Adaptive relay co-ordination using a busbar splitting approach for a system integrity protection scheme

Power system faults can often result in excessively high currents. If sustained for a long time, such high currents can damage system equipment. Thus, it is desirable to operate the relays in the minimum possible time. In this paper, a busbar splitting approach is used for adaptive relay setting and co-ordination purposes for a system integrity protection scheme (SIPS). Whenever a fault occurs, the busbar splitting scheme splits a bus to convert a loop into a radial structure. The splitting schemes are chosen such that the net fault current is also reduced. Busbar splitting eliminates the dependency upon minimum breakpoints set (MBPS) and reduces the relay operating time, thus making it adaptive. The proposed methodology is incorporated into the IEEE 14-bus and IEEE 30-bus systems with single and multiple fault conditions. The modeling and simulation carried out in ETAP, and the results of the proposed busbar splitting-based relay co-ordination are compared with the MBPS splitting-based relay co-ordination.


Introduction
Power system networks are prone to electrical faults which may occur for numerous reasons such as lightning, equipment failure, or other environmental reasons. High fault current is the ultimate consequence, irrespective of the type of electrical fault. Hence, protecting the power system network is of utmost concern and to be achieved by making sure that faults are cleared in the minimum possible time. Modern technologies such as the wide area monitoring system (WAMS), and using phasor measurement units (PMU), have led to the gradual development of system integrity protection schemes (SIPS), which oversee the complete network and undertake necessary actions to prevent system blackout.
Reference [1] describes the architecture of SIPS and its application in the power industry, while information from industrial engineering and research prospects is also included. In [2], a detailed literature review on the classification, purpose, implementation methods, and application of SIPS is presented, and the application of SIPS for improving system stability, avoiding distance relay mal-operation, reducing load shedding, etc. is briefly discussed. In [3], SIPS is used to avoid unwanted tripping of relays and prevent cascaded tripping which could result in system collapse. Two different algorithms are used to achieve this objective, i.e., changing the relay characteristics and identifying various points in the system, so as to help avoid relay mal-operation. In [4], a special protection scheme is proposed which depends on the various readings and measurements that are provided by the power system characteristics and the measurements from SCADA to protect the Kinmen power system from collapsing. A prevention scheme is proposed in [5] by analyzing and taking predetermined actions after severe disturbance conditions. In [6], an indices-based scheme is proposed to avoid the mal-operation of the distance relay, and synchronized data is used for the calculation of the proposed indices. Reference [7] develops a methodology that identifies the fault location and the bus in the nearest vicinity using PMU-based wide area protection. In [8], a technique based on WAMS is proposed to identify the faulty equipment in the protection system, in which the status signals from the circuit breakers, the voltage and current phasors, and decisions from relays, are used to find the faulty equipment. In [9], a methodology for the enhancement of SIPS is proposed that includes enhancement of its architecture by an algorithm involving the optimal power flow of the AC and DC. The requirement for testing the SIPS and its architecture before installing in the field is also provided. In [10], a method is proposed to overcome the challenges seen by SIPS because of high penetration of renewable energy and stressed conditions that lead to relay maloperation. The proposed method is classified into two techniques, with the first technique based on the security index of the relay to achieve a faster operation of zone 3 and the second technique based upon the stability index to foresee the conditions of system stability and take corrective action. A model-based approach for calculating the fault location is proposed in [11], one that uses control and processing software, along with simulation results of the power system. The proposed method overcomes the limitation of fault location calculation based on impedance. In [12], a methodology to find the fault location in an AC meshed microgrid is proposed, in which the fault location is identified by using support vector machines (SVM). Reference [13] proposes a mechanism to improve the zone 3 protection of the distance relay, where PMU data is used to develop an index that is derived from the rate of change of active power. In [14], a methodology is proposed that makes use of PMU data to form an index relating to the critical clearing angle for the transient stability analysis. An integrated early warning system is proposed in [15] to detect any kind of voltage instability in the system. By extrapolating trends, the violations of limits in the reactive powers of generators and the busbar voltages can be estimated to prevent the system from breakdown.
The difficulty in clearing a fault increases with the increase in the complexity of the network. In a multiloop transmission system, fault clearing depends heavily upon the MBPS of relays. The MBPS are those sets of relays that can cut a network in both anti-clockwise and clockwise directions [16]. In a multi-loop system, the relays which are common to more than one loop have different sets of relay settings [17]. Hence, to have a common setting for such relays, relay co-ordination starts from MBPS points. In [16], a methodology is proposed to find the MBPS using graph theory and processes such as the formation of a branch incidence matrix. The methodology is incorporated in the IEEE-14 bus system to find its breakpoints. The results of an MBPS-based relay coordination method are better and comparable with the other methods of relay coordination. In [18,19], an algorithm is proposed to find the breakpoints using integer linear programming and a graph theory approach that considers the cardinality of a relation as the breakpoints, whereas in [20], a new methodology is proposed to identify the MBPS-based upon the power quality and the zones which are at risk of cascaded tripping. In [21], a depth-first search algorithm is proposed to determine the MBPS of the given multi-loop network. Determination of the MBPS is a time-consuming process along with the concern that different methodologies give a distinct set of results for a single network. For example, for the same network, the methods used in [16] and [19] give 10 and 9 relays as a set of minimum breakpoints, respectively. Hence, the need arises for either eliminating the requirement of MBPS or at least minimizing its use to a great extent. Also, even if the MBPS is determined for a multi-loop system, to achieve the optimum relay settings for a coordination purpose, a large number of iterations are needed.
The relay setting mainly depends upon the maximum fault current and pick-up current of a relay (which in return depends on the minimum fault current value). Hence, to have a relay setting that can reduce the fault clearing time, it is necessary to modify the maximum and minimum fault current values. In [22], superconducting fault current limiters (SFCL) are proposed to limit the level of the fault current, though in the event of SFCL failure, the fault current will not be limited. In [23], a series reactor is used to limit the fault current and act as a compensating agent, although the methodology proposed increases the impedance of the lines and affects the voltage profile in normal scenarios. Devices such as IS-limiter, or a solid-state based fault current limiter [24][25][26], can be used to restrict the fault current. However, all the available techniques use external devices to manage the fault current level, and in the case where they are damaged, it may lead to severe maloperation of the protection system.
In addition, power system transmission networks are too complex to be able to analyze the effects of changes within the network. References [27,28] propose methodologies to determine the exact reduced equivalent networks to effectively examine the possible effects of events occurring in the original network. In [29], sensitivity analysis and busbar splitting techniques are used to maintain the voltage profile of the network, while such busbar splitting techniques are also used in [30,31] for managing the islanding processes, line overloading, and violation in voltage levels, and in [32,33] for managing the fault current level. This paper proposes a busbar splitting technique that clears a fault in a reduced time compared to the conventional MBPS-based relay co-ordination. It does this by minimizing the dependency on MBPS, thereby enhancing the protection objective of SIPS. The remainder of the paper is organized as follows. Section 2 details conventional MBPS-based relay co-ordination, whereas Sect. 3 describes the busbar splitting technique. In Sect. 4, the busbar splitting methodology for adaptive relay co-ordination is implemented. Section 5 gives the results and comparative analysis of the proposed and conventional techniques of relay co-ordination and Sect. 6 draws conclusions and reflects on future perspectives.

MBPS determination
To determine the minimum breakpoint set of any given network the following steps are conducted as given in [16].
A power system network A = (V , E) is considered which has m buses and n branches, with V being the set of buses given by v 1 , v 2, v 3, . . . , v m and E denoting the set of branches given by e 1 , e 2, e 3, . . . , e n .
Step 1: The bus to branch admittance matrix A is formed by assigning the matrix elements ( y ij ) as: + 1 when bus v j is the starting point of branch e i . -1 when bus v j is the ending point of branch e i . 0 when bus v j is neither the starting nor the ending point of branch e i Step 2: The rows of matrix A that correspond to the parallel branches in the given network are combined to obtain a reduced branch admittance matrix Ã , such that Ã ∈ R n×m .
Step 3: The rows of Ã that are combined are permuted to the bottom of the matrix.
Step 4: The columns of Ã are arranged in ascending order of their respective 2-norms.
Step 5: LU-factorization of Ã is performed and terminates after r steps, where r is the rank of Ã in order to determine the lower trapezoidal matrix, L, and upper trapezoidal matrix, U.
Step 6: The non-zero elements of U are used to sectionalize the given network into m − r number of connected graphs.
Step 7: Considering that the qth connected part of the network has m q buses and n q branches, a reduced matrix of L is determined, one that gives L such that L ∈ R n q × m q . In L , the column consisting of the maximum number of non-zero elements in the last n q + m q − 1 rows is determined. Considering all the non-zero elements in the chosen column, if the non-zero element is 1, the start of the branch corresponding to the element is chosen as a break point, whereas the end is chosen as a break point if the non-zero element is -1. In the case where the non-zero matrix elements correspond to the rows of combined branches, the starts of both branches are chosen as break points, if the element is 1, whereas if the element is -1, the ends of the parallel branches are considered to be the break points. The step is repeated untill there exists at least a single non-zero element in the chosen column.
Step 8: The steps above are repeated for all the connected networks with the MBPS being the combination of the MBPS of all connected graphs.
The above-mentioned steps are implemented in the IEEE 14-bus system to determine its minimum breakpoint set [16]. Figure 1 shows the MBPS of the IEEE 14-bus system.
The relays marked in black are the MBPS for the IEEE -14 bus system. In a power network with defined MBPS, relay co-ordination always starts from the breakpoints even for particular fault conditions.

Relay co-ordination with MBPS
A 5-bus test system is considered as shown in Fig. 2, with blackened relays depicting the breakpoints. These  breakpoints are determined by implementing the steps discussed in Sect. 2.1, for a fault at bus 5.
In this case, the co-ordination of relays starts from the breakpoint relays R4, R5, and R6 simultaneously, instead of relay R1 that is nearer to the fault. This makes the TMS of relay R1 more than those of relays R4, R5, and R6 respectively. Consequently, it makes the relay operating time t op long, which would otherwise have been shorter if relay R1 had operated first. The relays used in this work are assumed to have IEC standard inverse characteristics given in [17] as: where TMS j is the time multiplier setting of relay 'j' , I F MAX and I S are the maximum fault current and the pick-up value of current seen by the relay, respectively. As the pick-up value varies between the range of maximum load current to 2 3 of the minimum fault current [17], its value used in this work for co-ordination purpose is 2 3 of the minimum fault current. This refers to the worstcase analysis. Also, the co-ordination time interval (CTI) between the operation of two consecutive relays is considered to be 0.3 s. Knowing the fault currents seen by each relay, the TMS and t op for each relay can be obtained as in [17] using (1).

Busbar splitting-based relay co-ordination
Busbar splitting schemes are created by the use of Ward's equivalent method for network reduction. It works by transforming a star network into an exact equivalent reduced meshed network, such that any analysis carried out in the reduced equivalent model applies to the original network. According to Ward's equivalent method [27], the transformation proceeds by eliminating the desired buses and then the boundary buses are interconnected with each other by lines having impedance given as: where Z ab is the impedance between buses a and b, Z a , Z b , Z 1 , Z 2 . . . are the impedances between the bus and the neutral. The buses connected directly to the bus under consideration are known as boundary buses and the rest are considered as external buses. With Q representing the boundary buses and R representing the external buses, the voltage equation of a power system network can be given as: In the case where bus P is considered to be eliminated to reduce the network, using the Gauss elimination technique, the equivalent boundary bus admittance matrix and impedance matrix are respectively given as: Splitting the busbar P into bus P and P ′ , and applying the same technique as above, the after splitting equivalent admittance matrix and impedance matrix are respectively given as: Comparing (5) and (7), the boundary bus whose impedance is increased can be found when bus P is split into buses P and P ′ [34], i.e., it will give the information regarding which boundary bus should be split in order to have an increased impedance at the faulty bus. The flowchart in Fig. 3 depicts the sequence in which the steps need to be followed for assessment. Hence, in the case of a fault at one of the buses, splitting the boundary bus will result in an increase in the impedance and reduction in the fault current [34]. This change in the fault current along with the busbar splitting results in better adaptability of the network for relay coordination with reduced relay operating time.

Effect of busbar splitting on relay co-ordination
Considering the example network in Fig. 2 and assuming a fault at bus 5, bus 2 is split according to the proposed scheme as shown in Fig. 4.
With the splitting of bus 2, bus 5 is no longer a part of any loop and thus makes the faulty bus radial. With relays R11 and R12 seeing no fault current, relay R1 behaves as the primary relay and hence has the least TMS as coordination of relays now starts from relay R1. This results in reduced fault clearing time, which otherwise would have been longer in the case where the primary relays were relay R4, R5, and R6, as in the case before splitting.

Effect of fault current modification on relay operating time
From (1), it is clear that t op depends upon the ratio of (I FMAX /I s ) . If the ratio increases, t op reduces. Even if the net fault current at a bus decreases after busbar splitting, there is the possibility that the fault current seen by the relay nearest to the fault location increases. Thus along with minimizing the TMS for the relay nearest to fault location, if the ratio of (I FMAX /I s ) is increased, there will be a significant dip in the operating time of the relay that is nearest to the fault. Thus, the methodology that has been described for SIPS reduces the fault clearing time.

Relay co-ordination in the IEEE 14-bus system with MBPS
The location of breakpoints for the IEEE 14-bus system and the fault location considered are shown in Fig. 5. For relay co-ordination, it is considered that bus 11 is under fault. The TMS and t op for all the relays are calculated using (1) with the initial assumption of TMS = 0.025 for breakpoint relays and CTI of 0.3 s. Table 1 shows the operating time of the relays nearest to the fault and their backup relays, before incorporating splitting scheme.

Managing fault level with busbar splitting in the IEEE 14-bus system
Since the relay parameters such as TMS and t op depend upon the fault level, it is necessary to modify the maximum and minimum fault current so t op can be reduced, as given in Sect. 4.2. Table 2 shows the results of the applied busbar spitting methodology to the IEEE 14-bus system, where all the buses have been considered under fault with one bus at a time, with all possible splitting scenarios resulting in reduction in fault currents.

Consider fault at bus 'I'
Calculate ẐII for bus I according to equation (5) Split boundary bus bn of bus I, for n=1 to m, and m being the total number of boundary buses Calculate ẐII * for the splitting scheme according to equation (7) Is ẐII>ẐII * ? n=n+1 Implemented splitting scheme corresponding to maximum Zii Start relay co-ordination

Relay co-ordination in IEEE 14-bus system with busbar splitting
The same case is considered as that for MBPS-based relay co-ordination, i.e., a fault occurs at busbar 11. The splitting scheme of busbar 11 is presented in Fig. 6, and the fault currents seen by each relay are given in Table 3 after splitting. It can be seen that after the busbar splitting, relays R14 and R15 do not see any fault current, and the section of the network starting from busbar 11 becomes radial. This takes the concept of MBPS out of consideration with the nearest relay, i.e., relay R1, being the primary relay and the starting point for relay co-ordination. It is assumed that the initial setting of relay R1 is TMS = 0.025 and the CTI for relay co-ordination is 0.3 s. For relay R1, before busbar splitting, and after splitting of busbar 11, Table 4 shows the TMS and t op of the relay nearest to a fault and also of their backup relay, while Table 5 depicts the TMS and t op for each relay after the busbar splitting, with the blank places indicating no fault current for the corresponding relays. From Tables 1 and 4, it is deduced that the time of operation, t op , of the relay nearest to the fault, i.e., relay R1, has reduced from 1.5943 s (before splitting) to 0.4169 s (after splitting). Also, t op for the  back-up relay R1 has reduced from 1.846 s (before splitting) to 0.7169 s (after splitting). This shows that, after splitting, (I FMAX /I s ) increases and thus leads to a reduction in t op as in (1). Figure 7 shows the flowchart for relay co-ordination with busbar splitting approach. Figure 7 shows the flowchart for the relay co-ordination with busbar splitting approach.

Relay co-ordination in IEEE 30-bus system with MBPS
A similar analysis has been done for the IEEE 30-bus system using the proposed method, and Fig. 8 shows the breakpoints of the IEEE 30-bus system (darkened relays). In this case, a LLL fault is considered at bus 17 and the fault current values shown in Table 6 are obtained from the simulation in ETAP. The TMS and t op for the required relays (R4 and R7 for fault at bus 17) are calculated using (1) with the initial assumptions of TMS = 0.025 for the breakpoint relays and CTI of 0.3 s. Table 7 shows the TMS and the t op of the relays nearest to the fault location. It can be seen that, in a normal scenario, i.e., without busbar splitting, the respective TMS and t op for relay R4 are 0.1011 and 1.204 s, and 0.03 and 0.378 s for relay R7.

Relay co-ordination in IEEE 30-bus system with busbar splitting
For the fault at bus 17, the boundary bus 10 is split such that the net fault current at bus 17 can be reduced. Figure 9 shows that bus 10 is split into bus 10 and 10 ′ respectively. Such a splitting results in reducing the maximum fault current at bus 17 from 15.579 to 14.957 kA and the minimum fault current from 13.331 kA to 12.866 kA (as can be seen in Table 6). Splitting of bus 10 also converts the part of the system containing bus 17 into radial form. Thus, the starting relay for co-ordination is the relay nearest to the fault rather than the breakpoint relays, i.e., relay R4. In addition, relay R7 doesn't see any fault current, and hence the TMS and t op for relay R7 have no value as can be seen in Table 7. The TMS and t op for R4 are found to be 0.025 and 0.312 s, as shown in Table 7.
It can be seen that, t op has reduced from 1.204 s under conventional method to 0.312 s when the busbar splitting scheme is incorporated.

Relay co-ordination in IEEE 30-bus system with MBPS for multiple faults
In this case, faults are considered at buses 17 and 23, i.e., at multiple locations, as shown in Fig. 10. For the case of   relay co-ordination with MBPS, the maximum and minimum fault currents seen at buses 17 and 23 are given in Table 6. Table 8 provides the TMS and t op of the relays nearest to the fault (before splitting cases), which have been calculated using (1) with an initial assumption of TMS = 0.025 for breakpoint relays and CTI of 0.3 s.

Relay co-ordination in IEEE 30 bus system with bus bar splitting for multiple faults
For multiple faults, bus bar splitting is considered for each fault location. In this case, bus 17 and bus 23 are under fault, and bus 10 and bus 15 are split as shown in Fig. 11. This results in reduced fault currents as can be seen in Table 6.
With the splitting of bus 10 into bus 10 and 10 ′ , and bus 15 into bus 15 and 15 ′ , the buses under fault are no longer part of any loop. This leads to the nearest relay being the starting point for relay co-ordination. As can be seen in Table 8, after splitting of busbars the t op for relay R4 has reduced from 1.204 s (before splitting) to 0.312 s and for relay R14, it is reduced from 1.52 s (before splitting) to 0.313 s. Also, relays R7 and R11 do not see any fault current.

Conclusion
An adaptive relay co-ordination scheme is mainly focused on reducing the operating time of the relays based on the busbar splitting scheme applicable for SIPS. The scheme is based upon the short-circuit level of the network. Relay co-ordination is tested for the IEEE14-bus and 30-bus systems considering different scenarios such as single LG fault, single LLL fault, and multiple LLL faults. The