استفاده از شاخص ریاضی برای اندازه گیری درجات خطای تغییر شکل سیم پیچ
Abstract
1- Introduction
2- Frequency response analysis
3- Measurement of transformer winding deformation based on frequency response method
4- Frequency response analysis based on mathematical statistical indicators
5- Conclusion and future work
References
Abstract
Since the power transformers are indispensable in power systems, their reliable operation are of significance. Winding deformation is one of the most common failures within transformers. At present, many methodologies have been proposed to detect these faults, and frequency response analysis (FRA) technique is most frequently used because of its low cost, convenience, simplicity and effectiveness. However, there is no standard and reliable code to interpret the mechanical deformations from FRA traces as so far. In this study, an actual transformer experimental platform was used to simulate variable winding deformation faults in order to standardize the interpretation of winding fault degree from FRA data. The measured FRA raw data are processed from the perspective of statistics, and some numerical indices are found to be more suitable for the diagnosis of transformer winding deformation degree.
Introduction
The stable, reliable, safe and high quality power supply concerns the lives of hundreds of millions of people. As one of the most important pieces of equipment in power system, the stable operation of power transformer plays a critical role in the safe operation of the power system [1,2]. Statistics indicates that 25 percent of transformer faults were caused by winding mechanical deformation. Since the low voltage impulse (LVI) method was first introduced by Lech and Tyminski from Poland in 1966, many methodologies have been successively proposed to detect transformer winding deformation, for instance, the short circuit impedance (SCI) method, frequency response analysis (FRA) method and vibration method. SCI is based on the comparison and analysis of two short circuit impedance tests. In FRA, the frequency response traces are compared with its fingerprint to diagnose the occurrence of internal faults within power transformers. The vibration diagnosis method utilizes the characteristics that the transformer tank vibration will change when the winding fails [3]. Recently, some scholars present other improved techniques, namely, the sweep frequency impedance (SFI) method, which is the combination of SCI and FRA [1], the online impulse frequency response analysis (online IFRA) [4], the electromagnetic wave method [5] and the Lissajous locus method [6]. Above all, FRA is an effective and economic diagnostic technique for detection of mechanical deformations inside a power transformer [7]. FRA is sensitive to failures in the windings and iron core. What's more, it can provide reliable information about the geometry of active parts of the power transformer without any needs to dismantle the unit. FRA uses input voltage as the input signal and another voltage measured at the other terminal of transformer winding as the output signal, besides, the measured current could also be used as output signal. In most cases, the output voltage is used. However, a standard and widely accepted code for interpretation of FRA traces can be improved; at present, experienced personnel are still needed to diagnose winding faults by using visual inspection [8], which is still a matter of great concern. At present, there are variable FRA interpretation methods, for instance, the equivalent electric model simulation method [9], the artificial intelligence method [10], and the mathematical statistics method [11]. In the first method, finite element analysis (FEA) of transformer winding is often used to build its equivalent electric model [12], to aid in interpreting the FRA curves. N. Hashemnia has studied the impact of winding axial displacement and radial deformation fault on equivalent electric parameters and FRA curves based on FEA [8,13], to improve the detection of winding faults. Besides, Z. W. Zhang has built a hybrid model based on FEA and multi-transmission line (MTL) to study the FRA curve beyond 1 MHz [14]. In the artificial intelligence method, M. Bigdeli has applied support vector machine (SVM) algorithm to classify the winding faults [15].