Directional element for faulty feeder identiﬁcation of high-resistance fault in high-surety power supply systems

The high-surety power supply systems are gaining great attention to enhance the reliability of uninterruptable power supplies. A high-resistance fault along a high-surety power supply feeder results in a low fault current, making the conventional high-surety power supply protection strategy ineffective. To address this problem, this paper develops a directional fault protection strategy for high-resistance fault detection and faulty feeder identiﬁcation. Using the intelligent electronic device, the feeder current is sampled and normalised. Then, the fault-imposed component of the current signal is calculated. This component is added to the input of the forced Helmholtz oscillator to increase the sensitivity of the proposed protection scheme for the detection of high-resistance faults. The output of the forced Helmholtz oscillator equation is adopted as the fault detection criterion because it is inﬁnity for reverse faults while it is lower than 1 for forward faults, facilitating the fault detection. The developed strategy is local and can detect and classify both pole-to-ground and pole-to-pole high-resistance faults. Also, it is effective for both unidirectional and bidirectional converters. The merits of the proposed protection strategy are demonstrated through several fault scenarios using a ± 375 V high-surety power supply system.


Motivation
Critical loads such as hospitals, banks, and internet data centres require continuous and high-quality power during normal and faulty conditions [1,2]. To this end, uninterruptable power supply (UPS) systems have been used for many years to make sure of continuous energy supply. To enhance the reliability of UPS systems, the concept of high-surety power supply (HSPS), also know as super UPS, has been recently introduced. An HSPS system is formed by several energy source and energy storage units such as photovoltaic (PV), battery, gas turbine, and super capacitor [3][4][5][6]. Certain attributes such as no harmonic and reactive power problems, no need for frequency control and complicated synchronisation, and low stages of power conversion make the DC network a proper selection for implementing the HSPS systems [7][8][9][10]]. An HSPS system can operate in both grid-connected and islanded modes to ensure the continuous energy supply. The lack of effective protection is one of the main barriers to the deployment of super UPS systems, jeopardising their offered resilience. Both pole-to-ground (PG) and pole-to-pole (PP) faults may occur in an HSPS system, by a pole coming in contact with the ground and another pole, respectively. DC nature of super UPS system makes its protection more challenging due to the following specific features: (i) in a DC network, there is no zero-crossing point in fault current, (ii) the capacitor discharge highly increases the magnitude and rate of change of fault current, (iii) the lines of DC networks are usually short, making the faulty feeder identification difficult.

Literature review
Due to the need for high levels of reliability, speed, and selectivity, usually, the protection strategies proposed for DC microgrids are not effective in an HSPS system as follows.
The current derivative-based protection schemes proposed in [11,12] are susceptible to noise. The overcurrent protection FIGURE 1 Study HSPS system strategies proposed in [13,14] cannot provide proper selectivity or require extra devices. The distance protection schemes proposed in [15,16] suffer from low operating speed. The directional protection schemes proposed in [17][18][19][20][21][22] and differential protection methods proposed in [23][24][25][26] as the communication strategies increase selectivity; however, dependence on communication infrastructure decreases their reliability, making them improper for a high reliable power supply. To address these shortcomings, only one HSPS protection scheme is proposed; Li et al. [2] present a local current direction-based protection strategy for the HSPS systems. However, the conventional HSPS protection suffers from two problems: (1) it does not consider the dynamics of primary energy sources of HSPS system and (2) in the case of a high-resistance fault, the current direction does not change and consequently the current direction-based protection strategy fails.

Aims and contributions
To address these problems, this paper proposes a protection scheme for the detection of high-resistance faults in super UPS systems. In the proposed local protection scheme that is a supplementary protection for the conventional HSPS protection, the current of each feeder is sampled using an intelligent electronic device (IED). Then, the fault-imposed component of the current is calculated to detect the direction change of high-resistance fault current. To increase the sensitivity of the proposed method, this component is added to the input of the forced Helmholtz oscillator (FHO). The advantage of the proposed directional protection scheme is that the FHO state variable is infinity for the faulty feeder while it has a finite value for the healthy one. Specifically, the contributions of this paper with respect to [2] are as follows: • It considers the dynamics of primary energy sources of the HSPS system; • It is effective for high-resistance faults; and • It classifies the fault condition.

Paper organisation
The rest of the paper is organised as follows. Section 2 describes the dynamic model, control system, and conventional protection scheme of the study HSPS system. The proposed two-stage protection strategy is developed in Section 3. Section 4 is dedicated to assessing the performance of the developed scheme. Finally, Section 5 concludes the paper. Figure 1 shows the single diagram of the study HSPS system that is the extended version of the test system of [2]. It is a bipolar (three-wire) network with the TN-S earthing system where the midpoint of the converter is connected to the ground. The operating voltage of the study HSPS is ±375 V. The resistance and inductance of the HSPS feeders are 4.8 mΩ and 3.8 H, respectively. A 100 kW voltage-sourced inverter (VSI) is employed as the interface between the HSPS system and the main AC grid. The study HSPS system consists of five distributed energy resource (DER) units including gas turbine, energy storage system, PV array, supercapacitor, and fuel cell. The critical load is a three-phase, 220 V, 60 Hz, 230 kW, and 150 kVAr load that is interfaced with the HSPS system using a VSI. Also, a bypass switch is employed to directly connect the critical load to the AC main grid in the critical conditions.

Dynamic modelling
The main grid is a 25 kV distribution system that is fed by a 120 kV transmission network with 2500 MVA short-circuit capacity.  [14]. The adopted battery type is lithium-ion with the discharge and charge models. The supercapacitor is modelled using the Stern equation by considering the self-discharge phenomenon. The gas turbine is equipped with the IEEE type 1 excitation system. The PV array and fuel cell are interfaced to the HSPS system using the DC-DC boost converters, while the interfaces of the battery and supercapacitor are the bidirectional DC-DC converters. The generated AC electricity of the synchronous generator of the gas turbine is converted to DC using a rectifier; then, a buck converter is used to interface this generation system to the HSPS system.

Control system
The control system of grid VSI is a two-loop control system. The objective of the external control loop is the regulation of DC link voltage while the inner control loop regulates the grid current real and reactive components. According to [2], each DER unit can operate in both voltage and current control modes. The control system of the DC-DC converters of DERs is shown in Figure 2. It consists of two control loops. The outer loop regulates the converter output voltage where the

Conventional current-based directional protection scheme
The basic principle of the directional protection proposed in [2] is the change of current direction in the case of a fault. In that method, the current is measured at the beginning of each feeder with a positive sign for the injected current to the HSPS system. When a fault occurs in one feeder, the main grid and DER units of other feeders contribute in fault current; thus, their current directions do not change and are positive, while for the faulty feeder, the current direction at the beginning of the feeder changes and is negative. There are two thresholds: one for negative currents and one for positive currents where the latter threshold is greater than the former one. If the measured current is negative and greater than the negative threshold, the feeder is faulty and is isolated from the HSPS system. If the primary protection fails, the fault persists in the HSPS system and the positive fault current of healthy feeders exceeds the latter threshold. In this condition, these feeders are isolated from the HSPS system as backup protection. Similarly, when a fault occurs in the main feeder of the HSPS system, all measured fault currents are positive and exceed the second threshold; thus all units are isolated from the HSPS system. In both cases of primary protection failure and main bus fault, the load voltage drops and the load module activates the bypass switch; in this condition, the load is directly supplied by the main grid.

PROPOSED DIRECTIONAL PROTECTION STRATEGY
The main drawback of the conventional protection system of HSPS is its inability to detecting high-resistance faults. When a high resistance fault occurs, the fault current is low and it cannot change the current direction in the faulty feeder. Thus, the fault persists in the HSPS system. Also, the conventional protection lacks a fault classification scheme. The remainder of this section is dedicated to developing a two-stage protection strategy of the HSPS system for addressing these problems.

First stage: Fault-imposed current
In the proposed local protection strategy, the injected current of each feeder to the HSPS system is measured using a current transducer. To reduce the scheme cost, instead of a relay, one IED is installed at the beginning of each feeder. The measurement noises are attenuated by a simple moving average filter (MAF) as [27] i feeder (kT where i meas feeder and i feeder are the measured and filtered feeder currents, respectively, and N is the length of MAF window.
The scalability of the proposed method is enhanced by normalising the filtered current i feeder as where i pu feeder and I base are the normalised feeder current and base current, respectively. The latter current is determined based on the nominal power and nominal voltage of the feeder DER. k and T s are the sampling step and sampling time, respectively.
To address the problem of the low sensitivity of the conventional protection system of the HSPS system, the superposition theorem [28] is used. According to this theory, the current signal consists of normal-running and fault-imposed components. The latter is independent of the former and poses a signature of fault. The superimposed network can be used to calculate the fault-imposed component as where i where T d is the time delay of the Delta filter. Due to the DC nature of the current signal, the fault-imposed component is nearly zero (due to the presence of normal disturbances is not exactly zero) during normal conditions while it is non-zero when a fault occurs,as shown in Figure 3. Thus, even a high-resistance fault occurs, the non-zero value of the faultimposed component can be used as a promising fault detection feature.

Stage 2: Forced Helmholtz oscillator
During a high-resistance fault condition, the fault current is not high enough to strongly increase the fault-imposed component. It may result in a malfunction of the fault-imposed current-based protection strategy during normal disturbances. Consequently, to more increase the sensitivity of the proposed scheme, the FHO is employed as follows.

Operating principle
The second-order differential equation of FHO as an asymmetrical nonlinear oscillator is expressed as [30] where os is the Helmholtz forcing signal frequency. and F are the damping level and forcing amplitude, respectively, and and are the positive constants. By defining the system output y as [30] one can write If F exceeds its critical values F c , the system state changes from normal motion to the chaotic motion. Indeed, F c determines the boundary of stable and unstable modes of FHO.
The promising feature of FHO is that by changing the system state to chaotic motion, its output reaches infinity because the system is unstable. By adding small turbulence, the proposed protection scheme is developed that operates based on the condition of the system state during various conditions. During a fault condition, the FHO output is infinity while during normal operation, its output is non-infinity; it facilitates fault detection.

FHO-based directional element
Using the superimposed component of current, the modified FHO equation is expressed as where F ′ is the modified forcing amplitude and is defined as The IED of each feeder calculates the output of the modified FHO equation. During the normal operation that the fault-imposed current is zero, F ′ = 0.5F c < F c and thus, the system output is in the stable mode. By solving (6) and (7), the maximum value of y during normal conditions is determined equal to 0.60 pu. When a fault occurs, the sign of fault-imposed current determines whether the system motion is normal or chaotic.
Since in an HSPS system, energy storage systems with the bidirectional current flow are available, the pre-fault current in that feeders can be positive or negative. If a DER injects current to the HSPS system, the pre-fault current is positive. When a high-resistance fault occurs in that feeder, the current direction does not change but the current magnitude at the beginning of the feeder decreases. Thus, the fault-imposed component is negative. If a DER is in the charging mode and thus, it absorbs the HSPS energy, the pre-fault current is negative. When a high-resistance fault occurs in that feeder, the current at the beginning of the feeder slightly increases with a negative sign. Therefore, in this case, the fault-imposed component is negative too. It should be noted that for a low-resistance fault, the fault-imposed component is negative in both aforementioned cases. Since the control system of DER units regulates their injected currents, the current of healthy feeders does not change during a fault condition. Thus, the fault-imposed component is nearly zero.
Consequently, for the faulty feeder that i pu feeder,FI is negative, F ′ becomes greater than F c and the maximum of system output of modified FHO equation calculated at beginning of the feeder increases to infinity while its peak at healthy feeders is equal to 0.60 pu because i pu feeder,FI is zero and F ′ < F c . The proposed FHO-based directional element FHO is defined as If FHO = 1, it interprets as a reverse fault condition and the faulty feeder IED sends the trip signal to its associated solidstate circuit breakers (SSCBs) for isolating this zone from the healthy sections of HSPS, as shown in Figure 1.
To classify the fault condition, the proposed directional element is determined in both positive and negative poles. It should be noted that the absorbed current from the HSPS system is measured in the negative pole. If the directional element in both poles + FHO and − FHO are 1, the fault is PP. If only one of these directional elements becomes one, the fault is PG.

Determination of critical forcing amplitude
To determine the boundary of stable and unstable modes of FHO, the Melnikov function M (t 0 ) = 0 is solved. To this end, unperturbed homoclinic coordinates (UHCs) are considered in the steady-state. In this condition, the total energy of the system should be zero. Hamiltonian H (x, y) is an operator that in most cases represents the total energy of the system; thus, in the unperturbed steady-state condition, H (x, y) = 0. The system Hamiltonian is expressed as [30] { dx∕dt = H (x, y)∕ y, On the other hand, since and F are zero in the unperturbed system, (6) and (7) are described in steady-state as [30] { dx∕dt = os y, To determine the homoclinic coordinates, the following equation can be acquired by substituting (11) in (12) and solving the resulting equation [30]: The homoclinic orbits q + 0 and q − 0 are determined by solving (13) for x and y as a function of t as [30] . (15) Thus, the UHCs are expressed as [30] x 0 (t ) = ± 3 2 sec 2 Using the Melnikov function, one can write [30] Solving (17) determines Melnikov function as [30] M (t 0 , , , , F, os ) = Since | cos(⋅)| is smaller than 1, one can write If (20) satisfies, FHO enters the unstable mode and its output reaches to infinity. Considering = = 1, the critical value of F as the boundary of stable and unstable modes of FHO is expressed as Considering = 0.5 and os = 1 pu, F c is equal to 0.37. Thus, F > 0.37 results in chaotic motion. Figure 4 shows the flowchart of the developed directional protection scheme. In the first stage, the noises of the measured current of each feeder are removed using the MAF. Then, the filtered current signal is normalised and the fault-imposed component of the feeder current is calculated by the IED. In the second stage, the output of the modified FHO equation is calculated by the feeder IED for both positive and negative poles. Based on this output, the proposed directional element is determined. If FHO in a feeder becomes one, the feeder is faulty and the IED sends the trip command to the SSCB of that feeder. Based on the values of the proposed directional element in positive and negative poles, the fault type is determined.

Practical implementation
IED is a key component of smart systems technology, facilitating the exchange of operational (status points and instantaneous values) and non-operational (waveforms and files) data [31]. According to IEC TS 61850-2, it consists of the processor(s) and can send (receive) data/control to (from) another electronic device such as meters and controllers. The architecture of an IED is shown in Figure 5. It has various functions such as protection, control and logics, metering and power quality analysis, and monitoring [32]. After measuring the analog current signals of both poles, they are converted to digital signals by the A/D converter. The discrete signals are processed and the proposed directional elements are calculated that have binary values of either 0 (no/forward fault) or 1 (reverse fault), based on (10). On the other hand, IED is equipped with a self-monitoring software that this watchdog (WD) service disables all IED functions in the case of an internal error (zero output of WD); WD signal is 1 when IED properly works. According to the logical diagram of the proposed method that is shown in Figure 5, if WD signal is 1 and a fault condition is detected in positive pole, the positive pole circuit breaker (PPCB) signal will be 1 and the trip command is sent to the SSCB of positive pole. Also, in the case of a detected negative pole fault while IED has no internal error, the negative pole circuit breaker (NPCB) signal will be 1. In the case of a detected PP fault, both PPCB and NPCB signals will be 1 provided that WD signal is 1.

PERFORMANCE EVALUATION
This section is dedicated to evaluating the effectiveness of the proposed protection strategy. To this end, the study HSPS system of Figure 1

Case 1: PG high-resistance fault in unidirectional converter
The first case study is dedicated to investigating the developed protection scheme for a PG high-resistance fault. To this end, a negative pole to ground high-resistance fault with fault FIGURE 5 Implementation of the proposed method in the IED

FIGURE 6
Feeder current and its normalised fault-imposed component for a PG high-resistance fault at gas turbine feeder resistance of 5 Ω is simulated at gas turbine feeder at t = 2 s. Figure 6 shows the feeder current and its normalised faultimposed component during this fault condition. When the fault occurs, the current amplitude reduces from 66.67 to 20.25 A. As the current direction does not change, the conventional protection system of HSPS fails to detect this fault condition. However, the negative value of i pu feeder,FI increases F ′ c based on (9), resulting in system becomes unstable. Figure 7 shows the output of the FHO equation calculated by IEDs of HSPS feeders for both positive and negative poles. For the healthy feeders, this output is lower than 1; thus, they are properly interpreted as healthy feeders. However, for the IED of gas turbine feeder (IED2), the FHO output for negative pole is infinity while this output for positive pole is lower than 1. Therefore, IED2 detects a reverse fault and sends the trip command to the FIGURE 7 FHO output for study HSPS system during a PG highresistance fault at gas turbine feeder FIGURE 8 Phase portraits of FHO for study HSPS system during a PG high-resistance fault at gas turbine feeder associated SSCB. The fault detection time is about 40 ms; this relatively long time for a DC network fault detection does not result in any problem because a high-resistance fault has a limited current amplitude and rate of change, and the fault current is not destructive. Thus, a short fault detection time is not a major priority for a high-resistance fault detection scheme. On the other hand, using advanced routing, the processing delay can be minimised [33] and is in the order of several microseconds that can not affect the performance of the proposed scheme. The fault condition is classified as a negative pole to ground fault. The phase portraits of FHO for all feeders are shown in Figure 8. The system state is the normal motion for all healthy feeders while it is the chaotic motion for the negative pole of IED2.

Case 2: PG High-resistance fault in bidirectional converter feeder
To evaluate the performance of the proposed method in the case of a high-resistance fault in feeders with bidirectional converters, a PG fault (negative pole to ground) is simulated in the battery feeder. Unlike other HSPS sources, battery is in charging mode and the pre-fault current is negative. The fault resistance is 10 Ω and the fault is initiated at t = 2 s. The results are shown in Figure 9. The FHO outputs calculated by IEDs of healthy feeders are lower than 1 and the system states are stable in both poles. Also, the system state for positive pole of IED5 is stable while it is unstable in the negative pole of battery feeder and y − IED5 increases to ∞. Thus, − FHO at battery feeder becomes 1. The fault detection time is about 40 ms. It verifies the effectiveness of the developed protection strategy for both unidirectional and bidirectional DC-DC converters.

Case 3: PP High-resistance fault
Although the high-resistance faults are usually pole to ground, the objective of this scenario is to evaluate the performance of the developed protection scheme in the case of a PP fault. For this purpose, a PP fault is simulated at the supercapacitor feeder with the fault resistance of 50 Ω. Figure 10 shows the FHO output calculated by IEDs of the study HSPS system. The system states for the healthy feeders are stable and y IED < 1. While for both positive and negative poles of supercapacitor feeder, the system states are unstable and y + IED4 and y − IED4 increase to infinity. Thus, the proposed method properly classifies this fault condition as a PP fault. Thus, a reverse fault condition is detected by IED4 and it sends the trip command

Case 4: Parameter uncertainty
To investigate the effect of parameter uncertainty on the performance of the proposed protection strategy, a PP fault with fault resistance of 20 Ω is simulated at the PV feeder. The uncertainty is considered in solar irradiance, environmental temperature, and feeder impedance. To this end, solar irradiance and temperature are decreased by 50% and 40%, respectively, while the feeder impedances are increased by 15%. The outputs of the FHO equation calculated by the IEDs of HSPS feeders are shown in Figure 11. The maximum of FHO outputs for healthy feeders are below 1 and based on (10), FHO = 0. However, at t = 2.03 s, y IED3 in both poles increases to infinity and the fault condition is detected. The results verify that the change of the network parameter does not affect the proper performance of the proposed strategy.

CONCLUSION
The conventional protection system of the HSPS system suffers from ineffectiveness for high-resistance faults, where the fault current direction does not change. This paper has proposed a supplementary directional protection strategy to address this problem. The developed high-resistance fault detection strategy relies on one IED at the beginning of each feeder, sampling the feeder current. By calculating the fault-imposed component of the feeder current, a reverse fault can be discriminated from a forward fault because in the former case, the fault-imposed current is negative while it is zero for the latter case. To increase the sensitivity of the proposed strategy, Extracting the unique features of high-resistance DC faults for very high-resistance fault detection can be considered as future work.