Sequence component based approach for fault discrimination and fault location estimation in UPFC compensated transmission line

Electric Power Systems Research 180 (2020) 106155

Contents lists available at ScienceDirect

Electric Power Systems Research journal homepage: www.elsevier.com/locate/epsr

Sequence component based approach for fault discrimination and fault location estimation in UPFC compensated transmission line

T

Biswapriya Chatterjeea,*, Sudipta Debnathb a b

Department of Electrical Engineering, Aliah University, Kolkata 700160, India Department of Electrical Engineering, Jadavpur University, Kolkata 700032, India

ARTICLE INFO

ABSTRACT

Keywords: Internal fault detection Unified power flow controller Transmission line Fault location Sequence component analysis

This study presents an integrated approach for discrimination of internal and external fault, and to obtain pinpoint location of the fault in a transmission line connected to unified power flow controller (UPFC). To detect internal fault, this method uses positive sequence component of voltage and current at the local and remote end bus. Fault location estimation technique requires both positive and negative sequence voltage and current components measured from both ends of the line. The performance of the proposed technique has been evaluated by simulating a 400 kV double circuit UPFC compensated transmission line. This technique can detect internal fault under stressed conditions, viz. high impedance fault (HIF), change in source strength, change in load angle, and power swing. Precise fault location has been achieved under wide variation of system and fault parameters. The comparative assessment has been provided to prove the efficacy of the proposed approach.

1. Introduction The ever increasing demand of power has drawn our attention to expand the existing power system networks by building new transmission networks. Nevertheless, this is not always feasible due to restrictions of right-of-way. Aiming to meet power demand, and to make steady expansion of the network, double circuit transmission lines with compensation have been utilized. The uninterrupted operation of power transmission lines is affected mostly by faults, which occur frequently. In the past few years, researches have been focussed to develop reliable algorithm for quick detection of faults, which helps to improve the system stability and economy. Discrete wavelet transform (DWT) is an effective mathematical tool to analyze the faulty voltage/current signals, which are mainly transient in nature. DWT technique has been successfully implemented in conjunction with artificial intelligent based methods for detection, classification, and location of faults in transmission lines [1–6]. The protection scheme of single circuit transmission line has been presented in [1,2] using artificial neural network (ANN) with DWT. These techniques utilize two ANNs for fault location, which increases execution time and computational burden. In [3], directional protection scheme for double circuit transmission line has been proposed using ANN and DWT. However, the accuracy of the scheme under high impedance fault has not been addressed. Deep neural network (DNN) is a better

⁎

alternative for complex non-linear optimization problem, and has recently been successfully applied for fault location in series compensated three-terminal transmission line [4]. The main disadvantage associated with DNN is that, it requires large number of hidden layers to attain reasonable accuracy. The supremacy of adaptive neuro fuzzy inference system (ANFIS) over ANN in fault location scheme has been established in [5]. Fault location scheme has been presented in [6] using extreme learning machine (ELM) and DWT, and its superiority over support vector regression and ANN has been established. Support vector machine (SVM) can be a lucrative classifier tool as compared with ANN, because SVM solves quadratic optimization problem, which does not have the problem of local minima. Isolation strategy of faulty branch, and pinpoint location of fault in an interconnected power system using SVM, has been described in [7]. Fast discrete orthogonal S-transform (FDOST) in conjunction with SVM has been successfully applied for the protection of compensated transmission line in [8]. The main concern with this scheme is that, it requires eleven SVM units, involving huge computer memory space. The protection scheme of a new hybrid model, consisting of a transmission line and an underground cable, has been described using FDOST and SVM in [9]. The main drawback of this scheme is the use of high sampling frequency. A scheme has been reported in [10], to differentiate between faulty condition and swing condition for a series compensated line using modified full cycle discrete Fourier transform and SVM. However, the variation of source

Corresponding author. E-mail address: [email protected] (B. Chatterjee).

https://doi.org/10.1016/j.epsr.2019.106155 Received 25 August 2019; Received in revised form 4 November 2019; Accepted 4 December 2019 Available online 14 December 2019 0378-7796/ © 2019 Elsevier B.V. All rights reserved.

Electric Power Systems Research 180 (2020) 106155

B. Chatterjee and S. Debnath

scheme for untransposed single circuit transmission line. The effect of variation of source strength has not been considered in this study. Fault location scheme for double circuit transmission line has been presented in [29], using the sequence components of voltage and current signals. The effect of mutual coupling between the parallel circuits, due to zero sequence networks, has not been considered. A novel method has been proposed in [30], where the fault location problem has been solved using an optimization technique. Faulty line identification and location scheme in multi-terminal transmission network using the voltage and current data has been proposed in [31]. However, the algorithm is vulnerable in a situation when, close-in fault is generated near a tapping node, or a far-end fault occurs at a point, where no phasor measurement unit (PMU) is present. Discrimination of faults and estimation of fault distance in a transmission line using the dynamic change in shunt admittance has been proposed in [32]. However, the effect of FIA on performance of this algorithm has not been reported. A new algorithm of fault location scheme in an untransposed, double-circuit transmission line having dissimilar voltage rating is presented in [33]. However, the effect of varying source strength is neglected. Impedance based fault location scheme for composite power system network using optimum number of PMU has been described in [34]. However, the response time for this algorithm is little bit on the higher side. Fault detection and localization scheme in double circuit three-terminal transmission network using sequence component of voltage and current has been proposed in [35]. However, the algorithm does not detect nongrounded faults. From the above literatures, it is found that, protective relaying scheme for an overhead transmission line can be addressed in different ways. First, transform based techniques, viz. wavelet transform [1–6], S-transform [8], Fourier transform [10], Hilbert–Huang transform [11], etc. can be adopted for effective analysis of faulty signals. These methods have been successfully applied for detection, classification and location of faults. Further, machine learning tools, viz. ANN [1–3], ANFIS [5], DT [14], SVM [10], FIS [13,15,16] in conjunction with some transform based techniques can be another choice for detection, classification and location of faults. The time required for relays to respond remains a genuine issue, as these methods need proper training to attain the desired accuracy. In the recent past, few methods [18,33–35] have been developed without using any transform or artificial intelligent based methods. These methods are better in the sense that, they require less memory storage, as they need no expensive training or complex mathematical computation. This paper proposes an integrated approach for detection and location of asymmetrical faults in a UPFC compensated transmission line. The proposed scheme utilizes time synchronized voltage and current signals from two ends of the line. It can successfully discriminate internal faults from external faults. Precise fault location scheme has also been presented. The performance of the proposed scheme has been evaluated under diversified simulation studies. The scheme is robust, as it correctly detects internal faults under challenging conditions, like power swing, HIF, pre-fault power angle variation, weak infeed, etc. The remaining part of this manuscript has been organized mainly in eight sections. Section 2 illustrates the proposed fault identification and location schemes. Section 3 presents the test power system. The results of detection and location scheme have been discussed in Sections 4 and 5 respectively. Discussion, comparative study and concluding remarks are presented in Sections 6, 7 and 8 respectively. The list of references has been given at the end of this paper.

Nomenclature 1, 2 Up1, Up2 Un1, Un2 Ip1, Ip2 In1, In2 Sp1, Sp2 z m

Bus numbers connected to the relaying line Positive sequence voltage at bus 1 and 2 respectively Negative sequence voltage at bus 1 and 2 respectively Positive sequence current at bus 1 and 2 respectively Negative sequence current at bus 1 and 2 respectively Complex power at bus 1 and 2 respectively Impedance of transmission line Distance in per unit at which fault occurred

strength remains unaddressed in this study. Faulty phase selection using Hilbert–Huang transform (HHT), for shunt compensated transmission line has been described in [11]. However, the classification accuracy under swing condition has remained untouched. Cross correlation based detection and classification of faults in transmission line have been described in [12], in which the performance of the proposed scheme under critical scenarios, viz. HIF, swing, and load fluctuation have not been discussed. The authors of [13] have presented fault classification scheme in a transmission line with a high penetration of distributed generation using cross correlation technique. Decision tree (DT) based fault classification scheme for transmission line, utilizing magnitude of differential power, has been presented in [14]. Fuzzy inference system (FIS) based protection schemes utilizing positive sequence component of voltage and current signals have been reported in [15,16]. The main disadvantage associated with these techniques, viz. ANN, SVM, ANFIS, DT is that, they require a huge training data. Moreover, time required for making any decision may not be suitable for real time applications, as a very prompt response is desirable for digital relays. Wavelet analysis and fuzzy systems are sensitive to frequency variation; and hence, multilevel digital filtering becomes ubiquitous. On the other hand, algorithms, which are developed without using any transform based techniques, have drawn attention for their simplicity and low computational burden. Directional protection scheme for single and double circuit line, using positive sequence superimposed current signal, has been proposed in [17]. Back-up protection scheme for UPFC connected transmission line; utilizing magnitude of differential power has been discussed in [18]. Identification of fault zone and faulty line in wide area power system has been described in [19], using complex power. However, in [18,19], the performance under close-in and far-end faults have not been investigated. A new notion of integrated moving sum approach has been successfully applied in [20], for classification of faulty phase/s in transmission lines. A new algorithm of syntactic pattern recognition of power system signals has been introduced in [21,22], and has been successfully implemented in transmission line fault detection [23].Travelling wave based fault location schemes have been described in [24,25]. The primary concern of the travelling wave based schemes is that, it requires large sampling frequency for desired accuracy. The energies of the sequence components of voltage signals during fault have been utilized in [26], for locating fault in distribution systems. Aforementioned schemes have contributed significantly to the various protection schemes of transmission/distribution lines; however, the dependency of line parameters on the accuracy of these schemes remains a genuine issue. Due to ageing, loading, and atmospheric conditions, line parameters change over the time, which may affect the accuracy of several relaying algorithms. The works reported in [27–30], propose a one step forward solution, making fault location scheme independent of line parameters. In [27], fault location scheme for single circuit transmission line has been proposed using positive and negative sequence voltage and current data. However, the effect of fault inception angle (FIA) on location accuracy has not been reported. The method presented in [28], describes impedance based fault location

2. Proposed scheme Detection of various faults in a transmission line is essential for reliable operation of power system. However, a practical power system network consists of several interconnected lines. Proper detection of fault in an interconnected power system network is a challenging task. The proposed scheme has been addressed in two different stages. First, 2

Electric Power Systems Research 180 (2020) 106155

B. Chatterjee and S. Debnath

the scheme for internal fault detection has been presented, and segregation of internal and external fault has been done. Later, estimation of fault distance has been proposed. This algorithm has been developed for all unsymmetrical faults, viz. single line to ground (SLG), double line (LL) and double line to ground (LLG) faults. Hence, triple line to ground (LLLG) fault has not been considered in this case.

Sp1 =

Sp2 =

A threshold based algorithm has been developed for identifying the faulty line. A fault index (FI) has been formulated from the complex power of the two ends of transmission line. The equivalent network of a transmission line compensated with UPFC is depicted in Fig. 1. Circuitry involved between bus 1 and 3 is the equivalent circuit of UPFC. Ise and Ish are currents flowing through series impedance Zse and shunt impedance Zsh of the compensator respectively. In this study, the positive sequence voltage and current has been considered for fault detection. KVL equation of the line connected between bus 1 and 2 is shown in Eq. (1).

mz. Ip1 = Up2

(1

m ) z . Ip2

Up1 = Up2 + mz. Ip1

(1

m ) z . Ip2

=

m) z . (|Ip2 |2 ))

(3)

Sum and difference of positive sequence complex power at bus 1 and 2 can be obtained as,

Sp1 + Sp2 = (Up1. Ip2 + Up2. Ip1) + mz. (|Ip1 |2

|Ip2 |2 ) + z. Ip2 . (Ip2

Sp1

Ip1) (4)

= C1

Up1. Ip2) + mz. (Ip1 + Ip2 )2

Sp2 = (Up2. Ip1

|Sp1

Sp2 |

=

|C1| |C2|

Ip2 =

Up1

Up2

Up2

Up1

Up21

Up22

(12)

z

|Sp1 + Sp2 | |Sp1

Sp2 |

=

|Up1

Up2 |

|Up1 + Up2 |

(13)

(1

m) z . Ip2 = Up1

Up2

mz. In1

(1

m) z . In2 = Un1

Un2

mz =

(Up1

Up2 ). In2 Ip1In2

(Un1

Un2). Ip2

Ip2In1

Using (7) and (8), complex power at bus 1 and 2 can be calculated as follows,

Sp1 = (Up1) × (Ip1)* = (Up1) ×

Up1

Up2

(15)

Solving Eqs. (14) and (15), we get,

(8)

z

(14)

Similar relation can be obtained for negative sequence network also.

(7)

z

(11)

Sp2 =

mz. Ip1

(6)

In the event of an external fault, the equivalent network is depicted in Fig. 2. Hence, positive sequence current of both the terminals can be mathematically expressed as,

Ip1 =

Up2 ) 2

The fault location algorithm utilizes voltage and current information from both ends of the transmission line. Under faulty condition, the same network as depicted in Fig. 1, can be observed as positive and negative sequence networks of the transmission line [27]. Rearranging Eq. (1),

z. Ip2 . (Ip1 + Ip2 ) = C2 (5)

|Sp1 + Sp2 |

2Up1 Up2 z

2.2. Location estimation for internal fault

Using Eqs. (4) and (5), FI value has been formulated as shown in Eq. (6),

FI =

Up21 + Up22

From Eqs. (6), and (13), it is observed that, FI value for internal fault condition is always greater than that of external fault, or no fault situation. This is because, as per Eq. (13), FI value for external fault or no fault condition is a function of voltage only. As it appears from (13), numerator would yield a very small quantity, as voltages recorded from the two ends appear to be almost same. Hence, FI value for these conditions is very small, and close to zero. On the contrary, FI value for internal fault condition is a function of both voltage and current as shown in Eq. (6). A wide range of internal fault has been studied, and the FI value has been calculated. The FI value for internal fault condition has been found to be always greater than external fault under all conditions. Hence, a threshold value can be determined for segregating internal and external or no fault condition. Determination of the threshold value has been done by simulating various faults under diversified conditions.

(2)

(mz. Ip1. Ip2)

(10)

z

z

FI =

Similarly,

Sp2 = (Up1. Ip2) + ((1

Up1. Up2

Using (11) and (12), FI value has been formulated in Eq. (13),

(1)

m) z . Ip1. Ip2)

Up22

(Up1

Sp1

Sp1 = Up1 × (Ip1)* ((1

(9)

z

Sp1 + Sp2 =

Complex power at bus 1 and 2 can be represented by Eqs. (2) and (3), respectively,

= (Up2. Ip1) + (mz. (|Ip1 |2 ))

Up1. Up2

Sum and difference of complex power at bus 1 and 2 can be obtained as,

2.1. Identification of faulty line

Up1

Up21

*

z

Fig. 1. Equivalent network configuration during internal fault. 3

(16)

Electric Power Systems Research 180 (2020) 106155

B. Chatterjee and S. Debnath

Table 1 Transmission line, UPFC parameters.

Transmission Line

Fig. 2. Equivalent network configuration during external fault.

(1

m) z =

(Up1

Up2). In1 Ip1In2

(Un1

Un2). Ip1

UPFC

(17)

Ip2In1

Parameters

Value

Length of the line Nominal voltage Nominal frequency Positive sequence impedance Zero sequence impedance Positive sequence capacitance Zero sequence capacitance Voltage Power Type

100 km 400 kV 50 Hz 0.03273 + j0.3184 Ω/km

Hence, the distance at which fault occurred can be found from Eq. (18),

m=

mz = mz + (1 m ) z (Up1

(Up1

0.2587 + j1.174 Ω/km 0.013 μF/km 0.00768 μF/km 400 kV 100 MVA 48 pulse GTO-based converters

4. Simulation results for fault detection

Up2). In2

(Un1

Un2). Ip2

Up2)(In1 + In2)

(Un1

Un2). (Ip1 + Ip2)

The detection of faults in transmission line has been decided based on a threshold value, which is derived from the complex power of two ends of the line. The detailed description has been made in the following sections.

(18) 3. System model The single line diagram of the system studied along with the relaying algorithm is depicted in Fig. 3 [18]. The system consists of a 400 kV, 50 Hz transmission line connected between bus 1 and 2. The line is compensated with UPFC connected between bus 3 and 1. The transmission line is modelled with equivalent pi-sections of transmission line block in MATLAB/Simulink environment. The relevant parameters of transmission line and UPFC are reported in Table 1. In this paper, UPFC is simulated in automatic power flow control mode. The UPFC used in this study, consists of two 100 MVA, three-level, 48 pulse GTO based voltage source converters (VSC) connected by a dc coupling capacitor. The first converter is the shunt converter, which acts as STATCOM. It controls reactive power flow (generation or absorption) through it, while allowing active power flow to the second converter through the coupling capacitor. The second converter is the series converter, which acts as static synchronous series compensator (SSSC). The function of SSSC is to regulate voltage at the transmission line end (bus 1), by controlling the reactive power flow through it.

4.1. Determination of robust threshold One of the prime criteria of any relaying scheme is to segregate the faulty portion of the network from the healthy part. The criterion for internal fault detection is that, FI value must be greater than a threshold value, i.e. FI > ε, where, ε is threshold value. On the other hand, the criterion for external fault as well as no fault situation has been set as FI < ε. Comparing (6) and (13), it has been observed that, FI value for internal fault is more than ε. Conversely, FI value is less than ε in external fault as well as in normal condition. It is worth mentioning here that, for identification of faulty line, appropriate selection of threshold value (ε) is very important. In order to segregate the internal fault from external fault, the entire dataset of FI values for both the situations, have been analyzed. A diversified simulation study has been conducted to obtain a large data set. The input fault parameters, which have been varied to obtain data, are fault resistance (Rf) (0–200 Ω), FIA (0°–90°) and fault distance (1–99

Fig. 3. Simulated power system and relaying algorithm. 4

Electric Power Systems Research 180 (2020) 106155

B. Chatterjee and S. Debnath

km). Using these parameters, SLG, LL and LLG faults have been simulated in the line between bus 1 and 2. The readings were taken at a gap of 10 km in the line, as well as near the relaying points of both the buses. In addition to this, challenging situations, such as swing, strong/ weak infeed and line parameter variation have been simulated. Moreover, external faults have been generated in the line between bus 2 and 4. During the simulation of external fault, the ranges of input parameters are kept same, as that in the case of internal faults. For proper selection of robust threshold, the FI values have been validated through analytical as well as simulation study.

and current values have been computed as follows:

Up1 = 323.8 9.6°kV Ip1 = 3173.1 51.7°A

Up2 = 246.4

Up1 = 326.9 9.2°kV Up2 = 325.3

1.5°kV

FI value comes out to be 0.0673 per unit in this case. Case IV: An SLG fault in phase-A is created in the transmission line connected between bus 2 and 4 at a distance of 1.5 km from bus 2 with Rf = 0 Ω and FIA = 0°. In this condition, the positive sequence voltage and current values have been computed as follows:

Up1 = 324.5 9.6°kV Up2 = 265.1

0.2°kV

FI value has been computed as 0.1323.

Up1 = 326.6 9.5°kV Ip1 = 2137.8 8.8°A

4.1.2. Validation of ε by simulation studies To obtain the behavior of FI values in different simulating conditions, simulation of all fault (internal/external) cases as well as normal condition have been carried out. The trajectory of FI values along the length of the line for different fault impedances has been analyzed. SLG, LL and LLG faults have been created at t = 0.04 s in the line connected between bus 1 and 2 at different locations (from 1 km to 99 km in steps of 10 km) with varying Rf (0–200 Ω), and FIA (0°–90°). FI values have been recorded just after the inception of fault for SLG, LL, LLG faults, and have been plotted in Fig. 4(a)–(c) respectively. On the contrary, an SLG fault has been created in the line connected between bus 2 and 4 at

166.3°A

z = 0.03273 + j 0.317 /km m = 0.5 Using Eqs. (4) and (5), C1 and C2 have been calculated as follows:

C1 = 355.25

87.4°A

Using the same equations, FI value comes out to be 1.1914. Case III: An SLG fault in phase-A is created at the mid-point of the line connected between bus 2 and 4 with Rf = 100 Ω and FIA = 0°. In this condition, the positive sequence voltage at bus 1 and 2 have been recorded as,

4.1.1. Analytical validation of ε Few cases of both internal and external faults have been considered for analytical calculation of FI. Case I and II correspond to internal fault case and, case III and IV correspond to external fault situation. Case I: An SLG fault in phase-A is created at the mid-point of the line connected between bus 1 and 2 with Rf = 100 Ω and FIA = 0°. In this condition, the positive sequence voltage and current values have been computed as follows:

Up2 = 326.3 0.4°kV Ip2 = 1109.1

0.1°kV Ip2 = 42529.2

j29.41MVA C2 = 1021.7 + j255.1MVA

Finally, FI value comes out to be 0.3385 per unit, as per Eq. (6). Case II: An SLG fault in phase-A is created in the transmission line connected between bus 1 and 2 at a distance of 1 km from bus 2 with Rf = 0 Ω and FIA = 0°. In this condition, the positive sequence voltage

Fig. 4. Selection of ε for (a) SLG internal fault, (b) LL internal fault, (c) LLG internal fault, (d) external fault. 5

Electric Power Systems Research 180 (2020) 106155

B. Chatterjee and S. Debnath

different locations with varying Rf (0–200 Ω), and FIA (0°–90°). FI values computed for this case have been depicted in Fig. 4(d). After analyzing the entire data set of FI values, one interesting feature can be observed. There is a clear segregation of FI values between internal and external fault situation. Hence, a threshold value has been considered as ε = 0.15. The variation of FI values in case of internal fault condition as reported in Fig. 4(a)–(c), has been observed to be more than the threshold value for fault resistance up to 200 Ω. On the other hand, Fig. 4(d) reveals that, FI value is always less than the threshold for all external fault cases. Table 2 illustrates, the values of FI obtained from analytical calculation as well as from simulation studies. It has been observed that, FI values obtained by analytical and simulation study are in close proximity.

the FI values in this case are found to be more than the threshold. 4.2.3. Effect of change in pre-fault power transfer angle In this work, FI value is evaluated from the complex powers of two ends. Since, power flow in the line is dependent on pre-fault power transfer angle; hence, to test the performance of the scheme, pre-fault power transfer angle is varied in the range of −30° to +30°. The variation of FI value due to the change in pre-fault power transfer angle has been depicted in Fig. 6(b), which shows that, FI values under all the conditions are more than the threshold. It is to be mentioned here that, Fig. 6(b) shows SLG fault with Rf = 200 Ω and FIA = 0° at 10%, 50% and 90% line length from bus 1. 4.2.4. Effect of change in line length The performance of the proposed scheme has been tested by changing the length of the line. The actual length of the line is 100 km. A ± 5% and ± 10% change in line length has been introduced to check the validity of the proposed scheme. Fig. 6(c) has been shown for Rf = 200Ω and FIA = 0°, which demonstrates that, the proposed scheme correctly detects all internal faults.

4.2. Evaluation of the proposed scheme The fault detection scheme has been evaluated over wide variation of fault parameters, i.e. fault resistance, fault inception angle and fault location. This scheme does not involve any transform based techniques or artificial intelligent based methods in its algorithm. This is advantageous over the methods in [1–5,8,16,25], in terms of computer memory usage. The proposed scheme is better in terms of relaying speed compared to [18], whose fault detection time varies from 5 to 17.5 ms. The fault detection schemes in [2,7,16] can detect fault for maximum Rf of 100 Ω. The scheme proposed in this paper can detect fault accurately with maximum Rf of 200 Ω. The detection is based on FI value, which does not need complex mathematical calculation. The sampling frequency is 4 kHz for data acquisition, which is quite less as compared to the methods in [24,25].

4.2.5. Effect of change in source strength Source strength plays an important role in any relaying algorithm. The strength of the source connected at bus 3 has been made stronger. Similarly weak infeed condition has been introduced to check the robustness of the algorithm. A plot of variation of FI under both strong and weak infeed condition is shown in Fig. 6(d) (with Rf = 200 Ω and FIA = 0°), the X-axis of which is represented in log scale. The result shows that, the proposed algorithm clearly detects fault under both strong and weak infeed condition.

4.2.1. Variation of fault impedance, fault inception angle, fault location SLG, LL and LLG faults have been simulated in the line connected between bus 1 and 2 at t = 0.04 s. It has been observed that, fault detection is found to be sensitive to fault impedance variation. An attempt has been made to check the validity of proposed scheme under the variation of fault impedance. Fig. 5(a) gives result of fault detection for all types asymmetrical faults with Rf = 0 Ω. It is found that, FI value crosses the threshold just after the fault inception. Similar studies have been performed for Rf = 50 Ω and 200 Ω, the results of which, are shown in Fig. 5(b) and (c) respectively. All external faults have been simulated in the line connected between bus 2 and 4, the result of which is shown in Fig. 5(d), which reveals that, FI value for external/no fault condition is less than the threshold. Hence, the criterion set for detection of internal and external/ no fault condition is valid. Hence, FI value computed utilizing the complex powers of two ends of the line, is found to be effective for discriminating the external and internal fault condition. It is worth mentioning here that, the sampling frequency for data acquisition is kept at 4 kHz.

4.2.6. Effect of power swing Power swing occurs when a sudden load is added/ increased to the power system [20]. It can also occur after the fault clearance. Simulation has been carried in the line between bus 1 and 2 to demonstrate the effect of the swing condition. A 150 MW load is suddenly connected at bus 2 at t = 0.04 s. Fig. 7(a) reveals that, FI value during swing condition deviates from no fault condition. However, it remains below the threshold value. On the other hand, an SLG fault has been created in the same line at t = 0.04 s, and cleared at t = 0.1 s. This brings the power system into swing mode. This is depicted in Fig. 7(b), where after t = 0.1 s, swing condition has taken place due to the removal of fault. The proposed scheme clearly segregates no fault condition under swing mode of operation. 5. Simulation results for fault location estimation The location scheme has been developed for internal faults, and it is applied on the line connected between bus 1 and 2. The fault location can be estimated using Eq. (18), which is a function of voltage and current. Hence, it is expected to give good accuracy over wide variation of fault resistance. The performance of fault location algorithm is judged by evaluating percentage error, which is computed using (19).

4.2.2. Effect of change in line parameter Transmission line conductors are exposed to the vagaries of adverse atmospheric conditions. Moreover, ageing effect leads to the change in line parameters. To observe the performance of the proposed scheme, the line parameters are varied in the range of ± 5% and ± 10%. The variation of FI value for SLG fault with Rf = 200 Ω and FIA = 0° at 10%, 50% and 90% line length from bus 1 is depicted in Fig. 6(a). All

%error =

|Dest

Dactual | × 100 D

(19)

where, Dest, Dactual, D are the estimated distance, actual distance of fault

Table 2 Comparison of FI value obtained from analytical calculation and simulation studies. Fault type

Condition

FI value obtained from analytical condition

FI value obtained from simulation study

Internal

Rf Rf Rf Rf

0.3385 1.1914 0.0673 0.1323

0.3397 1.1929 0.0627 0.1219

External

= = = =

100 Ω, FIA = 0° at 50 km from bus 2. 0 Ω, FIA = 0° at 1 km from bus 2. 100 Ω, FIA = 0° at 50 km from bus 2. 0Ω, FIA = 0° at 1.5 km from bus 2.

6

Electric Power Systems Research 180 (2020) 106155

B. Chatterjee and S. Debnath

Fig. 5. Detection of faults (a) internal fault with Rf = 0 Ω, (b) internal fault with Rf = 50 Ω, (c) internal fault with Rf = 200 Ω, (d) external/no fault.

Fig. 6. Variation of fault index under (a) change in line parameter, (b) change in pre-fault power transfer angle, (c) change in line length, (d) change in source strength. 7

Electric Power Systems Research 180 (2020) 106155

B. Chatterjee and S. Debnath

Fig. 7. Fault detection during swing due to (a) sudden load increased, (b) fault clearance.

and total length of the line respectively. Different simulations have been carried out in order to evaluate the accuracy of this algorithm. All unsymmetrical faults, viz. SLG, LL and LLG faults have been simulated in the line. The faults have been created randomly along the whole length of the line. Fault impedance and fault inception angle has been varied in the range of 0–600 Ω and 0–90° respectively. Percentage error computed for various lengths (10 km–90 km in steps of 10 km) has been represented in Fig. 8 (with Rf = 0 Ω and FIA = 0°), which reveals that, the error becomes maximum near the buses and minimum at the middle of the line. The highest value of error is observed to be 0.041%. Moreover, it can be observed that, percentage errors computed at a particular length for all three faults are almost same. This is illustrated in Table 3 (with Rf = 300 Ω and FIA = 0°), which reveals percentage errors computed for all three faults are in close proximity. Hence, the fault locator gives accurate results irrespective of the fault type. An interesting observation can be made from Fig. 8 and Table 3 as, the percentage errors are same for different faults created at same location with different fault impedances. For an

example, error for SLG fault at 10 km is 0.039% for both 0 Ω and 300 Ω faults. Hence, the fault location is independent of fault impedance. The average error for all types of unsymmetrical faults is found to be 0.0248%. The supremacy of this proposed fault locator can be established by comparing this scheme with some reported works. Most reported works [2,3,8,30] use fault resistance up to 100 Ω. This proposed scheme is system independent; hence, a fair amount accuracy is obtained when fault is created even with a very high impedance (Rf = 600 Ω). Further, the accuracy of fault locating schemes [1,16], which have been developed using artificial intelligent tools, depends on type of the fault. This will increase the computational burden. The proposed scheme is independent of fault type. In addition, maximum and mean error is found to be quite less as compared with other fault locating schemes [1,2,5]. Table 4 shows the location accuracy of different fault location schemes. The proposed scheme requires two-end voltage and current measurement; hence, synchronization of the two end signals is very crucial for synchronized tripping of the circuit breakers connected at bus 1 and

Fig. 8. Distribution of mean % error in (a) SLG fault, (b) LL fault, (c) LLG fault. 8

Electric Power Systems Research 180 (2020) 106155

B. Chatterjee and S. Debnath

Table 3 Fault location accuracy for all asymmetrical faults. Line length from bus 1 (km)

5 10 30 60 90 95

SLG fault

LL fault

LLG fault

Estimated distance (km)

% Mean error

Estimated distance (km)

% Mean error

Estimated distance (km)

% Mean error

5.023 10.039 30.027 59.986 89.974 94.975

0.023 0.039 0.027 0.014 0.026 0.025

5.024 10.032 30.034 59.974 89.974 94.974

0.023 0.032 0.034 0.026 0.026 0.025

5.027 10.041 30.025 59.975 89.981 94.98

0.027 0.041 0.025 0.025 0.019 0.02

2 of the line. A study has been conducted, to examine the effect of synchronization delay on location accuracy by delaying remote end signal, i.e. signals measured at bus 2 by 2, 4, 6, 8 and 10 ms. It has been observed that, the error in fault location estimation due to synchronization delay is negligible. This scheme is found to be robust against this time delay of two-end signals. It is worth mentioning here that, this synchronization of the two end voltage and current signals is achieved by the GPS network, which ensures to transmit the synchronized data to the remote end by high speed broadband system.

literatures, which address the protective relaying schemes for double circuit/compensated transmission line, have been chosen for comparative study. In this paper, algorithms for detecting internal fault along with localization of internal fault have been proposed. The key advantage of this proposed scheme is that, it does not require any transform based technique [3,4,25], or any artificial intelligent based methods [3,4,15] to detect or locate fault. This ensures quite less computational burden as compared to other methods. This proposed scheme requires sampling frequency of 4 kHz, which is quite less compared to the scheme proposed in [25], which has used sampling frequency of 200 kHz. The fault location methodology for compensated transmission line has not been discussed in [15,18,20]. The schemes presented in [33,34] require the information about the series and shunt parameters of the line to locate faults; hence, the schemes are parameter dependent, whereas, the proposed technique does not require any information about the line parameters. The maximum error for this proposed scheme has been found to be 0.041%, which is less compared to other reported fault localization schemes in [3,4,25,33,35]. The synchronization delay of 10 ms increases the fault detection time considerably in [18], and deteriorates the performance. The effect of this delay has been found negligible in this work. The works reported in [3,33] use fault impedance up to 100 Ω to locate fault. This proposed fault location algorithm is independent of fault impedance; hence, it gives accurate result even with a very high value of Rf = 600 Ω. The proposed algorithm has been found to be robust, as the detection of fault for all cases has been found correct under critical scenarios, such as weak infeed, change in power transfer angle, power swing, etc.

6. Discussion The fault detection and location algorithm used in this paper does not require very high sampling frequency, as it is developed on phasor domain. A sampling frequency of 4 kHz has been used, which reduces the cost of the protection system. Pre-fault parameters have not been used in the development of the algorithm. The fault location algorithm is independent of system parameters, such as line impedance and bus impedance matrix. The fault location scheme does not need the information about the type of fault, and it is immune to fault resistance and fault inception angle. The effect of synchronization delay on the location accuracy of the system is negligible. Since this algorithm does not require any zero sequence parameter, the mutual coupling between double circuit lines can be avoided. The fault is detected using the complex power of two ends of the line. The fault is detected, if the fault index calculated from the complex power is more than the threshold value. Hence, this technique does not require any training to attain the desired accuracy. The performance of this algorithm has been tested over wide variation of parameters. It produces accurate results under critical scenarios, such as power swing and change in power flow angle. The short circuit level of the sources sometimes may not be strong enough to activate the overcurrent relays. Under such weak infeed condition, the performance of the proposed scheme is reliable, as indicated by the simulation results. The accuracy of this algorithm is not affected by the change in line parameter and change in line length. Hence, it can be applied in real time environment.

8. Conclusion In this paper, a new technique has been developed for detection of internal/external faults in a 400 kV transmission line compensated with UPFC. This scheme also estimates the location of fault. It utilizes the positive sequence complex power from two ends of the line for fault detection. Subsequently, the location estimation has been done using positive and negative sequence voltage and current. The robustness of the proposed scheme has been tested by simulating the power system model in MATLAB/Simulink software. The simulation result reveals that, the proposed scheme can detect internal faults under diversified conditions. Moreover, location scheme is found to be independent of transmission line parameters and fault resistance. The accuracy of fault location is observed to be only 0.0248%. The proposed scheme has been

7. Comparison with existing schemes A comparative analysis of various reported works with this proposed scheme has been performed, and tabulated in Table 5. The Table 4 Fault location error of existing and proposed schemes. Ref.

Technique used

Fault type

Fault resistance

Maximum error in %

Mean error in %

[1] [2] [3] [5] [16] [25] Proposed scheme

DWT, ANN DWT, ANN DWT, ANN DWT, ANFIS Sequence component, FIS FDST, Travelling wave Sequence component

Dependent Independent Independent Independent Dependent Independent Independent

– 15 Ω 99 Ω Independent 100 Ω Independent Independent

2.3 1.517 0.6665 9.9 5 0.185 0.041

0.068 0.24 – – – 0.078 0.0248

9

Electric Power Systems Research 180 (2020) 106155

B. Chatterjee and S. Debnath

Table 5 Comparative assessment. Scheme proposed by

Technique used

Signal used

Task performed

Compensator

Localization error

Yadav and Swetapadma [3] Mirzaei et al. [4] Jena and Pradhan [15] Kumar and Yadav [18] Biswal [20] Sahoo and Samantaray [25]

DWT, ANN DWT, DNN Sequence component, FIS Differential power IMSUM FDST, travelling wave

V, I I I V, I I I

Location Location Detection Classification Classification Location

No TCSC Series capacitor UPFC Series capacitor TCSC

Saber [33]

Transmission line theory, Taylor series Least square estimation Sequence component Sequence component

V, I

Location

No

Maximum % error is 0.6665 Mean % error is 0.0458 – – – Maximum error is 0.37 km Mean % error is 0.078 Maximum % error is 0.996

V, I V, I V, I

Location Detection and location Detection and location

No No UPFC

Barman and Roy [34] Gaur and Bhalja [35] Proposed scheme

found very effective under stressed conditions like power swing, HIF, weak infeed etc.

[17]

Conflict of interest

[18]

Nothing declared.

[19]

Reference

[20]

[1] A.G. Shaik, R.R.V. Pulipaka, A new wavelet based fault detection, classification and location in transmission lines, Int. J. Electr. Power Energy Syst. 64 (2015) 35–40. [2] B. Rathore, A.G. Shaik, Wavelet-alienation based transmission line protection scheme, IET Gener. Transm. Distrib. 11 (4) (2017) 995–1003. [3] A. Yadav, A. Swetapadma, A single ended directional fault section identifier and fault locator for double circuit transmission lines using combined wavelet and ANN approach, Int. J. Electr. Power Energy Syst. 69 (2015) 27–33. [4] M. Mirzaei, B. Vahidi, S.H. Hosseinian, Fault location on a series-compensated three-terminal transmission line using deep neural networks, IET Sci. Meas. Technol. 12 (6) (2018) 746–754. [5] M.J.B. Reddy, D.V. Rajesh, P. Gopakumar, D.K. Mohanta, Smart fault location for smart grid operation using RTUs and computational intelligence technique, IEEE Trans. Syst. J. 8 (4) (2014) 1260–1271. [6] M. Shafiullah, M.A. Abido, Z. Al-Hamouz, Wavelet-based extreme learning machine for distribution grid fault location, IET Gener. Transm. Distrib. 11 (17) (2017) 4256–4263. [7] P. Gopakumar, M.J.B. Reddy, D.K. Mohanta, Transmission line fault detection and localisation methodology using PMU measurements, IET Gener. Transm. Distrib. 9 (11) (2015) 1033–1042. [8] Z. Moravej, M. Pazoki, M. Khederzadeh, New pattern-recognition method for fault analysis in transmission line with UPFC, IEEE Trans. Power Deliv. 30 (3) (2015) 1231–1242. [9] B. Patel, A new FDOST entropy based intelligent digital relaying for detection, classification and localization of faults on the hybrid transmission line, Electr. Power Syst. Res. 157 (2018) 39–47. [10] U.J. Patel, N.G. Chothani, P.J. Bhatt, Sequence-space-aided SVM classifier for disturbance detection in series compensated transmission line, IET Sci. Meas. Technol. 12 (8) (2018) 983–993. [11] S. Biswal, M. Biswal, O.P. Malik, Hilbert Huang transform based online differential relay algorithm for a shunt-compensated transmission line, IEEE Trans. Power Deliv. 33 (6) (2018) 2803–2811. [12] A. Dasgupta, S. Debnath, A. Das, Transmission line fault detection and classification using cross-correlation and k-nearest neighbor, Int. J. Knowl. Based Intell. Eng. Syst. 19 (10) (2015) 183–189. [13] B. Chatterjee, S. Debnath, Cross correlation aided fuzzy based relaying scheme for fault classification in transmission lines, Eng. Sci. Technol. Int. J. (2019), https:// doi.org/10.1016/j.jestch.2019.07.002. [14] M.M. Taheri, H. Seyedi, B. Mohammadi-ivatloo, DT-based relaying scheme for fault classification in transmission lines using MODP, IET Gener. Transm. Distrib. 11 (11) (2017) 2796–2804. [15] P. Jena, A.K. Pradhan, An integrated approach for directional relaying of the double-circuit line, IEEE Trans. Power Deliv. 26 (3) (2011) 1783–1792. [16] A. Yadav, A. Swetapadma, Enhancing the performance of transmission line

[21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35]

10

Not reported Maximum error is 1.3% Maximum error is 0.041% Mean % error is 0.0248 %

directional relaying, fault classification and fault location schemes using fuzzy inference system, IET Gener. Transm. Distrib 9 (6) (2015) 580–591. A.R. Adly, R.A. El-Sehiemy, A.Y. Abdelaziz, A directional protection scheme during single pole tripping, Electr. Power Syst. Res. 144 (2017) 197–207. B. Kumar, A. Yadav, Backup protection scheme for transmission line compensated with UPFC during high impedance faults and dynamic situations, IET Sci. Meas. Technol. 11 (6) (2017) 703–712. S.S. Mirhosseini, M. Akhbari, Wide area back up protection algorithm for transmission line based on fault component complex power, Int. J. Electr. Power Energy Syst. 83 (2016) 1–6. M. Biswal, Faulty phase selection for transmission line using integrated moving sum approach, IET Sci. Meas. Technol. 10 (7) (2016) 761–767. C. Pavlatos, V. Vita, L. Ekonomou, Syntactic pattern recognition of power system signals, Proceedings of the 19th WSEAS Int. Con. Syst. (Part of CSCC’ 15) (2015) 73–77. C. Pavlatos, V. Vita, Linguistic representation of power system signals, Electricity Distribution, Springer-Verlag, Berlin Heidelberg, 2016, pp. 285–295 Energy Syst. Series. C. Pavlatos, V. Vita, A.C. Dimopoulos, L. Ekonomou, Transmission lines’ fault detection using syntactic pattern recognition, Energy Syst. 10 (2) (2019) 299–320. P. Nan, C. Menghan, L. Rui, F. Zare, Asynchronous fault location scheme for halfwavelength transmission lines based on propagation characteristics of voltage travelling waves, IET Gener. Transm. Distrib. 13 (4) (2019) 502–510. B. Sahoo, S.R. Samantaray, An enhanced fault detection and location estimation method for TCSC compensated line connecting wind farm, Int. J. Elect. Power Energy Syst. 96 (2018) 432–441. T.F. Moraes, L. Lovisolo, L.F.C. Monteiro, Fault location in distribution systems from analysis of the energy of sequence component waveforms, IET Gener. Transm. Distrib. 12 (9) (2018) 1951–1960. V. Terzija, Z.M. Radojevic, G. Preston, Flexible synchronised measurement technology-based fault locator, IEEE Trans. Smart Grid 6 (2) (2015) 866–873. A.S. Dobakhshari, Noniterative parameter-free fault location on untransposed single-circuit transmission lines, IEEE Trans. Power Deliv. 32 (3) (2017) 1636–1644. C.A. Apostolopoulos, G.N. Korres, A novel fault-location algorithm for double-circuit transmission lines without utilizing line parameters, IEEE Trans. Power Deliv. 26 (3) (2011) 1467–1478. M. Davoudi, J. Sadeh, E. Kamyab, Parameter-free fault location for transmission lines based on optimisation, IET Gener. Transm. Distrib. 9 (11) (2015) 1061–1068. Y. Cai, A.D. Rajapakse, N.M. Haleem, N. Raju, A threshold free synchrophasor measurement based multi-terminal fault location algorithm, Int. J. Electr. Power Energy Syst. 96 (2018) 174–184. S. Gangolu, P. Raja, M.P. Selvan, V.K. Murali, Effective algorithm for fault discrimination and estimation of fault location in transmission lines, IET Gener. Transm. Distrib. 13 (13) (2019) 2789–2798. A. Saber, Fault location algorithm for multi-terminal mixed lines with shared tower and different voltage levels, IET Gener. Transm. Distrib. 12 (9) (2018) 2029–2037. S. Barman, B.K.S. Roy, Detection and location of faults in large transmission networks using minimum number of phasor measurement units, IET Gener. Transm. Distrib. 12 (8) (2018) 1941–1950. V.K. Gaur, B. Bhalja, New fault detection and localisation technique for doublecircuit three-terminal transmission line, IET Gener. Transm. Distrib. 12 (8) (2018) 1687–1696.