Application of Back-Propagation Neural Network to ... - Semantic Scholar
Application of Back-Propagation Neural Network to Power Transformer Insulation Diagnosis Po-Hung Chen1 and Hung-Cheng Chen2 1
St. John’s University, Department of Electrical Engineering, Taipei, Taiwan, 25135, R.O.C. [email protected] 2 National Chin-Yi University of Technology, Department of Electrical Engineering, Taichung, Taiwan, 411, R.O.C. [email protected]
Abstract. This paper presents a novel approach based on the back-propagation neural network (BPNN) for the insulation diagnosis of power transformers. Four epoxy-resin power transformers with typical insulation defects are purposely made by a manufacturer. These transformers are used as the experimental models of partial discharge (PD) examination. Then, a precious PD detector is used to measure the 3-D (φ-Q-N) PD signals of these four experimental models in a shielded laboratory. This work has established a database containing 160 sets of 3-D PD patterns. The database is used as the training data to train a BPNN. The training-accomplished neural network can be a good diagnosis system for the practical insulation diagnosis of epoxy-resin power transformers. The proposed BPNN approach is successfully applied to practical power transformers field experiments. Experimental results indicate the attractive properties of the BPNN approach, namely, a high recognition rate and good noise elimination ability.
1 Introduction More than half of the breakdown accidents of power apparatuses are caused by insulation deterioration. The reliability of a power apparatus is affected significantly by the presence of insulation defects. Partial discharge (PD) pattern recognition has been regarded as an important diagnosis method to prevent power apparatuses from malfunction of insulation defect. However, the measurement of PD needs heavy power equipment and precious instruments. The Heavy Power Lab in St. John’s University installed a set of precious instrument (Hipotronics DDX-7000 Digital Discharge Detector). Therefore, the school has the capability of doing insulation diagnosis related researches. Power transformer is one of the most important apparatuses in a power delivery system. Breakdown of a power transformer can cause an interruption in electricity supply and result in a loss of considerable profits [1]. Therefore, detecting insulation defects in a power transformer as early as possible is of priority concern to a power transformer user. PD phenomenon usually originates from insulation defects and is an important symptom to detect incipient failures in power transformers. Classification of different types of PD D. Liu et al. (Eds.): ISNN 2007, LNCS 4493, Part III, pp. 26–34, 2007. © Springer-Verlag Berlin Heidelberg 2007
Application of BPNN to Power Transformer Insulation Diagnosis
27
patterns is of importance for the diagnosis of the quality of power transformers. PD behavior can be represented in various ways. Because of the randomization of PD activity, one of the most popular representations is the 3-D (φ-Q-N) distribution [2], [3], i.e., the PD pattern is described using a pulse count N versus pulse height Q and phase angle φ diagram. Previous experimental results have adequately demonstrated that φ-Q-N distributions are strongly dependent upon PD sources, therefore, the 3-D patterns can be used to characterize insulation defects [4]. This provides the basis for pattern recognition techniques that can identify the different types of defects. Previous efforts at PD pattern recognition have applied various identification techniques to make the problem solvable. Various pattern recognition techniques such as fuzzy clustering [5], expert system [6], extension theory [7], [8], and statistical analysis [9] have been proposed. These techniques have been successfully applied to PD pattern recognition. However, these conventional approaches not only require human experiences but also have some difficulties in acquiring knowledge and maintaining the database of decision rules. In recent years, a biologically artificial intelligence technique known as artificial neural network has emerged as a candidate for the pattern identification problem. The neural network can directly acquire experience from the training data to overcome the shortcomings of the conventional approaches. The neural network can quickly and stably learn to categorize input patterns and permit adaptive processes to access significant new information [10]. In this paper, a novel insulation diagnosis approach based on the PD and back-propagation neural network (BPNN) is proposed for epoxy-resin power transformers. Four cast-resin transformers with typical insulation defects, which were purposely made by a manufacturer, are used as the models of the PD examination. The DDX-7000 Digital Discharge Detector is then used to measure the 3-D PD signals of these transformers. So far, this work has established a database which contains 160 sets of 3-D PD patterns. Then, this database is used as the training data to train a BPNN. Finally, the training-accomplished neural network can be a good diagnosis system for the practical insulation diagnosis of epoxy-resin power transformers. Experimental results show that different types of insulation defects within power transformers are identified with rather high recognition rate.
2 Partial Discharge Experiments 2.1 Partial Discharge Measurement PD is a forced phenomenon occurring in insulation parts in a power apparatus. When the intensity of electric field exceeds the breakdown threshold value of a defective dielectric, PD occurs and results in a partial breakdown in the surrounding dielectrics. PD is a symptom of insulation deterioration. Therefore, PD measurement and identification can be used as a good insulation diagnosis tool to optimize both maintenance and life-risk management for power apparatuses. The new standard IEC60270 [11] for PD measurement has been published in 2001, which establishes an integral quality assurance system for PD measurement instead of the old standard IEC60060-2 published in 1994. The standard IEC60270 ensures accuracy of measuring results, comparability and consistency of different instruments
28
P.-H. Chen and H.-C. Chen
and measuring methods. Moreover, the new standard provides digital PD measuring recommendations as well as the analog measuring. In this work, all PD experiments are based on the new standard IEC60270. A PD experiment laboratory, including a set of precious instrument (Hipotronics DDX-7000 Digital Discharge Detector), has been set up in the St. John’s University Shielded Room
Test Object (Transformer)
Coupling Capacitor
Calibration Capacitor
Step-Up Transformer
Isolation Transformer
Safety Door
High Voltage Control Panel DDX-7000 PD Analyzer
Control Room
Fig. 1. Block diagram of PD experiment
Fig. 2. Practical PD field measurement
Fig. 3. PD analyzer and control panel
Application of BPNN to Power Transformer Insulation Diagnosis
29
according to the standard IEC60270 recommendations. Fig. 1 shows the block diagram of the PD experiment laboratory. The constitution of the PD experiment laboratory includes a PD analyzer, a high-voltage control panel, an isolation transformer, a high-voltage generator (step-up transformer), a calibration capacitor, and a coupling capacitor. Fig. 2 shows a practical PD field measurement in the shielded room. Fig. 3 shows the PD analyzer and high-voltage control panel. 2.2 Experimental Models In this work, the experimental objects are power transformers which use epoxy-resin as HV insulation materials. The rated voltage and capacity of the transformers are 12 kV and 2kVA, respectively. For testing purposes, four experimental models of power transformers with typical insulation defects were purposely manufactured by a power apparatuses manufacturer. These typical PD models include a low-voltage coil PD (Type A), a high-voltage coil PD (Type B), a high-voltage corona discharge (Type C), and a healthy transformer (Type D). Fig. 4 shows the pictures of these four experimental models. The voltage step-up procedure of the PD experiment, according to the standard IEC60270 recommendations, is shown in Fig. 5. First, the high-voltage generator generates a rising-voltage from 0V to 18kV, which is the 1.5 times of the rated voltage, in 50 seconds. This high-voltage will be maintained for 1 minute to trigger discharges.
(a) Low-voltage coil PD (Type A).
(b) High-voltage coil PD (Type B).
(c) High-voltage corona discharge (Type C).
(d) Healthy transformer (Type D).
Fig. 4. Four experimental models with typical insulation defects
P.-H. Chen and H.-C. Chen
18kV
Test Voltage (kV)
30
12kV
50
60
20
120
T e st T im e (s e c o n d )
Fig. 5. Voltage step-up procedure
Then, the voltage will descend from 18kV to the rated voltage in 20 seconds. The 12kV rated voltage will be kept and the PD detector starts to measure and record the PD signals for 2 minutes. During the experimental process, all of the measuring analog data are converted to digital data in order to store them in a computer.
3 BPNN Solution Methodology In recent years, a biologically artificial intelligence technique known as artificial neural network (ANN) has emerged as a candidate for the feature identification problems. The basic conception of ANN is intended to model the behavior of biological neural functions. An ANN is generally modeled as a massively parallel interconnected network of elementary neurons. The original desire for the development of ANN is intended to take advantage of parallel processors computing than traditional serial computation. Fig. 6 shows the PD patterns of four typical defects. As shown in Fig. 6, obviously, the problem of transformer insulation diagnosis is essentially a PD patterns classification problem. Therefore, the ANN is an applicable solution tool to the problem of transformer insulation diagnosis.
(a) Low-voltage coil PD (Type A). Fig. 6. Typical PD patterns of four defect types
Application of BPNN to Power Transformer Insulation Diagnosis
31
(b) High-voltage coil PD (Type B).
(c) High-voltage corona discharge (Type C).
(d) Healthy transformer (Type D). Fig. 6. (cont.): Typical PD patterns of four defect types
From the literature survey, several models and learning algorithms of ANN have been proposed for solving the patterns classification problems [12]. In this paper, we establish a triple-layer feed-forward BPNN, as shown in Fig. 7, for solving the PD patterns classification problem. The number of neurons in the output layer is set at the number of defect types. The input data for the BPNN is the field measuring 3-D PD pattern. To fit the form of the input layer and accelerate convergence, each original PD
32
P.-H. Chen and H.-C. Chen
pattern is pre-translated into a 1600x1 matrix. Then, the only control parameter is the number of neurons in the hidden layer. In this work, the transfer function in the hidden layer is the hyperbolic tangent function, as shown in Fig. 8 [12]. The transfer function in the output layer is the sigmoid function, as shown in Fig. 9 [12]. In the training procedure, a faster back-propagation learning algorithm named “RPROP algorithm” is used as the learning rule. Riedmiller and Braun [13] showed that both convergence speed and memory requirement of the RPROP algorithm are better than traditional gradient-descent learning algorithms.
A 3-D PD Pattern
Recognition Result
Translate to 1600x1 Matrix
B C D
1
1
-1
I=1600×1 Input Layer
H Hidden Layer
O=4 Output Layer
Fig. 7. Structure of the BPNN insulation diagnosis system
Fig. 8. Hyperbolic tangent transfer function
Fig. 9. Sigmoid transfer function
4 Experimental Results The proposed BPNN approach was implemented on a MATLAB software and executed on a Pentium IV 2.8GHz personal computer. To illustrate the identification ability of the proposed approach, 160 sets of field measuring PD patterns are used to test the BPNN diagnosis system. The diagnosis system will randomly choose 80 sets of data as the training data, and the rest as the testing data. The structure of the proposed BPNN insulation diagnosis system is shown in Fig. 7. The number of neurons in the output layer (O) depends on the defect types to be identified, which is O=4 in this work. Since the original 3-D PD pattern is translated into a 1600x1 matrix in this work, the
Application of BPNN to Power Transformer Insulation Diagnosis
33
number of neurons in the input layer (I) is set at I=1600. Then, for a multi-layer BPNN, the main control parameter is the number of neurons in the hidden layer (H). Table 1 shows the test results of the proposed BPNN diagnosis system with different number of neurons in the hidden layer from H=20 to H=100. Test results show that H=60 has the highest recognition rate of 90%, 90%, 93%, and 80% for defect type A, B, C, and D, respectively. The proposed system achieves a high average recognition rate of 88.25%. In a practical PD field measurement, the obtaining data would unavoidably contain some noise. The sources of noise include instrumental noise and environmental noise. This work takes noise into account in order to study the noise tolerance ability of the proposed approach. In this example, 240 sets of noise-contained testing data are generated by adding ±10%, ±20%, and ±30%, respectively, of randomly distributed noise into the training data to test the BPNN diagnosis system. Table 2 shows the test results with different percentage of noise. As shown in Table 2, the descending of recognition rate generally follows the raising of noise percentage, a finding that is consistent with general expectations. However, experimental results show a good tolerance to noise interference of the proposed approach, which achieves a high recognition rate of 81.75% in noise of 20% and 75.75% in extreme noise of 30%. Table 1. Recognition rate with different neurons in the hidden layer (H) Defect Type
H
A
B
C
D
Average
20
80%
88%
85%
80%
83.25%
40
88%
88%
90%
80%
86.5%
60
90%
90%
93%
80%
88.25%
80
88%
80%
85%
77%
82.5%
100
88%
80%
77%
73%
79.5%
Table 2. Recognition rate with different percentage of noise Percentage of Noise
Average Recognition Rate
0%
±10%
±20%
±30%
88.25%
86.5%
81.75%
75.75%
5 Conclusions This paper presents a new methodology based on the back-propagation neural network for solving the insulation diagnosis problem of epoxy-resin power transformers. The
34
P.-H. Chen and H.-C. Chen
proposed approach is considered a general tool because it can be easy implemented on the popular MATLAB software. Moreover, the approach is considered a flexible tool because it can be also applied to other high-voltage power apparatuses such as current transformer (CT), potential transformer (PT), cable, and rotation machine. Experimental results indicate that the proposed approach has a high degree of recognition accuracy and good tolerance of noise interference. We expect this work providing useful reference to electric power industry.
Acknowledgment Financial support given to this research by the National Science Council of the Republic of China, Taiwan, under grant No. NSC 93-2213-E-129-015 is greatly appreciated.
References 1. Feinberg, R.: Modern Power Transformer Practice. John Wiley & Sons (1979) 2. Gulski, E., Burger, H.P., Vaillancourt, G.H., Brooks, R.: PD Pattern Analysis During Induced Test of Large Power Transformers. IEEE Trans. Dielectrics and Electrical Insulation 7 (2000) 95–101 3. Kim, C.S., Kondo, T., Mizutani, T.: Change in PD Pattern with Aging. IEEE Trans. Dielectrics and Electrical Insulation 11 (2004) 13–18 4. Krivda, A.: Automated Recognition of Partial Discharge. IEEE Trans. on Dielectrics and Electrical Insulation 2 (1995) 796-821 5. Tomsovic, K., Tapper, M., Ingvarsson, T.T.: A Fuzzy Information Approach to Integrating Different Transformer Diagnostic Methods. IEEE Trans. on Power Delivery 8 (1993) 1638–1643 6. Satish, L., Gururaj, B.I.: Application of Expert System to Partial Discharge Pattern Recognition. in CIGRE Study Committee 33 Colloquium, Leningrad, Russia, (1991) Paper GIGRE SC 33.91 7. Wang, M.H., Ho, C.Y.: Application of Extension Theory to PD Pattern Recognition in High-Voltage Current Transformers. IEEE Trans. Power Delivery 20 (2005) 1939–1946 8. Wang, M.H.: Partial Discharge Pattern Recognition of Current Transformers Using an ENN. IEEE Trans. Power Delivery 20 (2005) 1984–1990 9. Cavallini, A., Montanari, G.C., Puletti, F., Contin, A.: A New Methodology for the Identification of PD in Electrical Apparatus: Properties and Applications. IEEE Trans. Dielectrics and Electrical Insulation 12 (2005) 203-215 10. Salama, M.M.A., Bartnikas, R.: Determination of Neural Network Topology for Partial Discharge Pulse Pattern Recognition. IEEE Trans. Neural Networks 13 (2002) 446–456 11. IEC: High-Voltage Test Techniques-Partial Discharge Measurements. IEC 60270 (2001) 12. Wasserman, P.D.: Neural Computing, Theory and Practice. Van Nostrand Reinhold (1989) 13. Riedmiller, M., Braun, H.: A Direct Adaptive Method for Faster Back Propagation Learning- The RPROP Algorithm. IEEE International Conference on Neural Networks 1 (1993) 586-591
St. John’s University, Department of Electrical Engineering, Taipei, Taiwan, 25135, R.O.C. [email protected] 2 National Chin-Yi University of Technology, Department of Electrical Engineering, Taichung, Taiwan, 411, R.O.C. [email protected]
Abstract. This paper presents a novel approach based on the back-propagation neural network (BPNN) for the insulation diagnosis of power transformers. Four epoxy-resin power transformers with typical insulation defects are purposely made by a manufacturer. These transformers are used as the experimental models of partial discharge (PD) examination. Then, a precious PD detector is used to measure the 3-D (φ-Q-N) PD signals of these four experimental models in a shielded laboratory. This work has established a database containing 160 sets of 3-D PD patterns. The database is used as the training data to train a BPNN. The training-accomplished neural network can be a good diagnosis system for the practical insulation diagnosis of epoxy-resin power transformers. The proposed BPNN approach is successfully applied to practical power transformers field experiments. Experimental results indicate the attractive properties of the BPNN approach, namely, a high recognition rate and good noise elimination ability.
1 Introduction More than half of the breakdown accidents of power apparatuses are caused by insulation deterioration. The reliability of a power apparatus is affected significantly by the presence of insulation defects. Partial discharge (PD) pattern recognition has been regarded as an important diagnosis method to prevent power apparatuses from malfunction of insulation defect. However, the measurement of PD needs heavy power equipment and precious instruments. The Heavy Power Lab in St. John’s University installed a set of precious instrument (Hipotronics DDX-7000 Digital Discharge Detector). Therefore, the school has the capability of doing insulation diagnosis related researches. Power transformer is one of the most important apparatuses in a power delivery system. Breakdown of a power transformer can cause an interruption in electricity supply and result in a loss of considerable profits [1]. Therefore, detecting insulation defects in a power transformer as early as possible is of priority concern to a power transformer user. PD phenomenon usually originates from insulation defects and is an important symptom to detect incipient failures in power transformers. Classification of different types of PD D. Liu et al. (Eds.): ISNN 2007, LNCS 4493, Part III, pp. 26–34, 2007. © Springer-Verlag Berlin Heidelberg 2007
Application of BPNN to Power Transformer Insulation Diagnosis
27
patterns is of importance for the diagnosis of the quality of power transformers. PD behavior can be represented in various ways. Because of the randomization of PD activity, one of the most popular representations is the 3-D (φ-Q-N) distribution [2], [3], i.e., the PD pattern is described using a pulse count N versus pulse height Q and phase angle φ diagram. Previous experimental results have adequately demonstrated that φ-Q-N distributions are strongly dependent upon PD sources, therefore, the 3-D patterns can be used to characterize insulation defects [4]. This provides the basis for pattern recognition techniques that can identify the different types of defects. Previous efforts at PD pattern recognition have applied various identification techniques to make the problem solvable. Various pattern recognition techniques such as fuzzy clustering [5], expert system [6], extension theory [7], [8], and statistical analysis [9] have been proposed. These techniques have been successfully applied to PD pattern recognition. However, these conventional approaches not only require human experiences but also have some difficulties in acquiring knowledge and maintaining the database of decision rules. In recent years, a biologically artificial intelligence technique known as artificial neural network has emerged as a candidate for the pattern identification problem. The neural network can directly acquire experience from the training data to overcome the shortcomings of the conventional approaches. The neural network can quickly and stably learn to categorize input patterns and permit adaptive processes to access significant new information [10]. In this paper, a novel insulation diagnosis approach based on the PD and back-propagation neural network (BPNN) is proposed for epoxy-resin power transformers. Four cast-resin transformers with typical insulation defects, which were purposely made by a manufacturer, are used as the models of the PD examination. The DDX-7000 Digital Discharge Detector is then used to measure the 3-D PD signals of these transformers. So far, this work has established a database which contains 160 sets of 3-D PD patterns. Then, this database is used as the training data to train a BPNN. Finally, the training-accomplished neural network can be a good diagnosis system for the practical insulation diagnosis of epoxy-resin power transformers. Experimental results show that different types of insulation defects within power transformers are identified with rather high recognition rate.
2 Partial Discharge Experiments 2.1 Partial Discharge Measurement PD is a forced phenomenon occurring in insulation parts in a power apparatus. When the intensity of electric field exceeds the breakdown threshold value of a defective dielectric, PD occurs and results in a partial breakdown in the surrounding dielectrics. PD is a symptom of insulation deterioration. Therefore, PD measurement and identification can be used as a good insulation diagnosis tool to optimize both maintenance and life-risk management for power apparatuses. The new standard IEC60270 [11] for PD measurement has been published in 2001, which establishes an integral quality assurance system for PD measurement instead of the old standard IEC60060-2 published in 1994. The standard IEC60270 ensures accuracy of measuring results, comparability and consistency of different instruments
28
P.-H. Chen and H.-C. Chen
and measuring methods. Moreover, the new standard provides digital PD measuring recommendations as well as the analog measuring. In this work, all PD experiments are based on the new standard IEC60270. A PD experiment laboratory, including a set of precious instrument (Hipotronics DDX-7000 Digital Discharge Detector), has been set up in the St. John’s University Shielded Room
Test Object (Transformer)
Coupling Capacitor
Calibration Capacitor
Step-Up Transformer
Isolation Transformer
Safety Door
High Voltage Control Panel DDX-7000 PD Analyzer
Control Room
Fig. 1. Block diagram of PD experiment
Fig. 2. Practical PD field measurement
Fig. 3. PD analyzer and control panel
Application of BPNN to Power Transformer Insulation Diagnosis
29
according to the standard IEC60270 recommendations. Fig. 1 shows the block diagram of the PD experiment laboratory. The constitution of the PD experiment laboratory includes a PD analyzer, a high-voltage control panel, an isolation transformer, a high-voltage generator (step-up transformer), a calibration capacitor, and a coupling capacitor. Fig. 2 shows a practical PD field measurement in the shielded room. Fig. 3 shows the PD analyzer and high-voltage control panel. 2.2 Experimental Models In this work, the experimental objects are power transformers which use epoxy-resin as HV insulation materials. The rated voltage and capacity of the transformers are 12 kV and 2kVA, respectively. For testing purposes, four experimental models of power transformers with typical insulation defects were purposely manufactured by a power apparatuses manufacturer. These typical PD models include a low-voltage coil PD (Type A), a high-voltage coil PD (Type B), a high-voltage corona discharge (Type C), and a healthy transformer (Type D). Fig. 4 shows the pictures of these four experimental models. The voltage step-up procedure of the PD experiment, according to the standard IEC60270 recommendations, is shown in Fig. 5. First, the high-voltage generator generates a rising-voltage from 0V to 18kV, which is the 1.5 times of the rated voltage, in 50 seconds. This high-voltage will be maintained for 1 minute to trigger discharges.
(a) Low-voltage coil PD (Type A).
(b) High-voltage coil PD (Type B).
(c) High-voltage corona discharge (Type C).
(d) Healthy transformer (Type D).
Fig. 4. Four experimental models with typical insulation defects
P.-H. Chen and H.-C. Chen
18kV
Test Voltage (kV)
30
12kV
50
60
20
120
T e st T im e (s e c o n d )
Fig. 5. Voltage step-up procedure
Then, the voltage will descend from 18kV to the rated voltage in 20 seconds. The 12kV rated voltage will be kept and the PD detector starts to measure and record the PD signals for 2 minutes. During the experimental process, all of the measuring analog data are converted to digital data in order to store them in a computer.
3 BPNN Solution Methodology In recent years, a biologically artificial intelligence technique known as artificial neural network (ANN) has emerged as a candidate for the feature identification problems. The basic conception of ANN is intended to model the behavior of biological neural functions. An ANN is generally modeled as a massively parallel interconnected network of elementary neurons. The original desire for the development of ANN is intended to take advantage of parallel processors computing than traditional serial computation. Fig. 6 shows the PD patterns of four typical defects. As shown in Fig. 6, obviously, the problem of transformer insulation diagnosis is essentially a PD patterns classification problem. Therefore, the ANN is an applicable solution tool to the problem of transformer insulation diagnosis.
(a) Low-voltage coil PD (Type A). Fig. 6. Typical PD patterns of four defect types
Application of BPNN to Power Transformer Insulation Diagnosis
31
(b) High-voltage coil PD (Type B).
(c) High-voltage corona discharge (Type C).
(d) Healthy transformer (Type D). Fig. 6. (cont.): Typical PD patterns of four defect types
From the literature survey, several models and learning algorithms of ANN have been proposed for solving the patterns classification problems [12]. In this paper, we establish a triple-layer feed-forward BPNN, as shown in Fig. 7, for solving the PD patterns classification problem. The number of neurons in the output layer is set at the number of defect types. The input data for the BPNN is the field measuring 3-D PD pattern. To fit the form of the input layer and accelerate convergence, each original PD
32
P.-H. Chen and H.-C. Chen
pattern is pre-translated into a 1600x1 matrix. Then, the only control parameter is the number of neurons in the hidden layer. In this work, the transfer function in the hidden layer is the hyperbolic tangent function, as shown in Fig. 8 [12]. The transfer function in the output layer is the sigmoid function, as shown in Fig. 9 [12]. In the training procedure, a faster back-propagation learning algorithm named “RPROP algorithm” is used as the learning rule. Riedmiller and Braun [13] showed that both convergence speed and memory requirement of the RPROP algorithm are better than traditional gradient-descent learning algorithms.
A 3-D PD Pattern
Recognition Result
Translate to 1600x1 Matrix
B C D
1
1
-1
I=1600×1 Input Layer
H Hidden Layer
O=4 Output Layer
Fig. 7. Structure of the BPNN insulation diagnosis system
Fig. 8. Hyperbolic tangent transfer function
Fig. 9. Sigmoid transfer function
4 Experimental Results The proposed BPNN approach was implemented on a MATLAB software and executed on a Pentium IV 2.8GHz personal computer. To illustrate the identification ability of the proposed approach, 160 sets of field measuring PD patterns are used to test the BPNN diagnosis system. The diagnosis system will randomly choose 80 sets of data as the training data, and the rest as the testing data. The structure of the proposed BPNN insulation diagnosis system is shown in Fig. 7. The number of neurons in the output layer (O) depends on the defect types to be identified, which is O=4 in this work. Since the original 3-D PD pattern is translated into a 1600x1 matrix in this work, the
Application of BPNN to Power Transformer Insulation Diagnosis
33
number of neurons in the input layer (I) is set at I=1600. Then, for a multi-layer BPNN, the main control parameter is the number of neurons in the hidden layer (H). Table 1 shows the test results of the proposed BPNN diagnosis system with different number of neurons in the hidden layer from H=20 to H=100. Test results show that H=60 has the highest recognition rate of 90%, 90%, 93%, and 80% for defect type A, B, C, and D, respectively. The proposed system achieves a high average recognition rate of 88.25%. In a practical PD field measurement, the obtaining data would unavoidably contain some noise. The sources of noise include instrumental noise and environmental noise. This work takes noise into account in order to study the noise tolerance ability of the proposed approach. In this example, 240 sets of noise-contained testing data are generated by adding ±10%, ±20%, and ±30%, respectively, of randomly distributed noise into the training data to test the BPNN diagnosis system. Table 2 shows the test results with different percentage of noise. As shown in Table 2, the descending of recognition rate generally follows the raising of noise percentage, a finding that is consistent with general expectations. However, experimental results show a good tolerance to noise interference of the proposed approach, which achieves a high recognition rate of 81.75% in noise of 20% and 75.75% in extreme noise of 30%. Table 1. Recognition rate with different neurons in the hidden layer (H) Defect Type
H
A
B
C
D
Average
20
80%
88%
85%
80%
83.25%
40
88%
88%
90%
80%
86.5%
60
90%
90%
93%
80%
88.25%
80
88%
80%
85%
77%
82.5%
100
88%
80%
77%
73%
79.5%
Table 2. Recognition rate with different percentage of noise Percentage of Noise
Average Recognition Rate
0%
±10%
±20%
±30%
88.25%
86.5%
81.75%
75.75%
5 Conclusions This paper presents a new methodology based on the back-propagation neural network for solving the insulation diagnosis problem of epoxy-resin power transformers. The
34
P.-H. Chen and H.-C. Chen
proposed approach is considered a general tool because it can be easy implemented on the popular MATLAB software. Moreover, the approach is considered a flexible tool because it can be also applied to other high-voltage power apparatuses such as current transformer (CT), potential transformer (PT), cable, and rotation machine. Experimental results indicate that the proposed approach has a high degree of recognition accuracy and good tolerance of noise interference. We expect this work providing useful reference to electric power industry.
Acknowledgment Financial support given to this research by the National Science Council of the Republic of China, Taiwan, under grant No. NSC 93-2213-E-129-015 is greatly appreciated.
References 1. Feinberg, R.: Modern Power Transformer Practice. John Wiley & Sons (1979) 2. Gulski, E., Burger, H.P., Vaillancourt, G.H., Brooks, R.: PD Pattern Analysis During Induced Test of Large Power Transformers. IEEE Trans. Dielectrics and Electrical Insulation 7 (2000) 95–101 3. Kim, C.S., Kondo, T., Mizutani, T.: Change in PD Pattern with Aging. IEEE Trans. Dielectrics and Electrical Insulation 11 (2004) 13–18 4. Krivda, A.: Automated Recognition of Partial Discharge. IEEE Trans. on Dielectrics and Electrical Insulation 2 (1995) 796-821 5. Tomsovic, K., Tapper, M., Ingvarsson, T.T.: A Fuzzy Information Approach to Integrating Different Transformer Diagnostic Methods. IEEE Trans. on Power Delivery 8 (1993) 1638–1643 6. Satish, L., Gururaj, B.I.: Application of Expert System to Partial Discharge Pattern Recognition. in CIGRE Study Committee 33 Colloquium, Leningrad, Russia, (1991) Paper GIGRE SC 33.91 7. Wang, M.H., Ho, C.Y.: Application of Extension Theory to PD Pattern Recognition in High-Voltage Current Transformers. IEEE Trans. Power Delivery 20 (2005) 1939–1946 8. Wang, M.H.: Partial Discharge Pattern Recognition of Current Transformers Using an ENN. IEEE Trans. Power Delivery 20 (2005) 1984–1990 9. Cavallini, A., Montanari, G.C., Puletti, F., Contin, A.: A New Methodology for the Identification of PD in Electrical Apparatus: Properties and Applications. IEEE Trans. Dielectrics and Electrical Insulation 12 (2005) 203-215 10. Salama, M.M.A., Bartnikas, R.: Determination of Neural Network Topology for Partial Discharge Pulse Pattern Recognition. IEEE Trans. Neural Networks 13 (2002) 446–456 11. IEC: High-Voltage Test Techniques-Partial Discharge Measurements. IEC 60270 (2001) 12. Wasserman, P.D.: Neural Computing, Theory and Practice. Van Nostrand Reinhold (1989) 13. Riedmiller, M., Braun, H.: A Direct Adaptive Method for Faster Back Propagation Learning- The RPROP Algorithm. IEEE International Conference on Neural Networks 1 (1993) 586-591
