separated by the cooper plates using EPO-TEK 302-3M resin and kept in the oven for 3 hours at a temperature of 65°C for the cure. A double aluminum structure was used to accommodate the ceramic stack and the 82-mm-length

energy [Niewczas et al, 2005, Yao and Yi, 2006 and Allil and Werneck, 2011].

cost fully compatible with conventional CTs and VTs.

proportional to the applied voltage.

Olshansky, Maurer, 1976 and Kurkjian et al, 1989].

power industries.

span for these devices.

**3.4.2 Experimental setup** 

sensor was glued on the top of it.

2010].

Fig. 3.4.1. Schematic diagram of the FBG-Piezostack.

For improving isolation for high voltage the entire assembly was immersed in a bath of insulating oil. The FBG with central wavelength of 1532.9 nm was stretched to 1535.18 nm as shown in Fig. 3.4.2, before bonded to the aluminum structure to allow measurements in both directions, that is augmenting and retreating PZT thickness.

Fig. 3.4.2. (a) The FBG reflection spectrum before bonded to the aluminum base and (b) after bonded.

Notice that by bonding the FBG on the PZT stack as we described, we would have the strain on the FBG equals to the strain on the PZT. This is because, although the total displacement is bigger, so is the length of the fiber, yielding therefore the same strain.

Since the fiber is bonded to the ends of the stack, the displacement previewed by (8) will be transmitted to the fiber, so that

$$
\Delta \mathcal{L}\_{\text{FZT}} = \Delta \mathcal{L}\_{\text{FBG}\upharpoonright} \tag{1}
$$

Now combining (5), (8) e (9) and considering ΔT=0 (constant temperature environment), we achieve:

Optical Fiber Sensors 25

For the first experiment, only DC voltages were applied to the PZT in order to measure the Bragg displacement accurately by the interrogation system. Eq. 9 was used to calculate the maximum voltage to be applied to the PZT ceramic and do not exceed the allowed value,

By applying a DC voltage to the PZT and recording the respective Bragg shift we can see the linear relationship between the applied voltage and the central Bragg wavelength (Fig. 3.4.4). The results show that the measured sensitivity was of 91.5 pm/kV and the correlation

> 0 250 500 750 1000 1250 1500 1750 2000 2250 2500 **Vin (V)**

Fig. 3.4.5 represents the interrogation system for AC voltage measurements. Since the optical spectrum analyzer is too slow to respond to the 60-Hz line frequency, the central wavelength variation can be obtained by using a photo-detector. The light from the ASE illuminates the FBG-PZT sensor via an optical circulator. The reflected spectrum of the sensor pass through the Fabry-Perot tunable filter (FFP-TF) with 0.89 nm bandwidth, nominal finesse of 130 and 116 nm of free spectral range (FSR). The light signal enters an amplified photo-detector with designed for detection of light signal over 700 nm – 1800 nm. The AC output signal is monitored by an oscilloscope. The FFP filter was tuned in 1540.04 nm by applying a voltage of the 7.2 volts. This point indicates that the FFP filter is matched with the FBG sensor on the

y = 9,154E-05x + 1,537E+03 R2 = 9,990E-01

1536,7 1536,72 1536,74 1536,76 1536,78 1536,8 1536,82 1536,84 1536,86 1536,88 1536,9 1536,92 1536,94

Fig. 3.4.4. FBG-PZT sensor curve when a DC voltage is applied

stack and the intensity of light at the photo-detector is at maximum.

Fig 3.4.5. Schematic diagram of experiment setup for AC voltage.

**3.4.4 The optical setup with a AC voltage power supply** 

**Bragg wavelength shift (nm)**

V = E.dij (4)

accordingly with the Table 3.

coefficient (R2) were 0.999.

$$
\Delta\lambda\_{\rm B} = \lambda\_{\rm B}(1\text{-}\rho\_{\rm e}) \text{ nd}\lambda\_{\rm 3} \text{V/L}\_{\perp} \tag{2}
$$

Substituting the PZT constants of Table 3 in (10) we have the following sensitivity for the applied DC voltage:

$$
\Delta\lambda\_{\text{B}}/\Delta\text{V} = 128.3 \text{ pm/kV}\_{\square} \tag{3}
$$

Notice that the larger the LFBG, the greater the strain experienced by the FBG and consequently, the greater the sensitivity.


Table 3. FBG and PZT Parameters

#### **3.4.3 Optical setup for DC high voltage input**

A DC voltage was applied on the PZT crystal terminals by using a high voltage supply and the displacement of the PZT was converted into variations of the Bragg central wavelength. The interrogation system for DC voltage measurements is schematically illustrated in Fig. 3.4.3. The light from an amplified spontaneous emission (ASE) ranging from 1520 nm to 1610 nm was used to illuminate the sensor and a commercial interrogation system from FOS&S model Spectral Eye 400, with accuracy of 2.0 pm was used to measure the reflected FBG spectrum accordingly to the sensor displacement.

Fig. 3.4.3. Schematic diagram of experiment setup for DC voltage.

Substituting the PZT constants of Table 3 in (10) we have the following sensitivity for the

Notice that the larger the LFBG, the greater the strain experienced by the FBG and

Physical and dielectric properties Value

PZT type PZT4 Ceramic Shape Ring

Curie Temperature (Tc) 325°C Number of elements in stack n=10

Photo-elastic coefficient <sup>e</sup> =0.22

Piezoeletric strain constant d33=300 pm/V Thickness of ceramic w =4 mm Maximum allowed direct field strength 1-2 kV/mm Maximum allowed reverse field strength 350-500 V/mm

Bragg wavelength <sup>B</sup> = 1535.18 nm

Coefficient of thermal expansion =0.55 x 10-6/C Thermo-optic coefficient (dn/dT) =8.6 x 10-6/C Length of FBG L=28 mm

A DC voltage was applied on the PZT crystal terminals by using a high voltage supply and the displacement of the PZT was converted into variations of the Bragg central wavelength. The interrogation system for DC voltage measurements is schematically illustrated in Fig. 3.4.3. The light from an amplified spontaneous emission (ASE) ranging from 1520 nm to 1610 nm was used to illuminate the sensor and a commercial interrogation system from FOS&S model Spectral Eye 400, with accuracy of 2.0 pm was used to measure the reflected

applied DC voltage:

PZT

FBG

Table 3. FBG and PZT Parameters

**3.4.3 Optical setup for DC high voltage input** 

FBG spectrum accordingly to the sensor displacement.

Fig. 3.4.3. Schematic diagram of experiment setup for DC voltage.

consequently, the greater the sensitivity.

ΔλB=λB(1-ρe) nd33V/L� (2)

ΔλB/ΔV=128.3 pm/kV� (3)

For the first experiment, only DC voltages were applied to the PZT in order to measure the Bragg displacement accurately by the interrogation system. Eq. 9 was used to calculate the maximum voltage to be applied to the PZT ceramic and do not exceed the allowed value, accordingly with the Table 3.

$$\mathbf{V} = \mathbf{E}.\mathbf{d}\_{\ddot{\mathbf{j}}}\tag{4}$$

By applying a DC voltage to the PZT and recording the respective Bragg shift we can see the linear relationship between the applied voltage and the central Bragg wavelength (Fig. 3.4.4). The results show that the measured sensitivity was of 91.5 pm/kV and the correlation coefficient (R2) were 0.999.

Fig. 3.4.4. FBG-PZT sensor curve when a DC voltage is applied

#### **3.4.4 The optical setup with a AC voltage power supply**

Fig. 3.4.5 represents the interrogation system for AC voltage measurements. Since the optical spectrum analyzer is too slow to respond to the 60-Hz line frequency, the central wavelength variation can be obtained by using a photo-detector. The light from the ASE illuminates the FBG-PZT sensor via an optical circulator. The reflected spectrum of the sensor pass through the Fabry-Perot tunable filter (FFP-TF) with 0.89 nm bandwidth, nominal finesse of 130 and 116 nm of free spectral range (FSR). The light signal enters an amplified photo-detector with designed for detection of light signal over 700 nm – 1800 nm. The AC output signal is monitored by an oscilloscope. The FFP filter was tuned in 1540.04 nm by applying a voltage of the 7.2 volts. This point indicates that the FFP filter is matched with the FBG sensor on the stack and the intensity of light at the photo-detector is at maximum.

Fig 3.4.5. Schematic diagram of experiment setup for AC voltage.

Optical Fiber Sensors 27

variation will affect not only the sensor response, but also all parts of the transducer, producing unwanted drifts. However for DC only, a simple high-pass filter easily filters out temperature drifts from the output signal. A picture of the sensor on the high voltage rig is

This experiment provided information for the mathematical model developed in section IV and showed a good repeatability in sets of measurements and a correlation coefficient

> 0 250 500 750 1000 1250 1500 1750 2000 2250 2500 **Vin (V)**

In this work, it was presented the development of an optical high voltage transformer based in FBG and PZT piezoelectric ceramics for use on a 13.8 kV-Class electric power

y = 8,972E-05x + 1,537E+03 R2 = 9,975E-01

1536,7 1536,72 1536,74 1536,76 1536,78 1536,8 1536,82 1536,84 1536,86 1536,88 1536,9 1536,92 1536,94

Fig. 3.4.8. FBG-PZT sensor curve when a DC voltage is applied

**Bragg wavelength shift (nm)**

Fig. 3.4.9. Photograph of the FBG-PZT sensor.

**3.4.6 Conclusions** 

shown in Fig. 3.4.9.

R2=0,997.

From Fig. 3.4.6 we can see a linear relationship between the AC voltage applied to the FBG-PZT sensor and the output signal. A high voltage source was used to supply the input signal ranging from 0 kV to 2 kV at the terminals of the PZT electrodes, and on Figure 3.4.7 we can see the sequence of screens of the oscilloscope when an increment of AC voltage is applied to the PZT terminals.

Fig. 3.4.6. Relationship between the input AC voltage versus output signal

Fig. 3.4.7. Photodetector output signal for each increment of the AC voltage applied (Vrms)

#### **3.4.5 Mechanical and temperature stability**

All mechanicals parts are very rigid, including the PZT ceramics and the FBG itself which presents a Young modulus of 70 GPa, close to that of steel as measured in section II. In an OVT the vibrations are mainly of 60 Hz, due to magnetic movement of the transformer core. However, in the case of an optical CT, there is no iron core to vibrate and then this equipment is noiseless and does not present this kind the vibrations.

Figure 3.4.8 shows the results for several acquisitions employing the sensor, where a low dispersion of results when a DC voltage is applied on the terminals of the sensor can be observed. However, it is important to notice that one degree Celsius in temperature change will cause an approximately 14 pm Bragg wavelength displacement. Therefore, temperature compensation is important in DC/AC applications because the drift caused by temperature

From Fig. 3.4.6 we can see a linear relationship between the AC voltage applied to the FBG-PZT sensor and the output signal. A high voltage source was used to supply the input signal ranging from 0 kV to 2 kV at the terminals of the PZT electrodes, and on Figure 3.4.7 we can see the sequence of screens of the oscilloscope when an increment of AC voltage is

> 0 250 500 750 1000 1250 1500 1750 2000 2250 **Vin (V)**

Fig. 3.4.7. Photodetector output signal for each increment of the AC voltage applied (Vrms)

All mechanicals parts are very rigid, including the PZT ceramics and the FBG itself which presents a Young modulus of 70 GPa, close to that of steel as measured in section II. In an OVT the vibrations are mainly of 60 Hz, due to magnetic movement of the transformer core. However, in the case of an optical CT, there is no iron core to vibrate and then this

Figure 3.4.8 shows the results for several acquisitions employing the sensor, where a low dispersion of results when a DC voltage is applied on the terminals of the sensor can be observed. However, it is important to notice that one degree Celsius in temperature change will cause an approximately 14 pm Bragg wavelength displacement. Therefore, temperature compensation is important in DC/AC applications because the drift caused by temperature

y = 7,837E-04x - 7,341E-02 R2 = 9,955E-01

Fig. 3.4.6. Relationship between the input AC voltage versus output signal

0 0,25 0,5 0,75 1 1,25 1,5 1,75 2 2,25

**Vout (V)**

**3.4.5 Mechanical and temperature stability** 

equipment is noiseless and does not present this kind the vibrations.

applied to the PZT terminals.

variation will affect not only the sensor response, but also all parts of the transducer, producing unwanted drifts. However for DC only, a simple high-pass filter easily filters out temperature drifts from the output signal. A picture of the sensor on the high voltage rig is shown in Fig. 3.4.9.

This experiment provided information for the mathematical model developed in section IV and showed a good repeatability in sets of measurements and a correlation coefficient R2=0,997.

Fig. 3.4.8. FBG-PZT sensor curve when a DC voltage is applied

Fig. 3.4.9. Photograph of the FBG-PZT sensor.

#### **3.4.6 Conclusions**

In this work, it was presented the development of an optical high voltage transformer based in FBG and PZT piezoelectric ceramics for use on a 13.8 kV-Class electric power

Optical Fiber Sensors 29

failure of the generators or transformers. Paradoxally, since current and voltage maintain their values approximately the same all over the HEP, there are much more temperature

With the idea of decreasing the amount of copper wires, facilitating the maintenance, possibility of remote sensing and consequently decreasing costs, we designed a fiber-optic multiplexed temperature sensor for application in large air cooled HEP. The system has the objective to cover all temperature monitoring needs of a HEP that would also overcome some the disadvantages presented by the conventional RTD (resistive temperature detector)

The Eletronorte, the largest electric energy producer in Brazil, contracted the Instrumentation and Photonics Laboratory, at the Federal University of Rio de Janeiro to project, test and install a complete FBG system to monitor the temperatures inside hydro-

This paper relates the world's first real application, test and operation of a FBG temperature sensor array inside a fully operational and connected-to-the-grid hydro-electric power

Although hydro-generators are very reliable, the temperature monitoring of these machines is a well-established procedure. The reason for this is that the stator windings, cooper and insulation, age over time and tend to degrade when the machine operates at relatively high temperatures such as the in the range 100-120oC [Stone, G. C., 1999]. Keeping the temperature below these limits is not easy because large hydrogenerators stators and rotor can weigh as much as 1,000 tons and 1,700 tons, respectively, and as a consequence, these

For keeping the temperature below these limits, large hydro-electric machines of 40 MW or more are normally air cooled. These generators are supplied with a closed air-cooling circuit where the air is cooled by a water refrigerated radiator. In this type of generator the air

The temperature monitoring of the cooling air or directly inside the stator winding conductors are the most reliable methods of assuring the proper operation of power generator [Stone, G. C., 1999] and for these measurements, the most popular sensor is the Pt-100, meaning 100 Ω platinum resistance sensor, also known as RTD. These sensors are placed at various locations within the generator, for instance, in the cooling air passages, inside the lubricant and hydraulic oil pipes, in the bearings and also inserted into the slots of the stator core, summing up to about fifteen or more sensors for each

These reliable, accurate and relatively inexpensive sensors are in use by the industry for almost a century and perfectly fulfill all temperature monitoring needs of a HEP. They have disadvantages, though, that can be mentioned: a) sensitivity to electromagnetic interference (EMI), demanding low pass filters; b) tendency to carry the high voltage of the generator to the control room if short circuits occur; c) tendency to burn inside the slots of the stator winding where they cannot be replaced. Additionally, each sensor is driven by a three wire harness that needs to go all the way from the machine to the control room where a large rack with many modules receives each sensor harness. For larger distances it is necessary to

temperature is monitored before and after it passes through the radiators.

control points in a HEP than there are for current or voltage.

**3.5.2 Hydrogenerator temperature monitoring** 

machines have a big thermal time constant.

network.

generator.

generator.

machine.

transmission system The advantages of piezoelectric material with the characteristics of a sensor fiber Bragg grating is employed. For the assembly of the prototype used, the aluminum structure was designed in order to support a larger number of ceramic rings and thus increase the longitudinal displacement of the material by improving the resolution of the demodulation system. An aspect to be considered is related to the maximum field strength allowed according to the manufacturer's specifications restricting the voltage applied to the sensor, an aspect that can be solved with a capacitive divider. Despite this limitation, the results make it viable the use of this technology for monitoring power substations. In order to improve the system and increase accuracy, a more appropriate setup is under development. An increased longitudinal displacement can be obtained with a new prototype sensor based on ceramics with a higher piezoelectric charge constant and by encapsulation of the sensor by increasing the sensitivity.

#### **3.5 Fiber Bragg grating temperature sensor system applied to large air cooled hydrogenerators**

#### **3.5.1 Introduction**

This project describes the research, project, construction, calibration, installation and operation of a fiber Bragg grating based fiber-optic system applied to a hydro-electric generator to perform a continuous monitoring of temperature. After being deployed for two and a half years, the system has proved itself to be capable of reliably and accurately measuring and monitoring temperatures inside the generator, even taking into consideration the harsh environment of the stator. The results were considered satisfactory, demonstrating the usefulness of the fiber-optic system in power generation equipment.

The technology of power generation by hydro-electric plants (HEP) in Brazil has reached a high level of sophistication and investment. Nowadays about 73% of all electric energy produced in the country is from HEPs, including very large ones such as Itaipu and Tucuruí with 14 and 8.3 GW respectively. This figure will increase further with the already in construction Rio Madeira Complex whose 88 turbines will produce over 6.8 GW and Belo Monte (11.3 GW) in licensing processes.

This electric grid represents a very high capital invested which is also of very expensive maintenance. Each minute down time of any piece of equipment could cost the energy providers thousands of dollars from profit losses of undelivered energy and also from several types of fines applied by the National Electric Power Agency which they are subjected to.

For this reason the reliability of equipment has become a highest priority and many control systems have been designed to protect and perform real time diagnosing for prompt shutdown or warnings in case of faults.

The main control parameter in any HEP or substation is, of course, the electric current that can rise without limits in case a short-circuit or excess load occurs. The second parameter in importance is the voltage that may present surges or transients due to switching or atmospheric discharges. The third parameter, normally a consequence of the current, is the temperature that must be under severe observation since rises above 100oC can accelerate aging of the insulating material and conductors or even destroy them, causing a general

transmission system The advantages of piezoelectric material with the characteristics of a sensor fiber Bragg grating is employed. For the assembly of the prototype used, the aluminum structure was designed in order to support a larger number of ceramic rings and thus increase the longitudinal displacement of the material by improving the resolution of the demodulation system. An aspect to be considered is related to the maximum field strength allowed according to the manufacturer's specifications restricting the voltage applied to the sensor, an aspect that can be solved with a capacitive divider. Despite this limitation, the results make it viable the use of this technology for monitoring power substations. In order to improve the system and increase accuracy, a more appropriate setup is under development. An increased longitudinal displacement can be obtained with a new prototype sensor based on ceramics with a higher piezoelectric charge constant and by encapsulation of the sensor by increasing the

**3.5 Fiber Bragg grating temperature sensor system applied to large air cooled** 

This project describes the research, project, construction, calibration, installation and operation of a fiber Bragg grating based fiber-optic system applied to a hydro-electric generator to perform a continuous monitoring of temperature. After being deployed for two and a half years, the system has proved itself to be capable of reliably and accurately measuring and monitoring temperatures inside the generator, even taking into consideration the harsh environment of the stator. The results were considered satisfactory, demonstrating the usefulness of the fiber-optic system in power generation

The technology of power generation by hydro-electric plants (HEP) in Brazil has reached a high level of sophistication and investment. Nowadays about 73% of all electric energy produced in the country is from HEPs, including very large ones such as Itaipu and Tucuruí with 14 and 8.3 GW respectively. This figure will increase further with the already in construction Rio Madeira Complex whose 88 turbines will produce over 6.8 GW and Belo

This electric grid represents a very high capital invested which is also of very expensive maintenance. Each minute down time of any piece of equipment could cost the energy providers thousands of dollars from profit losses of undelivered energy and also from several types of fines applied by the National Electric Power Agency which they are

For this reason the reliability of equipment has become a highest priority and many control systems have been designed to protect and perform real time diagnosing for prompt

The main control parameter in any HEP or substation is, of course, the electric current that can rise without limits in case a short-circuit or excess load occurs. The second parameter in importance is the voltage that may present surges or transients due to switching or atmospheric discharges. The third parameter, normally a consequence of the current, is the temperature that must be under severe observation since rises above 100oC can accelerate aging of the insulating material and conductors or even destroy them, causing a general

sensitivity.

equipment.

subjected to.

Monte (11.3 GW) in licensing processes.

shutdown or warnings in case of faults.

**hydrogenerators 3.5.1 Introduction**  failure of the generators or transformers. Paradoxally, since current and voltage maintain their values approximately the same all over the HEP, there are much more temperature control points in a HEP than there are for current or voltage.

With the idea of decreasing the amount of copper wires, facilitating the maintenance, possibility of remote sensing and consequently decreasing costs, we designed a fiber-optic multiplexed temperature sensor for application in large air cooled HEP. The system has the objective to cover all temperature monitoring needs of a HEP that would also overcome some the disadvantages presented by the conventional RTD (resistive temperature detector) network.

The Eletronorte, the largest electric energy producer in Brazil, contracted the Instrumentation and Photonics Laboratory, at the Federal University of Rio de Janeiro to project, test and install a complete FBG system to monitor the temperatures inside hydrogenerator.

This paper relates the world's first real application, test and operation of a FBG temperature sensor array inside a fully operational and connected-to-the-grid hydro-electric power generator.

#### **3.5.2 Hydrogenerator temperature monitoring**

Although hydro-generators are very reliable, the temperature monitoring of these machines is a well-established procedure. The reason for this is that the stator windings, cooper and insulation, age over time and tend to degrade when the machine operates at relatively high temperatures such as the in the range 100-120oC [Stone, G. C., 1999]. Keeping the temperature below these limits is not easy because large hydrogenerators stators and rotor can weigh as much as 1,000 tons and 1,700 tons, respectively, and as a consequence, these machines have a big thermal time constant.

For keeping the temperature below these limits, large hydro-electric machines of 40 MW or more are normally air cooled. These generators are supplied with a closed air-cooling circuit where the air is cooled by a water refrigerated radiator. In this type of generator the air temperature is monitored before and after it passes through the radiators.

The temperature monitoring of the cooling air or directly inside the stator winding conductors are the most reliable methods of assuring the proper operation of power generator [Stone, G. C., 1999] and for these measurements, the most popular sensor is the Pt-100, meaning 100 Ω platinum resistance sensor, also known as RTD. These sensors are placed at various locations within the generator, for instance, in the cooling air passages, inside the lubricant and hydraulic oil pipes, in the bearings and also inserted into the slots of the stator core, summing up to about fifteen or more sensors for each machine.

These reliable, accurate and relatively inexpensive sensors are in use by the industry for almost a century and perfectly fulfill all temperature monitoring needs of a HEP. They have disadvantages, though, that can be mentioned: a) sensitivity to electromagnetic interference (EMI), demanding low pass filters; b) tendency to carry the high voltage of the generator to the control room if short circuits occur; c) tendency to burn inside the slots of the stator winding where they cannot be replaced. Additionally, each sensor is driven by a three wire harness that needs to go all the way from the machine to the control room where a large rack with many modules receives each sensor harness. For larger distances it is necessary to

Optical Fiber Sensors 31

Equivalently, a variation of temperature can also change both parameters, by thermal

With such a device, by injecting a spectrally broadband source of light into the fiber, a narrowband spectral component at the Bragg wavelength will be reflected by the grating. In the transmitted light, this spectral component will be missed but the remaining of this light can be used to illuminate other FBGs in the same fiber, each one tuned in a different wavelength. The final result of such arrangement is that we will have at the beginning of the fiber all Bragg peak reflections of each FBG, each one in its specific

Now, by designing the proper interface, measurands can be made to impinge perturbation on the grating resulting in a shift in the Bragg wavelength which can then be used as a

Starting from the theorem of the conservation of energy and momentum, after a series of algebraic manipulations, very well detailed in [Othonos and Kalli, 1999], one arrives to the following equation, which establishes the relationship between the Bragg wavelength, strain

1 *<sup>B</sup> e z*

Where, <sup>z</sup> is the longitudinal strain; T is the temperature variation; <sup>e</sup> is the photo-elastic coefficient; is the thermal expansion coefficient and is the thermo-optic coefficient, representing the temperature dependence of the refractive index (dn/dT). For materials with positive thermal expansion coefficient, the index of refraction normally decreases with temperature. These parameters have the following values for a silica fiber with a germanium

e=0.22; =0.55 x 10-6/C; and =8.6 x 10-6/C. Since we want to measure only the temperature, we must protect the fiber against strain by placing the grating portion of the fiber inside a protective tubing, for instance. Thus the sensitivity of the grating for temperature at the wavelength range of 1550 nm is, after

> <sup>0</sup> 14.18 *<sup>B</sup> pm T C*

This theoretical value, though, is not absolute as each FBG of the same fabrication batch will

Before installing the sensors into the generator they had to be calibrated because, as already mentioned, (3) is not valid for all FBGs as they may have different thermo-optic coefficients

present slightly different sensitivities, as we will see later in the following sections.

 

*T*

(2)

(3)

*B*

dilation and by the thermo-optic effect respectively.

wavelength range.

parameter transducer.

doped core:

and temperature applied to the FBG:

substituting the constants in (2):

**3.5.4 Calibration of sensors** 

and they are tuned into different wavelengths.

use a current loop to carry the information signals, therefore a terminal box must be installed close to the sensor location with amplifiers, filters and converters to 4-to-20 mA, for example. In a relatively large HEP with ten generators, there are many terminal boxes, harnesses, plug-in modules and racks all over the plant with hundreds of kilometers of electric wires. This is the principal aspect where a multiplexed sensor array can help, as with only a few fiber-optic cables the system can manage all temperature check points of the whole plant.

The feasibility of application FBG sensors in electric machines for temperature monitoring has been the theme of many recent works. One of them is the paper from a Siemens AG engineer team [Theune, et al., 2002] in which the authors investigate the application of FBG sensors embedded into the stator core of a generator on a test bench. This test demonstrated the viability of the FBG technique applied to generators. More recently, the internal temperatures of oil-immersed power transformers were measured by FBG arrays extending the application of this kind of fiber-optic sensor in electric machines [Kim et al., 2008, Weigen et al, 2008 and Ribeiro et al, 2008]

#### **3.5.3 FBG theory**

Fiber Bragg Grating (FBG) technology is one of the most popular choices for optical fiber sensors for strain or temperature measurements due to their simple manufacture (UV photo-inscribed) and relatively strong reflected signal strength. They are formed by a periodic modulation of the index of refraction of the fiber core along the longitudinal direction and can be produced by various techniques [Othonos and Kalli, 1999 and Meltz et al., 1989].

Since the strain or temperature measurands are encoded into wavelength shifts, these sensors are also self-calibrated because wavelength is an absolute parameter. Thus these sensors do not drift on the total light levels, losses in the connecting fibers and couplers or light source power. Additionally, the wavelength encoded nature of the output also allows the use of wavelength division multiplexing technique (WDM) by assigning each sensor to a different wavelength range of the available light source spectrum.

In the FBG, due to the periodic modulation of the index of refraction, light guided along the core of the fiber will be weakly backwards reflected by each grating plane. The contribution of the reflected light from each grating plane will add up with each other in the backward direction. This addition can be constructive or destructive, depending on whether the wavelength of the incoming light satisfies or not the Bragg condition, given by:

$$
\lambda\_{\rm B} = 2\mathbf{n}\_{\rm eff} \cdot \boldsymbol{\Lambda} \tag{1}
$$

Where, neff is the effective index of refraction of the fiber core and is the modulation period of the index of refraction.

Equation (1), also known as the Bragg reflection wavelength, is the peak wavelength of the narrowband spectral component reflected by each FBG of the array. The FWHM (full-widthhalf-maximum) or bandwidth of this reflection depends on several parameters, particularly the grating length. Typically, the FWHM is 0.05 to 0.3 nm in most sensor applications. Equation 30 also shows that the Bragg wavelength is a function of and neff. Thus we conclude that a longitudinal deformation due to an external force can change both and neff, the later by the photo-elastic effect and the former by increasing the pitch of the grating.

use a current loop to carry the information signals, therefore a terminal box must be installed close to the sensor location with amplifiers, filters and converters to 4-to-20 mA, for example. In a relatively large HEP with ten generators, there are many terminal boxes, harnesses, plug-in modules and racks all over the plant with hundreds of kilometers of electric wires. This is the principal aspect where a multiplexed sensor array can help, as with only a few fiber-optic cables the system can manage all temperature check points of the

The feasibility of application FBG sensors in electric machines for temperature monitoring has been the theme of many recent works. One of them is the paper from a Siemens AG engineer team [Theune, et al., 2002] in which the authors investigate the application of FBG sensors embedded into the stator core of a generator on a test bench. This test demonstrated the viability of the FBG technique applied to generators. More recently, the internal temperatures of oil-immersed power transformers were measured by FBG arrays extending the application of this kind of fiber-optic sensor in electric machines [Kim et al., 2008, Wei-

Fiber Bragg Grating (FBG) technology is one of the most popular choices for optical fiber sensors for strain or temperature measurements due to their simple manufacture (UV photo-inscribed) and relatively strong reflected signal strength. They are formed by a periodic modulation of the index of refraction of the fiber core along the longitudinal direction and can be produced by various techniques [Othonos and Kalli, 1999 and Meltz et

Since the strain or temperature measurands are encoded into wavelength shifts, these sensors are also self-calibrated because wavelength is an absolute parameter. Thus these sensors do not drift on the total light levels, losses in the connecting fibers and couplers or light source power. Additionally, the wavelength encoded nature of the output also allows the use of wavelength division multiplexing technique (WDM) by assigning each sensor to a

In the FBG, due to the periodic modulation of the index of refraction, light guided along the core of the fiber will be weakly backwards reflected by each grating plane. The contribution of the reflected light from each grating plane will add up with each other in the backward direction. This addition can be constructive or destructive, depending on whether the

Where, neff is the effective index of refraction of the fiber core and is the modulation

Equation (1), also known as the Bragg reflection wavelength, is the peak wavelength of the narrowband spectral component reflected by each FBG of the array. The FWHM (full-widthhalf-maximum) or bandwidth of this reflection depends on several parameters, particularly the grating length. Typically, the FWHM is 0.05 to 0.3 nm in most sensor applications. Equation 30 also shows that the Bragg wavelength is a function of and neff. Thus we conclude that a longitudinal deformation due to an external force can change both and neff, the later by the photo-elastic effect and the former by increasing the pitch of the grating.

B ff λ 2n*<sup>e</sup>* (1)

wavelength of the incoming light satisfies or not the Bragg condition, given by:

different wavelength range of the available light source spectrum.

whole plant.

**3.5.3 FBG theory** 

al., 1989].

gen et al, 2008 and Ribeiro et al, 2008]

period of the index of refraction.

Equivalently, a variation of temperature can also change both parameters, by thermal dilation and by the thermo-optic effect respectively.

With such a device, by injecting a spectrally broadband source of light into the fiber, a narrowband spectral component at the Bragg wavelength will be reflected by the grating. In the transmitted light, this spectral component will be missed but the remaining of this light can be used to illuminate other FBGs in the same fiber, each one tuned in a different wavelength. The final result of such arrangement is that we will have at the beginning of the fiber all Bragg peak reflections of each FBG, each one in its specific wavelength range.

Now, by designing the proper interface, measurands can be made to impinge perturbation on the grating resulting in a shift in the Bragg wavelength which can then be used as a parameter transducer.

Starting from the theorem of the conservation of energy and momentum, after a series of algebraic manipulations, very well detailed in [Othonos and Kalli, 1999], one arrives to the following equation, which establishes the relationship between the Bragg wavelength, strain and temperature applied to the FBG:

$$\frac{\Delta\lambda\_{\oplus}}{\lambda\_{\oplus}} = \left(1 - \rho\_{e}\right)\varepsilon\_{z} + \left(\alpha + \eta\right)\Delta T \tag{2}$$

Where, <sup>z</sup> is the longitudinal strain; T is the temperature variation; <sup>e</sup> is the photo-elastic coefficient; is the thermal expansion coefficient and is the thermo-optic coefficient, representing the temperature dependence of the refractive index (dn/dT). For materials with positive thermal expansion coefficient, the index of refraction normally decreases with temperature. These parameters have the following values for a silica fiber with a germanium doped core:

$$\begin{array}{c} \text{\(\mu\text{=}0.22\text{;} \cdot \text{ and } \cdot \text{ is } 10^{\text{-6}}\text{/}\text{\textdegree C;} \cdot \text{ and } \cdot \text{\(\mu\text{=}0.6\text{ or } 10^{\text{-6}}\text{/}\text{\textdegree C.}\text{\)} \\ \text{and } \eta\text{=} 8.6 \times 10^{\text{-6}}\text{/}\text{\textdegree C.} \end{array}$$

Since we want to measure only the temperature, we must protect the fiber against strain by placing the grating portion of the fiber inside a protective tubing, for instance. Thus the sensitivity of the grating for temperature at the wavelength range of 1550 nm is, after substituting the constants in (2):

$$\frac{\Delta\mathcal{L}\_{\text{B}}}{\Delta T} = 14.18 \frac{pm}{^{0}C} \tag{3}$$

This theoretical value, though, is not absolute as each FBG of the same fabrication batch will present slightly different sensitivities, as we will see later in the following sections.

#### **3.5.4 Calibration of sensors**

Before installing the sensors into the generator they had to be calibrated because, as already mentioned, (3) is not valid for all FBGs as they may have different thermo-optic coefficients and they are tuned into different wavelengths.

Optical Fiber Sensors 33

1 14.00 10.3 1530,534 2 14.10 11.6 1540,667 3 14.16 10.2 1547,027 4 14.21 10.5 1553,035 5 14.27 10.6 1559,063 6 14.32 9.9 1565,090

Notice in Table 4 that, as predicted by (3), the theoretical sensitivities are different from those obtained in the calibration experiment. But, since all FBGs were made out of the same optical-fiber reel, α, the silica coefficient of temperature should be the same for all FBGs produced from that fiber. The other parameter in (3) is , the thermo-optic coefficient, representing the temperature dependence of the refractive index (dn/dT). Equation (1) teaches us that λB is a function of neff, the average index of refraction between the pristine fiber core and that of the ultra-violet-irradiated core. Recall that during the FBG fabrication, the radiation time for each FBG inscription is not the same as the operator turns off the laser only when she observes the Bragg reflection above a certain level. Since the UV irradiation modifies the index of refraction of the fiber core, it is possible that it could also modify differently the values of in each FBG, ending up to the slightly dispersed sensitivities found above. However, to the best of our knowledge, there is no mention of this effect

The data obtained from Fig. 3.29 also allow us to calculate the linear relationship between wavelength and temperature for each FBG. These equations were fed into the software in

The second step of the calibration procedure was the comparison between the calculated temperatures by the optical interrogator software with the calibrated temperatures of each FBG, as given by a precision thermometer. From this experiment it was possible to calculate the inaccuracy of the measurement which was less than 0.5oC, quite sufficient for this application.

The correlation coefficient of the linear curve fitting was 0.9994 as shown in Fig. 3.5.3.

Measured Sensitivity (pm/oC) Wavelength @ 25oC (nm)

Table 4 summarizes the information acquired from the last experiment.

Sensitivity (pm/oC)

Table 4. Theoretical and measured sensitivities of each FBG.

Sensor # Theoretical

whatsoever in the literature.

Fig. 3.5.3. Calibration of FBG 2.

order to calculate the temperature of each sensor.

Fig. 3.5.1. Superimposed wavelength shift of each FBG as temperature varies from 25oC to 95oC.

The calibration procedure of the sensors followed two steps. In the first set of measurements, the six sensors were calibrated simultaneously by immersion into a temperature controlled bath and the Bragg wavelengths were monitored along with the temperature in order to calculate the sensitivity of each sensor, as predicted by (3).

This first set of measurements allowed us to observe and measure the Bragg shift of each FBG as a function of temperature in the range of 25oC to 95oC. Fig. 3.5.1 shows all Bragg reflection of each temperature superimposed. In this experiment it is important to make sure that each pulse does not enter the neighbor's range during its displacement.

From this data the software calculates the center wavelength of each Bragg reflection and plots the Bragg shift versus temperature for each FBG, producing the graph shown in Fig. 3.5.2.

Fig. 3.5.2. Wavelength shift versus Temperature for each FBG.

Fig. 3.5.1. Superimposed wavelength shift of each FBG as temperature varies from 25oC

The calibration procedure of the sensors followed two steps. In the first set of measurements, the six sensors were calibrated simultaneously by immersion into a temperature controlled bath and the Bragg wavelengths were monitored along with the

This first set of measurements allowed us to observe and measure the Bragg shift of each FBG as a function of temperature in the range of 25oC to 95oC. Fig. 3.5.1 shows all Bragg reflection of each temperature superimposed. In this experiment it is important to make sure

From this data the software calculates the center wavelength of each Bragg reflection and plots the Bragg shift versus temperature for each FBG, producing the graph shown in Fig.

temperature in order to calculate the sensitivity of each sensor, as predicted by (3).

that each pulse does not enter the neighbor's range during its displacement.

Fig. 3.5.2. Wavelength shift versus Temperature for each FBG.

to 95oC.

3.5.2.


Table 4 summarizes the information acquired from the last experiment.

Table 4. Theoretical and measured sensitivities of each FBG.

Notice in Table 4 that, as predicted by (3), the theoretical sensitivities are different from those obtained in the calibration experiment. But, since all FBGs were made out of the same optical-fiber reel, α, the silica coefficient of temperature should be the same for all FBGs produced from that fiber. The other parameter in (3) is , the thermo-optic coefficient, representing the temperature dependence of the refractive index (dn/dT). Equation (1) teaches us that λB is a function of neff, the average index of refraction between the pristine fiber core and that of the ultra-violet-irradiated core. Recall that during the FBG fabrication, the radiation time for each FBG inscription is not the same as the operator turns off the laser only when she observes the Bragg reflection above a certain level. Since the UV irradiation modifies the index of refraction of the fiber core, it is possible that it could also modify differently the values of in each FBG, ending up to the slightly dispersed sensitivities found above. However, to the best of our knowledge, there is no mention of this effect whatsoever in the literature.

The data obtained from Fig. 3.29 also allow us to calculate the linear relationship between wavelength and temperature for each FBG. These equations were fed into the software in order to calculate the temperature of each sensor.

The second step of the calibration procedure was the comparison between the calculated temperatures by the optical interrogator software with the calibrated temperatures of each FBG, as given by a precision thermometer. From this experiment it was possible to calculate the inaccuracy of the measurement which was less than 0.5oC, quite sufficient for this application. The correlation coefficient of the linear curve fitting was 0.9994 as shown in Fig. 3.5.3.

Fig. 3.5.3. Calibration of FBG 2.

Optical Fiber Sensors 35

within the existing cable trays along with other electric cables following all the way up, from the generator to the HEP control room where the optical interrogator and an industrial PC

The optical interrogation setup consists of a broad band optical source that illuminates all FBGs in the array. The return signal of each FBG is detected by an optical spectrum analyzer (OSA) that identifies the center wavelength of each FBG reflection pulse. The OSA communicates with an industrial PC via RS-232 interface, running a LabView software for calculation and storage of the temperatures. The PC publishes all data into the company's Intranet that automatically and instantaneously become available to the HEP central

However, the proposed system goes much further in ambition. After the approval of the current system, the proposed project planes to use this technology to fulfill all temperature needs of the HEP, including turbines, air, oil and water ducts and other electrical equipment as well at the substation (see Fig. 3.5.6). Since a single optical fiber cable can monitor about 16 or more sensors, it is just necessary one cable per equipment for all temperature measurements. The system is also intended to access the Internet so as to be able to be accessed remotely, even from another location. This is especially advantageous for

Fig. 3.5.5. Depiction of a cross-section of the hydro-generator (left). Generator in detail with sensors connected to the interrogation system (right).1 to 6) FBG sensors; 7) Radiator;

Shortly after the installation we noticed that the last two FBGs in the fiber-optic cable were not identified by the optical interrogator, probably due a malfunction of the optical connectors. But there was no time to open up again the inspection windows of the stator as the machine was programmed to start up immediately. Since then, the machine did not stop again as our requests for shutting down were not granted so far. Currently, at the time of writing this article, the machine is in operation for two and a half years and the fiber-optic system is monitoring normally four radiators. The results of the measurements are sent periodically to the company's head-quarters in Belém, some 1,800 km north and from there

8) Stator; 9) Machine room; 10) Bearing; 11) Kaplan turbine.

to our laboratory located in Rio de Janeiro, 2,400 km south.

software control. Fig. 3.5.5 shows the block diagram of the system.

were installed.

automatic unmanned substations.
