Reliable Impedance Islanding Detection for Power Distribution Systems with High Motor Penetration Nguyen Duc Tuyen, Goro Fujita Shibaura Institute of Technology, 3-7-5, Toyosu, Koto-ku, Tokyo, Japan Email: m609504@shibaura-it.ac.jp Abstract— Regarding to dispersed generation field, an islanding is known as a situation that those generators are disconnected from the utility feeders and a bunch of local loads being powered by them is isolated. Accordingly, anti-islanding schemes are used to immediately detect that phenomenon in facilitating the detachment the DG or adaptive controlling of autonomous operation. Negative-sequence impedance measurement is considered as a novel method of detecting electrical islands in DG networks. However, in the case of high induction motor penetration, this method might malfunction. Using two-round impedance measurement, this paper deals with that problem while making typical assessments on how the impedance tolerant values are selected and how these values vary with network configuration. The simulation in PSCAD/EMTDC and experiments validate the methodology. Keyword Islanding Detection, Impedance Measurement, Symmetrical Transformation, Clarke Transformation. I. I NTRODUCTION Recent high DG penetration into power distribution systems has created new challenges for protection engineers. One typical protection configuration that should be considered is when an open circuit condition causes the creation of an island of electrical load and generation. In 2003, the IEEE 1547 Standard[1] has the requirements for DG connection to detect and separate itself from the network within two seconds since islanding operation may damage existing equipment, human safety, power reliability, power quality and so forth. However, in 2011 July, IEEE Standard 1547.4[2] facilitates of islanded operation provisioning of multi-islanded operation in required cases and reconnected as smart grid of the future. Research on islanding detection (ID) using impedance measurement was reviewed in 2004 [3], and implemented in 2008 by Wrinch et al.[4]. However, this approach has not investigated thoroughly in the system with high motor penetration which probably makes it in a wrong operation. This paper outlines the ID technique and alleviates the disadvantages. II. I SLANDING D ETECTION BASED ON N EGATIVE - SEQUENCE I MPEDANCE M EASUREMENT FOR F IRST ROUND The ID method developed in this paper uses the theoretically accurate concept of impedance measurement when the negative-sequence components of voltage and current are available. This detection idea is shown in Fig.1. The premise is that the large impedance difference between an island and a grid-connected condition has a very low Non Detection Zone Z DG Z utility Z load E utility E DG Fig. 1. Principle of using negative-sequence impedance measurement for islanding detection Z 0sys Z 0load V0 Z -sys Z -load V- E sys Z +sys I0 I- V m I+ Z mload I0 ! I ! I+ Z +load V+ Fig. 2. Symmetrical-component currents flow in unbalanced-system expanded circuit (NDZ) for radial systems with stiff network connections, that is Z utility << (Z DG \ \Z load ). The unbalanced system is transformed into the symmetrical system described in Fig.2[5](subscript +, −, 0 represents for positive-, negative- and zero-sequence components). In this sys figure, the system equivalent voltage source, E+ remains sys sys as the only source and the current from the Z0 and Z− sys sys sequences are indeed supplied by the E+ . E+ is used because the normal resources are considered to generate only positive-sequence components. From the Fig.2, a current source that is produced at the load for negative- and zerosequence is a result of I+ being divided up at node Vm . Vm is the point where three symmetrical currents join together and m represents mutual components. From the current and voltage derivation in symmetrical components with this configuration, the current flows out of the unbalanced load and goes into the system. By using Ohm’s law for negative-sequence components at the point of the unbalanced load or source, the system sys Thevenin negative-sequence impedance of system, Z− can be calculated as in Eq.(1)[4]. 0.060 (1) Normal Operation Z +sys 0.020 III. P OSITIVE S EQUENCE I MPEDANCE M EASUREMENT FOR S ECOND ROUND Z DG + E sys As mentioned in the previous section, high motor penetration (around 60% of the utility load) poses limitation to the reliability of negative-sequence impedance ID. The following technique could be called the second round of ID based on positive-sequence impedance measurement. However, because of the second round is implemented simultaneously, then it does not increase detection time. After the first round, the system negative-impedance of post-islanding has been obtained and it is considered to be smaller to the positivesequence because of the assuming appearance of motors. When an islanding occurs, besides system negative-sequence impedance increases, the positive-sequence impedance also increases. On the other hand, while separating a bunch of motors, only negative-sequence impedance of system increases because of the difference of negative- and positive-sequence of motors themselves. If system positive-sequence impedance is measured during the transient, those different phenomena could be clarified. Clarke transformation is recommended for an impedance calculation. It provides more data points for results verification and it can determine impedance without requiring complete three-phase voltages. And real operators can be applied to instantaneous voltages and currents. Furthermore, excessive complex number operations tend to introduce errors to the voltage and current phase angles. The Clarke transformation has the following form: (2) The basic idea of this method is in Fig.3. During the normal operation, due to connecting to stiff grids, the DG unit components apparently are integrated in the entire system of which impedance is quite small. At islanded operation, there is only DG units with its high impedance. Contrarily, while disconnecting a motor, those positive-sequence components of system is kept connecting. System positive-sequence impedances can be determined using the following steps: V+s , I +s Z +load Motor Disconnecting Z +sys B. Clarke Transformation to Measure Positive Impedance Z +load -0.020 -0.060 The positive-sequence impedance of induction motors varies between and around 0.9 to 1.0 p.u, while the negative-sequence impedance remains essentially constant at approximately 0.15 p.u. This means the low negative-sequence impedance somehow creates confusion for the threshold window of first round of ID method because they can indicate a falsely stiff utility. V+s , I +s Islanding E DG 0.000 -0.040 A. Sequence Components for Induction Motors 2 √ −1 −1 V̇A V̇α √ V̇B V̇β = √1 0 √3 −√ 3 √ 6 2 V̇C V̇ i h 2 i2 h0 → V̇αβ0 = [T ] V̇ABC E sys Ibeta 0.040 y sys Z− V− ≈− I− Vbeta 1st cycle V+s , I +s Z +load 6th cycle Transient Fig. 3. Observation window to measure positive-sequence impedance 1) Use the “observation window” that can gather voltage and current values every 6 cycles as Fig.3. Record the waveforms of the voltages and currents at the chosen PoM as ”1st-transient” parameters (V̇s1st , I˙s1st -s represent α, β and 0 components) 2) Record the ”2nd-transient” values (V̇s2nd , I˙s2nd ) in 6 cycles after the 1st one. For the system impedance, the following system equations can be developed: V̇s1st = Ėssys − I˙s1st Żssys V̇s2nd = Ėssys − I˙s2nd Żssys (3) where Ėssys and Żssys are the internal system’s positivesequence voltage and system impedance, respectively. 3) Determine the system α − β − 0 impedances based on the Eq.(3) using the Eq. (4): sys Ż+s = 1st 2nd V̇+s − V̇+s I˙2st − I˙1nd +s (4) +s 4) Assuming again from the first round that the system impedance to be measured is balanced or symmetrical. The Clarke transformation will yield: V̇α I˙α i h sys V̇β = [T ] Ż+ T −1 I˙β V̇ I˙0 0 (5) Żs 0 0 I˙α = 0 Żs 0 I˙β 0 0 Żs I˙0 As a result, computing either Żα or Żβ is equivalent to computing positive-sequence component, Ż+ . IV. E VALUATE R ESULTS FROM S IMULATION The proposed ID technique was tested on IEEE 13-node system. A. Islanding The islanding phenomenon is by opening an utility circuit breaker at the 0.4 instance. Fig.4 shows step changes of negative-sequence impedance during an islanding and a motor disconnecting. Before islanding, the negative-sequence impedance was small and the system’s positive-sequence Beta Current and Voltage during an motor disconnecting transient Negative-sequence impedance of system during transients 6 Voltage[kV]-Current[kA] Impedance[Ohm] 60 40 20 Islanding Motor Disconnecting 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 0 -4 Fig. 7. 1 Impedance [Ohm] Voltage[kV]-Current[kA] 1st cycle th 6 cycle 0 -1 Voltage Vbeta Current Ibeta 0.38 0.4 0.42 0.44 6th cycle 0.38 0.4 0.42 0.46 0.48 0.5 0.52 Impedance [Ohm] Zsys + 200 Zsys mean + 100 0 6 8 10 12 14 16 18 20 Value number Fig. 6. Zsys mean + 1000 -500 -2000 -3500 2 4 6 8 10 12 14 16 18 20 Positive-impedance of system in case of motor disconnecting as more than 20[Ω] after islanding. The second round demonsys strates that the island’s positive-sequence impedance, Z+ was bigger than the negative-sequence one. This is understandable since the motor is still connected after the islanding. In comparison to the low system negative-sequence impedance measured by the first round (around 4[Ω] as in Fig.4), this step change is the evidence of an islanding phenomenon. B. Induction Motor Disconnecting The motor disconnecting is at the 0.4-instance as in Fig.4. The negative-sequence impedance change is also significant as of the case of an islanding. For the second round, positivesequence impedance is obtained in the similar way as described in previous subsection. And the calculated data is shown on Fig.8. Consequently, the data in Table II are based on the current and voltage waveforms of Vβ and Iβ shown in Fig.7. Subsequent to the transient, the positive-sequence impedance apparently is remained in small values. Table II synthesizes the entire statistics. sys The first round gave the negative-sequence impedance, Z− as more than 36[Ω] after islanding. The second round shows sys the system’s positive-sequence impedance, Z+ is still small (8.12[Ω]) which is supposed to be higher in the case of islanding. This value is still larger than the system’s negativesequence impedance before opening the motor which is shown in Fig.4. Therefore, the difference in those impedances of both rounds indicates that the islanding is not actually occurring. Positive-impedance of system in islanding case TABLE II AND Iβ FROM OBSERVATION WINDOW IN CASE OF ISLANDING t 0.41 0.51 0.54 Zsys + TABLE I Vβ 0.52 Value number Fig. 8. Vβ and Iβ in case of islanding 300 4 0.5 2500 -5000 400 2 0.48 5000 4000 0.54 impedance was a bit higher because of the connecting motor. Subsequent to islanding, negative-sequence impedance increases to much higher values. That quick change brings promise to the speed of detection. In the second round of impedance measurement, V̇β and I˙β waveforms are shown in Fig.5. From these waveforms, the parameter values of Clarke transformation are extracted by using an observation window. The investigated discrete values of voltage V̇β1st in the instance range [0.41 ÷ 0.43] with the step 0.001 and after six cycles of the 60Hz-system, those values of V̇β2nd in the instance range [0.51 ÷ 0.53]. The corresponding discrete values of current I˙β1st and I˙β2nd at the similar instance are also obtained. As a result, there are 21 impedance values to get the average positive-sequence impedance. Then the calculated data is shown on Fig.6. Then, the synthesized results is in Table I. sys The first round gave the negative-sequence impedance, Z− -100 0.46 Vβ and Iβ in case of motor disconnecting Time [s] Fig. 5. 0.44 Time [s] Beta Current and Voltage during an islanding transient -2 0.36 1st cycle -2 Negative-sequence impedance of system 2 Current Ibeta 2 -6 0.36 Time [s] Fig. 4. Voltage Vbeta 4 Vβ 1.4180 0.4708 Iβ 0.0283 0.0049 sys Z+ sys Z− 42.6843 20.0221 Vβ AND Iβ FROM OBSERVATION WINDOW IN CASE OF OPENING MOTOR t 0.41 0.51 Vβ 4.4782 4.8045 Iβ 0.0733 0.0776 sys Z+ sys Z− 8.1185 36.7681 3000 1 Zdiscrete positive 2 Esitmated Z+ [Ohm] 1 2 Zdiscrete mean positive 2000 MeanValue:163.06 1000 0 -1000 V,I -2000 Fig. 11. Fig. 9. 0 5 Esitmated Z+ [Ohm] Negative-sequence Impedance [Ohm] 15 20 25 30 35 40 45 50 Positive-sequence impedance in islanding by experiment Configuration of experiment system 50 0.5 Z- Islanding 0.45 10 Z- Motor Disconnecting 0.4 Zdiscrete positive Zdiscrete mean 40 positive 30 20 Mean Value: 9.08 10 0.35 0 0.3 0.25 Circuit Breakers open 0.2 0.02 0.04 0.06 0.08 0.1 0.12 0 5 10 15 20 25 30 35 40 45 50 Fig. 12. Positive-sequence impedance in motor disconnecting by experiment 0.14 Time Fig. 10. Negative-sequence impedance of experimental model V. E VALUATE R ESULTS FROM E XPERIMENT The three-phase experiment system is set up as described in Fig.9. At first, the utility supply power for two IMs. Simultaneously, the synchronous DG is initiated. Using the synchroscope, the DG unit is synchronized to supply a part of two IMs’ power demand. Subsequently, the experiment is focused on opening CB1 and CB2 imitating an islanding and motor disconnecting, respectively. The negative-sequence impedances in two events are shown on Fig.10. Initial conditions of two events are a bit different from each other which make their negativesequence impedance are not tightly identical. Both curves show negative-sequence impedances step up quickly. Those magnitudes keep fluctuating during operation. Fortunately, their mean value can be determined. It should be noted that magnitude of the negative-sequence impedance is too small because the measured negative-sequence voltage is much smaller compare to measured negative-sequence current. During gridconnected operation, the magnitude is about 0.3[Ω], then in the case of islanding it increases to 0.37[Ω]. In the case of motor disconnecting, it comes up with higher values which is about 0.45[Ω] and the negative-sequence impedance gets less fluctuation than the islanding case does due to the connection to stiff grid. This phenomenon affirms some disconnection happened, then in order to distinguish the difference between two events, the proposed second round need to be presented. Under the same data processing as the previous simulation analysis, the estimated positive-sequence impedance in the islanding case is shown in Fig.11. In this case, for a more precise calculation, 55 discrete points subsequent to islanding (6-cycle distance) are synthesized to get the mean value of estimated positive-sequence impedance. Estimated positivesequence impedance of system in the motor disconnecting case is presented in Fig.12. Those figures match the proposed ideas. In the case of islanding, because of disconnection of stiff utility, the system positive-sequence impedance reaches a very high subsequent value compared to grid-connected mode’s. On the contrary, in the case of motor disconnecting, the positivesequence impedance stays at small value which is due to the parallel connection with the stiff utility. Those values can be predetermined accurately based on through analysis of system power flow. And the load of islanded entity should be considered seriously. VI. C ONCLUSION A method of detecting electrical islands in DG networks with high motor penetration was presented. In this method, the first round measures negative-sequence impedance of distribution system while at the second round, only the positivesequence impedance is measured during transient periods. The theoretical and experimental analysis with promising characteristics has been proven. R EFERENCES [1] IEEE-SA-Standards-Board, IEEE STANDARD for interconnecting distributed resources with electric power systems, IEEEStd 1547 Std., 2003. [2] I. SCC21, IEEE Guide for Design, Operation, and Integration of Distributed Resource Island Systems with Electric Power Systems, IEEEStd 1547.4 Std., 2011. [3] S. Xu Wilsun, K.Mauch, “An assessment of distributed generation islanding detection methods and issues for Canada,” Natural Resources Canada, Tech. Rep., 2004. [4] J. Wrinch and M.Nagpal, “Negative sequence impedance based islanding detection for distributed generation (NSIID),” in Electric Power Conference EPEC 2008. IEEE Canada, 2008, pp. 1–6. [5] M. C.Wrinch, “Negative sequence impedance measurement for distributed generator islanding detection,” Ph.D. dissertation, niversity of British Columbia, 2008.