WO2019117337A1

WO2019117337A1 - Device and method for classifying appliances by using electric power data analysis

Info

Publication number: WO2019117337A1
Application number: PCT/KR2017/014595
Authority: WO
Inventors: 김호원; 김지현; 김해용; 김재현
Original assignee: 부산대학교 산학협력단
Priority date: 2017-12-12
Filing date: 2017-12-13
Publication date: 2019-06-20
Also published as: KR102049453B1; KR20190070147A

Abstract

The present invention relates to a device and a method for classifying appliances by using electric power data analysis, which classify appliances by analyzing electric power data that are collected from customers and extracting an impedance component of electric power, including: an electric power consumption data detection unit which detects electric power consumption data; a frequency component extraction unit which receives the input of the electric power consumption data as a stream and separates and extracts a low-frequency component and a high-frequency component; a low-frequency component analysis data output unit which outputs electric energy data from the low-frequency component and outputs electric power analysis data employing developments of the data; a high-frequency component analysis data output unit which outputs phase angle data from the high-frequency component and outputs frequency analysis data from the high-frequency component; and an impedance modeling unit which performs impedance modeling based on a change in an electric current amount and a change in a phase angle and classifies appliances by using same.

Description

Apparatus and method for classifying household appliances using power data analysis

The present invention relates to a household appliance classification system, and more particularly, to an apparatus and method for classifying household appliances using power data analysis for analyzing power data collected in a customer and extracting an impedance component of power to classify household appliances .

With the recent depletion of fossil fuels, interest in the efficient use of energy is increasing worldwide. As a result, Smart Grid technology that integrates ICT technology into the power grid is being realized, especially in advanced countries such as the US and Europe.

A smart meter is installed to measure and transmit the power consumption of each customer, so that monitoring and power prediction using power consumption is becoming possible.

However, it is difficult to predict accurate energy use simply by collecting the total power consumption of the customer. For more accurate prediction, it is necessary to collect power consumption information at the consumer electronics stage used by the customer.

NILM (Non-intrusive Load Monitoring) is a technique for realizing this.

NILM is a technology for analyzing the power data of a customer to find out which household appliances are being used. Therefore, it is possible to collect power consumption information at the consumer electronics stage of the consumer by using the NILM.

However, the home appliance classification models that have been researched and commercialized so far are not accurate enough to be commercialized.

In order to manage electric energy for each appliance, the service provider must secure the labeled data for a certain period of time.

For this purpose, it is practically impossible to collect data by attaching a sensor to each home appliance or to provide information on home appliance usage to a service provider whenever the user uses the appliance.

Alternatively, there is a method of generating a model through speculation by sending a list of household appliances and power consumption information of each appliance to the service provider when the user executes the initial service, but this is not only inconvenient for the user but also the generated model is accurate not.

Especially, in the case of a smart meter in general, raw data input through a power sensor is removed by using a low-pass filter to remove high-frequency components such as signal noise and phase change.

This method has a limit to obtain one data stream through one sensor. In the case of the classification model using such data, the modeling accuracy is inherently limited.

Particularly, as the accuracy of the modeling of a device having a similar power consumption or a multi-state device is lowered, a method of supplementing the same with an additional cognitive sensor or a group-based measurement using a plurality of current measuring sensors is used.

Therefore, it is required to develop a new home appliance classification method that can solve the problems of the home appliance classification model of the related art and increase the accuracy.

The present invention solves the problem of the conventional method of classifying household electrical appliances. The present invention analyzes electric power data collected from a customer and extracts an impedance component of electric power to classify household appliances using power data analysis for classifying household appliances And an object of the present invention is to provide an apparatus and a method for providing the same.

The present invention analyzes electric power data collected from a customer and analyzes the electric power data of the electric power using the filter before the machine learning and the latest technology Deep Learning are applied to the non-intrusive load monitoring (NILM) And an apparatus and method for classifying household appliances using power data analysis for classifying household appliances.

The present invention is not a simple current value measurement method used in a smart monitor but a frequency analysis using a high frequency component measured in a current sensor and a peak value in a cycle to model the impedance of a home appliance, And an apparatus and method for classifying household appliances using power data analysis.

The present invention can instantaneously monitor the change of the power data by extracting the impedance component of the electric power, so that the instantaneous power analysis and the device classification according to the power change of the individual electric appliance can be performed. And an object of the present invention is to provide a device and a method for the same.

The present invention can model changes in internal appliances and changes in RLC components through high frequency analysis, thereby extracting RLC component vectors of individual home appliances, and analyzing which combination of RLC components the entire power data is composed And an apparatus and method for classifying household appliances using power data analysis.

The present invention relates to a device and a method for classifying household appliances using power data analysis capable of extracting RLC component vectors of individual home appliances so that the electromechanical devices can be more finely classified into different two-dimensional configurations according to the configuration of internal components The purpose is to provide.

The present invention can be applied to a refrigerator, an air conditioner, a washing machine, a refrigerator, a refrigerator, a refrigerator, a refrigerator, a refrigerator, The present invention provides an apparatus and method for classifying household appliances using power data analysis that can be classified into a combination of a long-term power amount data transition and an RLC circuit according to a state change.

The objects of the present invention are not limited to the above-mentioned objects, and other objects not mentioned can be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention, there is provided an apparatus for classifying household appliances using power data analysis, comprising: a power consumption data detector for detecting power consumption data; A low frequency component analysis data output unit for outputting power amount data from a low frequency component and outputting power analysis data using a trend of the data, a phase angle data output unit for outputting phase angle data from a high frequency component, And an impedance modeling unit for performing impedance modeling on the basis of the change in the amount of current and the change in the phase angle and classifying the home appliances using the impedance modeling.

Here, the frequency component extraction unit may include: a low-pass filter that receives power consumption data as a stream from the power consumption data detection unit and extracts low-frequency components; a high-frequency filter that receives power consumption data as a stream from the power consumption data detection unit and extracts high- And a pass filter.

The low-frequency component analysis data output unit includes a power amount data output unit for outputting power amount data from low-frequency components of the low-pass filter, and a power analysis data output unit for outputting power analysis data using the transition of the low- And the like.

The high frequency component analysis data output section includes a phase angle data output section for outputting phase angle data from a high frequency component of the high frequency pass filter and a frequency analysis data output section for outputting frequency analysis data from a high frequency component of the high frequency pass filter .

The impedance modeling unit may further include an impedance analyzing unit for extracting information for classifying the home appliances secondarily on the basis of a change in the amount of current and a change in the phase angle and a home appliance classifying unit for classifying home appliances using the analysis result of the impedance analyzing unit .

Impedance modeling by the impedance modeling unit models RLC component vectors of individual home appliances by modeling changes in internal appliances and changes in RLC components through high frequency analysis and determines which combination of RLC components the entire power data is composed Thereby enabling analysis.

And a combination of all or a part of the power amount data, the power analysis data, the phase angle data, and the frequency analysis data by impedance modeling by the impedance modeling unit as learning and classification data of the NILM system.

The instantaneous change of the power data is monitored by extracting the impedance component through the impedance modeling of the impedance modeling unit, and instantaneous power analysis and device classification according to the power supply change of the individual household appliances are performed.

Impedance Modeling of Impedance Modeling to monitor the change of electrical appliances such as TVs, hair dryers, and lighting, where momentary ON / OFF changes occur, using instantaneous changes in power data, The refrigerator, the air conditioner, and the washing machine are classified into a combination of a long time energy amount data transition and an RLC circuit according to a change in state.

According to another aspect of the present invention, there is provided a method for classifying household appliances using power data analysis according to the present invention includes the steps of extracting a frequency component of a low frequency component and a high frequency component by receiving power consumption data as a stream from a power consumption data detector, Outputting power amount data from the low frequency component, outputting power analysis data using the transition of the data from the low frequency component, outputting phase angle data from the high frequency component, and outputting frequency analysis data from the high frequency component .

Here, the step of extracting the energy data set to be used in the household appliance classification model and providing all or a part of the power amount data, the power analysis data, the phase angle data, and the frequency analysis data as learning and classification data of the NILM system .

And performing impedance modeling based on a change in the amount of current and a change in the phase angle.

And, by impedance modeling, RLC component vectors of individual appliances are extracted by modeling changes of internal appliances and changes of RLC components through high frequency analysis, and it is possible to analyze which combination of RLC components the entire power data is composed .

The instantaneous change of the power data is monitored by extracting the impedance component through impedance modeling, and instantaneous power analysis and device classification according to the power supply change of the individual household appliances are performed.

The apparatus and method for classifying household appliances using power data analysis according to the present invention have the following effects.

First, by analyzing the power data collected from the customer, the impedance component of the power can be extracted to classify the household appliances and improve the accuracy.

Second, before applying machine learning and deep learning to non-intrusive load monitoring (NILM), impedance components of power are extracted using filters to achieve high accuracy and efficiency.

Third, the impedance of the household appliances is modeled through the frequency analysis using the high frequency component measured in the current sensor and the peak value in the cycle, rather than the simple current value measurement method used in the smart monitor, and the input data of the learning and classification model It is possible to increase the classification accuracy of the household appliance.

Fourth, when the impedance vector analysis method is further applied to the classification method using machine learning and deep learning, the accuracy and efficiency of the analysis method using only the current amount are improved.

Fifth, as it becomes possible to classify devices having similar power consumption or to model multi-state household appliances, they will have an advantage in detecting the status of equipment and classifying multiple equipment.

Sixth, NILM models can solve the problem of accuracy, classification speed, user inconvenience, and increase practicality by increasing classification accuracy of CE appliances.

In addition, it can be used for other systems or services by obtaining home appliance classification information as well as information about power consumption.

Seventh, it is possible to classify home appliances in embedded devices with limited environment, and it can be used as additional variables in the conventional classification methods using deep running and machine learning, thereby contributing to improvement of accuracy in the system.

1 is a block diagram of an apparatus for classifying household appliances using power data analysis according to the present invention

2 is a detailed block diagram of an apparatus for classifying household appliances using power data analysis according to the present invention

3 is a detailed configuration diagram of an impedance modeling unit according to the present invention.

4 is a flowchart illustrating a method for classifying household appliances using power data analysis according to the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of an apparatus and method for classifying household appliances using power data analysis according to the present invention will be described in detail.

The features and advantages of the apparatus and method for classifying household appliances using power data analysis according to the present invention will be apparent from the following detailed description of each embodiment.

1 is a block diagram of an apparatus for classifying household appliances using power data analysis according to the present invention.

The present invention analyzes electric power data collected from a customer and analyzes the electric power data of the electric power using the filter before the machine learning and the latest technology Deep Learning are applied to the non-intrusive load monitoring (NILM) To classify the household appliances.

First, data processing using a filter to analyze power data and a method of analyzing the processed data through deep learning and machine learning will be described.

Deep Learning is an approach that mimics the way in which neurons that make up the human brain work.

Providing the learning data enables the computer to learn, judge, or classify by itself.

Deep learning can be roughly divided into RNN (Recurrent Neural Network) and CNN (Convolutional Neural Network).

RNN is one of the deep learning algorithms that analyze and input data sequentially input. By providing the input / output data set for a specific time, the home appliance classification model can be generated by learning the relationship of the continuous data during the relevant time.

In addition, CNN is used in various fields such as image processing and face recognition. It can be used as learning data after graphing changes in power over time.

Typical data used for such machine learning and deep running are power data.

Representative methods for expressing power usage in an AC circuit include Apparent Power and Effective Power.

Apparent power is calculated on the assumption that the pure power resistances are not considered in the AC circuit. In the AC circuit, the impedance is a combination of a resistor having a real component and a capacitor and an inductor having a complex component. The apparent power does not take into account such a complex component.

On the other hand, the effective power is calculated by considering the phase difference between the voltage and the current appearing in the AC circuit, and represents the amount of electric power used by the actual household appliances. In order to measure the active power, the phase difference between the voltage and the current as well as the current value should be measured using a precision instrument such as a power analyzer.

The present invention includes a classification accuracy of a NILM technique for analyzing the power consumption of a customer and finding a currently used household appliance, and a power data extraction method applicable to various machine learning techniques.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an apparatus and method for classifying household appliances using power data analysis according to the present invention will be described in detail with reference to the accompanying drawings.

An apparatus and method for classifying household appliances using power data analysis according to the present invention can instantaneously monitor power data changes by extracting impedance components of power, And a configuration for classifying devices.

The present invention is capable of modeling changes in internal appliances and changes in RLC components through high frequency analysis to extract RLC component vectors of individual home appliances and analyzing which combination of RLC components the entire power data is composed .

The present invention includes a configuration for extracting RLC component vectors of individual home appliances and further finely classifying the electromechanical devices into different two-dimensional configurations according to the configuration of internal components.

The present invention can be applied to a refrigerator, an air conditioner, a washing machine, a refrigerator, a refrigerator, a refrigerator, a refrigerator, a refrigerator, Includes a configuration for classifying into a combination of a long-term power amount data transition and an RLC circuit according to a state change.

As shown in FIG. 1, an apparatus for classifying household appliances using power data analysis according to the present invention includes a power consumption data detecting unit 10 for detecting power consumption data, A low frequency component analysis data output unit 30 for outputting power amount data from low frequency components and outputting power analysis data using the trend of the data, A high frequency component analysis data output unit 40 for outputting data and outputting frequency analysis data from a high frequency component; an impedance modeling unit 40 for performing impedance modeling based on a change in the amount of current and a change in phase angle, (50).

The apparatus for classifying household appliances using power data analysis according to the present invention has the following effects by acquiring information for classifying household appliances through frequency analysis using high frequency components and periodic peak values.

The previous power sensor provides the power data for the cycle time by calculating the RMS using the output value of the power sensor. In order to classify the home appliance using the data, the classification method using the power consumption trend for a long time must be used, There is a problem that real-time analysis (within 1 to 2 seconds) of the device is impossible.

The present invention monitors the instantaneous change of the power data instead of the average value using the RMS by classifying the high frequency component and the peak value in the cycle so that power supply variation (ON / OFF, state change ), Enabling instantaneous power analysis and device classification.

In addition, due to the characteristics of home appliances, it is not a pure resistance component but a combination of components such as a coil and a capacitor. The position of the peak value in the cycle varies depending on the change of the RLC combination, So that it can monitor the change of the apparatus quickly.

The detailed configuration of the apparatus for classifying household appliances using the power data analysis according to the present invention having the above-described configuration is as follows.

FIG. 2 is a detailed configuration diagram of an apparatus for classifying household appliances using power data analysis according to the present invention, and FIG. 3 is a detailed configuration diagram of an impedance modeling unit according to the present invention.

2, the frequency component extraction unit 20 includes a low-pass filter 20a that receives power consumption data as a stream from the power consumption data detection unit 10 and extracts low-frequency components, a power consumption data detection unit 10, And a high-pass filter 20b that receives power consumption data as a stream and extracts high-frequency components.

The low frequency component analysis data output section 30 includes a power amount data output section 30a for outputting power amount data from the low frequency components of the low frequency pass filter 20a and a power amount data output section 30b for outputting the trend of the data from the low frequency components of the low frequency pass filter 20a And a power analysis data output unit 30b for outputting power analysis data used.

The high frequency component analysis data output section 40 includes a phase angle data output section 40a for outputting phase angle data from high frequency components of the high frequency pass filter 20b, And a frequency analysis data output unit 40b for outputting the frequency analysis data.

A combination of all or a part of the energy data set (power amount data, power analysis data, phase angle data, and frequency analysis data) generated and converted by the apparatus for classifying household appliances using the power data analysis according to the present invention having the above- And used as learning and classification data of the NILM system.

Energy data sets (power data, power analysis data, phase angle data, frequency analysis data) extracted in this manner can be used as learning and classification data of various NILM systems according to the situation. For example, instantaneous ON / The change of the home appliances such as the TV, the hair dryer, and the lighting in which the change occurs is monitored using the instantaneous change of the power data, and the refrigerator, the air conditioner, and the washing machine, And the RLC circuit according to the state change.

FIG. 3 illustrates a configuration for modeling an impedance available for adding a new device and detecting the state of a multi-state device using time variation of data extracted through the transform technique of FIG.

The impedance modeling unit 50 includes an impedance analyzing unit 50a for extracting information for classification of household appliances secondary to a change in the amount of current and a change in phase angle, rather than a total amount of consumed current and power, (50b).

The impedance modeling by the impedance modeling unit 50 models changes in internal appliances and changes in RLC components (internal resistances / coils / capacitors) through high frequency analysis. Using this, impedance values of RLC component vectors And it is possible to analyze which combination of RLC components the entire power data is made of.

This type of impedance modeling allows more detailed combinations than previous one-dimensional analysis using full power data.

For example, even in the case of a device using the same 10W, since it can be classified into another two-dimensional configuration according to the configuration of the internal component, it is possible to further narrow the device classification.

As described above, in the present invention, the impedance model that is secondarily extracted based on the change of current amount and phase angle instead of the total amount of consumed current and power, is a single variable, a simple class of machine learning technique such as SVM, It is possible to increase the accuracy of the classification of the apparatus.

Also, when the change of the ON / OFF state of the TV classified according to the state change does not appear in the long time transition, the data is re-learned to increase the accuracy, or to add a new device or notify the state of the TV failure May also be used for services.

The process of classifying household appliances in the apparatus for classifying household appliances using the power data analysis according to the present invention having such a structure will be described in detail as follows.

First, the power sensor receives power consumption data as a stream, and extracts frequency components of a low frequency component and a high frequency component (S401)

Next, the power amount data is output from the low frequency component, and power analysis data using the transition of the data is outputted from the low frequency component (S402)

Then, the phase angle data is outputted from the high frequency component, and the frequency analysis data is outputted from the high frequency component (S403)

Next, an energy data set to be used in the household appliance classification model is extracted (S404), and a combination of all or some of the power amount data, power analysis data, phase angle data, and frequency analysis data is used as learning and classification data of the NILM system. S405)

The apparatus and method for classifying household appliances using the power data analysis according to the present invention are classified into machine learning and non-intrusive load monitoring (NILM), which is a method of classifying household appliances in use by analyzing power data collected in a customer. Before applying Deep Learning, which is the latest technology, it is to classify household appliances by extracting the impedance component of power by using filter.

As described above, it will be understood that the present invention is implemented in a modified form without departing from the essential characteristics of the present invention.

It is therefore to be understood that the specified embodiments are to be considered in an illustrative rather than a restrictive sense and that the scope of the invention is indicated by the appended claims rather than by the foregoing description and that all such differences falling within the scope of equivalents thereof are intended to be embraced therein It should be interpreted.

Claims

A power consumption data detector for detecting power consumption data;

A frequency component extraction unit that receives power consumption data as a stream and separates and extracts a low frequency component and a high frequency component;

A low frequency component analysis data output unit outputting power amount data from a low frequency component and outputting power analysis data using a trend of the data;

A high frequency component analysis data output unit outputting phase angle data from a high frequency component and outputting frequency analysis data from a high frequency component;

And an impedance modeling unit for performing impedance modeling based on the change of the amount of current and the change of the phase angle and classifying the home appliance using the impedance modeling.
The apparatus as claimed in claim 1,

A low-pass filter for receiving power consumption data as a stream and extracting a low-frequency component from the power consumption data detector,

And a high-pass filter for receiving power consumption data from the power consumption data detector as a stream and extracting a high-frequency component.
The apparatus according to claim 1, wherein the low-

A power amount data output unit for outputting power amount data from low-frequency components of the low-pass filter,

And a power analysis data output unit for outputting power analysis data using a transition of the data from a low frequency component of the low pass filter.
The apparatus according to claim 1, wherein the high-

A phase angle data output unit for outputting phase angle data from a high frequency component of the high pass filter;

And a frequency analysis data output unit for outputting frequency analysis data from a high-frequency component of the high-pass filter.
The apparatus of claim 1, wherein the impedance modeling unit comprises:

An impedance analyzer for secondarily extracting information for classification of household appliances on the basis of a change in current amount and a change in phase angle;

And a household appliance classifying section for classifying the household appliance using the analysis result of the impedance analyzing section.
6. The method of claim 5, wherein the impedance modeling by the impedance modeling unit comprises:

The RLC component vectors of individual home appliances are extracted by modeling changes in internal appliances and changes in RLC components through high frequency analysis,

And to analyze which combination of RLC components the entire power data is composed of.
The method according to any one of claims 1 to 5, wherein a combination of all or a part of the power amount data, the power analysis data, the phase angle data, and the frequency analysis data is used as the learning and classification data of the NILM system by impedance modeling by the impedance modeling unit Apparatus for classifying household appliances using power data analysis.
6. The method of claim 1 or claim 5, further comprising the steps of: monitoring an instantaneous change of the power data by extracting an impedance component through impedance modeling of the impedance modeling unit; and performing instantaneous power analysis and device classification Wherein the power data analysis unit is configured to classify household appliances using power data analysis.
The impedance modeling method according to claim 1, wherein, through impedance modeling of the impedance modeling unit,

Changes in home appliances such as TVs, hair dryers, and lighting where momentary ON / OFF changes occur are monitored using instantaneous changes in power data,

Wherein the refrigerator, the air conditioner, and the washing machine which use a long time and change the state are classified into a combination of a long time power amount data change and an RLC circuit according to a state change.
Extracting a frequency component of a low frequency component and a high frequency component by receiving power consumption data as a stream from the power consumption data detector;

Outputting power amount data from a low frequency component and outputting power analysis data using a transition of the data from a low frequency component;

And outputting phase angle data from the high frequency components and outputting frequency analysis data from the high frequency components.
11. The method as claimed in claim 10, further comprising: extracting an energy data set to be used in the household appliance classification model and providing a combination of all or a part of the power amount data, power analysis data, phase angle data and frequency analysis data as learning and classification data of the NILM system Further comprising the step of analyzing the power data.
The method of claim 10, further comprising: performing impedance modeling based on a change in a current amount and a change in a phase angle.
13. The method of claim 12, wherein, by impedance modeling,

The RLC component vectors of individual home appliances are extracted by modeling changes in internal appliances and changes in RLC components through high frequency analysis,

And analyzing which RLC components the entire power data consists of. &Lt; Desc / Clms Page number 19 >
13. The power supply system according to claim 12, wherein the instantaneous change of the power data is monitored by extracting an impedance component through impedance modeling, and instantaneous power analysis and device classification are performed according to power supply variation of the individual household appliances A Method for the Classification of Home Appliances using Analysis.