MPE Mathematical Problems in Engineering 1563-5147 1024-123X Hindawi Publishing Corporation 10.1155/2015/195170 195170 Research Article Ultra-High-Speed Travelling Wave Protection of Transmission Line Using Polarity Comparison Principle Based on Empirical Mode Decomposition Wang Dong 1, 2 Gao Houlei 1, 2 Luo Sibei 1, 2 Zou Guibin 1, 2 Dabnichki Peter 1 Key Laboratory of Power System Intelligent Dispatch and Control of Ministry of Education Shandong University Jinan 250061 China sdu.edu.cn 2 School of Electrical Engineering Shandong University Jinan 250061 China sdu.edu.cn 2015 11102015 2015 30 07 2015 08 09 2015 09 09 2015 11102015 2015 Copyright © 2015 Dong Wang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The traditional polarity comparison based travelling wave protection, using the initial wave information, is affected by initial fault angle, bus structure, and external fault. And the relationship between the magnitude and polarity of travelling wave is ignored. Because of the protection tripping and malfunction, the further application of this protection principle is affected. Therefore, this paper presents an ultra-high-speed travelling wave protection using integral based polarity comparison principle. After empirical mode decomposition of the original travelling wave, the first-order intrinsic mode function is used as protection object. Based on the relationship between the magnitude and polarity of travelling wave, this paper demonstrates the feasibility of using travelling wave magnitude which contains polar information as direction criterion. And the paper integrates the direction criterion in a period after fault to avoid wave head detection failure. Through PSCAD simulation with the typical 500 kV transmission system, the reliability and sensitivity of travelling wave protection were verified under different factors’ affection.

1. Introduction

According to the protection principle, travelling wave protection methods include travelling wave differential protection, travelling wave distance protection, travelling wave amplitude comparison protection, and travelling wave polarity comparison protection [1, 2].

Travelling wave differential protection principle is simple and clear. But travelling wave has attenuation characteristic. There may be large unbalance current in the transmission line to cause wrong operation. And it is also affected by the bus structure . Travelling wave distance protection cannot protect the whole line and does not have direction discrimination ability. Same with differential protection, it is affected by bus structure, too . Travelling wave amplitude comparison protection principle has improved much compared to other protection principles. But it is affected by the bus structures, fault inception angles, and different thresholds, too. The traditional travelling wave polarity comparison protection principle has a lot of advantages: high operation speed, clear direction discrimination, and simple protection principle. But it is also affected by some things: fault initial angles, different bus structures, different fault locations, and even threshold. If those disadvantages can be overcome, a new travelling wave protection principle can be got .

The traditional travelling wave protection principle’s application is limited by the transformer technology. The traditional current transformer (CT) and voltage transformer (VT) cannot transfer the travelling wave signal correctly. Currently, Rogowski based electronic current transformer (R-ECT) and capacitive divider electronic voltage transformer (C-EVT) have been able to transfer current travelling wave and voltage travelling wave accurately. And the output of C-EVT and R-ECT is the differential signal of the input. By integration circuit, the original signal can be regained exactly. So there is no transformer technology limit in the travelling wave protection principle anymore .

This paper compares the polarity relationship between voltage travelling wave and current travelling wave with different fault directions. Combined with empirical mode decomposition algorithm (EMD), the new travelling wave polarity comparison protection principle based on the integration of amplitude is derived. It not only uses the initial travelling wave but also uses the travelling wave after fault happens. So it is a reliable protection principle with obvious direction discrimination. By the way, this new protection principle is not affected by different initial angles, different grounding resistance, different bus structures, and different fault locations. To verify the characteristics of the new principle, a simulation based on PSCAD/EMTDC is carried on. And the PSCAD simulation proved that this new protection principle has the characteristics mentioned above indeed.

The high operation speed is a very important advantage for travelling wave protection. Considering the different parts of the new travelling wave’s operation time, the new protection principle can determine if the fault is internal or external in 5 ms. Then it can send the signal to breaker to operate. So, it can be called an ultra-high-speed travelling wave protection.

2. Traditional Travelling Wave Polarity Comparison Protection

The direction element of the protection principle is a polarity comparison relay. It detects the initial voltage travelling wave and current travelling wave as comparison objects. When the voltage travelling wave and current travelling wave have opposite polarities of both sides, the internal fault can be determined. When the voltage travelling wave and current travelling wave have same polarities of any side, the external fault can be determined. The schematic of protection is shown in Table 1. (Superimposed voltage in the table appears at the moment that fault happens. It has same value and opposite polarity with the voltage on the transmission line just before the fault moment. And it is the voltage source in the circuit of the table indeed.)

Analysis of travelling wave polarity comparison protection.

Fault location Superimposed voltage polarity M side N side Schematic of fault superimposed state circuit
Voltage Current Voltage Current
Internal fault +  +  − + +  − +

External fault +  +  − + +  +

External fault +  +  +  +  − +
3. Empirical Mode Decomposition

The empirical mode decomposition algorithm can distinguish the different scale fluctuations or trends in the signal gradually. And the result of EMD is a series of different characteristic scales data called intrinsic mode functions (IMF) .

The result of EMD can be described as(1)St=i=1NIMFit+Rt,where S(t) is the EMD result, IMFi(t) is i order intrinsic mode function, and R(t) is trends signal.

Intrinsic mode function is a single component signal and it must meet the following two conditions: (1) difference between the number of extreme points and zero crossing points is not more than one over the entire length of the signal; (2) the envelope of IMF is symmetry about time axis.

The processes of EMD are described as the following steps.

(1) Find all the maxima of the original signal S(t). Then the maxima envelope E+(t) can be calculated based on cubic spline interpolation. Similarly, the minimum envelope E-(t) can also be calculated. Then, the average envelope M(t) can be defined as(2)Mt=E+t+E-t2.

(2) Let S(t) be minus M(t) to get a new signal H11(t):(3)H11t=St-Mt.

Then check if the following condition can be met:(4)H1kt-H1k-1t2H1k-1t2ε,where k is the cycle number and the value of ε is between 0.2 and 0.3. This paper selects 0.3.

If (4) can be met after k cycles, then IMF1(t) can be defined as(5)IMF1t=H1kt.

(3) Let original signal be minus IMF1(t) to get a residual signal R(t) as(6)Rt=St-IMF1t.

Repeat steps (1) to (3) to get another intrinsic mode function IMF2(t). Repeat steps (1) to (3) until residual signal R(t) is small enough or monotonic function.

Based on EMD, the first IMF of voltage travelling wave and current travelling wave can be calculated as Figure 1. The first figure and second figure are the voltage and current travelling wave before EMD, respectively. The third figure and fourth figure are the voltage and current travelling wave after EMD, respectively. As we can see, the similarity of the voltage and current travelling wave using EMD is more obvious than before. So it is more convenient to construct a protection principle using intrinsic mode function.

The comparison of travelling wave and travelling wave after EMD.

4. New Travelling Wave Polarity Comparison Protection 4.1. Derivation of Direction Criterion

As shown in Figure 2, fault component voltage appears between the fault location F in transmission line and the earth at time t=0. Then, the transmission line will be charged. After a short time Δt, a short length of transmission line Δx is charged to ΔQ=Cu0Δx (C is the capacitance value per unit length of transmission line). An electrical field E will appear surrounding this short transmission line. And the flow of current will form a magnetic field around the line. If Δx is small enough, the current i0 can be described as(7)i0=limΔx0ΔQΔt=limΔx0Cu0ΔxΔt=Cu0v,where v is the speed of the wave and C is the capacitance value per unit length of transmission line.

Equivalent circuit of single phase line.

Now the magnetic flux around Δx is ΔΦ=LiΔx. According to the law of electromagnetic induction, the electromotive force is described as(8)E=limΔx0ΔΦΔt=limΔx0LCu0vΔxΔt=LCu0v2,where v is the speed of the wave, C is the capacitance value per unit length of transmission line, and L is the inductance value per unit length of transmission line.

Because the voltage on capacitance cannot change suddenly and Δx is small enough, E equals voltage u0. Then the wave speed will be(9)v=1LC.

Take (9) to (7) to get(10)u0i0=LC.

As we can see, the ratio of voltage and current is a constant value called wave impedance.

Define voltage amplitude conditioning factor k:(11)k=iu.

And the value of k is approximately equal toC/L.

Considering the initial polarity of voltage and current travelling wave, define a factor λ to identify the fault direction:(12)λ=t=tFtF+dkutitdtt=tFtF+di2tdt,where tF is the arrival point of the travelling wave, d is the length of the integration time, and k is the voltage amplitude conditioning factor defined above.

The discretization of (12) can be described as(13)λ=t=tFtF+dkutitt=tFtF+di2t.

Considering different fault directions, the value of λ can be calculated.

(1) If the fault happens as Figure 2 shows, it will be forward fault type for R1. Assume the transmission line is lossless. When the travelling wave arrived, the travelling wave signal of R1 will be(14)u=u++u-=1+kufu-,i=i++i-=1+kifi-,where u is the voltage travelling wave, i is the current travelling wave, u+ is the forward voltage travelling wave, u- is the reverse voltage travelling wave, i+ is the forward current travelling wave, i- is the reverse current travelling wave, kuf is the voltage reflection coefficient at the bus, and kif is the current reflection coefficient at the bus. And the inequality relationship will be 0kuf,kif1.

Now the amplitude conditioning factor can be calculated:(15)k=itut=1+kif1+kufi-tu-t.

Because it is a forward direction fault for R1, the reverse voltage and current travelling wave have different polarities. So (15) can be simplified as(16)k=-1+kif1+kufi-tu-t.

Take (14) and (16) to (13):(17)λ=t=tFtF+dkutitt=tFtF+di2t=-1.

As we can see, the value of λ is a constant number −1 when it is a forward direction fault. And it is not affected by the reflection coefficient and the construction of the bus.

(2) If the fault happens as Figure 2 shows, it will be reverse fault type for R2. Assume the transmission line is lossless. When the travelling wave arrived, the travelling wave signal of R2 will be(18)u=u+=kuzu-,i=i+=kizi-,where u is the voltage travelling wave, i is the current travelling wave, u+ is the forward voltage travelling wave at R2, i+ is the forward current travelling wave at R2, u- is the reverse voltage travelling wave at R1, i- is the reverse current travelling wave at R1, kuz is the voltage refractive coefficient at bus, and kiz is the current refractive coefficient at bus. And there is an inequality relationship kuz,kiz0.

Now the amplitude conditioning factor can be calculated:(19)k=itut=kizi-tkuzu-t.

Because it is a reverse direction fault for R2 and also a forward direction fault for R1, the reverse voltage and current travelling wave have different polarities. And the forward direction of R2 is opposite to R1. So (19) can be simplified as(20)k=itut=kizkuzi-tu-t.

Take (18) and (20) to (13):(21)λ=t=tFtF+dkutitt=tFtF+di2t=1.

As we can see, the value of λ is a constant number 1 when it is a reverse direction fault. And it is not affected by the reflection coefficient and the construction of the bus.

Taking a variety of errors in the actual system into account, the fault direction discrimination schematic is shown in Figure 3. When forward fault happens, the value of λ is less than zero. When reverse fault happens, the value of λ is greater than zero. If two fault direction discrimination results of both ends are forward fault, an internal fault can be determined. If one of the fault direction discrimination results of both ends is reverse fault, an external fault can be determined.

Fault direction discrimination schematic.

4.2. Protection Scheme

The protection scheme is shown in Figure 4. First of all, the three-phase voltage and current should be decoupled using Clark transformation. Then amplitude conditioning factor, defined above, can be calculated point by point in d length of time. After that, the value of λ can be calculated. Because the other end of the transmission line needs the value of λ to identify the fault section, the value of λ should send to another end though fiber path. Then the value of factor λ can be checked to identify the external fault type. If it is a forward fault for the relay, the value of factor λ from another end will be received and checked to identify the external fault. If λ and λ are both less than zero, an internal fault can be determined. At last, the breaker will clear the transmission line fault.

Flow chart of travelling wave protection.

5. Simulation Analysis 5.1. Simulation Model in PSCAD

The 500 kV power transmission system is constructed in PSCAD/EMTDC as shown in Figure 5. The system includes three transmission lines whose lengths are 100 km, 200 km, and 100 km, respectively. R1 and R2 are two relays on the middle line. Now the new travelling wave polarity comparison protection can be studied by different fault locations and different fault types.

Model of 500 kV power transmission system.

The transmission line uses frequency-dependent model and has uniform transposition. The transmission line parameters for per km length are shown in Table 2. The bus stray capacitance to ground is set to CS = 0.01 μF. Taking the past studies into account, the sampling rate is set to 1 MHz. The integration time d is 0.1 ms.

Transmission line parameters.

R X G B
(Ω/km) (Ω/km) (S/km) (S/km)
Positive sequence 0.01798 0.29278 1 × 108 3.93905 × 10−6
Negative sequence 0.01798 0.29278 1 × 108 3.93905 × 10−6
Zero sequences 0.28662 1.08210 1 × 108 2.43767 × 10−6
5.2. Typical Fault Examples

In order to verify the protection principle’s operating characteristics, A phase to ground fault is set located at F3. The initial fault angle is 45 and the ground resistance is 50Ω. Using empirical mode decomposition algorithm, the first-order intrinsic mode function of voltage and current travelling wave of both sides can be calculated.

Take ku and i data in Figure 6 to (13) to calculate the λ=-0.8710. Then the forward fault of R1 can be determined.

Comparison chart of ku and i of M side.

Take ku and i data in Figure 7 to (13) to calculate the λ=0.9811. Then the reverse fault of R2 can be determined.

Comparison chart of ku and i of N side.

As we can see, the fault discrimination results of R1 and R2 are corrected. Taking the protection scheme of Figure 4 into account, an external fault type can be determined.

5.3. Relater Factors 5.3.1. Different Fault Location

Based on some different fault locations at F1 and F2, the fault discrimination factor λ of both M and N side is calculated.

Table 3 is the simulation results for different fault locations. The fault distance in the table is from the bus of M side to fault location. As can be seen, the fault principle based on EMD can identify fault direction correctly. Even at the beginning or end of the transmission line, it can still identify fault direction correctly.

Simulation results for different fault locations.

Fault location Fault distance/km M side N side Results
λ Direction λ Direction
F1 10 −0.8160 Forward −0.5732 Forward Internal
100 −0.8763 Forward −0.4071 Forward Internal
190 −0.9378 Forward −0.7965 Forward Internal

F2 10 0.9972 Reverse −0.9603 Forward External
50 0.9980 Reverse −0.3843 Forward External
90 0.9972 Reverse −0.9603 Forward External
5.3.2. Grounding Resistance

Based on some different grounding resistance at F1 (100 km away from the bus of M side) and F2 (10 km away from the bus of M side), the fault discrimination factor λ of both M and N side is calculated.

Table 4 is the simulation results for different grounding resistance. As can be seen, the fault principle based on EMD can identify fault direction correctly. With the increasing of grounding resistance, protection’s sensitivity will not change.

Simulation results for different grounding resistance.

Fault location Grounding resistance/Ω M side N side Results
λ Direction λ Direction
F1 1 −0.4944 Forward −0.8818 Forward Internal
100 −0.4829 Forward −0.8802 Forward Internal
300 −0.4984 Forward −0.8839 Forward Internal

F2 1 0.9978 Reverse −0.9438 Forward External
100 0.9963 Reverse −0.8730 Forward External
300 0.8522 Reverse −0.9033 Forward External
5.3.3. Fault Initial Angle

Based on some different fault initial angel at F1 (100 km away from the bus of M side) and F2 (10 km away from the bus of M side), the fault discrimination factor λ of both M and N side is calculated.

Table 5 is the simulation results for different fault initial angles. As can be seen, the fault principle based on EMD can identify fault direction correctly. Even with small fault angles, it can still identify fault direction correctly. And, with the decreasing of fault initial angle, protection’s sensitivity will reduce slowly.

Simulation results for different fault angles.

Fault location Initial angle/° M side N side Results
λ Direction λ Direction
F1 1 −0.3884 Forward −0.3256 Forward Internal
45 −0.4291 Forward −0.8732 Forward Internal
90 −0.4944 Forward −0.8818 Forward Internal

F2 1 0.9972 Reverse −0.2356 Forward External
45 0.9997 Reverse −0.9687 Forward External
90 0.9997 Reverse −0.9700 Forward External
5.3.4. Different Fault Types

Based on some different fault types at F1 (100 km away from the bus of M side) and F2 (10 km away from the bus of M side), the fault discrimination factor λ of both M and N side is calculated.

Table 6 is the simulation results for different fault types. As can be seen, the fault principle based on EMD can identify fault direction correctly.

Simulation results for different fault types.

Fault location Fault type M side N side Results
λ Direction λ Direction
F1 AG −0.4944 Forward −0.8818 Forward Internal
AC −0.9600 Forward −0.9602 Forward Internal
ABG −0.4471 Forward −0.8804 Forward Internal
ABCG −0.9664 Forward −0.7452 Forward Internal

F2 AG 0.9972 Reverse −0.9700 Forward External
AC 0.9991 Reverse −0.6022 Forward External
ABG 0.8275 Reverse −0.4659 Forward External
ABCG 0.9997 Reverse −0.9586 Forward External
5.3.5. Sampling Rate

Based on some different sampling rate and AG fault at F1 (100 km away from the bus of M side) and F2 (10 km away from the bus of M side), the fault discrimination factor λ of both M and N side is calculated.

Table 7 is the simulation results for different sampling rate. As can be seen, the fault principle based on EMD can identify fault direction correctly with the change of sampling rate.

Simulation results for different sampling rate.

Fault location Sampling rate/Hz M side N side Results
λ Direction λ Direction
F1 100 k −0.9794 Forward −0.6816 Forward Internal
500 k −0.7345 Forward −0.9637 Forward Internal
1 M −0.5018 Forward −0.8825 Forward Internal

F2 100 k 0.9751 Reverse −0.9807 Forward External
500 k 0.9994 Reverse −0.9055 Forward External
1 M 0.9979 Reverse −0.9684 Forward External
5.3.6. Bus Structure

Traditional travelling wave protection principle is affected by the number of transmission lines connected to the bus. To verify the new EMD based protection principle, a new power transmission system is constructed in PSCAD as Figure 8.

Model of 500 kV power transmission system with different bus structures.

Table 8 is the simulation results for different fault locations. As can be seen, the fault principle based on EMD can identify fault direction correctly with different bus structure.

Simulation results for different fault locations.

Fault location M side N side Results
λ Direction λ Direction
F1 −0.9731 Forward −0.9603 Forward Internal
F2 0.9980 Reverse −0.9779 Forward External
F3 −0.9718 Forward 0.9992 Reverse External
6. Operation Time of Protection

The operation time of the travelling wave protection using polarity comparison principle based on EMD includes three parts: algorithm time, detection time, and propagation time. This new travelling wave protection principle can determine if the fault is inside or outside of the protection region in 5 ms. Then it can send the signal to breaker to operate. So it can be called ultra-high-speed travelling wave protection. The following is the introduction and analysis of the three parts.

6.1. Algorithm Time

Algorithm time includes two parts: the integration time and calculation time of the principle. In this paper, integration time length (the factor d in (12) and (13)) is 0.1 ms. Considering the computing power of the protection unit now, the calculation time of algorithm is not longer than 0.5 ms. So, the algorithm time is not longer than 1 ms.

6.2. Detection Time

Detection time is the time difference of two sides’ travelling wave arrival point. As we can see, the fault may happen everywhere in the transmission line. Then the arrival times of two sides are different, except that the fault happens in the middle of the line. As a protection principle which needs two sides’ information to decide the operation of breaker, the time difference will delay the operation time. As shown in Figure 9, tf, tm, and tn are the fault time, M side’s arrival time, and N side’s arrival time, respectively. And the time difference can be described as(22)Δt=tm-tn=L1-L2v.

Schematic diagram of detection time.

And v is the travelling wave speed.

Because the transmission line is generally several hundred kilometers, this time is obviously not longer than 2 ms.

6.3. Propagation Time

After the direction discrimination of one side, as shown in Figure 10, the value of λ should transfer to another side. Propagation time is the time from one side to another side. As the length of transmission line is generally several hundred kilometers, propagation time is no longer than 2 ms.

Schematic diagram of propagation time.

7. Conclusion

Comparing with the traditional polarity comparison travelling wave protection, the new travelling wave protection combines the relationship between amplitude and polarity. Based on empirical mode decomposition, the derivation of the direction criterion is finished. And this protection criterion not only uses the initial travelling wave front but also uses short time’s (0.1 ms in the paper) travelling wave information after the initial travelling wave front. Through the integration of travelling wave, it can avoid the failure of the travelling wave’s detection. So it can increase the reliability of the protection principle.

To verify the new protection principle, a simulation based on PSCAD is carried on. Taking the simulation results into account, this new protection principle is not affected by different fault locations, different fault types, different initial angels, different grounding resistance, and different bus structures. So it is a reliable travelling wave protection.

Operation speed is an important advantage for travelling wave protection. Because the new protection principle can send the operation signal to breaker in 5 ms, it can be called ultra-high-speed travelling wave protection.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the republication of this paper.

Acknowledgment

This work was supported in part by the National Natural Science Foundation of China (51177094, 51277114).

Zou G. Gao H. A traveling-wave-based amplitude integral busbar protection technique IEEE Transactions on Power Delivery 2012 27 2 602 609 10.1109/tpwrd.2012.2184561 2-s2.0-84859745311 Chen W. Malik O. P. Yin X. Chen D. Zhang Z. Study of wavelet-based ultra high speed directional transmission line protection IEEE Transactions on Power Delivery 2003 18 4 1134 1139 10.1109/tpwrd.2003.817511 2-s2.0-0141986811 Zou G. Song S. Xu C. Liu D. Gao H. Fast busbar protection based on waveform integral of directional traveling waves Proceedings of the Chinese Society of Electrical Engineering 2014 34 31 5677 5684 10.13334/j.0258-8013.pcsee.2014.31.026 2-s2.0-84912019712 Li Z. Hua H. Traveling wave protection based on wide area travelling wave information Proceedings of the CSEE 2014 34 6238 6245 Duan J.-D. Zhang B.-H. Luo S.-B. Zhou Y. Transient-based ultra-high-speed directional protection using wavelet transforms for EHV transmission lines Proceedings of the IEEE/PES Transmission and Distribution Conference and Exhibition August 2005 Dalian, China 10.1109/tdc.2005.1547120 2-s2.0-33751378139 Wu Q. H. Zhang J. F. Zhang D. J. Ultra-high-speed directional protection of transmission lines using mathematical morphology IEEE Transactions on Power Delivery 2003 18 4 1127 1133 10.1109/tpwrd.2003.817513 2-s2.0-0141918387 Zou G. Study on internal based travelling wave directional protection for transmission line [Ph.D. thesis] 2009 Jinan, China Shandong University Öhrström M. Söder L. Fast protection of strong power systems with fault current limiters and PLL-aided fault detection IEEE Transactions on Power Delivery 2011 26 3 1538 1544 10.1109/tpwrd.2011.2107530 2-s2.0-79959707076 Zou G. B. Gao H. L. Extra high speed hybrid protection scheme for high voltage transmission line International Journal of Electrical Power & Energy Systems 2014 63 83 90 10.1016/j.ijepes.2014.05.061 Ha H. Yu Y. Novel scheme of travelling wave based differential protection for bipolar HVDC transmission lines Proceedings of the International Conference on Power System Technology (POWERCON '10) October 2010 Hangzhou, China IEEE 1 6 10.1109/POWERCON.2010.5666376 Duan J. D. Zhang B. H. Study of starting algorithm using travelling waves Proceedings of the Chinese Society for Electrical Engineering 2004 24 9 30 36 Zou G. B. Gao H. L. Algorithm for ultra high speed travelling wave protection with accurate fault location Proceedings of the IEEE Power and Energy Society General Meeting July 2008 Pittsburgh, Pa, USA 20 24 Marracci M. Tellini B. Zappacosta C. Robles G. Critical parameters for mutual inductance between rogowski coil and primary conductor IEEE Transactions on Instrumentation and Measurement 2011 60 2 625 632 10.1109/TIM.2010.2051591 2-s2.0-78651281606 Li W. B. Mao C. X. Lu J. M. Study of Rogowski coils for measuring pulse currents of the high-power laser source International Journal of Power and Energy Systems 2005 2 96 101 Liu Y. Lin F. C. Zhang Q. Zhong H. Design and construction of a Rogowski Coil for measuring wide pulsed current IEEE Sensors Journal 2011 11 1 123 130 10.1109/jsen.2010.2052034 2-s2.0-78149243810 Gao H. Yang R. Analysis and test for electronic voltage transducer's transfer characteristics Journal of Beijing Jiaotong University 2014 38 5 114 118 10.11860/j.issn.1673-0291.2014.05.020 2-s2.0-84921685333 Huang N. Wu M. A confidence limit for the empirical mode decomposition and Hilbert spectral analysis Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 2003 459 2037 2317 2345 10.1098/rspa.2003.1123 Huang N. E. Shen Z. Long S. R. Wu M. C. Shih H. H. Zheng Q. Yen N.-C. Tung C. C. Liu H. H. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis The Royal Society of London. Proceedings—Series A: Mathematical, Physical and Engineering Sciences 1998 454 1971 903 995 10.1098/rspa.1998.0193 MR1631591 Wu Z. Huang N. E. A study of the characteristics of white noise using the empirical mode decomposition method Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 2004 460 2046 1597 1611 10.1098/rspa.2003.1221 2-s2.0-2542525254