Training data generation device, electrical device monitoring system, training data generation method, and non-transitory storage medium

Patent metadata

Field	Value
Publication number	US-10466283-B2
Application number	US-201515300531-A
Country	US
Kind code	B2
Filing date	Mar 17, 2015
Priority date	Mar 31, 2014
Publication date	Nov 5, 2019
Grant date	Nov 5, 2019

Abstract

A training data generation device ( 10 ) includes: a measurement data acquisition unit ( 11 ) that acquires measurement data including waveform data including at least one of a current consumption waveform, an input voltage waveform, and a power consumption waveform along a time axis of an electrical device and power consumption value data indicating time-series power consumption values of the electrical device; an analysis unit ( 12 ) that analyzes a rate of occurrence of each power consumption value using the power consumption value data; a specifying unit ( 13 ) that specifies one or more specific power consumption value bands including a power consumption value with the higher rate of occurrence than that of a standard power consumption value and having a predetermined power value width; a feature amount extraction unit ( 14 ) that determines a representative power value for each specific power consumption value band and extracts a feature amount using the waveform data; and a storage unit ( 15 ) that stores a pair of the representative power value determined based on each specific power consumption value band and a training feature amount which is the feature amount in association with the electrical device.

First claim

The invention claimed is:

1. A training data generation device comprising:
a measurement data acquisition unit that acquires measurement data including waveform data including at least one of a current consumption waveform, an input voltage waveform, and a power consumption waveform along a time axis of an electrical device which is a target of a process of generating training data and power consumption value data indicating time-series power consumption values of the electrical device;
an analysis unit that analyzes a rate of occurrence of each power consumption value using the power consumption value data;
a specifying unit that specifies one or more specific power consumption value bands including a power consumption value with the higher rate of occurrence than that of a standard power consumption value and having a predetermined power value width;
a feature amount extraction unit that determines a representative power value and extracts a feature amount using the waveform data for each specific power consumption value band; and
a storage unit that stores a pair of the representative power value determined based on each specific power consumption value band and a training feature amount which is the feature amount in association with the electrical device.

2. The training data generation device according to claim 1 ,
wherein in the measurement data, a measurement timing of each power consumption value included in the power consumption value data is able to be specified in the waveform data.

3. The training data generation device according to claim 1 ,
wherein the feature amount extraction unit determines the representative power value based on the power consumption value included in the specific power consumption value band and extracts the feature amount using the waveform data at the measurement timing of the power consumption value included in the specific power consumption value band for each specific power consumption value band.

4. The training data generation device according to claim 1 ,
wherein the specifying unit specifies one or more power consumption values which are maximum values in a graph in which a horizontal axis represents the power consumption value and a vertical axis represents the rate of occurrence and specifies one or more specific power consumption value bands including each specified power consumption value.

5. The training data generation device according to claim 1 ,
wherein the one or more specific power consumption value bands specified by the specifying unit do not include some of the power consumption value bands consumable by the electrical device.

6. The training data generation device according to claim 5 ,
wherein the power consumption value band consumable by the electrical device is the power value band of greater than 0 W and equal to or less than a value of rated power of the electrical device.

7. The training data generation device according to claim 1 , further comprising:
a monitoring waveform data acquisition unit that acquires monitoring waveform data including at least one of the current consumption waveform, the input voltage waveform, and the power consumption waveform along the time axis of the electrical device which is a monitoring target;
a collation unit that performs a collation process using a monitoring feature amount which is a feature amount extracted from the monitoring waveform data and the training feature amount of the electrical device stored in the storage unit; and
an output unit that outputs the representative power value associated with the training feature amount matching the monitoring feature amount as the power consumption value of the electrical device.

8. An electrical device monitoring system comprising:
the training data generation device according to claim 1 ; and
a monitoring device that infers an operation state of each monitoring electrical device which is a monitoring target using the training feature amount of the electrical device generated by the training data generation device,
wherein the monitoring device includes
a monitoring waveform data acquisition unit that acquires monitoring waveform data including at least one of a current consumption waveform, an input voltage waveform, and a power consumption waveform of the electrical device,
a collation unit that performs a collation process using a monitoring feature amount which is a feature amount extracted from the monitoring waveform data and the training feature amount of the electrical device generated by the training data generation device, and
an output unit that outputs the representative power value associated with the training feature amount matching the monitoring feature amount as the power consumption value of the electrical device.

9. A non-transitory storage medium storing a program causing a computer to function as:
a measurement data acquisition unit that acquires measurement data including waveform data including at least one of a current consumption waveform, an input voltage waveform, and a power consumption waveform along a time axis of an electrical device which is a target of a process of generating training data and power consumption value data indicating time-series power consumption values of the electrical device;
an analysis unit that analyzes a rate of occurrence of each power consumption value using the power consumption value data;
a specifying unit that specifies one or more specific power consumption value bands including a power consumption value with the higher rate of occurrence than that of a standard power consumption value and having a predetermined power value width;
a feature amount extraction unit that determines a representative power value and extracts a feature amount using the waveform data for each specific power consumption value band; and
a storage unit that stores a pair of the representative power value determined based on each specific power consumption value band and a training feature amount which is the feature amount in association with the electrical device.

10. A training data generation method performed by a computer, the method comprising:
a measurement data acquisition step of acquiring measurement data including waveform data including at least one of a current consumption waveform, an input voltage waveform, and a power consumption waveform along a time axis of an electrical device which is a target of a process of generating training data and power consumption value data indicating time-series power consumption values of the electrical device;
an analysis step of analyzing a rate of occurrence of each power consumption value using the power consumption value data;
a specifying step of specifying one or more specific power consumption value bands including a power consumption value with the higher rate of occurrence than that of a standard power consumption value and having a predetermined power value width;
a feature amount extraction step of deciding a representative power value and extracting a feature amount using the waveform data for each specific power consumption value band; and
a storage step of storing a pair of the representative power value determined based on each specific power consumption value band and a training feature amount which is the feature amount in association with the electrical device in a storage device.

Assignees

• Nec Corp

Inventors

• Homma Kosuke
• Saneyoshi Eisuke
• Hashimoto Ryo
• TOIZUMI Takahiro

Classifications

• G01D4/002

  Remote reading of utility meters · CPC title

• G01R19/2509

  Details concerning sampling, digitizing or waveform capturing · CPC title

• G06N20/10

  using kernel methods, e.g. support vector machines [SVM] · CPC title

• G05B23/0272

  Presentation of monitored results, e.g. selection of status reports to be displayed; Filtering information to the user · CPC title

• G05B9/02

  electric · CPC title