Flashover Prediction of Pin-type Insulator Based on the Characteristics of Leakage Current using ANN Bala Kumaran.M, 2 Goutham.S, 3 Raja Sabari.T, 4 Vimalathithan.S, 5 Vijeesh.V,2,3,4 UG Scholar, Department of Electrical and Electronics Engineering, Velammal Institute of Technology, Chennai, India. 5 Assistant Professor I, Department of Electrical and Electronics Engineering, Velammal Institute of Technology, Chennai, India. ABSTRACT The phenomenon of flash over in polluted insulators has been continued by the study of the characteristics of contaminating layers deposited on the surface of insulators in high voltage laboratories. In this paper pin type insulator model is used for studying the flash over phenomenon due to pollution under wet conditions given at low voltage. Laboratory based test were carried out on the sample model under AC voltage at different pollution levels. Different concentration of salt solution has been prepared using NaCl, Kaolin & distilled water representing the various contaminations. Leakage current during the experimental studies were measured for various polluted conditions and the results are compared with the simulation results. A new model of Vc = f( V, I mean, I max and I σ ) based on artificial neural network has been developed to predict flashover from the analysis of leakage current. The input variable to the artificial neural network are Mean(I mean ), Maximum(I max ) and Standard Deviation(I σ ) of leakage current and the input voltage(v).the target obtained was used to evaluate the performance of the neural network model. The developed ANN model is well-suited for the analysis of leakage current to predict flashover on the pin-type insulator with high accuracy. Keywords: Insulator, Leakage current, Flash over, Pollution, MSE, Artificial Neural Network.. INTRODUCTION Insulators used in outdoor electric power transmission lines are exposed to outdoor environmental contaminations which enhances the chances of flash over. Depending on the nature and duration of exposure, deposits of wind carried industrial, sea and dust contaminants build up on the insulator surface as a dry layer. The leakage current path through a layer of dry contaminants on an insulator surface capacitive where in the current amplitude is small and sinusoidal. The dry contaminant layer becomes conductive when exposed to light rain or morning dews as wetting progresses the leakage current path changes from capacitive to resistive with simultaneous increase in current amplitude. The increase in leakage current dries the conducting layer and forms the dry band around the areas with the high current density. These dry bands interrupt the current flow and most of the applied voltages are impressed across these narrow dry bands. If the dry bands cannot withstand the voltage, localized arcing develops and the dry bands will be spanned by discharges. The arcs merge together and form a single arc which triggers the surface flashover [6]. The contamination severity determines the frequency and intensity of arcing and, thus the probability of flash over in favorable condition when the level of the contamination is low, layer resistance is high and the arcing continues until the sun or wind dries the layer and stops the arcing. Continuous arcing is harmless for ceramic insulators[3]. The mechanism described above shows that heavy contamination and wetting cause insulator flash over and service interruption. Contamination in dry condition is harmless. the method of dry band formation, with subsequent growth of discharges on the polluted surface []. Karady measured the peak leakage current and correlated the current with flash over voltage. He suggested the flash over is imminent the leakage current peaks 00 ma[2]. In practice, there is various contaminant type that settle on outdoor insulators. These contaminants can be classified as soluble and insoluble. Insulators located near coastal regions are typically contaminated by soluble contaminants, especially salt (or Nacl). Insulators located near cement or paper industries are typically contaminated by non-soluble contaminants such as calcium chloride, carbon and cement dust. Irrespective of the type of contaminant, flashover can occur as long as the salts in the contaminant are soluble enough to form a conducting layer on the insulators surface. Many researchers studied that the leakage current due to the contamination level is the main cause for flashover [7]. Artificial neural networks (ANNs) can be used in problems requiring function approximation, modeling, pattern recognition and classification, estimation and prediction, etc. Jingyan et al.[2] extracted three characteristics of the leakage current, namely the mean value, maximum value, and the standard deviation of the root-mean-square (RMS) value of the leakage current, from the recorded value. 2. MATERIALS AND METHODS A. Insulator Model The samples used in this experiment were porcelain 230V insulators. BF. Hampton investigated the voltage distribution along the wet, polluted surface of a pin type insulator and 359
parameters is done by analyzing and comparing experimental results and the simulation results [4][5]. A. Clean Insulator: C 0.p R R2 I VOFF = 0 VAMPL = 230v FREQ = 50 V 0k Fig : Pin type insulator B. Experimental Set up A pin type insulator model of dimension of 0cm of height and 25cm of diameter is used for the contamination flashover experiments. In artificial testing, a contaminant is usually substituted by a dissolved mixture of an inert binder-kaolin and Nacl salt. The inert binder is supposedly non-conducting and the quantity of salt represents the level of contamination. Contamination salt solution was prepared of various NaCl values of 5g,20g,30g,40g,60g,80g,00g,20g the mixture usually dissolved in distilled water is known as slurry which is thoroughly mixed as per IEC std[8]. Before coating, trough is initially washed and wiped clean and dry. The experimental setup to measure the leakage current is shown in figure 2. The slurry is poured so that it rolls off uniformly in the trough. The contamination can thus be classified as light, medium or heavy according to the IEC standard []. B. Polluted Insulator: VOFF = 0 VAMPL = 230v FREQ = 50 V 0 Fig 3: Clean insulator R2 2 3.5k 0 C 0.p R 0k R3 I Fig 4: Polluted insulator C. Polluted Insulator With Dry Band Discharge: R5 R6 3.5k 0 C5 0.5p C4 0.p R9 I R3 Fig 2: Experimental setup VOFF = 0 FREQ = 50 V9.75k 3. EQUIVALENT CIRCUIT MODEL AND SIMULATION OF LEAKAGE CURRENT The general equivalent circuit model for insulator based on the contaminations such as clean, polluted and polluted insulator with dry band discharge are shown in figure. It consists of capacitor(s) and nonlinear resistor(s). By using the equivalent circuit leakage current waveforms were simulated using Pspice simulation software. Validation of the proposed equivalent circuit model and its VAMPL = 230 Fig 5: Polluted insulator with dry band discharge By simulating the above three circuit we can get different leakage current, for the clean insulator we got the maximum amplitude of 30mA, for the polluted insulator the leakage current is 50mA and finally for the insulator with dry band discharge we got the maximum amplitude of 250mA.The simulated output waveforms of the equivalent insulator circuit model is shown in figure. 0 360
Table : Experimental Readings Fig 6: Leakage current in clean insulator NaCl content in solution(g) Maximum leakage current(ma) 5 3 20 42 30 6 40 84 60 05 80 48 00 20 20 39 The experimental results show that the amplitude of the leakage current increases with increase in NaCl content. The waveform of the Kaolin coated with different Nacl concentration such as 5g, 20g, 30g, 40g, 60g, 80g, 00g, 20g are shown in figure 9-6. Fig 7: Leakage current in polluted insulator Fig 9: 5g of NaCl Fig 8: Leakage current in polluted insulator with dry band discharge A test voltage of 230V, 50 Hz was applied across the terminals and the leakage current is monitored through the suitable measuring meter from the instant of application of voltage till the formation of dry band. The dry band was precisely located on the model. Its shape, contour of growth and locations were physically measured. The test results either in a flashover or withstand. The leakage current is measured using ammeter and also voltage waveforms are captured in digital signal oscilloscope (DSO). By connecting the decade resistance of kω we can also calculate the value of leakage current. The value of leakage current that we obtained during experimentation is shown in the table. Fig0: 20g of NaCl Fig : 30g of NaCl 36
Fig 2: 40g of NaCl Fig 6: 20g of NaCl Table 2: COMPARISON BETWEEN MEASURED AND SIMULATED WAVEFORMS IN TERMS OF AMPLITUDE Insulator Simulation Output Experimental Output CLEAN (low polluted) 22mA 3mA,42mA,6mA,84mA Fig 3: 60g of NaCl 50mA 05mA,48mA POLLUTED (Medium polluted) DRY BAND DISCHARGE (Highly polluted) 250mA 20mA,39mA 4. NEURAL NETWORK MODEL FOR PREDICTING FLASHOVER BASED ON CHARACTERISTICS OF LEAKAGE CURRENT Fig 4: 80g of NaCl Fig 5: 00g of NaCl A. Neural Network ANN is a computer model representing the biological brain. It consists of a set of interconnected simple processing units (neurons or nodes) bonded with weight connections which combine to output a signal to solve a certain problem based on the input signals it received. The process of training an ANN can be supervised and unsupervised learning. In supervised learning, the ANN is previously trained and the target is fixed for a particular task. The ANN are adjusted or trained, so that a particular input leads to a specific target output. In the output layer the data is compared with the predefined target and the error is calculated. This error is propagated back and the weight updation is done. The process is repeated to meet the minimum error value, the training task is deemed complete. In unsupervised learning, the target is not fixed; weights are adjusted autonomously until a balanced condition is reached when the weights do not change further. The back propagation learning algorithm [9,0] which is a generalization of widrow-hoff error correction rule is the most popular method in training the ANN and is employed in this work. 362
B. Characteristics of Leakage Current Table 2 shows the experimental results of varying leakage current magnitude for different level of contaminations.the input parameters selected for the artificial neural network were the three characteristics of the measured leakage current which were Maximum value of the leakage current(i MAX ),and Standard deviation of leakage current(σ) and the input supply voltage. The three characteristics, [2]]i.e., the mean value, maximum value and standard deviation of the leakage current, are proposed as follows: values of a group of data are used to normalize each single input using equation given below. I = Where I max and I min are maximum and minimum values of measured input data. D. Flowchart and Algorithm (4) Fig 7: Neural Network Structure I em = () I emax =max (I e (i)) (2) σ= (3) Where N is the total number of sampling points in the test time; I e (i) is the leakage current value in one sampling period; I em is the mean value of leakage current in the test time; I emax is the maximum value of leakage current in the test time; σ is the standard deviation of leakage current in the test time. The standard deviation represents the degree of deviation between each sampling value and the mean value and also means the discrete distribution degree among all sampling points during the test time. C. Normalization of Input variables Normalization is a transformation performed on a single data input to distribute the data evenly and scale it into an acceptable range for the neural network. Such preprocessing of data ensures a uniform statistical distribution of each input value. If the input and the output variables are not of the same order of magnitude, some variables may appear to have more significance than they actually do. Input and output data can be normalized in many ways. In this work, the minimum and maximum Fig 8: Flowchart of the prediction process The flowchart of fig.8 shows the entire prediction process of flashover. Step : The leakage current signals are obtained from the experimental results for different level of contaminations. Step 2: The three characteristics, i.e., the mean value, maximum value and standard deviation of the leakage current for the various level are 363
calculated from the measured leakage current. Table 3: R 2 And Mse Value For Various Numbers Of Nodes In Hidden Layers Step 3: The leakage current for the occurrence of the flashover or withheld for various contamination levels and input voltage are fed to ANN and trained. Step 4: Then the BP algorithm is repeatedly executed, while all parameters are simultaneously adjusted into their respective ranges, in order to select the combination that produces the minimum forecast error for the given set. Step 5: Next, the trained data is tested with the cases which are not used for training. Step 6: Finally, the output result in withheld or flashover with respect to the contamination level prediction. E. Training of ANN: For training the neural network to predict flashover, Four input neurons are used, namely three to indicate the extracted leakage current characteristics,(the mean value, maximum value and standard deviation) and another neuron for input voltage, four neurons in the first hidden layers, fifteen neurons in the second hidden layer and one neuron in the output layer. The hidden layer is using tangent sigmoid transfer function and output neuron is using linear transfer function. All the input data used in the proposed neural networks were actual collected data based on actual measurements. The normalized training patterns are fed to the model. Leven Marquardt algorithm is used for training and the function trainlm is invoked. Using trial and error the numbers of nodes in the hidden layer are determined. The selection of the number of hidden neurons will ensure correct learning, and the network is able to correctly predict the data it has been trained. With too few hidden neurons, the network may be unable to learn the relationships amongst the data and the error will fail to fall below an acceptable level. The optimization process has been carried out based on MSE and R 2 value by varying the number of hidden layers. The number of hidden layer neurons is varied from to 5. No. of nodes in Hidden layers MSE R 2.03 0.99998 2 3.96 0.99986 3 0.7 0.99986 4 0.358 0.99994 5 2.9 0.99994 6 0.483 0.99988 7 0.983 0.99986 8 2. 0.99986 9 2.33 0.99994 0 4.49 0.99996.8 0.99988 2 2.92 0.99996 3 8.6 0.99994 4.72 0.99988 5 0.75 Table 3 shows the test that was made in order to decide the number of neurons of the hidden layer. It was found that the minimum MSE appears for fifteen neurons in the hidden layer. The same tests were repeated in order to find which network architecture and which values in the parameters of the ANN give best results. Table 4 compares the effect of number of hidden layers on the convergence rate of the training process. Finally, the ANN network using three hidden layer with three neurons in first hidden layer, eight neurons in second hidden layer and eleven neurons in third hidden layer has better effect on the convergence based on the minimum RMSE than the single hidden layer. The training is carried out for 400 set of input data and the neural network is tested for 9 set of input data to attain the performance goal of 0.000. The goal is achieved in 52 epochs. Table 4: Test Concerning The Architecture Of The Network No. of Hidden layers No of nodes in hidden layers MSE R 2 4 / 3 0.45 0.99988 2 4 / 5 0.75 3 4/ 9 7.2 0.99986 4 4 / 5/ 0.65 0.9999 5 4/ 5/0/4 3.65 0.99996 364
5. RESULTS AND DISCUSSION The parameters obtained from training BP ANN is shown below Mean Square Error : 0.75 Number of Epochs : 52 Time of training : Sec Range of Error : 0-4 Gradient value : 0.00574 Regression coefficient R 2 : The target for no arcing and no flashover are trained to be 0, Initialization of arc and no flashover are trained to be and for severe arcing and approaching flashover, it is trained to be 2. values. The maximum value that the correlation can take is. That means that the closer the correlation is to, the better the network operates. Fig. 20 presents the correlation between the estimated and the real values for the training set to test the network. As it is shown, the correlation is R =. Fig 20: Correlations between real and estimated Values of flashover The occurrence of flashover i.e. the target noted from the measured leakage current during the experiment and the simulated ANN result have been plotted for nine tests chosen randomly from the data stored as shown in Table 5. Fig 9: Performance Graph The performance graph of the variations of mean square error in the training process is shown in fig.4 against the number of iterations. Table 5: Ann Simulation Results ANN Output (Flashover) 0.000 0 0.000 0.9999 2 2 2 Modes of Flashover Low pollution withstand Medium pollution arcing and no flashover Heavy pollution arcing with dry band discharge From the trained network, testing is carried out and test results are compared with experimental results are shown in table 5. Another way to test the accuracy of the ANN is to see the correlation between the real values and the values that come up as output of the ANN that is the estimated 6. CONCLUSION In this paper, a simple scheme is proposed to predict flashover of insulator model using ANN. To simplify the mathematical analysis, tests were done on a Pin-type insulator. The maximum leakage current is determined. The network was modeled as Vc = f( V, I mean, I max and I σ ) to predict the flashover accurately to assess the condition of the insulation system. Thus the flashover was predicted with an accuracy of 99.98%, regression coefficient of about. The high R 2 value indicates a high generalization capability of the developed model. The simulation result is experimentally verified and the simulated errors of this predicting model presented are satisfactory. The proposed technique shows an effective solution against the occurrence of a pollution flashover and indicates the need for cleaning of the insulators. REFERENCES [] B.F. Hampton, Flashover mechanism of polluted insulation. Proc.IEEE vol., pp.985-990, 964. [2] George Karady, Felix Amarh, Raji Sundararajan, Dynamic modeling of an insulator flashover characteristics, th International Symposium of High Voltage Engineering, Vol.4, pp.07-0, London, England, August 999. 365
[3] S.Kumagai,N.Yosimura, Leakage Current Characterization for Estimating the conditions of ceramic and polymeric insulating surfaces, IEEE Trans. Dielectric. Electric. Insul.,vol.,pp. 68-690,2004. [4] Suwarno and Ario Basuki Wibowo, Increasing the performances of various Types Outdoor Insulatorsby using RTV Silicone Rubber Coating, International Journal on Electrical Engineering and Informatics Volume 4, Number 4, December 202. [9] Simon S Haykins, Neural networks A comprehensive foundation, 3 rd edition Pearson education, reprint: 2009. [0] Ajith Abraham, Artificial neural networks, Handbook of Measuring System Design, edited by Peter H. Sydenham and Richard Thorn. 2005 John Wiley & Sons, Ltd. ISBN: 0-470-0243-8. [] IEC6085, Guide for the selection of insulators in respect of polluted conditions, 986. [5] Suwarno, Fari Pratomosiwi, Electrical Equivalent Circuit of Ceramic Insulators with RTV Silicone Rubber Coating and Computer Simulation of Leakage Currents, WSEAS TRANSACTIONS on CIRCUITS and SYSTEMS, Issue 4, Volume 8, April 2009 [6] Ebenezer Jeyakumar, Development of verisimilar juxtaposition model and study of physical phenomena on polluted insulators, PhD Dissertation, Department of Electrical Engineering, Anna University Madras,India,June 99. [7] Guan Zhicheng,Cui Guoshun, A study on the leakage current along the suface of polluted insulator, properties and applications of dielectric materials proceedings of 4 th international conference on volume 2,3-8 July,pp.495-498,994. [2] Jingyan Li,Caixin Sun,Wenxia Sima,Qing Yang, Jianlin Hu, Contamination Level Prediction of Insulators Based on the Characteristics of Leakage Current,IEEE Transaction on Power Delivary,vol25,no:,January200.DOI:0.09/TP WRD.2009.2035426. AUTHOR PROFILES. Bala Kumaran.M, Goutham.S, Raja Sabari.T, Vimalathithan.S are currently proceeding final year B.E. Electrical and Electronics Engineering at Velammal Institute of Technology, Tamilnadu,India. 2. Vijeesh.V received his Masters in Power Systems Engineering from Government college of Technology, Coimbatore in 20 and currently he is an Assistant professor I at Velammal Institute of Technology, Panchetti, Chennai, Tamilnadu, India. [8] IEC 60507, Artificial pollution tests on high voltage insulators to be used in ac system, Switzerland, 99. 366