High impedance fault detection and location accuracy

Patent metadata

Field	Value
Publication number	US-12013427-B2
Application number	US-202117337976-A
Country	US
Kind code	B2
Filing date	Jun 3, 2021
Priority date	Jun 15, 2020
Publication date	Jun 18, 2024
Grant date	Jun 18, 2024

Abstract

High impedance fault (HIF) detection and location accuracy is provided. An HIF has random, irregular, and unsymmetrical characteristics, making such a fault difficult to detect in distribution grids via conventional relay measurements with relatively low resolution and accuracy. Embodiments disclosed herein provide a stochastic HIF monitoring and location scheme using high-resolution time-synchronized data in micro phasor measurement units (μ-PMUs) for distribution network protection. In particular, a fault detection and location process is systematically designed based on feature selections, semi-supervised learning (SSL), and probabilistic learning.

First claim

What is claimed is:

1. A high impedance fault (HIF) monitoring system, comprising:
a feature extractor, configured to:
receive power measurements from a plurality of sensors in a power distribution system; and
extract an effective feature set (EFS) for detecting an HIF from the power measurements by implementing a wrapper method, wherein the wrapper method comprises:
searching for features of the power measurements;
evaluating the features based on an accuracy of the features to determine which of the features are to be in the EFS; and
performing an induction algorithm to determine the accuracy of the features; and

an HIF detector configured to determine occurrence of the HIF using machine learning and the EFS, wherein the machine learning of the HIF detector is trained using semi-supervised learning (SSL), the SSL comprising:
minimizing a loss function regarding a posterior probability and maximizing mutual information between unseen data and a label to generate a probability matrix that indicates a probability that a data set belongs to either the HIF or a non-HIF;
providing an estimation expectation of an indicator loss function with respect to correct labeled data in the data set where the label is reformulated based on a decision function that provide a closed form solution to the label to determine the occurrence of the HIF.

2. The HIF monitoring system of claim 1 , further comprising an HIF locator configured to determine a probable location of the HIF based on an output of the HIF detector.

3. The HIF monitoring system of claim 2 , wherein the HIF locator is configured to determine the probable location of the HIF by estimating a distance value of the HIF using fault current estimation.

4. The HIF monitoring system of claim 3 , wherein the HIF locator is further configured to determine the probable location of the HIF by using a moving window least square approach to identify a zone in the power distribution system where the HIF is more likely to be occurring than other zones.

5. The HIF detector of claim 1 , wherein:
each of the plurality of sensors is a micro phasor measurement unit (μ-PMU); and
the power measurements comprise phasor measurements from the plurality of μ-PMUs.

6. The HIF monitoring system of claim 5 , wherein:
the feature extractor is configured to extract a first portion of the EFS by applying a discrete Fourier transform (DFT) to the phasor measurements from the plurality of μ-PMUs; and
the first portion of the EFS comprises at least one of a zero sequence voltage (V 0 ), a positive sequence voltage (V 1 ), a negative sequence voltage (V 2 ), a zero sequence current (I 0 ), a positive sequence current (I 1 ), or a negative sequence current (I 2 ).

7. The HIF monitoring system of claim 6 , wherein the feature extractor is further configured to extract an angle difference between the negative sequence voltage and the zero sequence voltage (θ V 2 −θ V 0 ).

8. The HIF monitoring system of claim 6 , wherein:
the feature extractor is further configured to extract a second portion of the EFS by applying a Kalman filter (KF) to the phasor measurements from the plurality of μ-PMUs; and
the second portion of the EFS comprises at least one of an estimated in-phase component of a harmonic of voltage (KF V cos(H V1 ˜H V6 ) or an estimated in-quadrature component of a first harmonic of voltage (KF V sin(H V1 ˜H V6 )).

9. The HIF monitoring system of claim 1 , further comprising an HIF alarm configured to provide an indication of the HIF occurrence.

10. The HIF monitoring system of claim 9 , further comprising an HIF locator configured to determine a probable location of the HIF based on an output of the HIF detector;
wherein the HIF alarm is further configured to provide an indication of the probable location of the HIF.

11. The HIF monitoring system of claim 10 , wherein the HIF alarm is further configured to provide a tripping signal to the power distribution system to cause a circuit breaker corresponding to the probable location of the HIF to trip.

12. A method for detecting and locating a high impedance fault (HIF), the method comprising:
receiving power measurements from a power distribution system;
extracting a power feature indicative of HIF occurrence from the power measurements by implementing a wrapper method, wherein the wrapper method comprises:
searching for features of the power measurements;
evaluating the features based on an accuracy of the features in performing to determine which of the features is to be the power feature; and
performing an induction algorithm to determine the accuracy of the features;

determining occurrence of an HIF based on the power feature, wherein machine learning is trained using semi-supervised learning (SSL), the SSL comprising:
minimizing a loss function regarding a posterior probability and maximizing mutual information between unseen data and a label to generate a probability matrix that indicates a probability that a data set belongs to either the HIF or a non-HIF;
providing an estimation expectation of an indicator loss function with respect to correct labeled data in the data set where the label is reformulated based on a decision function that provides closed form solution to the label to determine the occurrence of the HIF; and

determining a probable location of the HIF based on the power measurements.

13. The method of claim 12 , wherein the power feature corresponding to the HIF occurrence is an angle difference between a negative sequence voltage and a zero sequence voltage (θ V 2 −θ V 0 ).

14. The method of claim 13 , wherein extracting the angle difference θ V 2 −θ V 0 comprises performing a discrete Fourier transform (DFT) of the power measurements.

15. The method of claim 12 , further comprising:
extracting an effective feature set (EFS) comprising an angle difference between a negative sequence voltage and a zero sequence voltage (θ V 2 −θ V 0 ) and at least one of a zero sequence voltage (V 0 ), a positive sequence voltage (V 1 ), a negative sequence voltage (V 2 ), a zero sequence current (I 0 ), a positive sequence current) (I 1 ), a negative sequence current (I 2 ), or a harmonic of a received voltage signal (KF V cos(H V1 ˜H V6 ) or KF V sin(H V1 ˜H V6 )); and
determining the occurrence of the HIF based on the EFS.

16. The method of claim 15 , wherein extracting the EFS comprises sequentially extracting features of the EFS by:
performing a set of discrete Fourier transforms (DFTs) on the power measurements;
taking derivatives of one or more features extracted with the set of DFTs; and
Kalman filtering the power measurements.

17. A phasor data concentrator (PDC) for a power distribution system, the PDC comprising:
a network interface;
a memory configured to store power measurements received over the network interface from a plurality of sensors in the power distribution system; and
a high impedance fault (HIF) monitor connected to the memory and comprising:
feature extraction logic, configured to extract an effective feature set (EFS) for detecting an HIF from the power measurements by implementing a wrapper method, wherein the wrapper method comprises:
searching for features of the power measurements;
evaluating the features based on an accuracy of the features in performing to determine which of the features are to be in the EFS; and
performing an induction algorithm to determine the accuracy of the features;

HIF detection logic configured to determine occurrence of the HIF using the E

Assignees

• Cui Qiushi
• Weng Yang
• Univ Arizona State

Inventors

• Cui Qiushi
• Weng Yang

Classifications

• G01R31/086

  in power transmission or distribution networks, i.e. with interconnected conductors · CPC title

• G01R23/16

  Spectrum analysis; Fourier analysis · CPC title

• G01R27/16

  Measuring impedance of element or network through which a current is passing from another source, e.g. cable, power line · CPC title

• Y04S10/22

  Flexible AC transmission systems [FACTS] or power factor or reactive power compensating or correcting units · CPC title

• Y02E60/00

  Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation · CPC title