You are on page 1of 10

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO.

2, APRIL 2012

953

Sensor Network for Monitoring the State of Pollution of High-Voltage Insulators Via Satellite
Eduardo Fontana, Member, IEEE, Joaquim F. Martins-Filho, Senior Member, IEEE, Srgio C. Oliveira, Member, IEEE, Fernando J. M. M. Cavalcanti, Ricardo A. Lima, Gustavo O. Cavalcanti, Thiago L. Prata, and Renato B. Lima

AbstractThis paper describes the design and development of a sensor system network for remote monitoring of the state of pollution on high-voltage insulators, its deployment among four states in Northeast Brazil and preliminary analysis of the data gathered by the system. Six monitoring units, each composed by leakage current, humidity, and temperature sensors, an electronic processing module, and a satellite-communication link, were built and installed on 230- and 500-kV transmission towers. Measured data are sent by the different systems via satellite to a database controlled by a web-based graphical user interface, incorporated into the supervision-management system of the electric utility company. Sample eld results on one of the sensor systems, both for short- and long-term measurements, show a direct correlation of the activity recorded by the system with the pollution level expected to occur near the corresponding monitoring point. Index TermsInsulators, leakage current, microcontrollers, optical signal detection, partial discharges (PDs), power transmission lines, satellite mobile communication, transducers.

I. INTRODUCTION HE accumulation of pollutants on the surface of high-voltage (HV) insulators is one of the main reasons for the decrease of the insulation between phase and ground in transmission-line towers [1][12]. Under high humidity, the pollutants form a conductive layer on the insulator surface. The nonuniform leakage current ow inside the conductive layer forms dry bands, concentrating the electric eld, leading to the creation of partial discharges (PDs) near the dry bands [9]. The PD phenomena can increase its rate and intensity until a complete discharge from line to ground, known as ashover, occurs. According to the Hydroelectric Company of San Francisco, CHESF, which is responsible for the energy generation and distribution in Northeast Brazil, ashovers are one of the main causes of failures in its transmission lines
Manuscript received July 29, 2011; revised November 16, 2011; accepted December 21, 2011. Date of publication February 20, 2012; date of current version March 28, 2012. This work was supported by Companhia Hidro Eltrica do So Francisco (CHESF), under Contract CT-I-2008.0210.00. Paper no. TPWRD00650-2011. E. Fontana, J. F. Martins-Filho, R. A. Lima, T. L. Prata, and R. B. Lima are with the Department of Electronics and Systems, Federal University of Pernambuco, Recife, PE, 50740-550, Brazil (e-mail: fontana@ufpe.br). S. C. Oliveira and G. O. Cavalcanti are with the Polytechnic School of Pernambuco, University of Pernambuco, Recife, PE, 52720-001, Brazil (e-mail: scampello@ecomp.poli.br). F. J. M. M. Cavalcanti is with the Companhia Hidro Eltrica do So Francisco (CHESF), Recife, PE, 50761-901, Brazil (e-mail: fernandj@chesf.gov.br). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TPWRD.2012.2183623

[12]. As a result, the company, similar to other electric utility companies, has to mobilize part of its maintenance personnel to observe, generally by periodic visual inspections, the amount and attributes of PDs near the insulators in order to determine what is the most adequate time for washing the affected units within an extended region. The leakage current signal consists of a sinusoidal waveform superposed by intermittent and short pulses correlated with the PDs. From previous experiments, it was observed that the amplitude, rate, and duration of these pulses were directly associated with the degree of pollution on the insulator strings [13]. It was found that one way to indirectly infer the state of pollution of insulators was to classify the amplitude and rate of the short pulses superimposed on the leakage current waveform [13]. In previous work, the development of a system for monitoring pulses within the leakage current waveform as well as environmental humidity was reported [14]. More recently, the integration of the measurement system with a satellite-communication system was proposed as a strategy for remote sensing [15], and preliminary one-month activity data in the eld were reported [16] from a network of six sensor systems spread among four states in Northeast Brazil. This paper describes all aspects for system development and laboratory characterization, including details of the work involved in the deployment of this sensor network as well as a more in-depth analysis of data from longerterm eld results recorded on one of the systems. II. SENSOR SYSTEM NETWORK Fig. 1 shows a diagram of the sensor system network for the real-time monitoring of the PDs on the surface of HV insulator strings as well as of the environmental temperature and humidity. At the present stage, the network comprises six monitoring systems, a private service network to acquire the information via satellite, and a client intranet network with access to the server database. Each terminal in the network has access to the server via a graphical user interface (GUI). In the following subsections, these three parts of the sensor system network are described in more detail. A. Optical Sensor and Processing Module The optical sensor illustrated in Fig. 2 consists of an 840-nm light-emitting diode (LED) in a pigtail conguration terminated in a ferrule-type ber-optic connector [13]. The optical sensor supports a maximum direct current of 300 mA and 1 A of maximum pulsed current amplitude. The optical current sensor is connected in parallel with the insulator closest to the ground

0885-8977/$31.00 2012 IEEE

954

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

Fig. 4. Processing module (PM) block diagram.

Fig. 1. Diagram of the satellite telemetry system for monitoring the state of pollution of HV insulators.

Fig. 2. Electric diagram of the optical sensor.

to the input of a comparator with hysteresis. The comparison stage is connected to a PIC 16F877A microcontroller. Moreover, a semiconductor temperature sensor and a capacitive humidity sensor are also connected to the microcontroller. The processing module also includes two BNC female connectors for real-time observation of the leakage current signal on two different ranges. The processing module is connected with the satellite modem by a DB-9 serial interface connector. The module also has a RJ-12 input connector for incircuit serial programming allowing a quick in-loco software update. Based on laboratory experiments, literature review, and previous eld experiments [13], the peak values of leakage current pulses are classied into four ranges, dened as for for for for mA mA mA mA (1) (2) (3) (4)

Fig. 3. Schematic of the sensor system conguration in the tower.

terminal in the tower, and shunts the leakage current through itself. To achieve this, the LED cathode is grounded to the tower structure and the anode is connected to the cap of the second insulator, so the system always records the positive excursions of the leakage current. The LED signal is transmitted through the optical-ber link that connects the sensor to the processing module, as shown in Fig. 3. The processing module, illustrated in more detail in Fig. 4, consists of detection, amplication, and comparison electronic stages. The detection stage has a silicon PIN photodetector with a 13-mm active area. The amplication stage consists of four fast response operational ampliers, each one with an independent adjustable gain. Each amplier is connected

The microcontroller program has interruption-based routines responsible for identifying the incoming leakage current pulses in through . Periodically, the control program all four ranges requests the universal time clock (UTC) register of the satellite modem to synchronize with its internal clock reference to maintain an accurate time reference. Pulse counts in each range are recorded in 18-bit temporary registers. Each nal count is stored in the PICs nonvolatile memory at the top of each hour and the corresponding temporary register is reset for the start of a new counting cycle, thus yielding a maximum recordable rate of 262 kpulses/h. The pulse counts, together with average temperature and humidity per hour, internally stored, are periodically sent to the satellite modem via an RS-232 serial communication port. B. Satellite System Conguration A satellite-communication approach was adapted for the monitoring system because most of the monitoring points in CHESFs transmission-line system are in remote areas without coverage of standard mobile or another wireless data-communication approach. In addition, although the transmission-line network has an accompanying optical ground wire (OPGW)

FONTANA et al.: SENSOR NETWORK FOR MONITORING THE STATE OF POLLUTION OF HV INSULATORS

955

Fig. 5. Photograph of the sensor system developed for eld experiments.

half hour. Under these conditions, the monthly fee per modem is approximately that of a commercial mobile-phone line. 2) Database and GUI Application in the Client Network: In the client network, an application developed in the Java 2 platform enterprise edition (J2EE) to the web, together with the programming languages for presentation, formatting, and organization, that is, HTML, CSS and JavaScript, respectively, with database and graphical interface attributes, was designed to store the data and to allow graphical representation of the information measured by the sensors installed in the eld. The application was named ADECI, a Portuguese acronym for Analysis of Electrical Performance of Insulator Strings. ADECI runs by conducting a secure Internet connection with the private satellite service network gateway, in turn, downloading the information sent via satellite by the sensor systems, as illustrated in Fig. 1. The information is stored in a relational database developed in the Oracle Database Management System (Oracle DBMS) platform. The information database is accessed in the intranet of the client network via the ADECI GUI, by means of a username and password. The ADECI system is responsible for registration, search, update, and removal of system users. It is possible to select the monitoring point for viewing the records as well as choose the time period of analysis. The events recorded through , along with temperature and within the ranges humidity, can be plotted graphically or presented in tabular form. In addition, the ADECI system allows the user to register specic inspection reports related to the HV tower where the monitoring system is installed. III. RESULTS

cable, the corresponding optical bers are dedicated for private use and could not be integrated with the monitoring system network. Under these restrictions, each sensor system was designed with an integrated satellite modem with a buit-in planar antenna, as illustrated in Fig. 5. The processing module and the satellite modem are powered by a photovoltaic system. As shown in Fig. 5, the processing module, the charge controller, and the battery of the photovoltaic power system are mounted inside a metal cabinet that provides protection to the system from the outer environment, as well as shielding from electromagnetic interference (EMI). The satellite modem is mounted on top of the cabinet. The cabinet is fabricated with handles to facilitate attachment to the transmission tower. 1) Satellite Service Network: The satellite service network consists of the satellite constellation, which communicates with an earth base station that sends information to a gateway server, where the data are temporally stored and downloaded to the client database server via the Internet. The satellite constellation is geostationary and the upload frequency ranges from 1525 to 1559 MHz, whereas the download frequency is between 1626.5 and 1660.5 MHz. This satellite service network is private and a monthly fee per modem is charged based on the satellite channel data allocation. In order to minimize the monthly fee as well as the demand for electric power per modem, the information was coded in the processing module to be transmitted by the satellite modem into two 64-bit packets, each one transmitted every

This section describes the results obtained in the three development phases, which include laboratory experiments, localnetwork experiments, and eld experiments. Measurements of latency, packet loss, humidity, temperature, and leakage current are also reported and analyzed. A. Laboratory Experiments A laboratory experiment was performed in order to test the system performance in a controlled environment. This experiment was aimed to check data consistency, to measure average transmission latency, as well as to determine whether the photovoltaic power-supply system was operating as designed. In addition, it was also important to determine if the power consumption of each sensor system under satellite data transmission was that as predicted. The solar power supply and the battery charge status were also measured. In these experiments, a sensor system was installed externally next to a window of the laboratory, located on the fourth oor of the ve-story, engineering building of the university, and the solar panel was placed on top of the building. To emulate the pulses typically observed in the leakage current signal, a current source generating rectangular pulses with a frequency of 10 Hz and 50-ms duration was used. The current amplitude could be dened in adjusted to a value larger than the highest level (4), when desired, in order to provide a data count on all of the current ranges recorded by the processing module. The optical sensor was then connected to the current source for recording.

956

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

Fig. 6. Time dependence of the average monthly latency of system 1.

Fig. 7. Time dependence of the daily percentage of packet loss for system 1.

Environmental humidity and temperature were also measured and recorded by the processing module, so that all data could be assembled in packets in the processing module for the periodic transmission of 30 min set for the system. For the sensor system network, the latency of a packet is dened by the parameter (5) with and representing the instant of time of arrival of the packet at the server and that of delivery of the packet by the microcontroller, respectively. The time is originally set within the microprocessor routine, and packets are delivered every 30 min. Therefore, the parameter always assumes values separated by 30-min intervals from 00:00 to 23:30 for a full 24-h period, corresponding to 48 values. In the plots reported in this paper, the time is that for the city of Recife, Brazil, which is located in a time zone 3 h earlier than the Greenwich mean time (GMT-3). An average latency at time was also dened to enable determining average uctuations during the day. This parameter is dened by (6) with representing the number of days over which the average is taken and is the latency measured at time of the th day. Fig. 6 shows the measured time dependence of the param30 in eter given by (6) for a 30-day average [i.e., for (6)]. As can be noticed from the plot, average latency increases from early morning, reaching a maximum near the middle of the day, decreasing to a minimum in the afternoon, when it starts to increase toward the night. The maximum average latency observed was approximately 4 min. This value is acceptable, given that it is much smaller than the time period of 30 min between transmitted packets. Transmission reliability is important to guarantee completeness of data, and a high rate of packet loss on the telemetry system could compromise the analysis of the insulation capacity

of the strings. Given that a total of 48 packets are sent every day, the percentage of packet loss on the th day is dened by (7) with representing the number of packets received on the th day. Fig. 7 shows a plot of the parameter % for a one-monthlong experiment. The largest percentage of packet loss recorded for this time period was approximately 4% (2 lost packets in 48 transmitted), with an average of 0.6% over one month. This value of maximum loss does not compromise the effectiveness of the telemetry system since the missing data can be obtained by interpolation. Calibration experiments were also carried out during a onemonth period in the laboratory to check if the current levels and pulse rate set by the source would indeed produce the correct pulse counts in the four ranges set in the processing module. In those experiments, a pulsed current source having a squarewave output with adjustable minimum and maximum current levels and frequency was used. The electric leads of the optical sensor were directly connected to the output of the current source. As the maximum pulse count per range is 262 kpulses/h, which corresponds roughly to 73 pulses/s, the maximum pulse rate of the source was set to 60 pulses/s, or 216 kpulses/h. Fig. 8 shows the curves representing the pulse rates set by the source in each range as solid lines, and the corresponding pulse counts recorded by the processing module, as open circles. In the very beginning of the calibration experiments, the minimum and maximum current levels were set to 0 and 50 mA, respectively, in order to produce identical counts in all ranges, as dened in (1)(4). However, a few bugs in the microcontroller software counting routines produced distinct counts in these ranges. This can be noticed in region I of Fig. 8, with the current source operating at 216 kpulses/h. Notice that the record and in this region were saturated at the maxin ranges imum register capacity of 262 kpulses/h, whereas random readings were recorded in the other ranges. Between the 5th and the 7th day of the experiment, during debugging of the code and

FONTANA et al.: SENSOR NETWORK FOR MONITORING THE STATE OF POLLUTION OF HV INSULATORS

957

tests, the current source was switched off and occasionally on, as can be observed in Fig. 8. After the 7th day of the experiment, the bugs were xed, and the source frequency was adjusted to 10 pulses/s, yielding 36 kpulses/h. Notice from Fig. 8, that after this point, identical counts at the source rate were registered in the four ranges on days 8 and 9, as can be seen in region II of Fig. 8. On the 10th day of the experiment, the current source was . This was accomplished congured to excite only the range by setting the minimum and maximum current levels of the source at 30 and 50 mA, respectively. By doing this, one could produce a transition across the threshold given in (4) for the range, but not across the lower current thresholds, given in (1)(3). As can be noticed in the plot of Fig. 8, the system registered correctly this feature on the 10th day. After this day, the minimum level of the current was set back to 0 mA and a few tests were performed by changing the source pulse rate, until day 12, and correct counts were recorded in all ranges, as shown in Fig. 8. In order to verify the noise rejection capacity of the electronic processing module, the current source was turned off on the 15th day. In this situation, the processing module did not count any spurious pulse, as shown in Fig. 8. Finally, as noticed in region III of Fig. 8, the current source was turned on again on the 18th and with day and was congured to activate only ranges the same rate of 10 pulses/s, which was correctly detected by the processing module. From the results of this sequence of experiments, it was concluded that the processing module was calibrated. The processing module and the satellite-communication system are powered by a photovoltaic system, consisting of a solar panel, a charge controller, and a 12-V battery. Laboratory tests were conducted and it was determined that the processing module had a continuous power consumption of 0.56 W and that the satellite modem consumption was about 1 W during standby and 9.6 W during each packet transmission, which lasted approximately 2 s. Thus, the expected total daily energy consumption of the sensor system in the eld was 38.2 Wh. Thus, for a supply voltage of 12 V, the solar panel should provide a daily charge of 3.2 Ah to the processing module and modem, in addition to a surplus of energy to the battery, to enable long-term operation of the system when installed in the eld. To determine the solar power-supply capability, the electric current owing through the battery was measured using a 1precision series resistor. Battery charge changes could thus be obtained by integrating the measured current over time. Fig. 9 shows the time dependence of the measured battery charge variations, obtained during three days of charging/discharging cycles during normal operation of the system (processing module and satellite modem). In these measurements, the battery was not completely charged in the beginning of the measurement. As can be seen in Fig. 9, the total amount of charge supplied to the battery during daylight is greater than that at night, except on the third day, when both contributions become approximately equal. This is because the battery was fully charged after 48 h, and under these conditions, the controller cut off any charge increase beyond the battery maximum limit. An examination of

Fig. 8. Time dependencies of measured parameters for sensor system 1 from laboratory tests. Regions I III indicate different experimental conditions. The dashed line in each plot indicates the pulse rate of the current source within the corresponding range.

Fig. 9. Time dependence of the battery charge for sensor system 1.

the plot indicates that, after 48 h, the battery charge was increased by approximately 1.75 Ah, and the increase after 72 h was about the same. Thus, these measurements indicated that the electric power-supply system, comprising the solar panel, battery, and charge controller, was adequately dimensioned.

958

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

TABLE I TRANSMISSION PERFORMANCE PARAMETERS DURING LOCAL NETWORK EXPERIMENTS

B. Local Network Experiments Local network experiments were aimed to investigate the dynamics of data reception from several systems acting simultaneously as well as the transmission efciency of the modems under interference of the metallic structures of the towers. These experiments also enabled verifying installation difculties and to design strategies to attach the equipment to the towers. Further observations of the oxidation of metallic cabinets and sealing against rain and moisture were performed in order to guarantee long-term reliability. Six systems were installed in towers located in or near CHESF substations in the city of Recife, Brazil. In order to emulate current pulses for the optical sensor in each system, a bi-stable-type oscillator circuit was constructed based on a standard integrated circuit (IC) 555. The operating frequency was set to produce 216 kpulses/h and the minimum and maximum pulse amplitudes were 0 and 50 mA, respectively, in order to produce records on all four ranges dened in (1)(4). Each oscillator module was connected to the optical sensor placed inside the corresponding cabinet. Experiments were carried out for approximately 30 days. Table I presents, for both satellite transmission packets, the main communication parameters measured during the local network experiments. Latency, average latency, and packet loss were calculated according to (5)(7), respectively. The parameter in (6) for a given system corresponded to the number of days of experimental observation, which could vary due to occasional operation interruptions for inspection in one or the other systems. According to Table I, systems 2, 4, and 6 presented less than 2% of packet loss. The average latency for these three systems was less than a minute. These results indicated that these three systems presented very good performance. In system 3, a high average latency of approximately 15 min for packet 1 was observed, and 21 min for packet 2. Moreover, the percentage of packet loss was above 70% for both packets, as can be observed from the numbers in Table I. A follow-up inspection on cabinet 3, after removal from the tower, indicated the presence of a large quantity of rain water in its interior, due to bad sealing between the modems bottom surface and the top of the cabinet. System 5 also exhibited high packet loss and latency, as can be observed in Table I. This low performance was also later conrmed to be

due to rain water penetration within the cabinet provoked by bad sealing. Systems 6 and 7 were installed on distinct parts of a single tower. System 6 showed adequate performance, as can be concluded from the numbers in Table I. System 7, however, presented a packet loss exceeding 30% for both packets and an average latency of 25 min for packet 2, an unusual behavior. On the 20th day of the experiment, an inspection was carried out in system 7, but no problems were found with respect to sealing or the presence of rainwater inside the cabinet or any failure in electrical connections. However, it was noticed that the satellite modem of this system was not positioned properly, since part of the metallic structure of the tower was obstructing the direct sight to the satellite constellation, in turn, reecting back most of the radiation emitted by the modem. To conrm this hypothesis, system 7 was placed in a new location on the tower, with a clear line of sight to the satellite constellation. After a new sequence of measurements at the new position, the data transmission of system 7 exhibited an average latency down by a factor of 10 from that previously recorded (i.e., down to 2.5 min) and a packet loss 30 times smaller (i.e., of approximately 1%) for both packets, as can be observed from the numbers in Table I. In summary, systems 2, 4, 6, and 7 showed satisfactory performance with latency well below the 30-min time period between packet transmissions and low packet loss. The bad performances of systems 3 and 5 were due to damages caused in the processing module by water inltration in the cabinet. In addition to these problems, an inspection on all cabinets after approximately one month of experiment in the eld, under the combined effect of sun and rain, showed that oxidation in the cabinet parts was so severe that it could result in malfunctioning of the unit for long-term use in the eld. To eliminate oxidation problems, all cabinets were submitted to an electroplating process. In addition, to avoid the rain water inltration, stable sealing between modem and cabinet was obtained with the use of an enhanced silicone adhesive for hightemperature applications. Also, canvas covers were designed to protect the cabinets from rainwater. After these modications, the six systems were reinstalled in the tower and a new one-week test was conducted to verify the behavior of data transmission. No signicant difference was observed in the latency or packet loss between systems with or without the canvas covers. In both cases, the measured average latency was approximately 50 s and the packet loss was less than 2%. To further test the quality of cabinet sealing and canvas cover protection, one unit was placed under a running shower at heavy ow for one hour. No water inltration was detected in the cabinet. C. Deployment of the Sensor System Network and Preliminary Results 1) Deployment of the Sensor System Network: Sensor systems set to withstand long-term operation in the eld, according to the safety measures described in the previous section, were prepared. The training of CHESFs personnel was conducted, which included instructions for optical sensor installation on the insulators, assembly methodology of each system in the corresponding tower, and the elaboration of checklists with step-by-

FONTANA et al.: SENSOR NETWORK FOR MONITORING THE STATE OF POLLUTION OF HV INSULATORS

959

Fig. 10. Geographical distribution of the sensor system network in Northeast Brazil. The distances in rounded boxes are measured on a straight path to CHESFs Headquarters in Recife.

step instructions on how to energize and initialize each processing module immediately after installation. A network of six sensor systems was deployed in six towers spread amongst four states in Northeast Brazil, as illustrated on the map of Fig. 10. The tower locations were chosen by CHESFs technicians based on the company records of pollution severity in these locations. The maximum distance between systems is approximately 815 km, and corresponds, as indicated in Fig. 10, to that between the towers located in the city of Sobral in the State of Cear, and So Miguel dos Campos, in the State of Alagoas. Four towers in the network were located in the open eld and two were located in nearby substations. The decision on where to install each optical sensor, either on suspension or anchorage strings, was taken by CHESFs maintenance team, and depended on the degree of complexity of the installation procedure. For each installation, the modem on the cabinet had a direct line of sight to the satellite constellation to avoid communication problems as described in Section III-B, and the solar panel active surface was always facing the northern hemisphere, in order to maximize the time of sunlight illumination during the day. The optical sensor was always installed on the insulator string closest to the ground terminal of the tower and connected to the processing module in the cabinet through a multimode optical-ber cable specied for outdoor applications. Verication of the correct energization and initialization procedures of each sensor system was always checked in loco, by direct observation of a sequence of blinks of display light-emitting diodes (LEDs) mounted on the processing module box, indicative of proper operation of the unity. Conrmation that the data were being transmitted correctly could be obtained 30 min to 1 h after installation by direct communication between the eld team and a second team of researchers on duty in the laboratory in Recife, who were observing the arrival of data through the ADECI man-

agement system. The total time required to install all six units was two weeks and typically 2 h were required to complete a single sensor system installation. A short documentary video showing a number of features of this work recorded during deployment of the network can be watched on the hyperlink of [17]. 2) Results From the 230-kV Tower in So Miguel Dos Campos: Preliminary data recorded for a 30-day period on all six monitoring points of the sensor network were reported and briey discussed in [16]. In this section, the results for an extended period of observation of 17 months on the 230-kV tower in So Miguel dos Campos are reported with additional discussion. A detailed analysis of the evolution of activities on all sensor systems for longer term monitoring and correlation of results with those obtained from in situ inspections by CHESFs maintenance teams will be reported elsewhere. Sensor system 3 was installed in October 2009 on an anchorage porcelain string of the 230-kV tower which is located in a sugarcane eld, in a region characterized by a hot and humid climate. Furthermore, the tower is located near a sand road which is the route to trucks to and from the sugarcane eld, as can be watched directly in the video in [17]. The heavy trafc of vehicles and burning of sugarcane elds during harvest, a standard practice in Brazil, are the main causes of pollutants deposition on insulator strings. Fig. 11 shows the activity recorded during the rst 90 days after installation of the sensor system on the 230-kV tower. There are four regions indicated in Fig. 11 with particular features. Regions 1 and 2 are those before and after the 52nd day of observation, respectively, as dened by the dashed line in Fig. 11. One can notice that in region 1 there are, in general, activities recorded in all four ranges. On the other hand, in region 2, there is a substantial decrease in activity, with only occasional being observed. In addition, a closer inrecords on range spection of the activities in region 1 shows a variable pulse rate, with a time scale that cannot be inferred directly from Fig. 11. Moreover, there are regions without any signicant activities in all four ranges, as indicated by regions 3 and 4. To further elucidate the variable pulse rate and the existence , of regions without substantial activities within ranges a more detailed view of the recorded data between days 20 and 26, which includes region 3 of Fig. 11, is shown in Fig. 12. The upper plot of Fig. 12 is the time dependence of both humidity and temperature recorded by the sensor system. As can be seen from this plot, both parameters vary somewhat periodically, with humidity and temperature reaching maximum and minimum levels, respectively, at a time just before 6:00 A.M. A reversal of this behavior occurs near 12:00 P.M. One can clearly notice in the activities recorded in the four ranges shown in Fig. 12 that activities increase with the increase of humidity during night and early morning hours, and extinguish during the day, when humidity is low. This would be the expected standard behavior of the activities registered on a polluted insulator. One can also notice from this analysis that the degree of pollution is appear increasing toward the 23rd day, as activities in range on the 23rd day. The lack of activities, such as those in regions 3 and 4 of Fig. 11, is analyzed for the case of region 3 in the plot of Fig. 12.

960

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

Fig. 11. Activity registered in the 230-kV tower in So Miguel dos Campos, during the rst 90 days of observation.

Notice from Fig. 11 that this region starts on the 23rd day, and Fig. 12 indicates that indeed it starts close to 12:00 P.M. of this day. As can be noticed from the temperature and humidity traces in Fig. 12, there is a reversal of behavior of these curves near 12:00 P.M. (i.e., an increase in humidity with a maximum being reached right at this time point), with a simultaneous minimum being reached by the temperature curve. This is a clear indication of the presence of rain starting at a time close to 12:00 P.M. on the 23rd day. Notice that near this time of the day, when no , a large substantial activities were expected in ranges and with some activity pulse rate is recorded in ranges . Activities cease to exist after this time registered in range point, indicating that cleaning of the insulators took place due to rain. It is important to notice that maintenance personnel report that ashovers on polluted insulators, in many situations, take place during the start of rain. The sudden increase in pulse rate shown in Fig. 12 may be a strong indication that this type of effect may indeed occur. A second important feature observed in the plot of Fig. 11 is the sudden drop in activities registered in all four levels after the 52nd day of observation. To further elucidate this point, a longer term observation period is shown in Fig. 13. The plots in this gure represent the pulse rate recorded in levels during 17 months, approximately. One can notice the sudden drop in activity after the 52nd day and no substantial activity is registered for a time period of approximately one year. Then, starting again on the 375th day, in September/October 2010, a and in clear onset of activity takes place in ranges

Fig. 12. Activity registered in the 230-kV tower in So Miguel dos Campos, between the 20th and 26th day of observation.

range near the 390th day. The time period of the two groups of strong activities shown in Fig. 13 is 50 to 60 days. This behavior is totally correlated with the harvest cycle of sugarcane in Northeast Brazil. Harvest takes place between October and December, with small variations every year. During this period, the pollution level is expected to increase as a result of the heavy ow of trucks along the sand road and burning activities taking place on the sugarcane eld around the 230-kV tower. Plantation and growth of crops take place during the remaining part of the year, when a decrease of pollution level, together with natural washing due to rain, yield a substantial reduction in the activities recorded by the sensor system. IV. CONCLUSION In this paper, a detailed report was made on the development and deployment of a sensor network for monitoring, via satellite, the pollution state of insulator strings of HV transmission lines. A satellite-communication approach proved to be a feasible strategy because most of the monitoring points under supervision could not be covered by standard mobile or optical data-communication methods. Measured transmission latency well below the 30-min time required for transmission of data, and the percentage of packet loss of less than 2% were standard in each sensor system. A detailed discussion on the intermediate

FONTANA et al.: SENSOR NETWORK FOR MONITORING THE STATE OF POLLUTION OF HV INSULATORS

961

REFERENCES
[1] R. J. Van Brunt, Physics and chemistry of partial discharge and coronarecent advances and future challenges, IEEE Trans. Dielect. Elect. Insul., vol. 1, no. 5, pp. 761784, Oct. 1994. [2] G. Montoya, I. Ramirez, and J. I. Montoya, Correlation among ESDD, NSDD and leakage current in distribution insulators, Proc. Inst. Elect. Eng., Gen., Transm. Distrib., vol. 151, no. 3, pp. 334340, 2004. [3] M. Youssef and A. A. El-Alayly, Remote monitoring of surface current over insulators, in Proc. IEEE Transm. Distrib. Conf., Atlanta, GA, 2001, pp. 113116. [4] E. O. Abdelaziz, M. Javoronkov, C. Abdelaziz, G. Fethi, and B. Zohra, Prevention of the interruptions due to the phenomena of the electric insulators pollution, in Proc. 1st. Int. Symp. Control, Commun. Signal Process., 2004, pp. 493497. [5] T. Suda, Frequency characteristics of leakage current waveforms of a string of suspension insulators, IEEE Trans. Power Del., vol. 20, no. 1, pp. 481487, Jan. 2005. [6] S. Shihab, V. Melik, L. Zhou, G. Melik, and N. Alame, On-line pollution leakage current monitoring system, in Proc. IEEE Properties Appl. Dielect. Mater., 1994, pp. 538541. [7] S. Moorthy, S. Shihab, and X. cheng, Practical on-line partial discharge measuring system for high voltage apparatus, in Proc. IEEE Prop. Appl. Dielect. Mater., Jul. 1994, pp. 586589. [8] R. W. S. Garcia, N. H. C. Santiago, and C. M. Portela, A mathematical model to study the inuence of source parameters in polluted insulator tests, in Proc. IEEE Prop. Appl. Dielect. Mater., Jul. 1991, pp. 953956. [9] D. L. Williams, A. Haddad, A. R. Rowlands, H. M. Young, and R. T. Waters, Formation and characterization of dry bands in clean fog on polluted insulators, IEEE Trans. Dielect. Elect. Insul., vol. 6, no. 5, pp. 724731, Oct. 1999. [10] M. Khalifa, A. El-Morshedi, O. E. Gouda, and S. E.-D. Habib, A new monitor for pollution on power line insulatorsPart 2: Simulated eld tests, Proc. Inst. Elect. Eng., Gen., Transm. Distrib., vol. 135, no. 1, pp. 2430, Jan. 1988. [11] K. Isaka, Y. Yokoi, K. Naito, R. Matsuoka, S. Ito, K. Sakanishi, and O. Fujii, Development of real-time system for simultaneous observation of visual discharges and leakage current on contaminated dc insulators, IEEE Trans. Dielect. Elect. Insul., vol. 25, no. 6, pp. 11531160, Dec. 1990. [12] F. J. M. M. Cavalcanti, Control and performance analysis of overhead transmission lines insulation on polluted environments, searching for preventive actions, (in Portuguese) M.Sc. dissertation, Graduate Program Elect. Eng., Federal Univ. Pernambuco, Recife, Pernambuco, Brazil, 2004. [13] E. Fontana, S. C. Oliveira, F. J. M. M. Cavalcanti, R. B. Lima, J. F. Martins-Filho, and E. Meneses-Pacheco, Novel sensor system for leakage current detection on insulator strings of overhead transmission lines, IEEE Trans. Power Del., vol. 21, no. 4, pp. 20642070, Oct. 2006. [14] S. C. Oliveira and E. Fontana, Optical detection of partial discharges on insulator strings of high-voltage transmission lines, IEEE Trans. Instrum. Meas., vol. 58, no. 7, pp. 23282334, Jul. 2009. [15] R. A. Lima, E. Fontana, J. F. Martins-Filho, T. L. Prata, G. O. Cavalcanti, R. B. Lima, S. C. Oliveira, and F. J. M. M. Cavalcanti, Satellite telemetry system for pollution detection on insulator strings of high-voltage transmission lines, in Proc. SBMO/IEEE MTT-S Int. Microw. Optoelectron. Conf., 2009, pp. 574577. [16] R. A. Lima, E. Fontana, J. F. Martins-Filho, T. L. Prata, G. O. Cavalcanti, R. B. Lima, S. C. Oliveira, and F. J. M. M. Cavalcanti, Remote monitoring of the degree of pollution of high voltage insulator strings via satellite with a sensor system network, in Proc. IEEE Sensors Conf., 2010, pp. 11131117. [17] Programa de ps-graduao em engenharia eltrica da UFPE, PPGEEUFPE. (2010), Satellite telemetry system for pollution detection of high-voltage insulator strings. [Online]. Available: http://www.youtube.com/watch?v=sU2ZCD-fsYU Eduardo Fontana (M08) was born in Rio de Janeiro, Brazil, in 1957. He received the B.Sc. degree in electrical engineering and the M.Sc. degree in physics from the Federal University of Pernambuco (UFPE), Recife, Brazil, in 1980 and 1983, respectively, and the Ph.D. degree in electrical engineering from Stanford University, Stanford, CA, in 1989. Currently, he is a Professor in Electrical Engineering with the Electronics and Systems Department, UFPE. From 1991 to 1992, while on

Fig. 13. Activity registered in the 230-kV tower in So Miguel dos Campos for a time period of approximately 17 months.

tests, experiments, and system improvements that led to the reliable long-term (17 months) continuous eld experiments was provided. A network with six sensor systems was installed in locations exhibiting six different weather conditions in Northeast Brazil. In the present stage, no lightning effects and/or ashover-induced damages or even unexpected records due to those more severe conditions have been observed in the six units installed in the eld. A preliminary analysis performed for short- and long-term behavior of the activities registered on one of the systems, installed on a 230-kV tower in So Miguel dos Campos, in a place surrounded by a sugarcane eld, indicated a strong correlation of the pollution level inferred by the sensor system with the harvest activities expected to take place in the sugarcane eld. At the moment, the sensor network is under expansion on CHESFs transmission-line system in order to allow a broader coverage of pollution activities. The need for additional surge protection and improved environmental shielding of the electronic circuits of each sensor system will be analyzed prior to fabrication of the new units. A detailed analysis of the evolution of activities on the sensor network for longer term monitoring and correlation of results with those obtained from in situ inspections by CHESFs maintenance teams will be reported elsewhere.

ACKNOWLEDGMENT The authors would like to thank the Brazilian Research Council (CNPq) and Coordination of Advanced Studies of the Ministry of Education (CAPES) for scholarships and grants.

962

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

leave from UFPE, he was a Senior Scientist at Adelphi Technology, Inc., Palo Alto, CA, coordinating research-and-development activities in the eld of surface plasmon spectroscopy. His past research activities have also included microwave ferrite devices, low-temperature magnetic materials, magnetic semiconductors, optics, and free-electron lasers. He is the Financial Director of the Advanced Center of Engineering and Services of Recife (CAESER) and a Research Fellow with the National Research Council of Brazil (CNPq). Prof. Fontana is a member of the Photonics Group of the Electronics and Systems Department, UFPE, engaged in research and teaching activities in the elds of electromagnetic theory and applications, bioelectromagnetism, optical instrumentation, optical-ber sensors and devices, and surface plasmon spectroscopy applications. He is a member of the Optical Society of America, Brazilian Microwaves and Optoelectronics Society, and Brazilian Biomedical Engineering Society.

deployment of sensor systems for pollution monitoring of insulator strings of overhead transmission lines. Mr. Cavalcanti is a Professional Engineer with Companhia Hidro Eltrica do So Francisco (CHESF), Recife, and is currently coordinating CHESFs program for widespread access to electrical energy in Northeast Brazil.

Joaquim F. Martins-Filho (M95SM11) was born in Recife, Brazil, in 1966. He received the B.Sc. degree in electronics engineering and the M.Sc. degree in physics from the Federal University of Pernambuco (UFPE), Recife, in 1989 and 1991, respectively, studying nonlinear optics in optical bers, and the Ph.D. degree in electronics engineering from the University of Glasgow, Glasgow, U.K., in 1995, investigating ultrafast optical pulse generation from modelocked semiconductor quantum well lasers. Since 1998, he has been an Associate Professor in the Photonics Group, Department of Electronics and Systems, UFPE, Recife, Brazil. His current interests are in devices, subsystems, transmission systems, and networking for optical communications and optical sensors. Prof. Joaquim Martins-Filho is a member of the Optical Society of America, Brazilian Microwave and Optoelectronics Society, Brazilian Telecommunication Society, Brazilian Physics Society, and is a Research Fellow with the National Research Council of Brazil (CNPq).

Ricardo A. Lima was born in Recife, Brazil, in 1983. He received the B.Sc. degree in electronics engineering and the M.Sc. degree in electrical engineering from the Federal University of Pernambuco (UFPE), Recife, Brazil, in 2008 and 2010, respectively, where he is currently pursuing the Ph.D. degree in electrical engineering. Currently, he is with the Department of Electronics and Systems, Federal University of Pernambuco, Recife, in the laboratory of optical sensors. His interests are related to embedded electronics, optical sensors, satellite communications, and biomedical engineering.

Gustavo O. Cavalcanti was born in Recife, Brazil in 1981. He received the B.Sc. and M.Sc. degrees in electrical engineering from the Federal University of Pernambuco (UFPE), Recife, Brazil, in 2005 and 2008, respectively, where he is currently pursuing the Ph.D. degree in electrical engineering. He is a permanent faculty member of the Electrical Engineering Department, University of Pernambuco, Recife, Brazil. His research activities are in the elds of surface plasmon resonance for hydrogen detection and he is currently engaged in research and development of optical sensors for high-voltage equipment. Mr. Cavalcanti is a member of the Systems and Applied Electromagnetics Group, Electrical Engineering Department, UPE, engaged in research and teaching activities in the elds of electromagnetic theory and applications.

Srgio C. Oliveira (M06) was born in Recife, Brazil, in 1977. He received the B.Sc., M.Sc., and D.Sc. degrees in electrical engineering from the Federal University of Pernambuco (UFPE), Recife, Brazil, in 2002, 2004, and 2008, respectively. Currently, he is a permanent faculty member with the Polytechnic School of Pernambuco of the University of Pernambuco, Recife, Brazil, where he has been since 2007. His past research activities are in the elds of ber sensors, optical networks, and dense wavelength-division multiplexing technologies and he is currently engaged in the research and development of optical sensors, optical instrumentation, and embedded systems. Over the last few years, he has been the coordinator of the graduation on Computer Engineering of the Polytechnic School of Pernambuco. Prof. Srgio is a member of two post graduate programs on computer engineering and systems engineering.

Thiago L. Prata was born in Recife, Brazil, in 1987. He received the B.Sc. degree in computer engineering from the Federal University of Pernambuco (UFPE), Recife, Brazil, in 2010, where he is currently pursuing the M.Sc. degree in computer science. Currently, he is a Software Engineer in the SOPHIA Solues Company, Porto Digital, Recife. He was a System Engineer at Enfaltec, Recife, developing and designing embedded systems. His interests are related to computer vision, signal processing, machine learning, and articial intelligence. Mr. Prata is currently a member of the VIISAR Research Group, Center of Informatics (CIn), UFPE, a research group focused on the area of computer vision.

Fernando J. M. M. Cavalcanti was born in Recife, Brazil, in 1957. He received the B.Sc. degree in electrical engineering as well as the academic and professional M.Sc. degrees in electrical engineering from the Federal University of Pernambuco, Recife, Brazil, in 1980, 2000, and 2004, respectively. He is the Coordinator for Research Development and Innovation of CHESFs Operation Management Division. He has extensive experience in the maintenance of high-voltage transmission systems and is currently engaged in the development and

Renato B. Lima, photograph and biography not available at the time of publication.

You might also like