FAST FAULT DETECTOR

Abstract

A system and method for estimating fault current using a sliding observation window that is shorter than one cycle. The method may also include providing a pickup level that defines a current threshold for opening a switch in response to detecting the fault current; estimating the fault current from current measurement signals; accumulating time from a reset zero position during the time that the estimation of the fault current is greater than the pickup level; subtracting time from the accumulated time during the time that the estimation of the fault current is less than the pickup level after time has been accumulated from when the estimation of the fault current is greater than the pickup level; detecting the fault current if the accumulated time reaches a predetermined accumulation time; and opening the switch if the fault current is detected.

Claims

1. A method for estimating current in a power system, the method comprising: providing a pickup level that defines a current threshold for opening a switch in response to overcurrent; and estimating current in the power system from current measurement signals, wherein estimating the current is performed using a sliding observation window that is shorter than one cycle.

2. The method according to claim 1 further comprising: accumulating time from a reset zero position during the time that the estimation of the current is greater than the pickup level; subtracting time from the accumulated time during the time that the estimation of the current is less than the pickup level after time has been accumulated from when the estimation of the current is greater than the pickup level; detecting the overcurrent if the accumulated time reaches a predetermined accumulation time; and opening the switch if the overcurrent is detected.

3. The method according to claim 2 wherein subtracting time from the accumulated time includes stopping subtracting time if the accumulated time reaches the reset zero position.

4. The method according to claim 2 further comprising providing a no-delay threshold that defines a current threshold above the pickup level, wherein detecting the overcurrent includes immediately detecting the overcurrent if the estimated current goes above the no-delay threshold.

5. The method according to claim 4 wherein the no-delay threshold is in the range of 1% - 100% greater than the pickup level.

6. The method according to claim 1 wherein the switch is part of a recloser.

7. The method according to claim 6 wherein the recloser employs pulse closing technologies.

8. The method according to claim 1 wherein estimating current in the power system includes using observation windows that are about one-quarter of a current cycle.

9. The method according to claim 1 wherein estimating current in the power system includes using the time derivative of the current.

10. The method according to claim 1 wherein estimating current in the power system includes modeling the current as: $\tilde{x}\left(k\right)=Acos\left(2\pi \text{}Fk\right)+Bsin\left(2\pi \text{}Fk\right),$ where x̃(k) is the current waveform, F is the ratio of system frequency and sampling frequency, k is sample time, and A and B are unknown amplitudes.

11. A method for estimating current in a power system, the method comprising: providing a pickup level that defines a current threshold for opening a switch in response to the overcurrent; estimating current in the power system from current measurement signals; accumulating time from a reset zero position during the time that the estimation of the current is greater than the pickup level; subtracting time from the accumulated time during the time that the estimation of the current is less than the pickup level after time has been accumulated from when the estimation of the current is greater than the pickup level; detecting the overcurrent if the accumulated time reaches a predetermined accumulation time; and opening the switch if the overcurrent is detected.

12. The method according to claim 11 wherein subtracting time from the accumulated time includes stopping subtracting time if the accumulated time reaches the reset zero position.

13. The method according to claim 11 further comprising providing a no-delay threshold that defines a current threshold above the pickup level, wherein detecting the overcurrent includes immediately detecting the overcurrent if the estimated current goes above the no-delay threshold.

14. The method according to claim 13 wherein the no-delay threshold is in the range of 1% - 100% greater than the pickup level.

15. The method according to claim 11 wherein estimating current in the power system includes using a sliding observation window that is shorter than one cycle.

16. A system for protecting a power system from overcurrent, the system comprising: means for providing a pickup level that defines a current threshold for opening a switch in response to the overcurrent; and means for estimating current in the power system from current measurement signals, wherein the means for estimating the current uses a sliding observation window that is shorter than one cycle.

17. The system according to claim 16 further comprising: means for accumulating time from a reset zero position during the time that the estimation of the current is greater than the pickup level; means for subtracting time from the accumulated time during the time that the estimation of the current is less than the pickup level after time has been accumulated from when the estimation of the current is greater than the pickup level; means for detecting the overcurrent if the accumulated time reaches a predetermined accumulation time; and means for opening the switch if the overcurrent is detected.

18. The system according to claim 17 wherein the means for subtracting time from the accumulated time stops subtracting time if the accumulated time reaches the reset zero position.

19. The system according to claim 17 further comprising means for providing a no-delay threshold that defines a current threshold above the pickup level, wherein the means for detecting the overcurrent immediately detects the overcurrent if the estimated current goes above the no-delay threshold.

20. The system according to claim 17 wherein the means for estimating current in the power system uses observation windows that are about one-quarter of a current cycle.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0011] FIG. 1 is a simplified schematic illustration of an electrical power distribution network;

[0012] FIG. 2 is a simplified block diagram of a recloser used in the network shown in FIG. 1;

[0013] FIG. 3 is a graph with time on the horizontal axis and current magnitude on the vertical axis showing start of a fault;

[0014] FIG. 4 is a timing disc turning in a positive direction for accumulating time; and

[0015] FIG. 5 is a timing disc turning in a negative direction for removing time.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] The following discussion of the embodiments of the disclosure directed to a system and method for estimating fault current using a sliding observation window that is shorter than one cycle is merely exemplary in nature, and is in no way intended to limit the invention or its applications or uses.

[0017] FIG. 1 is a schematic type diagram of an electrical power distribution network 10 including an electrical substation 12 that steps down high voltage power on a high voltage power line (not shown) to medium voltage power, such as 12.47 kV, provided on a substation bus 14. A three-phase feeder 16 is connected to the bus 14 and a recloser 18 is provided proximate to the connection point between the feeder 16 and the bus 14. The recloser 18 is intended to represent any reclosing or fault interrupting device suitable for the purposes described herein. FIG. 2 is a simplified block diagram of a representative example of the recloser 18, and includes a vacuum interrupter 20 for opening and closing the recloser 18 to allow or prevent current flow therethrough on the feeder 16, sensors 22 for measuring the current and/or voltage of the power signal propagating on the feeder 16, a controller 24 for processing the measurement signals and controlling the position of the vacuum interrupter 20, and a transceiver 26 for transmitting data and messages to a control facility (not shown) and/or to other reclosers and components in the network 10.

[0018] The network 10 includes single-phase lateral lines 30 coupled to the feeder 16 usually at a utility pole 32 and a secondary service lines 34 coupled to each lateral line 30 usually at a utility pole 36, where a lateral fuse 38 is provided at the connection point between each lateral line 30 and the feeder 16 and a primary fuse 40 is provided at the connection point between each lateral line 30 and each service line 34. A distribution transformer 42 is provided at the beginning of each service line 34 that steps down the voltage from the medium voltage to a low voltage to be provided to loads 44, such as homes.

[0019] This disclosure proposes a current estimator and an overcurrent detection and protection algorithm operating, for example, in the controller 24 that is similar to the traditional 1-cycle DFT current estimation method that may use the time derivative of the current waveform or the current waveform. The estimator uses a parametric method, which means that the observed waveform is modeled, and the parameters of the model are estimated that best fit the model to the observed waveform. The algorithm uses a sliding observation window that is shorter than one cycle, such as ¼-cycle, which may cause the estimator to produce larger overshoots, such as greater than 10%. The amount of overshoot depends on the validity of simplifying assumptions, such as a large X/R ratio, where the X/R ratio is the ratio of total inductive impedance X (that comes from that net inductance L) to the total resistive impedance R along the fault path, and may thus trigger false fault detections at current levels lower than the predetermined pickup level. In order to reduce the false detections, the detection algorithm employs two mechanisms, namely, a time-delay qualifier and a no-delay threshold that are described in detail below. It is noted that the time-delay mechanism described herein is not used for purposes of coordination in the context of this disclosure, which typically use delays in protection, but for security reasons, to reduce the probability of false detections.

[0020] Assuming that the current in a conductor is i(t), and using the standard model of a fault, the instantaneous current is modeled as:

[00001]$\begin{array}{cc}i\left(t\right)=\sqrt{2}{I}_{rms}\cdot \cos \left(2\pi ft+\theta \right)+A_{DC}\cdot e^{-\frac{t}{T}}+N\left(t\right),& \text{­­­(1)}\end{array}$

where f is the system frequency (50 Hz or 60 Hz),

[00002]$\sqrt{2}{I}_{rms}\cdot \cos \left(2\pi ft+\theta \right)$

is the system power current,

[00003]$A_{DC}\cdot e^{-\frac{t}{T}}$

is the decaying DC component and N(t) represents other nuisance components such as harmonics, switching or other transients, and sensor noise. In general, noise is produced by active components such as A/D converters, amplifier stages in the conditioning circuits, filters and other electronic components. The constant T is the decaying time constant dependent on the X/R ratio of the circuit during the fault. It is assumed at this point that the system frequency f is known a priori. The frequency is assumed to be stable at the system frequency or is tracked independently by an external algorithm.

[0021] If the sensors 22 do not directly measure the current waveform itself, but the time derivative of the current, then the observed waveform di/dt (t) has a form of:

[00004]$\begin{array}{cc}\raisebox{1ex}{$di$}\!\left/ \!\raisebox{-1ex}{$dt$}\right.\left(t\right)=2\pi f\cdot \sqrt{2}{I}_{rms}\cdot \sin \left(2\pi ft+\theta \right)-\frac{A_{DC}}{T}\cdot e^{-\frac{t}{T}}+N^{\prime }\left(t\right).& \text{­­­(2)}\end{array}$

[0022] Note that after differentiation, the form of equation (2) is the same as equation (1) except for the scaling factors, and the nuisance term N(t) is differentiated into N′(t). During a fault, it is reasonable to assume that the nuisance parameter N(t) or N′(t) is negligibly small because the current magnitude is very large. In documenting and illustrating the algorithm in the discussion below, the time derivative of the current waveform is used, but the process applies equally to the waveforms that are the current itself.

[0023] The proposed algorithm estimates the model parameters by processing sampled waveform data, sampled at frequency F.sub.s and denoted from this point onward as

[00005]$x_k\triangleq x(t_k),$

where

[00006]$t_k=\raisebox{1ex}{$k$}\!\left/ \!\raisebox{-1ex}{$F_s$}\right.,$

using a moving window containing N samples. It achieves fast fault detection by operating with a sampling window that is shorter than one full cycle, such as ¼ cycle. For a window that is a fraction of one cycle, it is assumed that the amount of decay in one such processing window is negligibly small. In general, this assumption holds well for large X/R ratios, and worsens as X/R ratios decrease. Much like the effect of decay is considered negligibly small, so is the effect of the average DC offset within the processing window. Taking these assumptions into consideration, it is clear that the simplified model obtains an approximation to the true waveform, which is denoted as x̃(k). Using the definitions introduced above, the final model has the form:

[00007]$\begin{array}{cc}\tilde{x}\left(k\right)=Acos\left(2\pi Fk\right)+Bsin\left(2\pi Fk\right),& \text{­­­(3)}\end{array}$

where the model has two unknown parameters, namely, amplitudes A and B of the cos(.Math.) and sin(.Math.) components.

[0024] Digital frequency F is defined as the ratio of the system frequency and the sampling frequency as F = .sup.ƒ/F.sub.s Once these two components are known, the desired quantities can be computed as:

[00008]$\begin{array}{cc}{I}_{rms}=\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$2\sqrt{2}\pi f$}\right.\sqrt{A^2+B^2},& \text{­­­(4)}\end{array}$

[00009]$\begin{array}{cc}\theta ={\tan }^{-1}\left(\raisebox{1ex}{$B$}\!\left/ \!\raisebox{-1ex}{$A$}\right.\right).& \text{­­­(5)}\end{array}$

[0025] The model in equation (4) can be represented as an N component vector because the moving observation window has N samples, and can thus be written in the matrix-vector form x̃ = Ap. Vector p is a two-component vector such that p.sub.1 = A and p.sub.2 = B. Since the model is an approximation to the true observation .sub.Xk, the relationship between .sub.Xk and x̃(k) can be written as .sub.Xk = x̃(k) + ε(k), where ε(k) represents the unknown errors between the measurement and the model at instance t.sub.k. In vector notation, and using the definition for x̃(k) from above, the relationship becomes:

[00010]$\begin{array}{cc}\stackrel{.fwdarw.}{x}=A\stackrel{.fwdarw.}{p}+\stackrel{.fwdarw.}{\epsilon },& \text{­­­(6)}\end{array}$

where the quantities A and x are:

[00011]$\begin{array}{cc}\stackrel{.fwdarw.}{x}=\left[\begin{array}{c}{x}_k\\ x_{k+1}\\ .Math.\\ x_{k+N-2}\\ x_{k+N-1}\end{array}\right],& \text{­­­(7)}\end{array}$

[00012]$\begin{array}{cc}A=\left[\begin{array}{c}\cos \left(2\pi f{t}_k\right)\\ \cos \left(2\pi f{t}_{k+1}\right)\\ .Math.\\ \cos \left(2\pi f{t}_{k+N-2}\right)\\ \cos \left(2\pi f{t}_{k+N-1}\right)\end{array}\begin{array}{c}\sin \left(2\pi f{t}_k\right)\\ \sin \left(2\pi f{t}_{k+1}\right)\\ .Math.\\ \sin \left(2\pi f{t}_{k+N-2}\right)\\ \sin \left(2\pi f{t}_{k+N-1}\right)\end{array}\right].& \text{­­­(8)}\end{array}$

[0026] From the definition above it follows that x.sub.k is the earliest sample in the current window, and x.sub.k+N-1 is the most recent sample in the window. The optimal vector p, is solved to minimize the total error between x and x̃ as:

[00013]$\begin{array}{cc}\stackrel{.fwdarw.}{p_{\ast }}=\underset{\stackrel{.fwdarw.}{p}}{min}{\displaystyle {.Math.}_{k=0}^{N-1}{\left|\epsilon \left(k\right)\right|}^2}.& \text{­­­(9)}\end{array}$

[0027] This formulation is a least-squares problem for which the classic solution is p, = (A.sup.TA).sup.-1A.sup.Tx. This solution is easily implemented with a microprocessor in a digital relay. To simplify the implementation, the matrix [(A.sup.TA).sup.-1A.sup.T] can be precomputed, processed and stored using timestamps

[00014]$t_k,t_{k+1},t_{k+2},.Math.,t_{k+N-2},t_{k+N-1}=0,\frac{1}{F_s},\frac{2}{F_s},.Math.,\frac{N-2}{F_s},\frac{N-1}{F_s}.$

[0028] The time-delay qualifier referred to above is an integrator of time that works with a prescribed delay, and starts integrating when the current estimator first detects a current magnitude that crosses a predefined pickup level. This process of detection can be described using a rotating disk analogy that spins in one direction (positive, clockwise) when the current estimator is above the pickup level, and the other direction (negative, counterclockwise) when the current estimator is below the pickup level. The disk starts spinning from a reset zero position when the current estimator reaches the pickup level the first time, and detects a fault when current estimator reaches a predefined position governed by the predefined delay period.

[0029] FIG. 3 is a graph with time on the horizontal axis and current magnitude on the vertical axis showing a fault scenario and illustrating the time-delay qualifier discussed above. Waveform 50 is fault current, curve 52 is the RMS current estimate from the current estimator and line 54 is the pickup level that identifies when the vacuum interrupter 20 should open because of the fault current. In one example, the current estimator samples the current 128 times per current cycle and estimates the fault current in a manner so that it overshoots the pickup level a couple of times before settling below the pickup level. Thus, since the actual RMS fault current did not exceed the pickup level, the vacuum interrupter 20 should not be triggered open. However, for the traditional system, the vacuum interrupter 20 would be triggered open when the current estimator estimated the current above the pickup level during the first overshoot.

[0030] FIG. 4 illustrates a timing disk 60 positioned at an initial or vertical reset position that occurs before the fault current is detected. The current estimator as described above first estimates the fault current to go above the pickup level at a 0.3-cycle point at point 56. Instead of detecting a fault and opening the vacuum interrupter 20 as soon as the current estimate reaches the pickup level at the point 56, the fault detection algorithm starts integrating time while the current estimate is above the pickup level. As the detection algorithm integrates time when the current estimate is above the pickup level, the disk 60 rotates in a clockwise or positive direction, as shown in FIG. 4. Since the current estimate goes above the pickup level at the start, the disk 60 starts rotating in the positive direction from the reset zero position. If the current estimate goes below the pickup level, such as at point 58, the detection algorithm subtracts time from the already accumulated time while the current estimate is below the pickup level. This is illustrated in FIG. 5 where the disk 60 is shown rotating in a counter-clockwise or negative direction from a point where some time has been accumulated from the current estimate being above the pickup level. When the current estimator drops below the pickup level, the disk 60 spins in the negative direction, but it stops if it reaches the reset zero position. A certain time accumulation threshold value is used for each system that provides a point where the estimate of the fault current indicates that the actual fault current is above the pickup level, which is illustrated by line 62. This functionality can be implemented in software using an integrator.

[0031] Additionally, a no-delay threshold line 64 is provided at the maximum overshoot level above the pickup level, which is a parameter specified as a percentage above the pickup level, for example, in the range of 1% - 100%. If the current estimator estimates the fault current to be above the no-delay threshold, a fault is detected instantaneously because this level could not be reached by overshoot from the current level that is below the pickup level. Using the no-delay threshold mechanism speeds up the detection process in cases when the true current level is much higher than the pickup level.

[0032] The foregoing discussion discloses and describes merely exemplary embodiments of the present disclosure. One skilled in the art will readily recognize from such discussion and from the accompanying drawings and claims that various changes, modifications and variations can be made therein without departing from the spirit and scope of the disclosure as defined in the following claims.