Journal of Power and Energy Engineering, 2014, 2, 432437 Published Online April 2014 in SciRes. http://www.scirp.org/journal/jpee http://dx.doi.org/10.4236/jpee.2014.24058 How to cite this paper: Du, F., et al. (2014) A New Method of Early Short Circuit Detection. Journal of Power and Energy Engineering, 2, 432437. http://dx.doi.org/10.4236/jpee.2014.24058 A New Method of Early Short Circuit Detection Feng Du1, Weigang Chen1, Yue Zhuo1, Michael Anheuser2 1Siemens Ltd. China, Shanghai, China 2Siemens AG, Amberg, Germany Email: feng .du @si emens. co m Received January 2014 Abstract To reduce the pressure on contacts and circuit breaker and realize the zone selective interlocking (ZSI) function above the instantaneous protection threshold (e.g., >10In), the short circuit current needs to be early detected. The stateof–art of early short circuit detection (ESCD) method is reviewed. Based on the equivalent model of the short circuit, a new method based on the current and its integration is proposed. The prospective current value can be detected in the early stage of the short circuit. According to the evaluation result, the short circuit current can be early forecasted with the proposed method. Keywords Early Short Circuit Detection; Zone Selective Interlocking; Instantaneous Protection 1. Introduction As known, low voltage protection device, e.g., circuit breaker, are designed to handle with the problems of the short circuit. Electrical systems (e.g. conductor lines, cables, bus bar systems, contacts) and loads (e.g. machines) are stressed electrodynamically and thermally by shortcircuit currents. The amount of stress is affected primar ily by the amplitude of the shortcircuit current and the time from shortcircuit occurrence until switch off. In some cases, the short circuit current with the range of 10 to 150 kA (at 440 V) can be expected in lowvoltage networks. What’s more, realization of the zone selectivity interlocking (ZSI) function above the instantaneous protection threshold (e.g., >10In) is required. Thus, it is vitally important to isolate the fault as soon as possible to minimize downtime and damage. Naturally, the concept of early short circuit detection which detects the short circuit current in its early stage can greatly facilitate the protection action of the low voltage protection de vice to start the current limiting as soon as possible. Therefore faster and reliable detection algorithms are needed to realize fault detection. At short circuit (SC) current, the prospective current value or peak value will be bigger than normal case. The early detection of the peak value is proposed as the evaluation criterion for forecasting short circuit current in this paper. In the second section, the working principle is analysed. In the third section, the evaluation results are presented. Finally, the conclusion is summarized in the final part.
F. Du et al. 2. Working Princ iple 2.1. StatofArt of Early Short Circuit Detection To realize the early short circuit detection, the stateofart can be classified into three types. The first kind of ESCD is called as locus curves criteria. The locus curve of (i, i’) is proposed in [1]. However, the initial current should be zero which is not always same as the practical condition. The improved locus methods of (i, i’) are also motioned in [2] [3]. However, there is no concrete definition of the locus curve. The method of the extrapo lated LocusCurves based on (i,i’) is proposed in [4], which has the clear definition of locus curve. However, it has the high dependence on the practical power factor. And the Cubical criterion based on (i, i’, i’’) is proposed to improve the time performance in [5] [6]. However, the relatively high hardware complexity is necessary. The second kind of method is the regression method, which aims to calculate the prospective peak current value be fore it reaches. The Estimation of the current peak based on the sequence of (i,i’) is mentioned in [7]. And the eestimation of the current peak based on the sequence of i is proposed in [8], which is only suitable for the sinu soidal wave at given frequency. The other methods include the methods based on the Complex Impedances or Powers [9], or ShortTime Admittance Spectrum Analysis [10], or Neural Networks [11], or the Locu sCurves in Combination with MorphologyWaveletFiltering [12]. They have a strong dependence on the practical net works. What’s more, all the above methods involving the operation of i’or i’’ are often sensitive to the high fre quency noise and harmonic i nterference. In this paper, the regression method based on the sequence of (i, ∫idt) is proposed to solve the mentioned drawbacks. 2.2. Regression Method Based on the (i, ∫idt) The typical electric network can be equivalent to a circuit as in Figure 1, which comprises at least one voltage source connected to at least one resistive load R and one inductive load L, where Vm is maximum voltage of the power source, w is the frequency of the power source in rad, φ is the switchingon angle, cosθ is the power fac tor which is determined by the inductance L and resistance R in the model. The Formulas (1) and (2) below can be got according to the model. (1) (2) (3) (4) Figure 1. Equivalent circuit.
F. Du et al. where Ipeak is called as the short circuit current peak value. With the integration operation on the Formula (1), the Formula (5) can be got, ( ) 00 cos()() ()(0) tt m VwtdtRi tdtLiti θ +=⋅+⋅ − ∫∫ (5) And Formula (2)(5) can be converted to the Formula (6) by dividing the whole impedance, sin( )cos(cos( )1)sin ( )(0)/(()(0))cos cos peak peak II wt wt ItItgwitiww θθ ϕϕϕ − − =−⋅−++ (6) Further, the Formula (7) can be got by defining the For mu l as (8)(11). cos sin ()()()() cos cos peak peak II Rttg AtBtCt θθ ϕϕϕ ⋅⋅ =−⋅+⋅+ ⋅ (7) (8) (9) (10) (11) With the definition of the (12), (13), (14), the Formula (7) can be rewritten as Formula (15). (12 ) (13) (14) ()()()() RtAtP BtQCt γ =−⋅+⋅+⋅ (15) Or in the matrix formation, [ ] ()[,,][(),(),()]'RtPQAtBt Ct γ = ⋅ (16) It can be found that the (12)(14) can be solved as parameter estimation method in mathematics. For example, with three points of (i, ∫idt), [ ] (1), (2), (3) (1),(2),(3)[,,](1),(2),(3) (1), (2), (3) AA A RR RPQBB B CC C γ =⋅ To get a more reliable parameter estimation, the regression method is suggested. And the Formula (17) can be calculated to solve the peak value of the short circuit current. (17) Further, the power factor can also be calculated according to Formula (12) when (15) or (16) is solved. 2.3. Realization According to the above analysis, the proposed algorithm is realized in the following steps, which can be shown in the Figure 2.
F. Du et al. a measuring the values of the instantaneous current flowing in the load and calculating the integral of said current; b performing N successive samplings of the values of the instantaneous current; c estimating the short circuit current peak value in the sampling period according to the sampled current values and the integral values; and, d generating a shortcircuit detection signal when the values of a peak current exceed an assigned threshold level. 3. Evaluation In order to validate the proposed method, both the simulated short circuit current and the practical tested short circuit current are taken as the input of the algorithm. 3.1. Evaluation Based on Simulated Short Circuit Current As shown in the Fig ure 3, there are 13 simulated short circuit current (prospective current is 52 kA) with dif ferent switchingon angles from 0˚ to 180˚ with a step of 15˚. The ESCD only makes sense if it can forecast the short circuit current in the initial stage. Thus, the closer to 52 kA in the initial stage, the better performance of the prediction algorithm is. The average of first four calcula tion values is taken as the forecast result. With different regression points N = 5, N = 12, N = 16, the detection time is 80 μs, 150 μs, 190 μs respectively as shown in Figur e 4. The statistical results according to Forecast Error (FE, is defined in Formula (18)) is shown in Table 1. Figure 2. Realization structure. Figure 3. 13 Simulated short circuit current. letthrough current vs. time @ 52kA 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 00.0005 0.001 0.0015 0.0020.0025 0.003 0.0035 0.004 φ=0 φ=15 φ=30 φ=45 φ=60 φ=75 φ=90 φ=105 φ=120 φ=135 φ=150 φ=165 φ=180
F. Du et al. 52 FE(Forecast error)100% 52 forecast_prospective I kA kA = ⋅ (18) 3.2. Evaluation Based on Practical Short Circuit Current There are 11 practical short circuit current data with the prospective values of 153 KA (3 currents with a sampling frequenc of 50 kHz), 105 KA(1 current with a sampling frequenc of 50 kHz ), 52 KA (3 currents with a sampling frequenc of 50 kHz), 24.9 KA (3 currents with a sampling frequenc of 50 kHz) and 52 KA (1 currents with a sampling frequenc of 200 kHz) which have been the input of the proposed algorithm. The Figure 5 of the one current with 153 KA and Figure 6 of the 52 KA show the ealuation results as examples. The green line is the forecasted prospective value and the blue line is the orignal current. Per to the results, all the fault currents can be early detected within 200 us. It can be found that the faulted prosepective value can be forecasted in the early stage of the short circuit, which is very helpful to realize the ZSI function, especially for high short circuit. Figure 4. Results of the detected prospective current values. Figure 5. Results of 153 KA with fs = 50 kHz.
F. Du et al. Table 1. Statistical forecast error with different regression points. Points FE ≤ 15% 15% < FE ≤ 26% FE > 26% N = 16 13 0 0 N = 12 10 3 0 N = 5 4 3 6 Figure 6. Results of 52 KA with fs = 200 kHz. 4. Conclusion In this paper, an early short circuit detection method based on regression method is proposed. And some evalua tion results based on simulated short circuit current and practical short circuit current are provided. The time performance of the early detection is suitable for the ZSI function in power distribution system. References [1] Patent DE 36 42 136 A1, 1986. [2] Patent EP838887B1, 1996. [3] Patent WO2009019096, 2009. [4] Patent US6313639, 19 95. [5] Mü tzel, T., Berger, F. and Anheuser, M. (2006) New Algorithm for Electronic ShortCircuit Detection. 52nd IEEE Holm Conference on Electrical Contacts, Montreal, 4247. [6] Du, F., Che n, W.G. and Anheuser, M. (2012) Polyhedron Method for Early Short Circuit Detection. IEEE Power & Energy Society (APPEEC), Shanghai. [7] Patent US6437576, 1996. [8] Mü tzel, T., Berger, F. and Anheuser, M. (2008) Methods of Early ShortCircuit Detection for LowVoltage Systems. The 54 th IEEE Holm Conference on Electrical Contacts, Orlando. [9] Patent US4811154, 1986. [10] Öhrström, M. (20 03 ) Fast Fault Detection for Power Distribution Systems. Thesis, Royal Institute of Technology, Stockholm. [11] Patent DE19954950, 1999. [12] Li, S. (20 02 ) Wavelet Transform Applications in Power Systems. Diss., University ErlangenNuremberg.
