FACTORS OF INFLUENCE OVER INFRARED THERMAL INSPECTION IN OUTDOOR INDUSTRIAL SUBSTATIONS

Edson C Bortoni, Senior Member, IEEE, Laerte dos Santos, Guilherme S Bastos, Luiz E Souza
Electrical and Energy Systems Institute, Systems Engineering and Information Technology Institute Itajub Federal University Av. BPS, 1303 37501-903 Itajub-MG, Brazil bortoni@unifei.edu.br

Abstract This paper presents some difficulties encountered when evaluating information from infrared thermal inspections conducted in uncovered industrial substations. Procedural, technical and environmental are the main factors of influence identified. Based on field data and in laboratory tests, preliminary mathematical models are derived, which are suitable either to forecast the system behavior under specified conditions or to remove the influence such components. Keywords Industrial substations, Infrared thermal inspections. I. INTRODUCTION Infrared (IR) thermal inspection is a valuable tool to determine the operating conditions of substation components. Problems such as high resistance contacts, short- and opencircuits, inductive heating, harmonics, load imbalance and overloads can often be detected through IR thermal inspections. Applications of such technology to power and industrial systems are presented since the sixties [1-2]. Despite thermal inspection seems to be a simple task, there are a number of limitations and exogenous influences that conduct to erroneous diagnosis and eventually impede the failure detection [3]. Low emissivity of the components under inspection, load current variation and small dimensions of the inspected object located at large distances are examples of drawbacks that must be overcame in an IR thermal inspection. Environmental quantities such as the solar radiation, atmospheric attenuation, wind speed, precipitation, and environment temperature and humidity variation are uncertainty factors that must be added when inspecting uncovered substations. The works of Madding and Lyon [4] and Snell [5] consider the loading conditions and environmental components influence of IR thermal inspections. Lyon Jr. et al. [6] evaluate the relationship between the current loading and the temperature rise in a faulty connector. In addition, the papers discusses about the limitations of techniques of condition evaluation based only in the absolute temperature

or in the temperature rise, which can conduct to wrong diagnosis. This work tries to contribute with the understanding of the influence of environmental and technical quantities over the IR thermal inspection, by presenting actual data obtained in field and developing mathematical models that allows not only to consider the environmental influences over the results of a thermal inspection, but also to remove the effect of some of these quantities to forecast the system behavior when operating in specific conditions of interest. Therefore, the determination of the expected component temperature under extreme load conditions, environment temperature and other factors of influence becomes possible. II. INFLUENCE FACTORS IN IR OUTDOOR INSPECTIONS Infrared (IR) thermal inspection is a valuable tool to determine the operating conditions of substation components. Problems such as high resistance contacts, short- and opencircuits, inductive heating, harmonics, load imbalance and overloads can often be detected through IR thermal inspections. Applications of such technology to power and industrial systems are presented since the sixties [1-2]. Such factors of influence can be characterized as procedural, technical or environmental factors. The influence of Procedural factor is minimized when certified personnel is employed [7]. This work is concerned with the technical and environmental influence factors. Figure 1 shows a typical scene of an IR inspection in a high-voltage substation. The main elements are the inspector, the thermal-camera, the equipment under analysis and the environment. Low emissivity of the component under analysis, load current variation, small dimensions components at large distances and employed equipment are examples of technical factors of influence. In addition, for outdoor environments, there are other factors of influence such as solar radiation, atmospheric attenuation; wind velocity, ambient temperature changes, rain and humidity are some of the environmental factors that can turn the evaluation of an IR inspection in a difficult task.

This work presents some initial results with the application of the acclimatized chamber. Two models for characterization of loading current influence was developed and applied in laboratory and in field, which are presented in the following section. Figure 3 presents the installation of a connector to be tested at the acclimatized chamber. III. INFLUENCE OF LOADING CURRENT It is well-known that a component operating temperature is proportional to the square of the operating current. On the other hand, in order to evaluate a component behavior, it is desirable to know its temperature when working at the worst condition, i.e., with the maximum current. Nevertheless, to find this condition when carrying out inspections in field is not guaranteed. The common procedure is to apply a correction factor over the temperature rise that is the square of the maximum current to operating current ratio, in order to estimate the temperature at the worst condition. A test was conducted employing the developed chamber where a load current was applied to a connector under test. The temperature was recorded and presented at figure 4. Notice in that picture that there are three stages at 600 (A), with three different temperatures (!), 30C, 54C and 39C. If one applies the criterion to correct the temperature rise values with the square of the maximum current to operating current ratio, to obtain the temperature for 800 (A), would find 17.7C, 60.4C or 33.7C, none of them equal to 51C, which is the right value obtained by test (fig. 5). It is well known that this technique must be applied for constant currents; nevertheless varying current is what happens in power systems, mainly in industrial substations. In order to overcome the limitations of this method, two temperature models suitable to estimate the temperature for the heaviest current are presented as follows.

Fig. 1. Procedural, technical and environmental sources of influence.

In order to evaluate the extent of the influence of such factors, an acclimatized chamber was developed to conduct tests in laboratory under controlled conditions. The chamber was designed to accommodate the component under test and to carry out the same current loading conditions observed in actual operation. Temperature is measured both through contact sensor and thermal-camera. All the information is recorded using a data acquisition system and ready for use in the analysis. Figure 2 shows the main components of the developed chamber.

Fig. 2. Sketch of the developed acclimatized chamber.

Fig. 3. Developed acclimatized chamber.

Fig.4. Load current and temperature of the connector under test.

A. Thermal model parameters estimation Accurate thermal modeling is a complex task. Nevertheless, a simple thermal model can be constructed under the assumption that the temperature rise is a function of the square of the operating current and that the temperature rise over the environment temperature is the main variable concerned to the heat exchange. Therefore, considering the object under analysis as a homogeneous body, the temperature rise over the environmental temperature () in a time period is the result of the summation of two components: An increasing component due to the present loading period and a decreasing component of the final temperature of the last period:

Such results allow obtaining the equation that describes the dependence of the final temperature rise with the square of the applied current.

F = 2.152 + 0.781 10 4 I 2

(2)

This expression is of great value since it presents the maximum temperature elevation for any applied current at this component. B. Auto-regressive model Despite the former method allows the knowledge of the final temperature rise for any applied current flowing in the component under test, it is not useful for practical purposes, as long as the current varies according to the system load. For such situations another model is proposed, that is an auto-regressive model. The idea behind this model is that, neglecting environmental influence, the present temperature rise is not only a function of the present current, but it is also influenced by currents that of the past, as presented in equation (3).

= F ( 1 e t / TA ) + 0 e t / TA

(1)

Where F is the final temperature rise that, at the operating loading condition of the present period, the component would reach in the steady state (C), 0 is the final temperature rise of the last period (C), t is the duration of the studied period (s), TA is the heating time constant (s). This model was applied to identify the component thermal characteristics, i.e., its time constant and the temperature rise dependence with the square of the operating current. A test was carried out in the lab using the acclimatized chamber. In this case, only the operating current was object of variation, remaining constant all of the other variables. Figure 5 presents graphical results of the applied current and obtained temperature.
90,0

t = a0 I t2 0 t + a1 I t2 1 t + ... + a n I t2 n t
In a general form:

(3)

t = ai I t2 it
i =0

(4)

80,0

70,0 60,0

50,0

40,0 Temperature (C) Current / 10 (A)

30,0 20,0

Where t (C) is the temperature rise at instant t (s), ai (C/A) are constant coefficients obtained by a least square algorithm, I (A) is the operating current in a time period tit. The model is suitable to determine the temperature rise for any current. If one considers that the loading current is constant and equal to the current of interest, the final temperature rise for a given current, I, is given by:

10,0

0,0

F = I 2 ai
i =0

(5)

Fig.5. Applied current test at laboratory.

By inspection and employing a least square algorithm it was possible to obtain the time constant of the connector under analysis. The final temperature rise for each applied current is showed in table 1. TABLE I Connector thermal characteristics
I (A) 200 400 600 800

F (C)
4.70 15.2 30.6 51.9

TA (s) 39.1 36.7 33.5 33.9

The correct definition of the time interval (t) and the number of backward intervals (n) deserves some considerations. It depends on the available data and on the time constant of the device under analysis. The time constant depends not only on the body mass and material, but also on where the heating is generated. As long as thermal-cameras detects the temperature based on the surface emissivity, if the heat is internally generated the necessary time to get the thermal information becomes higher. Otherwise, a heat generated at the surface will result in a smaller time constant. After some testing it was noticed that three intervals within the period of a time constant is sufficient to obtain reliable results.

Figure 6 presents the behavior of current and temperature during the tests in laboratory. Data was collected at each minute during approximately five hours.
90,0

80,0

This same methodology was applied to data obtained in field, as an actual situation. A survey was carried out in a high-voltage uncovered substation. Information about load current, target and ambient temperature, solar radiation and wind speed was simultaneously acquired during approximately 36 hours at every 5 minutes. The results are presented in figure 7.

70,0

60,0

50,0

40,0

30,0 Temperature (C) Current / 10 (A) 20,0

10,0

0,0

Fig. 6. Current and temperature during the tests in laboratory.

As long as the average time constant of the studied device is 35.8 (min), two auto-regressive models were tested considering 6 and 3 backward intervals of 10 minutes. The first will consider previous currents of about 60 minutes (about twice the time constant) and the former will take into account currents that occurred in 30 minutes. The resulting coefficients are presented in table 2. TABLE II Determined coefficients for the auto-regressive models
Model 6x10 3x10 a0 0.044 0.236 a1 0.185 0.212 a2 0.185 0.104 a3 0.099 0.237 a4 0.146 a5 0.046 a6 0.144 a 0.848 0,789

Fig. 8. Obtained field data record.

Auto-regressive models 1x30, 2x30 and 4x30 was applied to the information of temperature and current in the interval between the last 8PM and 5AM, where solar radiation and wind speed presented few influence to the target temperature rise. Table 4 presents the determined coefficients of such models. TABLE IV Determined coefficients for the auto-regressive models
Model 1x30 2x30 4x30 a0 0.381 0.187 0.096 a1 1.093 0.173 -0.093 a2 1.119 0.142 a3 -0.039 a4 1.373 a 1.475 1.479 1.478

Table 3 presents a comparison of the final temperature rise calculated according to the studied models, while figure 7 presents the observed and estimated temperatures. TABLE III Final temperature rise comparison
I (A) 200 400 600 800 Thermal model 4.70 15.2 30.6 51.9 AR Model 6x10 3.39 13.5 30.5 54.2 AR Model 3x10 3.15 12.6 28.4 50.5

With great agreement, the summation of the autoregressive models coefficients is around an average value of 1.477, i.e., the final temperature rise for any current can be obtained by multiplying this constant to the square of the given current. The comparison of the estimated temperature rise from de models with the measured one is presented in figure 9.

Fig. 7. Observed and estimated temperatures.

Fig. 9. Observed and estimated temperature rise from 7PM to 7AM.

IV. INFLUENCE OF ENVIRONMENTAL FACTORS The main environmental factors considered in this work are the temperature, humidity, rain, wind velocity and solar radiation. A simple multivariable linear model to estimate the temperature rise over the environmental temperature is proposed:

The field data recorded is presented in figure 11. Humidity information was not used because it was observed a strong inverse correlation with the environmental temperature.

= I +

wj xj
jJ

(6)

Where (C) is the temperature rise over the environmental temperature, I (C) is the temperature rise due to the load current, obtained from the previous presented models, J is the set of considered environmental factors, wj is the weight of influence of the xj environmental factor. This equation is rewritten for each instant of time allowing for the obtaining of the weight coefficients by using the least squares method. This methodology was applied to a different set of data obtained in field in order to estimate the influence of the rain, wind velocity and solar radiation. A setup was prepared with a thermal-camera and a meteorological data acquisition system as presented in figure 10.

Fig. 11. Obtained field data record.

A 1x45 autoregressive model was used in this case as long as it was observed a delay of 45 minutes between the current and the temperature rise. Employing the presented formulation, a model was adjusted to obtain the temperature rise as a function of load current (I), wind speed (xWS), solar radiation (xSR) and rain (xR), that is:
= 4.906 105 I 2 1.241 xWS + 6.626 103 xSR 1.428 xR + 13.079 (7)

As long as this equation relates the temperature rise as a function of several quantities, it can also be employed to obtain the temperature rise for the worst conditions, i.e., maximum load current and solar radiation, and minimum wind speed and rain, as well. Figure 12 presents the estimated absolute temperatures for regular operation, which showed excellent agreement with the field obtained in field, and that calculated for extreme conditions. (a)

(b)
Fig. 10. Connector under analysis.

Fig. 12. Calculated temperatures for regular and extreme conditions.

CONCLUSIONS The work presented the first results of a study under development which aims at obtaining more consistent procedures for IR inspections in outdoor environments, under the influence of environmental factors of influence. The work makes use of an acclimatized chamber that allows simulating several environmental and loading conditions. Two methodologies were presented to estimate the temperature rise of a component working in any conditions; they are the thermal model and the auto-regressive model. Both models showed very good agreement when applied both in laboratory and in field. It was presented a proposal of technique to include environmental factors of influence in the analysis. The method was capable to catch the influence of such factors, allowing for replicate the tested conditions and also for determining the absolute temperature for extreme conditions. Field tests were carried out with a system capable to acquire meteorological quantities in order to validate de developed methodologies. REFERENCES
[1] G. Ferreti, A. Giorgi, A New Type of Pyrometer Employed for Preventive Maintenance in Electric Utilities, L`Energia Elettrica, N 12, 1969. [2] C.W. Brice, Infrared detection of hot spots in energized transmission and distribution equipment, Electric Power Systems Research, Volume 1, Issue 2, April 1978, pp 127-130. [3] J. Snell, R.W. Spring, Developing Operational Protocol for Thermographic Inspection Programs, SPIE Vol. 1682, 1992. [4] R. Madding, B.R. Lyon Jr., Environmental Influences on IR Thermography Surveys, Maintenance Technology 1999. [5] J. Snell, A Different Way to Determine Repair Priorities Using a Weighted Matrix Methodology, Snell Infrared 2001. [6] B.R. Lyon Jr, G.L. Orlove, L.P. Donna, The Relationship between Current Load and Temperature for Quasi-Steady State and Transient Conditions, Infrared Training Center 2002.

[7] L. Santos, E.C. Bortoni, L.C. Barbosa, R.A. Arajo, Centralized vs. decentralized thermal IR inspection policy: Experience from a major Brazilian electric power company, Conference 5782 Thermosense XXVII Proceedings of SPIE, vol. 5782, 2005.

BIOGRAPHIES
Edson da Costa Bortoni (S1994, M1996, SM2005) was born in Maring, Brazil, on December 1, 1966. He graduated from Itajub Federal University (UNIFEI), Itajub, Brazil, in 1990, received the M.Sc. degree in energy systems planning from University of Campinas (UNICAMP), Brazil, in 1993, and the D.Sc. degree in energy and electrical automation from the University of So Paulo (USP), So Paulo, Brazil, in 1998. Presently he is a Professor at UNIFEI. His areas of interest include instrumentation, power generation, and energy systems. He was a professor at So Paulo State University (FEG-UNESP) and University of Amazonas, Brazil. Dr. Bortoni is also Senior Member of ISA and SPIE. Laerte dos Santos was born in Passos, Brazil, on March 18, 1964. He graduated in Computer Technology from the State University of Minas Gerais (UEMG) and received the M.Sc. degree in Energy Engineering from UNIFEI in 2006. He is with FURNAS electric company since 1982 and is working with infrared thermography since 1996. Currently he is working towards his D.Sc. degree in Power Systems at UNIFEI. Mr. Santos is a Level III Infrared thermographer. Guilherme Sousa Bastos was born in Volta Redonda, Brazil, on December 22, 1977. He graduated in Electrical Engineering from Itajub Federal University (UNIFEI), Itajub, Brazil, in 2001, and has the M.Sc. degree in Industrial Systems Automation from UNIFEI, 2004. Currently he is working towards his D.Sc. degree in Mobile Cooperative Robotics at Technological Institute of Aeronautics (ITA), So Jos dos Campos, Brazil. Presently he is a Professor at UNIFEI. His areas of interest include robotics, automation, instrumentation, optimization and artificial intelligence. Luiz Edival de Souza was born in Itanhandu, Brazil, on April 1, 1957. He graduated from UNIFEI (1978), received the M.Sc. degree in mechanical engineering from UNIFEI (1981) and the D.Sc. degree in automation from UNICAMP (1987).