A fault location criterion for MTDC transmission lines using transient current characteristics

Electrical Power and Energy Systems 61 (2014) 647–655

Contents lists available at ScienceDirect

Electrical Power and Energy Systems journal homepage: www.elsevier.com/locate/ijepes

A fault location criterion for MTDC transmission lines using transient current characteristics Jingzhou Cheng a,⇑, Minyuan Guan b, Lv Tang c, Hongyang Huang a a

Department of Electrical Engineering, Zhejiang University, Hangzhou, Zhejiang 310027, PR China Huzhou Power Supply Company of State Grid Zhejiang Electric Power Company, Huzhou, Zhejiang 313000, PR China c Zhejiang Electrical Power Corporation, Hangzhou, Zhejiang 310007, PR China b

a r t i c l e

i n f o

Article history: Received 8 August 2013 Received in revised form 30 March 2014 Accepted 5 April 2014

Keywords: MTDC Transmission line DC filter Fault location Protection Complex wavelet algorithm

a b s t r a c t Multi-terminal DC (MTDC) transmission system is applied in modern power systems due to its technical features. In an MTDC system, the converter stations are generally connected in parallel. To smooth the DC current output from the converter, the smoothing reactors are integrated in the converter outputs, instead of in the DC transmission line. When a fault occurs on the DC transmission line of the MTDC system, the fault current will flow along the whole DC line. As a result, the protection scheme may not locate the fault accurately. Based on the characteristics of DC filter, this paper analyzes the impedance characteristics of the combinational circuits, and proposes a method to identify the fault location by the characteristic harmonic. A protection criterion of the non-integer harmonics comparison is composed with phase-frequency characteristics curve derived from complex wavelet transform. The inevitable blind zone is analyzed in theory which exists on the terminal of transmission lines just like distance protection in AC systems. The algorithm is successfully tested and compared with other techniques. Ó 2014 Published by Elsevier Ltd.

Introduction The electric power transmission of large capacity and long distance is an indispensable need of modern power system. Considering the separate distribution of the sending end power source and the stability of the receiving end power system, multiple distributed sending ends transmit power to multiple distributed receiving ends by Ultra High Voltage DC Transmission Line (UHVDC) becomes a feasible method. Currently, Multi-terminal DC (MTDC) has been applied in engineering, such as the NEA800 Project in India and the MTDC Project in Quebec-New England [1–3]. Meanwhile, the VSC-based MTDC has been researched and proposed for applications [4–7]. However, LCC-based MTDC system is more economical and reliable than VSC-based MTDC system which is still in development. Recently, VSC-based MTDC transmission line faults locating method and its protection are studied in [8,9]. Some faults locating method is involved in the DC fault current direction, but it cannot identify the fault location in the circumstance of AC/DC conductor contact fault or power reversal caused by the control system action after fault.

⇑ Corresponding author. Tel.: +86 0571 87952074. E-mail address: [email protected] (J. Cheng). http://dx.doi.org/10.1016/j.ijepes.2014.04.009 0142-0615/Ó 2014 Published by Elsevier Ltd.

HVDC transmission line fault location methods are presented in [10–12]. However, they are mainly for two-terminal transmission lines. The key factor is the characteristics of the boundary phenomenon. The smoothing reactor is usually applied to smooth the ripples of the DC current. Meantime, to prevent the shock wave generated by the DC line or the switch station from entering into valves, smoothing reactor is often positioned at the outlets of the converter in MTDC as shown in Fig. 1. (i.e., closer to the converter side in the T connection composed by the converter outgoing line and the DC transmission line). The traveling wave caused by the DC line fault will flow along the whole DC line since it cannot be suppressed by the smoothing reactor effectively, and the identification of which line the fault occurs fails. Therefore, the method based on the traveling wave boundary information cannot work in MTDC system. In addition, the method based on the distributed parameter line model in [11] and that based on the symmetrical component in [12] may be influenced by the noise of the DC transmission lines and the differences of the attenuation coefficients. The DC line protection of MTDC system based on Line Fault Location (LFL) has been applied in engineering already [1]. This protection compares the arrival times of the initial traveling waves on each terminal by the synchronizing clock function of the GPS system to determine the exact location of the fault on the DC line. In [13,14], the same method is also adopted. Since the fault

J. Cheng et al. / Electrical Power and Energy Systems 61 (2014) 647–655

DCF

DCF

LIB

DC smoothing reactor

DCF

DCF

DCF

LIA

Fault analysis of MTDC transmission line

D DCF

LRB

F3

C DCF

LRA

F2

B DCF

F1

A

DC line Rectifier-I

Rectifier-II

Inverter-I

Inverter-II

Fig. 1. A schematic of a bipolar MTDC system.

traveling wave transmits slightly slower than light at fixed speed, the difference of the arrival times of the traveling waves to every station is not significant when the stations are not far away from each other in distance, and the GPS time errors are not taken into account. Moreover, this scheme only utilizes the information in the limited time of the wave front, so the protection cannot work reliably if this information is not captured. In [15], the fault impedance used by fault locating scheme is unstable due to the variation in power system frequency. The protection methods based on wavelet transform is presented in [16,17]. Hilbert–Huang transform methods are also mentioned in [18]. There are advantages to detect the surge of the transient currents in these methods, but they could not be extended to detect the phase information. The fault location method contains two parts. The first part is the fault distance measurement which is presented in [19]. The second part is to detect the faulted transmission line. For MTDC control strategy, detecting the fault line section is more important. The fault line section selection is important to the MTDC control strategy. The characteristics of the DC filter are used in this paper for the fault locating of MTDC transmission lines. By analyzing the differences between the harmonics at the characteristic frequency and that at the other frequency when internal or external faults happen, a fault locating criterion based on complex wavelet transform is proposed. The inevitable one-end protection blind zones are deduced in the distributed parameter model. The algorithm is successfully tested and the results are compared with those from other algorithms. MTDC system structure and its control strategy after fault Take a four-terminal MTDC system as an example (as shown in Fig. 1). Four-terminal system is a paralleled extension of two-terminal system. The control strategy of CIGRE DC transmission test system is taken from [20]. In Fig. 1, traveling wave will flow along the whole line between Rectifier-I (point A) and Inverter-II (point D) when a fault happens on DC transmission line. The smoothing reactor in Rectifier-II (point B) and Inverter-I (point C) cannot be taken as the boundaries for the traveling wave after fault. The structure of NEA800 is similar to Fig. 1 and the control strategy is provided in [1]. When fault happens on line BC, the rectifiers will be retarded for about 205 ms for deionization of the fault followed by a restart attempt. If the fault on line BC remains, it has three attempts to restore operation [1]. When fault happens on line AB, only one restart attempt is proposed. If the first restart attempt fails, the REC-I will control the DC line current to zero, and will be disconnected from the MTDC system by high speed switch. Therefore, the fault control logics at different location on the DC transmission line are different. The correct control strategy must be chosen by identifying the fault location.

Fault zones of DC transmission line Assume a line-to-ground fault happens to the positive pole of the DC transmission line. Fig. 1 shows the line fault is divided into three fault conditions denoted by F1 (between AB), F2 (between BC) and F3 (between CD). According to the analysis above, the MTDC system cannot work when a fault occurs in F2. When faults happen in F1 or F3, the rest of the MTDC system can maintain the power transmission after the fault part is disconnected. Therefore, this paper mainly analyzes the difference between the faults F2 and F1 (or F2 and F3). DC protection configuration and DC filter characteristic Analysis Fig. 2 shows the smoothing reactor and the DC filter (DCF) which are installed in the output of the converter. The DC filter is composed of a set of three-tuned filter. The DC transmission line protection zone is denoted by the area c. The impedances of the smoothing reactor and the DCF are respectively expressed as

Z L ¼ jxLD

ð1Þ jxL2

Z DCF ¼

jxL3

1 jxC 2 jxC 3 þ jxL1 þ þ jxC 1 jxL2 þ jx1C 2 jxL3 þ jx1C3

ð2Þ

At the initial of fault happening, fault currents on DC line are varying in different frequencies due to the nonlinear effect of the fault arc and the circuit parameters. In AC system, fault analysis is involved in the symmetrical components of the steady-state circuit. Similarly, the steady-state circuit is also applicable to the DC fault analysis approximately. The fault current discharging path is shown in Fig. 3. The fault current can circulate not only through the DCF but also through the converter valves. The AC system commutation inductance LC and the smoothing inductance LD are in the same branch of the converter valves. The high-frequency current of this branch is small because of the characteristic of LD. Assuming the parameters of smoothing reactor, DCF and AC commutation inductance are the same in every converter station, the impedance of the whole converter station is:

Z eq ¼ ðjxLD þ jxLC Þ==Z DCF

ð3Þ

The amplitude-frequency characteristic of Zeq is shown in Fig. 4, which adapts some practical engineering parameters.

(c)

IdL1

IdL2

DC transmission line

(b) C1

LD IdH1

(a)

DCF

648

L1 C2

L2

C3

L3

IdN1

Fig. 2. The configuration of DC filter and the protections in converter station. (a) Converter protection. (b) DC bus bar protection. (c) DC transmission line protection.

649

J. Cheng et al. / Electrical Power and Energy Systems 61 (2014) 647–655

Fig. 5. Harmonic currents in fault conditions. (a) External fault current uninfluenced by DCF. (b) Internal fault current influenced by DCF.

Fig. 3. Discharging paths under line to ground fault.

"

U_ 1 I_1

#

" ¼ T1

U_ 2 I_2

#

 ¼

" _ # U2 D1 I_2

A1

B1

C1

ð5Þ

where A1 = D1 = chcl, B1 = ZCshcl, C1 = shcl/ZC. U_ 1 and I_1 denote the electric phasors at the initial end of the DC line, while U_ 2 and I_2 denote the electric phasors at the fault point of the DC line. The whole converter station (i.e. Zeq) is assumed as another twoport network and connected to the initial end of the DC line which is one port of the network in Eq. (5). We can get:

"

As seen in Fig. 4, the impedance–frequency characteristics of the converter station are very close to that of the DC filter. The turning frequencies are 600 Hz, 1200 Hz and 1800 Hz. In contrary to the impedance value at 600 Hz, the impedance value at 750 Hz is much larger. Therefore, the impedance values at 600 Hz and 750 Hz will be used to establish the protection criterion in this paper. In practical, DC filter cannot remove all the characteristic harmonics in transient periods. Document [21] presents a noniterative algorithm based on nodal equation formulation involving the arc model. In Fig. 5, the PACAD simulation results show the comparison of fault transient current with and without the DCF. If the fault happens near by the converter station, the comparison is more significant. From Fig. 5, the characteristic harmonics appear in spite of the DCF. However, the wave forms are obviously different with each other. Hence, the characteristic harmonics can be applied in the protection criterion. Fault characteristic analysis of DC line section DC line is considered as a uniform transmission line. Distributed parameter line can be written as:

U_ 1 ¼ U_ 2 chcl þ Z C I_2 shcl _ I_1 ¼ U2 shcl þ I_2 chcl

#

" ¼ T0

U_ 1 I_1

#

 ¼

" _ # U1 D0 I_1

A0

B0

C0

ð6Þ

According to the characteristics of two-port network and the impendence Zeq, A0 = 1, B0 = 0, C0 = 1/Zeq, D0 = 1. From Eqs. (5) and (6), we can get:

Fig. 4. Impedance characteristic curves of converter station.

(

U_ 0 I_0

ð4Þ

"

U_ 0 I_0

#

" ¼ T 0T1 cl

#

" ¼

chcl chcl Z eq

cl

Z C shcl

þ shZCcl cl

Z C shcl Z eq

þ chcl

#

U2 I_2

 ð7Þ

cl

where chcl ¼ e þe , shcl ¼ e e . 2 2 From Eq. (7), Zeq is zero at the DCF characteristic frequency. Hence, in the transfer matrix T0T1, we can get(chcl/Zeq + shcl/ ZC) ? 1, (ZC shcl/Zeq + chcl) ? 1. The matrix inversion becomes singular. Therefore, at characteristic frequency (i.e. 600 Hz), I_0 is zero thanks to the DC filter. In Eq. (5), A1 = D1. The long line is a symmetrical two-port network, and we can get A1D1 – B1C1 = ch2cl – sh2cl = 1. Hence, the two-port network of the transmission line (i.e. Eq. (5)) can be represented by a T network:

8 < Z T ¼ A1 1 ¼ Z C chcl1 shcl C1

ð8Þ

: Y T ¼ C 1 ¼ shcl ZC

The T network is shown in Fig. 6. DC line in Fig. 1 can be simplified as a T network. When DC line fault happens in F1 and F2, the fault equivalent network is shown in Fig. 7. The fault current measured by the CT, which is installed on the converter station of REC-II, is researched. When a fault happens in F1 as shown in Fig. 7(a), the two-port network parameters on the side of REC-I (at the left of the fault in Fig. 7(a)) and on the side of other converter stations (at the right of the fault in Fig. 7(a)) are:



ZC

From Eq. (4), this long line can be considered as a two-port network. The network can be expressed as:

U_ 2 I_2

T F1L ¼

AF1L

BF1L

C F1L

DF1L



 ¼

AT11

BT11

C T11

DT11



A0

B0

C0

D0

 ð9Þ

650

J. Cheng et al. / Electrical Power and Energy Systems 61 (2014) 647–655

T0

The network impedance from the CT to the right side is ZF1-(2–3) = AT2–3/CT2–3. Therefore, the current measured by the CT:

T1

I0

I1

ZT1

I2

ZT1

DCF

ICTF1 ¼

U0

U1

U2

YT1

DC converter station

DC transmission lines

K1 ¼

Fig. 6. Equivalent circuit of cascaded two-port network.

YT 2

DCF

YT 12

DCF

DCF

DCF

CT

Rf

YT 3

+U

LC

f

If

REC-I

REC-II

1=Y T12 Z eq  Z T12 þ Z eq ==Z F1ð23Þ þ 1=Y T12 Z eq þ Z F1ð23Þ

INV-II

INV-I

(a)

Uf

YT 22

DCF

YT 21

DCF

DCF

YT 1

Rf

YT 3

1 Y T21 Z T21 þ 1

f

If

REC-I

ICTF2 ¼

INV-II

INV-I

REC-II

ð14Þ

Supposing l is the distance between REC-II to INV-I, and l0 is the distance between the fault point and the CT. From Eq. (8), Eq. (14) can be written as:

+U

LC

Z T22 T22 Y T22 þ1 Z T21 Z Z T21 þZ Y þ1þZ T22 þZ YT22 þ1 T21 T21 T22 T22

Z T22 þZ

Rf þ ðZ T21 þ ZT21ZYT21 Þ==ðZ T22 þ ZT22ZYT22 Þ T21 þ1 T22 þ1 

CT

LD

ICTF2 ¼

ZT 3 ZT 3

ZT 21 ZT 21 ZT 22 ZT 22

DCF

ZT 1 ZT 1

(b)

Uf

tanh cðll0 Þ tanh cl0 þtanh cðll0 Þ 0

1  0 0 Rf þ Z C ðtanh cl == tanh cðl  l ÞÞ chcl cx

ZT 21

ZT 21

ZT 22

ZT 22

Rf

CT

YT 21

...

YT 22

+ Uf

...

If

REC-II

INV-I

(c) Fig. 7. Equivalent circuit under a DC transmission line fault. (a) Fault happens at F1. (b) Fault happens at F2. (c) Simplified circuit when a fault happens at F2.

T F1R ¼



AF1R BF1R



C F1R DF1R

 ¼

AT12 BT12



C T12 DT12

   AT23 BT23  D0 C T23 DT23

A0 B0 C0

cx

ð10Þ Blind zone of CT1 Effective zone of CT1

AT23

BT23

C T23

DT23



 ¼

AT2

BT2

C T2

DT2



A0

B0

C0

D0



AT3

BT3

C T3

DT3



A0

B0

C0

D0

ð15Þ

shcx e e where tanh cx ¼ ch cx ¼ ecx þecx . From Eq. (15), when the fault point is in the terminal of the DC line between BC (i.e. near INV-I converter station), l0 = l, ICT-F2 = 0. Otherwise, ICT-F2 > 0. Therefore, fault location can be identified by measuring the current component at the characteristic frequency. In external fault, the characteristic frequency current measured by CT is smaller compared with the non-characteristic frequency current. In contrary, the characteristic frequency current is larger than the noncharacteristic frequency current in internal fault. If the current is measured by single-side CT, a blind zone exists at the end of the line as Eq. (15). This is similar as the distance protection in AC system. The effective protection zone is shown in Fig. 8.

where



ð13Þ

When the frequency of the fault current is equal to the characteristic frequency of the DCF, then Zeq = 0, and K1 = 0. In Eq. (12), ICT-F1 = 0. When the frequency of the fault current is different from the characteristic frequency of the DCF current, then K1 – 0, and ICT-F1 > 0. A fault occurs in F2 is shown in Fig. 7(b). When the fault current frequency is different from the characteristic frequency of the DCF current, ICT-F2 > 0, which will not be analyzed in detail. When the fault current frequency is equal to the characteristic frequency of the DC filter, Zeq = 0. The simplified circuit is shown in Fig. 7(c). The current measured by the CT is:

ZT 3 ZT 3

ZT 2 ZT 2

ZT 11ZT 11 ZT 12 ZT 12

YT 11

ð12Þ

where K1 is the shunt coefficient, ‘‘//’’ denotes the paralleled value of the two impedances. K1 is given by:

Zeq

LD

Uf Z F1L   K1 Rf þ Z F1L ==Z F1R Z F1L þ Z F1R

 ð11Þ

According to the meaning of two-port network transfer parameters, from the view of the fault point, the impedances of the left side and the right side are: ZF1-L = AF1-L/CF1-L, ZF1-R = AF1-R/CF1-R

B CT1

CT2

C

Effective zone of CT2

Blind zone of CT2

Inverter-I Fig. 8. Effective zone of DC transmission lines protections on CT1 and CT2.

651

J. Cheng et al. / Electrical Power and Energy Systems 61 (2014) 647–655

Complex wavelet

ws;w ðkT; TÞ ¼ T

5 6 X pffiffiffiX di sðk  iÞT  kj ws;w ½ðk  jÞT; f  f i¼1

The theory of complex wavelet 3

Fast Fourier Transform (FFT) is widely used in the signal processing in the power system area. But its frequency band aliasing phenomenon in different scale is still a problem. According to the analysis in the above sections, the mutual relationship detection between 12nd and 15th harmonics is a challenge. If 600 Hz is the fundamental frequency for the analysis, then 750 Hz is 1.25th harmonic to the fundament frequency. This paper adapts a recursive complex wavelet algorithm which is well applied to the detection of non-integer harmonics. Complex wavelet has the advantages of smooth and symmetry which make it easy to match the signal phase-frequency characteristic [22]. Assume eix0 t is frequency shift factor. A rapid attenuation complex mother wavelet is taken as:

  r3 t3 r4 t4 r5 t5 ðrþix0 Þt e wðtÞ ¼    uðtÞ 3 6 15

ð16Þ

where x0 is the fundamental frequency. To satisfypthe ffiffiffi admissibility condition w(0) = 0 of the mother wavelet, x0/r ¼ 3. Supposing the sampling period of input signal s(nT) is T, and the translation parameter b = kT, then the wavelet transform is taken as:

ws ðkT; f Þ ¼

X pffiffiffi  ðnT  kTÞT sðnTÞ f w½f

ð17Þ

n

 is the conjugate complex of w, and f is the test frequency. where w To satisfy the real-time requirement, the rapid recursive algorithm obtained from the Z transform of Eq. (17) is:

4

ð18Þ

j¼1 5

3

4

5

3

5

B where d1 ¼ ½B3  B6 þ 15 A, d2 ¼ ½2B3  5B3 þ 26B A2 , d3 ¼ ½ 6B3 þ 22B A3 , 15 5 3

4

5

3

4

5

B A4 , d5 ¼ ½B3 þ B6 þ 15 A5 , A ¼ efTðrjx0 Þ , B = rfT, d4 ¼ ½2B3 þ 5B3 þ 26B 15

k1 ¼ 6A, k2 ¼ 15A2 , k3 ¼ 20A3 ; k4 ¼ 15A4 , k5 ¼ 6A5 , k6 ¼ A6 The wavelet transform coefficient ws,w(kT,T) is the characterization of the frequency f in the input signal at the time kT. Detection harmonics by complex wavelet s1(t) s2(t) are assumed to be two testing signals:



s1 ðtÞ ¼ sinðx1 tÞ þ k1 sinð12x1 tÞ s2 ðtÞ ¼ sinðx2 tÞ þ k2 sinð1:25x2 tÞ

ð19Þ

where x1 = 2p, x2 = 24p. s1(t) is taken to investigate the relationship between 50 Hz and 600 Hz currents and s2(t) is taken to investigate the relationship between 600 Hz and 750 Hz currents. k1 is selected as 0.001, 0.005, 0.01, and 0.02. k2 is selected as 0.1, 0.3, 0.5, and 0.8. The phase angles of the wavelet transform coefficient are shown in Fig. 9. As shown in Fig. 9, when 50 Hz is selected as the reference fundamental frequency, the phase spectrum of complex wavelet is highly sensitive in detecting 600 Hz harmonic signal in s1(t). When 600 Hz is selected as the reference fundamental frequency, the phase spectrum of complex wavelet is less sensitive in detecting 750 Hz harmonic (1.25th harmonic) signal in s2(t). According the characteristics shown in Fig. 5, detecting the 750 Hz harmonic (1.25th harmonic) signal is more suitable for the protection relay.

Fig. 9. Variation of the phase angle under different harmonic signals. (a) s1(t) with k1 = 0.001. (b) s1(t) with k1 = 0.005. (c) s1(t) with k1 = 0.01. (d) s1(t) with k1 = 0.02. (e) s2(t) with k2 = 0.1. (f) s2(t) with k2 = 0.3. (g) s2(t) with k2 = 0.5. (h) s2(t) with k2 = 0.8.

652

J. Cheng et al. / Electrical Power and Energy Systems 61 (2014) 647–655

Fig. 10. Variation of the phase angles under different harmonic signals. (a) s3(t) with k3 = 0.5. (b) s3(t) with k3 = 2.

A new criterion is presented in this paper. The fundamental frequency is set as 600 Hz (x0 = 24p). The test frequency is set as 750 Hz (1.25th). To supervise the change times of the phase curve in the time window of t1, the number of the phase angles greater than 3 radians will be counted.

g1 ¼ numbert1 ðdðtÞ > 3rad Þ

ð22Þ

g1 6 m1 (m1 is the setting value) indicates the percentage of 750 Hz harmonic is small compared to 600 Hz harmonic. Therefore, it is considered as an internal fault. Impact of MTDC control strategy and measurement

Fig. 11. Response characteristicses of each converter in MTDC system after a fault happen at F2. (a) DC current in each converter. (b) Ignition delay angle (firing angle) in each converter.

Since some frequency harmonics are not strict sine waves, fluctuation and attenuation will happen in these harmonics. Supposing a fixed oscillation frequency and a fixed attenuation coefficient, a testing signal is defined as:

s3 ðtÞ ¼ ½sinðx2 tÞ  sinðxC tÞ þ k3 sinð1:25x2 tÞ  sinðxC tÞ  ert

In Fig. 11, the response time of the DC system after the fault is caused by the PI controller. The influence of the control system should be reduced in designing the DC line protection. The time window of 5–10 ms after the fault can be used. Considering that the highest characteristic frequency is 750 Hz, the signal sampling frequency can be set at 20 kHz which is common in modern protection engineering (the sampling interval is 0.05 ms). The criterion monitoring time window is 5–10 ms after the fault, therefore the transient status of the initial traveling wave can be avoided in a large scale. As a result, the saturation factors of the measurements such as that of the CT can be neglected. Furthermore, the phase spectra of the complex wavelets in Figs. 9 and 10 are less sensitive in detecting 750 Hz harmonic (1.25th harmonic) signal in s2(t). Thus, the measurement impacts are not the dominate factor. Protection logic

ð20Þ where x2 = 24p, xC is the fixed oscillation frequency, and r is the attenuation coefficient. Suppose xC = 50p, r = 2.5, k3 is 0.5 and 2. The phase angle is shown in Fig. 10. Fig. 10 shows the phase angle of complex wavelet coefficient is changed by the oscillation frequency xC. But the variation of the phase angle caused by k3 with xC is the similar as that caused by k2 without xC. When the value of k3 increases, the influence of the higher frequency 750 Hz on the phase angle of s3(t) became dominant and the curve is denser.

The whole DC line protection is divided into the protection startup and fault locating components. Protection startup component uses the principle of traveling wave which is a mature technology. Adding the fault locating component, the overall protection criterion logic is shown in Fig. 12.

Δ1 du dt

Fault location and protection

>

YES

>

YES

Δu

NO

Δuset

Fault location criterion

reset

Staring component by traveling wave

And And

YES Timer

The phase angle curve of complex wavelet coefficient can be expressed as:

dðtÞ ¼ angleðws;w ðkT; TÞÞt

NO

du max( ) dt

ð21Þ

where ws,w(kT, T) is the wavelet transform coefficient of the transient current from Eq. (18).

Δid

Δiset

id

η1 ≤ m1

>

And

NO

NO YES

Distance measurement

Fault location

Fig. 12. The protection scheme of MTDC transmission lines.

653

J. Cheng et al. / Electrical Power and Energy Systems 61 (2014) 647–655 Table 2 Different fault locations on MTDC transmission lines.

Simulation verification Simulation model The parameters of the MTDC simulation model are listed in Table 1. Distributed parameter model of the transmission lines is applied. The tower parameters are shown in Fig. 13. It can be calculated that the positive-sequence wave impedance is 258 X and the zero-sequence wave impedance is 486 X by the line parameters under bipolar operation. Simulation results

Fault locations

Fault sections

Comments

F11 F12 F13 F14 F21 F22 F23 F24 F31 F32 F33 F34

REC-IREC-II (AB) REC-IREC-II (AB) REC-IREC-II (AB) REC-IREC-II (AB) REC-IIINV-I (BC) REC-IIINV-I (BC) REC-IIINV-I (BC) REC-IIINV-I (BC) INV-IINV-II (CD) INV-IINV-II (CD) INV-IINV-II (CD) INV-IINV-II (CD)

0% of AB from B 50% of AB from B 80% of AB from B 100% of AB from B 0% of BC from B 50% of BC from B 80% of BC from B 100% of BC from B 0% of CD from C 50% of CD from C 80% of CD from C 100% of CD from C

To verify the effectiveness of the protection, the fault locations are shown in Fig. 14, and the fault details are introduced in Table 2.

Table 1 Main circuit parameters of the MTDC model. Items

Values

Comments

Nominal power P Transformer MVA Transformer ratio Transformer leakage reactance Direct voltage DC reactor DC current of rectifier I and II Control strategy of rectifier I and II DC current of inverter I and II Control strategy of inverter I Control strategy of inverter II Control parameters Distance of AB/BC/CD

3200 MW 1000 MVA 525/313.6 kV

1.0 p.u.

±800 kV 0.29/0.27H 1 kA and 3 kA CC(in stable) 2 kA and 2 kA CC(in stable) CEA(in stable) a = 15°, c = 17° 500/1000/500 km

Y0/Y, and Y0/D 0.16 p.u. Rectifier/inverter MFA(auxiliary) CV (auxiliary) CC(auxiliary) Fig. 15. The phase angle by complex wavelet. (a) Fault happens at F11. (b) Fault happens at F21.

Fig. 16. The phase angle by complex wavelet. (a) Fault happens at F12. (b) Fault happens at F22.

Fig. 13. The tower structure of HVDC line.

Fig. 14. Different fault locations on MTDC transmission lines.

The operation performance of the protection using the transient current detected by the CT in Fig. 14 is investigated. For the protection at this CT, F11  F14 on AB are taken as external faults, and F21  F24 on BC are taken as internal faults. By complex wavelet transformation, the variations of the phase angles in all kinds of fault types are shown in Figs. 15–18. From Figs. 15–17, the criterion is sensitive in detecting every kind of fault. The phase angle curve is sparse when a fault happens in the protection zone, while the phase angle curve is denser during the external faults. Fig. 15 shows the phase angle curves under

654

J. Cheng et al. / Electrical Power and Energy Systems 61 (2014) 647–655

Fig. 17. The phase angle by complex wavelet. (a) Fault happens at F13. (b) Fault happens at F23.

Fig. 19. The phase angle using criterion I after adding 30 dB noise. (a) Fault happens at F13. (b) Fault happens at F23.

Table 3 The results calculated by two criteria in different faults. Fault condition

Fault location

Criterion I

g1

Fig. 18. The phase angle by complex wavelet. (a) Fault happens at F14. (b) Fault happens at F24.

the internal faults and the external faults nearby the REC-II. These two phase angle curves are significantly different. When faults happen on 50% and 80% of total lines (as shown in Figs. 16 and 17), the phase angle curves under the internal faults and the external faults are different. But the difference becomes less evident under the conditions when faults happen on 0% of total lines (as shown in Fig. 18). When the faults happen on the end of the lines (as shown in Fig. 18), the phase angle curves of the criterion cannot be distinguished. Therefore, the protection is in the blind zone. When 30 dB noise is added to the MTDC system after fault, the phase angle by the criterion is shown in Fig. 19. It shows the criterion is effective in noise circumstance. To compare the results obtained by the criterion in this paper and those obtained by the references, Table 3 shows the different results during the different faults. Criterion I represents the proposed criterion of this paper. Criterion II is the criterion of Ref. [9] which uses the direction classification. Criterion III is the criterion of Ref. [19] which calculates the fault distance. Criterion IV is the criterion of Ref. [10] which calculates the coefficients based on wavelet. Two fault conditions are simulated. One is the ground fault which transition resistance is 100 X, and the other is the AC/DC conductor contact fault (ac instantaneous voltage is 1000 kV, dc voltage is ±400 kV which is at reduction voltage). From Table 3, it can be summarized as follows:

Groundfault, Rf = 100 X

F11 F12 F13 F14 F21 F22 F23 F24 F31 F32 F33 F34

AC/DC conductor contact fault

F11 F12 F13 F14 F21 F22 F23 F24 F31 F32 F33 F34

p 24/ p 21/ p 23/ p 23/ p 9/ p 15/ p 18/ 25/ p 25/ p 24/ p 25/ p 25/ p 25/ p 23/ p 24 p 25/ p 10/ p 14/ p 18/ 26/ p 26/ p 23/ p 24/ p 25/

Criterion II Fault direction

Criterion III Fault distance (km)

Criterion IV Eb(108)

p / p / p / p / p +/ p +/ p +/ p +/ p +/ p +/ p +/ p +/

±10.4 ±239.1 ±419.4 ±489.6 ±7.6 ±491.8 ±815.3 ±1018.7 ±1014.3 ±1229.7 ±1394.2 ±1523.8

1.1301 0.4217 0.2845 0.1132 1.1202 0.1024 0.0341 0.0198 0.0197 0.0124 0.0085 0.0064

+/ +/ +/ +/ / / / / / / / /

±15.4 ±241.1 ±424.4 ±486.6 ±18.6 ±509.9 ±817.8 ±1022.5 ±1023.2 ±1222.4 ±1379.6 ±1531.4

1.2243 0.4654 0.2972 0.1241 1.1320 0.1269 0.0356 0.0178 0.0182 0.0118 0.0094 0.0072

p

is correct,  is wrong. + is positive direction, + is negative direction

Take m1 = 18, when faults happen at F21, F22 and F23, g1 6 m1. Hence, they are internal faults. The minimum values g1 of the other faults are larger than 21, which are considered as external faults. It also shows the blind zones when faults happen at F24. It is verified that Criterion I is effective except the fault at F24. It is shown that Criterion II is effective in detecting ground faults. However, the fault directions are misdiagnosed when AC/ DC conductor contact fault happen. For instance, when fault F22 happens, the fault location can be identified by the current direction. This is because that the currents flow into the fault point. However, the fault current directions are different in the fault of AC/DC conductor contact, when the peak value of the AC phase voltages are higher than that of the DC voltages. The direction

J. Cheng et al. / Electrical Power and Energy Systems 61 (2014) 647–655

655

simulation validates the criterion can locate the fault by identifying the relationship between the 600 Hz and 750 Hz harmonics. Acknowledgements This work was supported by the National High Technology Research and Development Program of China (863 Program) (No. 2011AA05A119). References

Fig. 20. Fault current directions in different fault types.

Table 4 Reliability assessments under different faults.

kp

Ground fault Rf = 200 X

AC/DC conductor contact fault

Adding 40 dB noise

DC line to line fault

1.167

1.278

1.222

1.167

changes of the fault current superposition components are illustrated in Fig. 20. It is shown that Criterion III can determine the fault distance effectively, but it cannot find whether the faults locate in the positive direction or in the negative direction. It is shown that each value of Eb is above 0.005*108 under Criterion IV. The fault can be detected very well, but may not be accurately located. The comparisons of the four criteria verify that Criterion I is relatively more effective in fault locating. After locating the fault line section by Criterion I, Criterion III is suitable for fault distance measurement. To evaluate the criterion I’ validity and efficiency, the assessment is defined as

kp ¼ minðgout Þ=m1

ð23Þ

where gout can be get from (21) in external faults (happened in AB and CD). The reliability of criterion I is given in Table 4. From Table 4, kp > 1 means that the outside faults can be detected efficiently. Conclusions Smoothing reactors in MTDC system cannot be taken as the boundary for transmission protection. The DCF can filter the characteristic harmonic of the transient currents to a certain extent. Therefore, the difference between the external and the internal faults can be detected. This paper presents a novel criterion using complex wavelet to identify the characteristic and non-characteristic harmonics. This single end protection criterion has blind zone which need to be compensated by the protection at the other end. The

[1] Kumar A, Lescale VF, Junlin LE, Chauhan RK. Concept to design multi-terminal at 800 kV HVDC: NER/ER-ER/WR interconnector – I project in India. [C]//CIGRE Session. Paris, France: CIGRE, 2010. [2] Kundur P. Power system stability and control, ser. EPRI Power System Engineering Series. New York: McGraw-Hill; 1994. [3] Sood VK. HVDC and FACTS controllers: applications of static converters in power systems. Kluwer Academic Pub., 2004. [4] Flourentzou N, Agelidis VG, Demetriades GD. VSC-based HVDC power transmission sytems: an overview. IEEE Trans Power Electron 2009;24(3): 592–602. [5] Chaudhuri NR, Chaudhuri B. Adaptive droop control for effective power sharing in multi-terminal DC (MTDC) Grids. IEEE Trans Power Syst 2011;28(1): 21–9. [6] Lu W, Ooi BT. Optimal acquisition and aggregation of offshore wind power by multiterminal voltage-source hvdc. IEEE Trans Power Deliv 2003;18(1):201–6. [7] Cole S, Beerten J, Belmans R. Generalized dynamic VSC MTDC model for power system stability studies. IEEE Trans Power Syst 2010;25(3):1655–62. [8] Candelaria J, Jae-Do P. VSC-HVDC system protection: a review of current methods. In: Proc IEEE/PES power systems conference and exposition (PSCE) 2011. p. 1–5. [9] Tang LX, Ooi BT. Locating and isolating DC faults in multi-terminal DC systems. IEEE Trans Power Deliv 2007;22(3):1877–84. [10] Liu XL, Osman AH, Malik OP. Hybrid traveling wave/boundary protection for monopolar HVDC line. IEEE Trans Power Deliv 2009;24(2):569–78. [11] Suonan JL, Gao SP, Song GB, Jiao ZB, Kang XN. A novel fault-location method for HVDC transmission lines. IEEE Trans Power Deliv 2010;25(2):1203–9. [12] Zhang Y, Tai NL, Xu Bin. Fault analysis and traveling-wave protection scheme for bipolar HVDC lines. IEEE Trans Power Deliv 2012;27(3):1583–91. [13] Zhu YL, Fan XQ. Fault location scheme for a multi-terminal transmission line based on current traveling waves. Int J Electr Power Energy Syst 2013;53(12): 367–74. [14] Dewe MB, Sankar S, Arrillaga J. Application of satellite time references to HVDC fault location. IEEE Trans Power Deliv 1993;8(3):1295–302. [15] Ngu EE, Ramar K. A combined impedance and traveling wave based fault location method for multi-terminal transmission lines. Int J Electr Power Energy Syst 2011;33(10):1767–75. [16] He ZY, Gao SB, Chen XQ, Zhang J, Bo ZQ, Qian QQ. Study of a new method for power system transients classification based on wavelet entropy and neural network. Int J Electr Power Energy Syst 2011;33(3):402–10. [17] Gafoor SA, Rao PVR. A transient current based busbar protection scheme using wavelet transforms. Int J Electr Power Energy Syst 2011;33(4):1049–53. [18] Bernadic A, Leonowicz Z. Fault location in power networks with mixed feeders using the complex space-phasor and Hillbert–Huang transform. Int J Electr Power Energy Syst 2012;42(1):208–19. [19] Mamis MS, Arkan M, Keles C. Transmission lines fault location using transient signal spectrum. Int J Electr Power Energy Syst 2013;53(12):714–8. [20] Szechtman M, Wess T, Thio C. First benchmark model for HVDC control studies. Electra 1991;135(4):54–67. [21] Mamis MS. Non-iterative nodal solution for transient simulation of nonlinear and time-varying elements. Int J Electr Power Energy Syst 2011;33(7): 1256–61. [22] Charri O, Meunier M, Brouaye F. Wavelets: a new tool for the resonant grounded power distribution systems relaying. IEEE Trans Power Deliv 1996;11(3):1301–9.