Systems and methods of bad data identification and recovery for electric power systems

Patent metadata

Field	Value
Publication number	US-11480594-B2
Application number	US-202017091792-A
Country	US
Kind code	B2
Filing date	Nov 6, 2020
Priority date	Nov 7, 2019
Publication date	Oct 25, 2022
Grant date	Oct 25, 2022

Abstract

Systems and methods for processing measurement data in an electric power system include acquiring the measurement data by a phasor measurement unit (PMU) coupled to a line of the electric power system, and inputting a plurality of the measurement data within a predetermined time window into a K-nearest neighbor (KNN) for identifying bad data among the plurality of the measurement data, wherein when one of the plurality of measurement data contains a bad datum, the machine learning module sends the bad datum to a denoising autoencoder module for correcting the bad datum, wherein the denoising autoencoder module outputs a corrected part corresponding to the bad datum, and when one of the plurality of measurement data contains no bad datum, the machine learning module bypasses the denoising autoencoder module and outputs the one of the plurality of measurement data as an untouched part.

First claim

What is claimed is:

1. A method for processing measurement data in an electric power system, the method comprising:
acquiring the measurement data from the electric power system; and
inputting a plurality of the measurement data within a predetermined time window into a machine learning module for identifying bad data among the plurality of the measurement data, the plurality of the measurement data being vector data acquired by a plurality of phasor measurement unit (PMU) coupled to a plurality of lines of the electric power system, wherein
when one of the plurality of measurement data contains a bad datum, the machine learning module sends the bad datum to a denoising autoencoder module for correcting the bad datum, wherein the denoising autoencoder outputs a corrected part corresponding to the bad datum; and
when one of the plurality of measurement data contains no bad datum, the machine learning module bypasses the denoising autoencoder module and outputs the one of the plurality of measurement data as an untouched part,
wherein the denoising autoencoder module includes a magnitude recovery denoising autoencoder and an angle recovery denoising autoencoder, wherein when the bad datum contains only a bad magnitude, the bad datum is only sent to the magnitude recovery denoising autoencoder for the correction;
when the bad datum contains only bad angle, the bad datum is only sent to the angle recovery denoising autoencoder for the correction; and
when the bad datum contains both bad magnitude and bad angle, the bad datum is sent to both the magnitude recovery denoising autoencoder and the angle recovery denoising autoencoder for the correction.

2. The method of claim 1 , wherein the plurality of the measurement data are arranged in a matrix with vector data arranged in columns.

3. The method of claim 1 , wherein the predetermined time window slides over time for inputting measurement data at different time.

4. The method of claim 1 , wherein the machine learning module includes a K-nearest neighbor (KNN) algorithm.

5. The method of claim 4 , wherein the machine learning module identifies a bad datum by weighted majority vote of a predetermined number of nearest data in terms of Euclidean distance.

6. The method of claim 1 , wherein the denoising autoencoder module includes a denoising autoencoder with symmetrical layers of neural network that are trained to reproduce input data at an output thereof.

7. The method of claim 1 further comprising combining the untouched part with the corrected part to form a recovered data stream.

8. The method of claim 1 , wherein the measurement data received by the machine learning module are always from a predetermined PMU.

9. The method of claim 1 , wherein the measurement data received by the machine learning module are from a first PMU at a first time and a second PMU at a second time different from the first time via a data bus.

10. A system for processing measurement data in an electric power system, the system comprising:
measurement devices including a plurality of phasor measurement unit (PMU) coupled to lines of the electric power system for measuring state information at the lines;
a processor; and
a computer-readable storage medium, comprising:
software instructions executable on the processor to perform operations, including:
acquiring the measurement data from the measurement devices; and
inputting a plurality of the measurement data within a predetermined time window into a machine learning module for identifying bad data among the plurality of the measurement data, wherein
when one of the plurality of measurement data contains a bad datum, the machine learning module sends the bad datum to a denoising autoencoder module for correcting the bad datum, wherein the denoising autoencoder module outputs a corrected part corresponding to the bad datum; and
when one of the plurality of measurement data contains no bad datum, the machine learning module bypasses the denoising autoencoder module and outputs the one of the plurality of measurement data as an untouched part,
wherein the denoising autoencoder module includes a magnitude recovery denoising autoencoder and an angle recovery denoising autoencoder, wherein when the bad datum contains only a bad magnitude, the bad datum is only sent to the magnitude recovery denoising autoencoder for the correction;
when the bad datum contains only bad angle, the bad datum is only sent to the angle recovery denoising autoencoder for the correction; and
when the bad datum contains both bad magnitude and bad angle, the bad datum is sent to both the magnitude recovery denoising autoencoder and the angle recovery denoising autoencoder for the correction.

11. The system of claim 10 , wherein the predetermined time window slides over time for inputting measurement data at different time.

12. The system of claim 10 , wherein the machine learning module includes a K-nearest neighbor (KNN) algorithm.

13. The system of claim 10 , wherein the measurement data received by the machine learning module are always from a predetermined PMU.

14. The system of claim 10 , wherein the measurement data received by the machine learning module are from a first PMU at a first time and a second PMU at a second time different from the first time via a data bus.

15. A method for processing measurement data in an electric power system, the method comprising:
acquiring the measurement data by a phasor measurement unit (PMU) coupled to a line of the electric power system; and
inputting a plurality of the measurement data within a predetermined time window into a K-nearest neighbor (KNN) for identifying bad data among the plurality of the measurement data, wherein
when one of the plurality of measurement data contains a bad datum, the machine learning module sends the bad datum to a denoising autoencoder module for correcting the bad datum, wherein the denoising autoencoder outputs a corrected part corresponding to the bad datum; and
when one of the plurality of measurement data contains no bad datum, the machine learning module bypasses the denoising autoencoder module and outputs the one of the plurality of measurement data as an untouched part,
wherein the denoising autoencoder module includes a magnitude recovery denoising autoencoder and an angle recovery denoising autoencoder, wherein when the bad datum contains only a bad magnitude, the bad datum is only sent to the magnitude recovery denoising autoencoder for the correction;
when the bad datum contains only bad angle, the bad datum is only sent to the angle recovery denoising autoencoder for the correction; and
when the bad datum contains both bad magnitude and bad angle, the bad datum is sent to both the magnitude recovery denoising autoencoder and the angle recovery denoising autoencoder for the correction.

16. The method of claim 15 further comprising combining the untouched part with the corrected part to form a recovered data stream.

Assignees

• Global Energy Interconnection Res Inst Co Ltd
• State Grid Corp Of China Co Ltd
• State Grid Jiangsu Electric Power Co Ltd
• State Grid Shanxi Electric Power Co

Inventors

• Gu Yingzhong
• Tian Guanyu
• Xu Chunlei
• Wu Haiwei
• Yu Zhe
• Shi Di

Classifications

• G06N3/088 Primary

  Non-supervised learning, e.g. competitive learning · CPC title

• G01R21/133 Primary

  by using digital technique · CPC title

• G06F2218/04

  Denoising · CPC title

• G06F18/24147

  Distances to closest patterns, e.g. nearest neighbour classification · CPC title

• G06N3/047

  Probabilistic or stochastic networks · CPC title