Detection of the initial region of the current transformer core saturation

Abstract

In the first part of the paper the detailed analysis of the existing current transformer (CT) saturation detection methods with their classification is presented. A new saturation detection method has been proposed in second part of the paper. Mathematical description of the method is given. In the next section of the paper a CT model with test scheme is presented. To identify the reliability of proposed method against nose and remanent flux density experiments have been produced in the fourth section. Comparative analysis between proposed and existing methods is also given. Finally, in conclusion a detailed description of the method is given regarding its behavior with respect to remanent flux and noise

Detection of Current Transformer Saturation Based on Machine Learning

“…Two methods have become the most common for solving the CT saturation problem: saturation detection with subsequent restoration of distorted current values [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23] and saturation detection with subsequent blocking of the PS and EA algorithms [24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41].…”

“…The methods related to the second approach have found widespread practical development due to their reliability and ease of implementation. They are based on the application of mathematical analysis methods [24][25][26][27][28][29][30][31][32], statistical signal processing [33][34][35][36][37][38], and the use of neural networks for classification problems [39][40][41].…”

“…The methods of statistical signal processing are based on the analysis of data extremes from one period of the measured current [33], correlation analysis of the CT secondary current with the calculated current [34], and the CT secondary currents installed on both sides of the power transformer [35]. This group also includes the analysis of the differential current form in the differential protection circuit of the power transformer [36], and the dispersion analysis of the measured current data obtained from one cycle [37] as well as the calculated values of the density [38]. Robustness to noise and the presence of harmonic components are among the main advantages of the methods in this group.…”

“…Approach Advantages/Disadvantages [24][25][26][27][28][29][30][31][32] mathematical analysis (+) low computational costs, time efficiency (−) sensitivity to noise and harmonics, disable to detect light CT saturation [33][34][35][36][37][38] statistical signal processing (+) robustness to noise and harmonic components (−) delay in processing the measurement signal [39][40][41] machine learning (+) high flexibility and range of coverage of complex current shapes under fault conditions (−) severe delay due to scaling methods, computational effort All these methods, as it was noted, have their advantages and disadvantages, but they can also complement each other. For example, with deep saturation, methods based on the current derivative are not able to identify CT saturation, because in the measured current, a sharp cut occurs too early-in the region near zero current values.…”