The importance of the research on insulator pollution has been increased considerably with the rise of the voltage of transmission lines. In order to determine the flashover behavior of polluted high voltage insulators and to identify to physical mechanisms that govern this phenomenon, the researchers have been brought to establish a modeling. In this paper, a dynamic model of AC flashover voltages of the polluted insulators is constructed using the least square support vector machine (LS-SVM) regression method. For this purpose, a training set is generated by using a numerical method based on Finite Element Method (FEM) for several of common insulators with different geometries. To improve the resulting model's generalization ability, an efficient optimization algorithm known as the grid search are adopted to tune parameters in LS-SVM design. In addition, two different testing set, which are not introduced to the LS-SVM during the training procedures, is used to evaluate the effectiveness and feasibility of the proposed method. Then, optimum LS-SVM model is firstly obtained and the performance of the proposed system with other intelligence method based on ANN is compared. It can be concluded that the performance of LS-SVM model outperforms those of ANN, for the data set available, which indicates that the LS-SVM model has better generalization ability.

Prediction of flashover voltage of insulators using least squares support vector machines

Assessment of ESDD on high-voltage insulators using artificial neural network

This paper reports an investigation into the leakage current harmonic components of 45 units of glass insulator samples. The insulator units comprise 5 groups of insulators: new insulators and naturally field-aged insulators with different ageing periods->30, 20, 15, and <;10 years from a 132 kV transmission line. Experiments were conducted at different electrical stresses (0.17, 0.24, and 0.34 kV/cm) and salt-fog conductivities (900 and 1500 μS/cm). The results suggest that the insulator samples' leakage current (LC) waveform and the total harmonic distortion (THD) correspond well with degree of ageing. Fast Fourier Transform (FFT) analysis performed on the stored waveforms showed a strong correlation between the ratio of the third to fifth harmonic amplitudes with the degree of ageing and in-service period of the naturally field-aged insulators. Furthermore, this investigation revealed a pattern between the groups of insulator samples in respect to their ageing. These findings were further validated with traditional Dielectric Dissipation Factor tests carried out on the insulators, and the results also correlated well with the samples' degree of ageing as well as service period. In addition, neural networks were trained to recognize the insulator samples' degree of ageing. This study found that, alternative to the diagnosis of insulator ageing through the measurement of loss angles or tangent delta, LC harmonics can be used as diagnostic tools and for the on-line monitoring of in-service aged transmission line ceramic glass insulators based on odd harmonics and third to fifth harmonic ratios.

Odd harmonics and third to fifth harmonic ratios of leakage currents as diagnostic tools to study the ageing of glass insulators

http://users.ntua.gr/vkont/EPSR_77_2007.pdf

Design of an artificial neural network for the estimation of the flashover voltage on insulators

Leakage current (LC) measurement is one of the effective methods for analysis of polluted insulators. But the traditional evaluation methods do not combine well the time-domain with frequency-domain characteristics of LC. In order to evaluate the safety condition of polluted insulators more effectively, an artificial neutral network (ANN) based method was developed in this paper. Firstly, a large number of artificial pollution tests for IEC standard suspension insulators were investigated under different relative humidity (RH) and salt deposit density (SDD). Then, based on the experimental data, the characteristics of LC were analyzed in both time-domain and frequency-domain. The results showed that the peak values of the LC (Ih ) has no relation with SDD when the RH is low ,but when the RH is high, the Ih increases with the increase of SDD; the phase difference (?) between LC and applied voltage decreases with the increase of RH, the LC becomes inductive when strong local arc occurs; the total harmonic distortion (THD) of LC increases slowly with the increase of RH when the pollution is light, but decreases firstly, then increases with the increase of RH when the pollution is middle or heavy level, the value reaches to minimum when the RH is about 80%. Consequently, Ih, ? and THD were proposed as characteristic parameters to evaluate the safety condition of polluted insulators. Finally, an ANN with fuzzy output was developed to evaluate the safety condition of polluted insulators. The input parameters of the ANN were Ih, ? and THD, while the output parameters were fuzzified into four fuzzy subsets, the capability of ANN was validated by 8 pairs of testing samples, and the effects of SDD and RH on the performance of insulators were discussed.

Evaluating the safety condition of porcelain insulators by the time and frequency characteristics of LC based on artificial pollution tests

A major field of neural networks (NN) application is function estimation, because the useful properties of NN such as adaptivity and nonlinearity are well suited to function estimation tasks where the equation describing the function is unknown. In this paper the prerequisite trainiing data are obtained from experimental studies performed on a flat plate model for a polluted insulator under power frequency voltage. Detailed studies have been carried to deterimine the NN parameters which give the best results. Studies have also been carried out to assess the effect of the presence of inadequate data in the training set on modeling accuracy. It is found that, when training is completed, NN is capable of estimating the function t = f(V,L,p) very efficiently and effectively even when the inadequate data are incorporateld in the training set.

S. Chlakravorti

Papers

Estimation of time-to-flashover characteristics of contaminated electrolytic surfaces using a neural netlwork