KR102004905B1 - Frequency measuring device including a plurality of voltage sampling devices using low sampling frequency and appliance monitoring system including the same - Google Patents

Abstract

A frequency measuring device for measuring a voltage signal of a plurality of devices receiving power from a distribution board according to the present invention measures the voltage signal at a specific location on an internal wiring connected to the devices based on a first sampling frequency, A first meter for converting a first voltage signal measured in accordance with the first sampling frequency into a frequency domain to generate a first peak frequency spectrum; A second meter for measuring based on a second sampling frequency different from the sampling frequency and for converting a second voltage signal measured according to the second sampling frequency into a frequency domain to generate a second peak frequency spectrum, Based on the second peak frequency spectrum, the original peak frequency included in the used voltage signal Comprising: a unit for extracting a peak pattern, the first and the second sampling frequency is less than twice the one of the peak frequencies contained in the frequency pattern of the original peak.

Description

TECHNICAL FIELD [0001] The present invention relates to a frequency measuring apparatus including a plurality of voltage meters using a low sampling frequency, and a home appliance monitoring system including the same. [0002]

More particularly, the present invention relates to a frequency measurement apparatus including a plurality of voltage meters using a low sampling frequency, and a home appliance monitoring system including the same.

In order to solve the problem of energy supply and demand in the future, it is necessary to secure technology for energy saving and efficient use. Accordingly, there is an increasing interest in building a smart grid environment for efficient use of energy. In addition, understanding and participation of the consumers as well as the producers is important for the successful introduction of Smart Grid in homes and buildings. In order to save energy on the customer side, the user can be aware of the current usage of power so that the user can participate in the usage reduction by himself.

A system that measures and monitors the amount of electricity used by each appliance is a straightforward and effective method, but has a disadvantage of high construction cost. Therefore, using non-intrusive load monitoring (NILM) technology that measures power consumption using a single measurement device can reduce facility complexity and cost. The NILM system uses a single measuring device to monitor the power consumption of each device and can monitor the signals connected to the measuring device to detect the electrical change caused by the operation of the individual device and notify the user.

NILM systems differ in the applicable algorithms depending on which power-related signal is used. When measuring power, the NILM system can use a steady-state variation of the power level measured in the meter in real-time or use a pattern that changes in transient state. When measuring current, the NILM system can use the RMS current value differential. When measuring the voltage, the NILM system can utilize the frequency spectrum of the Switched Mode Power Supply (SMPS) signal generated in the switching circuit within the appliance.

When the NILM system uses the voltage, since the configuration of the internal switching circuit is different for each home appliance, the frequency spectrum of the SMPS signal generated when the device is in operation is analyzed and the peak component is observed in different frequency bands , The NILM system can characterize this peak frequency to identify which appliances are operating.

However, when using voltage, the NILM system should measure the SMPS signal generated in tens of kHz band. According to the Nyquist theorem, the voltage measurement device should use a sampling frequency that is at least twice as large as the frequency of the device-specific SMPS signal in order to accurately determine the device-specific voltage usage pattern. Therefore, the NILM system needs to use an expensive voltage measuring device that uses a sampling frequency much higher than the frequency of the SMPS signal.

Korean Patent No. 10-1219416

It is an object of the present invention to provide a frequency measurement device for identifying a frequency of an SMPS signal for each device using a plurality of meters of low cost (using a low sampling frequency) And a home appliance monitoring system including the same.

It is also an object of the present invention to lower the performance of a computing device required in a frequency measuring device by using a low sampling frequency to measure a voltage signal and to reduce the data processing cost.

A frequency measuring device for measuring a voltage signal of a plurality of devices receiving power from a distribution board according to the present invention measures the voltage signal at a specific location on an internal wiring connected to the devices based on a first sampling frequency, A first meter for converting a first voltage signal measured in accordance with the first sampling frequency into a frequency domain to generate a first peak frequency spectrum; A second meter for measuring based on a second sampling frequency different from the sampling frequency and for converting a second voltage signal measured according to the second sampling frequency into a frequency domain to generate a second peak frequency spectrum, Based on the second peak frequency spectrum, the original peak frequency included in the used voltage signal Comprising: a unit for extracting a peak pattern, the first and the second sampling frequency may be characterized in that less than one or two times the frequency of the peaks contained in the peak frequency of the original pattern.

In an embodiment, the first and second sampling frequencies may be smaller than the minimum frequency of the peak frequencies included in the original peak frequency pattern.

In an embodiment, the first and second sampling frequencies may be prime numbers.

In an embodiment, the first and second meters may be normalized to the intensity of the entire spectrum to produce the first and second peak frequency spectra.

As an embodiment, the peak computing unit may determine peak frequency components having a similar relative size in the first and second peak frequency spectrums to corresponding peak frequency pairs.

As an embodiment, when the first peak frequency of the first peak frequency spectrum and the second peak frequency of the second peak frequency spectrum are determined as the peak frequency pair, the effective frequency spectrum according to the sampling frequency based on the Nyquist theorem Wherein the peak computing unit computes a first candidate frequency set of the original peak frequency included in the original peak frequency pattern matched with the first peak frequency and a second candidate frequency set of the original peak frequency included in the original peak frequency pattern matched with the first peak frequency, A second candidate frequency set of the original peak frequency matched with the frequency can be calculated.

As an embodiment, the peak computing unit may select the original peak frequency based on the frequency spreading pattern in the intersection of the first and second candidate frequency sets.

In one embodiment, the first and second meters filter a specific frequency domain through an analog filter to measure the use voltage signal, and the peak computing unit may calculate the use voltage signal from the intersection of the first and second candidate frequency sets, The original peak frequency can be selected by removing the candidate frequency belonging to the first peak frequency.

As an embodiment, the use voltage signal may be measured at a specific location in synchronism with the first and second meters based on a third sampling frequency different from the first and second sampling frequencies, and in accordance with the third sampling frequency And a third meter for converting the measured third voltage signal into a frequency domain to generate a third peak frequency spectrum.

As an embodiment, when it is determined that the third peak frequency of the third peak frequency spectrum corresponds to the peak frequency pair in accordance with the relative magnitude of the peak frequency, the peak computing unit may calculate the peak value of the original A third candidate frequency set of peak frequencies may be calculated and a frequency commonly included in the third candidate frequency set at the intersection of the first and second candidate frequency sets may be determined as the original peak frequency.

In an embodiment, when the first sampling frequency is m1, the second sampling frequency is m2, the first peak frequency is a1, and the second peak frequency is a2, the first candidate frequency set is m1 * n + a1 (n is a natural number), and the second set of candidate frequencies includes frequencies less than the least common multiple of the first and second sampling frequencies, and the second candidate frequency set includes values of m2 * n + a2 And may include frequencies less than the least common multiple of the second sampling frequency.

A household appliance monitoring system for measuring a use voltage signal of a plurality of appliances receiving power from a distribution board according to the present invention and identifying a use state of the appliances, And a plurality of meters for measuring the use voltage signal using the plurality of meters, wherein the voltage signals measured by the meters are Fourier transformed to generate peak frequency patterns, and based on the peak frequency patterns, A frequency measurement device for extracting an original peak frequency pattern and a monitoring device for comparing the original peak frequency pattern with a reference frequency pattern to identify a state of use of the devices, Peak frequency included in Of one of the can is smaller than two times.

In one embodiment, the frequency measuring apparatus measures the use voltage signal at the specific position based on the first sampling frequency, converts the first voltage signal measured according to the first sampling frequency into a frequency domain, A first meter for generating a peak frequency spectrum; a second meter for measuring the voltage at the specific location in synchronization with the first meter based on a second sampling frequency different from the first sampling frequency; A second meter for converting the measured second voltage signal into a frequency domain to produce a second peak frequency spectrum and a second peak frequency spectrum for converting the original peak frequency pattern included in the used voltage signal to a second peak frequency spectrum based on the first and second peak frequency spectra, And a peak computing unit for extracting the peak.

In an embodiment, the monitoring device includes a peak frequency determination system for determining whether the reference frequency pattern is consistent by applying a weight to the original peak frequency pattern, wherein the peak frequency determination system comprises a multi- Can be learned through a regression method (Multinomial Logistic Regression), a Recurrent Neural Network, or a Convolutional Neural Network.

According to an embodiment of the present invention, there is provided a frequency measurement device for identifying a frequency of an SMPS signal for each device currently in use by using a plurality of meters of low cost (using a low sampling frequency), and an appliance monitoring system including the same .

Further, according to the embodiment of the present invention, since the frequency measuring device can reduce the amount of data processed in the calculating device by using a low sampling frequency to measure the voltage signal, the performance of the calculating device can be lowered, The data processing cost can be reduced.

1 is a block diagram illustrating a home appliance monitoring system according to an embodiment of the present invention.
2 is an exemplary illustration of a peak frequency spectrum generated by the first and second meters of FIG. 1;
Figure 3 is a block diagram illustrating an exemplary monitoring device of Figure 1;
FIG. 4 is a block diagram illustrating a home appliance monitoring apparatus according to another embodiment of the present invention.
FIGS. 5A and 5B are diagrams for explaining a learning algorithm used in the frequency measuring apparatus and the monitoring apparatus of FIG. 1 or FIG. 4;

It is to be understood that both the foregoing general description and the following detailed description are exemplary and should provide a further description of the claimed invention. Reference numerals are shown in detail in the preferred embodiments of the present invention, examples of which are shown in the drawings. Wherever possible, the same reference numbers are used in the description and drawings to refer to the same or like parts.

The terms first, second, or the like may be used to describe various elements, but the elements should not be limited by the terms. The terms may be named for the purpose of distinguishing one element from another, for example without departing from the scope of the right according to the concept of the present invention, the first element being referred to as the second element, Similarly, the second component may also be referred to as the first component.

It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between. Expressions that describe the relationship between components, for example, "between" and "immediately" or "directly adjacent to" should be interpreted as well.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, the terms "comprises ", or" having ", and the like, are used to specify one or more of the features, numbers, steps, operations, elements, But do not preclude the presence or addition of steps, operations, elements, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the meaning of the context in the relevant art and, unless explicitly defined herein, are to be interpreted as ideal or overly formal Do not.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, the scope of the patent application is not limited or limited by these embodiments. Like reference symbols in the drawings denote like elements.

≪ Discrete Fourier Transform (DFT) >

The Fourier transform decomposes the signal in the time domain into frequency components that make up the signal. Thus, the Fourier transform of a function over time is a function of the frequency domain. In the voltage-based NILM algorithm used in the present invention, the SMPS signal acquired in the time domain is Fourier-transformed and the frequency component of the SMPS signal is analyzed. The obtained voltage SMPS signal is discontinuous since it is sampled data. To perform Fourier transform on the voltage SMPS signal, a discrete Fourier transform (DFT) is used. The formula of discrete Fourier transform can be expressed as Equation (1).

In Equation (1), n is a variable for discriminating data in the time domain, and k is a variable having the same role in the frequency domain. The discrete Fourier transform of the discrete signal x (n) in the time domain becomes the discrete function X (k) in the frequency domain, and N means the number of data of the frame performing the discrete Fourier transform. That is, the discrete Fourier transform serves to transform data in the N time domains into N relative frequency components. In this case, in order to obtain the actual frequency corresponding to each frequency component of the signal, a frequency resolution factor must be multiplied. Using this, a relationship between the actual frequency f and k of the signal can be expressed as Equation 2 .

In Equation (2), m is the sampling frequency. Since k has a value of 0 to N-1, f has a value of 0 to m (N-1) / N. Therefore, if discrete Fourier transform is performed on N time-domain signals sampled at a sampling frequency of m, frequency components corresponding to frequencies of 0 (DC value) to m included in the signals can be observed. In this case, k can have a total of N values in the spectrum, so that the larger the number N of data in the frame, the better the resolution of the frequency spectrum.

<Nyquist Theorem>

Nyquist theorem states that the signal can be completely reconstructed when the sampling frequency is more than twice the signal bandwidth. Which can be derived from the discrete Fourier transform. The following equations (3) and (4) can be derived in the discrete Fourier transform equation of Equation (1).

Using the equations (3) and (4), the function X (k) in the frequency domain can be extended to a region of a larger frequency. In this case, X (k) is a periodic function with N as a period, and k (n-1) N to nN, based on the median value in each cycle (i.e., (2n-1) N / 2 for natural number n) Is a symmetric function. As a result, a meaningful k in X (k) is a value of 0 to N / 2, and the remaining X (k) is a repetition of this interval (k = 0 to N / 2) or a symmetric repetition of this interval. If the effective interval (k = 0 to N / 2) is multiplied by the frequency resolution m / N, then f is a period of 0 to m / 2. Therefore, the effective frequency conversion spectrum of the signal observed at the sampling frequency of m is f = 0 to m / 2. This means that in order to observe the frequency component of a specific frequency for a signal in a certain time domain, the signal should be measured at a sampling frequency at least twice that of the specific frequency, which is consistent with the contents of the Nyquist theorem.

&Lt; Mapping of a high frequency signal at a low sampling frequency >

The frequency conversion spectrum is valid only up to half of the sampling frequency, and larger frequency components are overlapped and mapped within the valid frequency spectrum range. The reason can be deduced from Equations (3) and (4). When X (k) is extended to a larger frequency region using Equations (3) and (4), it can be seen that the effective frequency band (f = 0 to m / 2) is repeatedly expanded. That is, high-frequency components corresponding to the same X (k) value are combined to form a component value of a specific X (k) within the effective bandwidth. For example, the frequency components corresponding to X (k), X (Nk), X (N + k), X (2N-k), X will be. Considering the actual frequency, it means that all frequency signals such as f, mf, m + f, 2m-f, and 2m + f are mapped to the frequency f with respect to the frequency f with f <m / 2. Thus, when measuring the high frequency of a signal at a low sampling frequency, the high frequency component can be mapped to a specific k value in a frequency spectrum composed of a low frequency. The low frequency k ^* to which the high frequency k is mapped can be obtained as shown in Equation (5).

The mapping in the frequency dimension by multiplying the frequency resolution is shown in Equation (6).

In Equations 5 and 6, k ^* and f ^* are specific values within the domain of the low-frequency sampling spectrum, which are smaller than N / 2 and m / 2, respectively. Hereinafter, the characteristics of the present invention will be described based on discrete Fourier transform, Nyquist theorem, and mapping of a high frequency signal at a low sampling frequency.

&Lt; Method of combining plural low frequency sampling >

The frequency measuring apparatus according to an embodiment of the present invention can extract an original high frequency component corresponding to a voltage signal measured using a plurality of low frequency sampling combination methods. For example, since the original high-frequency component corresponding to the measured voltage signal is mapped based on Equations 5 and 6 when a low sampling frequency is used, a frequency component mapped to a specific value in the frequency spectrum It can be inferred from what high frequency actually occurred. The high frequency k that can be mapped to the specific low frequency k ^* with respect to the signal X (k) obtained by performing discrete Fourier transform on the signal sampled at the sampling frequency m is expressed by Equation (7).

The high frequency f that can be mapped to a specific low frequency f ^* in the frequency dimension by multiplying the frequency resolution is estimated as Equation (8).

The estimated frequency f through Equation (8) is myriad. However, it is possible to select one of the estimated frequencies by measuring the same points at different sampling frequencies using a plurality of low-frequency sampling devices and analyzing the results of the same. And applying it to the voltage-based NILM field, we can observe high-frequency peak frequency even at low sampling frequency and extract it as a feature for NILM classification algorithm.

When the signal is measured at different sampling frequencies, the high frequency included in the signal can be mapped to a specific value of each frequency spectrum according to Equations (5) and (6). On the other hand, based on Equations (7) and (8), it is possible to obtain a set of possible high frequencies corresponding to each spectrum. By comparing these two sets, the original high frequency can be extracted by selecting the overlapped frequency.

For example, in Equation (8), the high frequency f is mathematically expressed as m and the remainder is f ^* or mf ^* . Thus, using two sampling frequencies, the problem of high-frequency estimation can be understood as an algebraic problem in which the remainder is f1 or m1-f1 when divided by m1, and the remaining number f2 or m2-f2 is divided by m2. . Mathematically, according to the Chinese Remainder Theorem, when dividing by m1, the remainder is f1, and when dividing by m2, there exists a maximum of 1 in the least common multiple of m1 and m2. If one of the two residuals is selected, a total of four combinations can be made. Therefore, a total of four frequencies can be obtained within the least common multiple of the two sampling frequencies. Therefore, if the sampling frequency is made to be a prime number or the number of common parts is made to be as few as possible, the least common multiple can be increased so that the effect of high frequency estimation can be maximized. However, if the sampling frequency is excessively low, it is necessary to select an appropriate sampling frequency because a large number of high frequency components are superimposed on the frequency spectrum and analysis may be difficult.

1 is a block diagram illustrating a home appliance monitoring system according to an embodiment of the present invention. 2 is an exemplary illustration of a peak frequency spectrum generated by the first and second meters of FIG. 1; 1, the home appliance monitoring system 100 may include a frequency measuring device 110, a monitoring device 120, and a database 130. [ The database 130 is optional and may be included in the home appliance monitoring system 100 or may use an external database. The frequency measuring apparatus 110 can extract a high frequency component included in the use voltage signal Vsig through the plurality of low frequency sampling combination methods described above.

The devices 21 to 2n used in the home can be connected to the internal wiring 11 to receive power. For example, the devices 21 to 2n may include household appliances that are continuously used for a long time, such as a refrigerator and an air purifier, and appliances that are used intermittently such as a TV, a light, a washing machine, and an air conditioner. The distribution board 10 can transmit electric power supplied from an external power supply source to the internal wiring 11. [

Referring to FIGS. 1 and 2, the frequency measuring apparatus 110 may measure a use voltage signal Vsig of the devices 21 to 2n at a specific point P1. For example, the use voltage signal Vsig may have a form in which the switching voltage signals of the devices currently in use are synthesized. Each of the devices 21 to 2n may have a unique switching voltage signal. Therefore, by detecting the peak frequency pattern of such a switching voltage signal, the device currently in use can be identified.

In Fig. 2, it is assumed that the use voltage signal Vsig has peak frequency components f1, f2, f3. According to the Nyquist theorem, in order to restore a specific signal perfectly, a particular signal must be measured using a sampling frequency that is at least twice as high as the frequency of the particular signal. Therefore, in order to perfectly detect the peak frequency components f1, f2, f3 of the use voltage signal Vsig using one meter, the meter has the largest peak frequency component f3 of the use voltage signal Vsig, The use voltage signal Vsig should be measured using a sampling frequency that is greater than twice the sampling frequency. However, a meter using a sampling frequency that is greater than twice the largest peak frequency component (f3) generally requires significant cost. On the other hand, the frequency measuring apparatus 110 according to the embodiment of the present invention includes a plurality of gauges (low-cost) using a sampling frequency that is smaller than twice the peak frequencies f1, f2, and f3 of the use voltage signal Vsig Can be used to detect the peak frequency components f1, f2, f3 of the voltage signal Vsig.

The frequency measuring apparatus 110 may include a peak computing unit 111, a first meter 112, and a second meter 113. The frequency measuring device 110 can measure the used voltage signal Vsig of the devices 21 to 2n through a plurality of meters having different sampling frequencies at a specific point P1. For example, the first meter 112 may measure the use voltage signal Vsig using the first sampling frequency m1. The second meter 113 may measure the used voltage signal Vsig using the second sampling frequency m2 different from the first sampling frequency m1. The first meter 112 and the second meter 113 can measure the used voltage signal Vsig at the same time in synchronism with each other.

The first meter 112 may generate the first peak frequency spectrum Fsig1 by performing a discrete Fourier transform on the voltage signal measured at the first sampling frequency m1 into the frequency domain. The second meter 113 may generate a second peak frequency spectrum Fsig2 by performing a discrete Fourier transform on the voltage signal measured at the second sampling frequency m2 to the frequency domain. The first and second sampling frequencies m1 and m2 are frequencies less than twice the original peak frequencies f1, f2 and f3 included in the voltage signal Vsig. In one embodiment, the first and second sampling frequencies m1 and m2 may have values less than the smallest peak frequency f1 included in the usable voltage signal Vsig.

On the other hand, when the usable voltage signal Vsig is sampled using a frequency smaller than twice the peak frequencies f1, f2 and f3 included in the usable voltage signal Vsig, the first and second peak frequency spectra The peak frequencies f1 to f3 of the usable voltage signal Vsig may be mapped to frequencies a1 to a3 or b1 to b3 that are lower than the original frequency. For example, the first peak frequency spectrum Fsig1 may include peak frequencies a1, a2, a3. The second peak frequency spectrum Fsig2 may include peak frequencies b1, b2, b3.

The first and second meters 112 and 113 normalize the first and second peak frequency spectra Fsig1 and Fsig2 to the intensity of the entire spectrum to obtain a peak frequency component Fsig1 included in the first peak frequency spectrum Fsig1, May be adjusted to be equal to or similar to the magnitude ratio between the peak frequency components included in the second peak frequency spectrum Fsig2.

Therefore, the peak computing unit 111 can compare the first peak frequency spectrum Fsig1 with the second peak frequency spectrum Fsig2 to match peak frequency components having similar relative sizes. For example, the peak frequency component a3 and the peak frequency component b1 may correspond to the peak frequency component f1. The peak frequency component a2 and the peak frequency component b3 may correspond to the peak frequency component f2. The peak frequency component a1 and the peak frequency component b2 may correspond to the peak frequency component f3. That is, the peak computing section 111 can detect the peak frequency f1 based on the peak frequency a3 and the peak frequency b1. The peak computing section 111 can detect the peak frequency f2 based on the peak frequency a2 and the peak frequency b3. The peak computing section 111 can detect the peak frequency f3 based on the peak frequency a1 and the peak frequency b2.

Referring to Equation (8), the original peak frequency that can be observed at the peak frequency (A) in the peak frequency spectrum subjected to the discrete Fourier transform of the voltage signal after measuring the voltage signal using the sampling frequency (M) * n) &lt; / RTI &gt; + A. Where n is a natural number.

In one embodiment, the original peak frequency f1 measured at the first sampling frequency m1 and observed at the peak frequency a3 is one of (m1 * n) +/- a3 values. Also, the original peak frequency f1 observed at the peak frequency b1 measured by the second sampling frequency m2 is one of the values of (m2 * n) + b1. According to a plurality of low frequency sampling combination methods as described above, the original peak frequency f1 is one of the intersection elements of (m1 * n) ± a3 array and (m2 * n) ± b1 array. When the first and second sampling frequencies m1 and m2 are selected to be values that are smaller than each other, the candidate frequency set for the original peak frequency f1 is a minimum of the first and second sampling frequencies m1 and m2 Can be determined to be four values smaller than the common multiple (m1 * m2). Similarly, the candidate frequency set for the original peak frequency f2 is determined based on a2 and b3, and the candidate frequency set for the original peak frequency f3 can be determined based on a1 and b2.

As an example, in the case of FIG. 2, when m 2 = 1100 Hz, m 1 = 113 Hz, a 2 = 30 Hz, m 2 = 81 Hz and b 3 = 34 Hz, m 1 and m 2 are small. That is, the least common multiple of m1 and m2 is 9153 Hz. The candidates of the original peak frequency f2 possible for the peak frequency component a2 are (113 * n) 30 Hz and the candidates of the original peak frequency f2 for the peak frequency component b3 are 81 * n) &lt; / RTI &gt; Thus, the candidate frequency set for the original peak frequency (f2) less than 9153Hz in the intersection of the (113 * n) ± 30Hz array and the (81 * n) ± 34Hz array is four frequencies (196Hz, 1100Hz, 8053Hz, and 8957Hz ). &Lt; / RTI &gt;

On the other hand, the peak computing unit 111 can extract each of the original peak frequencies f1, f2, and f3 from each candidate frequency set through the determination algorithm. For example, if there is an element of the candidate frequency set coinciding with the peak high-frequency data of the already-learned appliances stored in the database 130, the peak computing unit 111 selects the matching candidate frequency as the original peak frequency . That is, if 1100 Hz is stored in the database 130, the peak computing unit 111 can select 1100 Hz as the original peak frequency f2 in the candidate frequency sets (196 Hz, 1100 Hz, 8053 Hz, and 8957 Hz).

In one embodiment, the peak computing unit 111 can extract the original peak frequency using the frequency spreading pattern. For example, the peak frequency component of an SMPS signal is not mapped exactly to a specific value due to noise inevitably caused by measurement of a voltage signal, and is mapped by spreading around a specific frequency. Since the pattern of frequency spread varies from frequency to frequency, it is possible to extract the corresponding original peak frequency by finding a similar portion of the pattern in which the peak frequency spreads in the low frequency sampling spectrum. That is, when the frequency spreading pattern information corresponding to 1100 Hz is measured at the first and second sampling frequencies (m1, m2) in the database 130, the peak computing unit 111 stores the stored frequency spreading pattern information and The first and second peak frequency spectra Fsig1 and Fsig2 can be compared to select the original peak frequency f2 at 1100 Hz in the candidate frequency set (196 Hz, 1100 Hz, 8053 Hz, and 8957 Hz).

In one embodiment, the frequency measurement device 110 may remove a frequency band of interest in advance using a filter. For example, the frequency measuring apparatus 110 further includes an analog filter circuit (not shown) such as a low-pass filter, a high-pass filter, or a band-pass filter at the front end of the first and second meters 112 and 113 . Since SMPS voltage signals typically occur in the frequency range of several to several tens of kHz, removal of components below 1 kHz and above 100 kHz through the analog filter circuit can prevent uninteresting frequency components from being mapped onto the low frequency sampling spectrum. In addition, if there is a frequency corresponding to the removed region through the analog filter circuit used in the estimated candidate frequency set (196 Hz, 1100 Hz, 8053 Hz, and 8957 Hz), the original peak frequency (1100 Hz) can be extracted have.

On the other hand, the frequency measuring device 110 can remove the determined peak frequency from the frequency spectrum and extract the next peak frequency to smoothly extract the original peak frequencies. For example, the frequency measuring device 110 determines the original peak frequency f2 based on the relatively largest peak frequencies a2 and b3 in the first and second peak frequency spectra Fsig1 and Fsig2 . The frequency measuring device 110 may then smooth out and remove the components of the peak frequencies a2 and b3 in the first and second peak frequency spectra Fsig1 and Fsig2. The frequency measuring device 110 measures the peak frequencies a1 and b2 based on the relatively largest peak frequencies a1 and b2 in the first and second peak frequency spectra Fsig1 and Fsig2 from which the components of the peak frequencies a2 and b3 are removed, The peak frequency f3 can be determined. Likewise, the frequency measuring apparatus 110 is configured to measure the first and second peak frequency spectra Fsig1 and Fsig2 at which the components of the peak frequencies a1, a2, b2, and b3 are removed, the original peak frequency f1 can be determined on the basis of b1. That is, the frequency measuring apparatus 110 can extract the original peak frequencies f1, f2, and f3 while sequentially eliminating the peak frequencies determined in the first and second peak frequency spectra Fsig1 and Fsig2. Therefore, the frequency measuring device 110 can more accurately extract the original peak frequencies f1, f2, and f3.

The frequency measuring apparatus 110 can generate peak pattern data PEAK for the original peak frequency components f1, f2 and f3 based on the first and second peak frequency spectra Fsig1 and Fsig2 have. Meanwhile, the frequency measuring apparatus 110 can learn a judgment algorithm through a learning algorithm. For example, the frequency measurement device 110 may extract the original peak frequency from the candidate frequency set through the decision algorithm. Learning algorithms can include Multinomial Logistic Regression, Recurrent Neural Network (RNN), and Convolutional Neural Network (CNN). The multiple logistic regression method is described in detail in FIGS. 5A and 5B.

The monitoring device 120 can identify the devices currently in use by comparing the peak pattern data (PEAK) with the reference peak pattern data stored in the database 130 through the peak pattern determination system. For example, the peak pattern determination system can search the database 130 for reference peak pattern data that matches the peak pattern data (PEAK). When there is reference peak pattern data that matches the peak pattern data (PEAK), the peak pattern determination system can identify the devices currently in use based on the reference peak pattern data. If there is no reference peak pattern data that matches the peak pattern data (PEAK), the peak pattern determination system may store the peak pattern data (PEAK) in the database 130 as new device information.

Meanwhile, the monitoring apparatus 120 can learn a peak pattern determination system through a learning algorithm. For example, the peak pattern determination system may apply a weight to the peak pattern data (PEAK) to compare with the reference peak pattern data. The learning algorithm can determine the weight using a plurality of learning data. Learning algorithms can include Multinomial Logistic Regression, Recurrent Neural Network (RNN), and Convolutional Neural Network (CNN). The multiple logistic regression method is described in detail in FIGS. 5A and 5B.

The database 130 stores the reference peak pattern data for each device, the data used in the determination algorithm of the peak computing unit 111, the learning data for learning the peak pattern determination system of the monitoring apparatus 120, the weight used in the peak pattern determination system And so on.

As described above, the frequency measuring apparatus 110 according to the embodiment of the present invention can measure the use voltage signal Vsig through a plurality of meters using a sampling frequency smaller than the peak frequency included in the use voltage signal Vsig Can be measured. In addition, the frequency measuring device 110 can extract the original peak frequency pattern (high frequency) of the use voltage signal Vsig based on the peak frequency spectra generated by the plurality of meters. Therefore, the household appliance monitoring system 100 according to the embodiment of the present invention can identify the devices currently in use by using a plurality of low-cost voltage meters.

3 is a block diagram illustrating an exemplary monitoring device 120 of FIG. Referring to FIG. 3, the monitoring device 120 may include a comparing unit 121 and an output unit 122. The comparing unit 121 may identify the devices currently in use based on the peak pattern data (PEAK) received from the frequency measuring device 110. For example, the comparison unit 121 may include a peak pattern determination system. The peak pattern determination system may apply a weight to the received peak pattern data PEAK and compare the peaked peak pattern data PEAK with the reference peak pattern data received from the database 130 to determine whether they match . The peak pattern determination system can identify the devices currently in use based on the result of the match.

The comparison unit 121 can learn a peak pattern determination system through a learning algorithm. For example, the comparison unit 121 may receive learning data for learning a peak pattern determination system. The comparison unit 121 can determine the weight used in the peak pattern determination system based on the learning data through the learning algorithm. The determined weight may be stored in the database 130.

The output unit 122 may display the usage state of each device identified through the comparison unit 121 on the display or store the usage state in the database 130. [ The output unit 122 may output the usage history corresponding to each device using the data stored in the database 130 at the request of the user.

4 is a block diagram illustrating a home appliance monitoring system according to another embodiment of the present invention. 4, the home appliance monitoring system 200 may include a frequency measuring device 210, a monitoring device 220, and a database 230. The database 230 is optional and may be included in the home appliance monitoring system 200 or may use an external database. The configuration of the appliance monitoring system 200 of FIG. 4 is substantially similar to that of the appliance monitoring system 100 of FIG. 1, so that the following description will focus on other configurations.

The frequency measuring device 210 may include a peak calculating unit 211, a first meter 212, a second meter 213 and a third meter 214. The frequency measuring device 210 can measure the used voltage signal Vsig of the devices 21 to 2n through a plurality of meters having different sampling frequencies at a specific point P1. For example, the first meter 212 may measure the use voltage signal Vsig using the first sampling frequency m1. The second meter 213 may measure the used voltage signal Vsig using a second sampling frequency m2 different from the first sampling frequency m1. The third meter 214 may measure the use voltage signal Vsig using a third sampling frequency m3 different from both the first and second sampling frequencies m1 and m2. The first meter 212, the second meter 213 and the third meter 214 can measure the used voltage signal Vsig at the same time in synchronization with each other.

The first meter 212 may generate a first peak frequency spectrum Fsig1 by discrete Fourier transforming the voltage signal measured at the first sampling frequency m1 into a frequency domain. The second meter 213 may generate a second peak frequency spectrum Fsig2 by performing a discrete Fourier transform on the voltage signal measured at the second sampling frequency m2 to the frequency domain. The third meter 214 may generate a third peak frequency spectrum Fsig3 by performing a discrete Fourier transform of the voltage signal measured at the third sampling frequency m3 into the frequency domain. Here, the first, second and third sampling frequencies m1, m2, and m3 are frequencies less than twice the original peak frequency components included in the usable voltage signal Vsig. In one embodiment, the first, second, and third sampling frequencies m1, m2, and m3 may have values less than the smallest peak frequency component included in the usable voltage signal Vsig. 1, the first, second, and third peak frequency spectra Fsig1, Fsig2, and Fsig3 may include different peak frequency components.

The peak computing unit 211 calculates the peak value of the original voltage signal Vsig included in the used voltage signal Vsig based on the plurality of low frequency sampling combination methods in accordance with the method described in Fig. 1 based on the first and second peak frequency spectra Fsig1 and Fsig2. The first candidate frequency set for the peak frequency components (high frequency) can be extracted. The peak computing unit 211 further performs a plurality of low frequency sampling combination methods to extract a second candidate frequency set common to the candidate frequencies based on the third peak frequency spectrum Fsig3 in the first candidate frequency set . The second candidate frequency set may include fewer elements than the first candidate frequency set. Therefore, the peak computing unit 211 can extract the peak pattern data PEAK more accurately than the peak computing unit 111 in FIG.

FIGS. 5A and 5B are diagrams for explaining a learning algorithm used in the frequency measuring apparatus and the monitoring apparatus of FIG. 1 or FIG. 4; 5A is a diagram showing a basic system of a logistic regression. FIG. 5B is a diagram showing a Sigmoid function used in a logistic regression method. FIG.

Logistic regression is a method of predicting the likelihood of an event through linear combination of independent variables. The fact that output (dependent variable) appears as a linear combination of inputs (independent variable) is similar to linear regression, but differs in that the dependent variable is divided into two categories and can be used for binomial classification.

In Fig. 5A, the correction factor B may be interpreted as a weight W0 for one input. The logistic regression is simply a weighting function to classify the input data into a desired value (Y) through equations dividing the points of the n-dimensional space consisting of n independent variables into two regions (two-dimensional lines and three- (W1, W2, ..., Wn). Since the linear combination (Z) of the input variable and the weight has a value in a range that is not constant, it is transformed to have a value between 0 and 1 for convenience of calculation. This process is called activation. The activation of the logistic regression method generally uses the sigmoid function of FIG. 8b. The sigmoid function can be expressed as Equation (9).

The sigmoid function has a value of 0.5 at x = 0, and is a function that is a circle-symmetric based on this. Using the sigmoid function, A has a value between 0 and 1 after the activation process (A = f (Z)) regardless of the Z value. If you convert the resulting value of A to a value closer to 0 and 1, the result will be the output of the system, Y '. If you know the output Y for a particular input X, you can adjust W so that Y 'equals Y. This process is called Back-Propagation, which allows the system to perform Supervised Learning . If you then enter a new input X, you can guess which of the two groups 0 and 1 the result belongs to. At this time, if you learn the system using a sufficient number of I / O data, the W value becomes quite elaborate and the guess value becomes closer to the correct answer. In this way, logistic regression produces a binomial classification system that responds to one of two groups for the input factor.

If the output of the basic logistic regression method is expressed as one of two groups, Multinomial Logistic Regression (or Softmax Regression) is an algorithm that classifies the output into three or more groups by applying it. W Combination One output classifies the output into two types, so if the W combination is equal to the number of groups (g) to be classified, the output can be classified into 2g regions. For example, when the output should be divided into A, B, and C groups, if there are three W combinations, each combination is divided into A or Not A, B or Not B, and C or Not C, of the ²³ zones will be divided into one place. Therefore, when there are m input variables and there are g groups, the weighting array can be expressed as shown in Equation (10).

In this case, the 0th factor of W is a simple expression of the correction value B as a weight for one input. In this case, a response Z to n inputs of data composed of m input variables can be expressed by Equation (11).

The activation process of polynomial logistic regression can use softmax function instead of sigmoid function. If the result of the sigmoid function is a value between 0 and 1, the result of the soft max function is g (number of groups), each of which has a value between 0 and 1, If the soft max function is represented by a matrix, it can be expressed as Equation (12).

By definition, each element of A matrix is a value between 0 and 1, and the sum of each column is 1. Where A _{i, j} means the probability that the j-th input is an i-th group. The object of the present invention is to match one of g devices with respect to each input, so the largest value should be selected for each column of the matrix A. Therefore, in the polynomial logistic regression method, the 'One-Hot Encoding' technique should be applied instead of the binomial classification of FIG. 5A. Applying this technique to each column yields a matrix with the largest value of 1 and the remainder of 0, which is the output of the logistic regression system, Y '. In summary, we assume that Y _{i, j} = 1 in the output Y 'of the g * n matrix when inputting n data with m input variables into the system and guessing that the j-th input corresponds to the i group Is the Forward Propagation of the polynomial logistic regression.

Now, the system constructed by the polynomial logistic regression method should obtain the weighting matrix W that accurately maps each input to one of g groups. Since W is not known at the beginning, it is set to an arbitrary value. Compute Y with the input X here and compare it with the Y created by the actual corresponding value, and modify W so that the two matrices are equal. This process is called Back-Propagation. In the polynomial logistic regression method, a cross entropy technique is used as a back propagation algorithm.

A cost function is a numerical representation of a difference between Y and Y '. In the cross entropy technique, a cost function H can be expressed by Equations (13) and (14).

In equation (13), -logA _{i, j} is 0 when A _{i, j} is 1, and diverges to positive infinity when it is 0. If we add Y _{i, j} to this, we can see from the equation that the cost function (H) is 0 when Y _{i, j} and A _{i, j} are equal, and the cost function (H) .

In the back propagation process, the matrix X is a constant that is not changed as an input, so it can be thought of as a constant multiplied by the weighting factors, and the cost function (H) is a function composed of independent variables Wa0 to Wgm. Therefore, it is possible to take a gradient for W in the cost function H (equation 15).

Since ∇H is a vector that increases the cost function H, to reduce the cost function H, let W0 be a point on the g * (m + 1) dimension produced by the W matrix. _{W = W0} direction. Therefore, if the learning rate is α, W after the back propagation process can be changed as shown in equation (16).

This implies that each factor W _{i, j} of the W matrix changes as shown in equation (17).

The learning rate α is reduced to 0.1, 0.01, and 0.001, and the cost function can be adjusted to decrease the fastest. If you put the input X back into the system where W is adjusted, the output A and hence the Y 'will be a little closer to the Y value.

When the polynomial logistic regression method and the cross entropy method are sufficiently performed, ∇H reaches almost 0, and the W and the cost function (H) hardly change. In the present invention, this point is defined as a point at which the sum of the absolute values of all the element values of | Wnew-Wo | becomes smaller than?. As a system parameter, ε is assumed to be the same as the learning rate α. At this point, we can determine that the cost function (H) reaches a local minimum, where A approaches the local closest to Y at local. In addition, Y 'coincides with Y in an ideal case. Y 'may not be equal to Y if the cost function H is strictly concave or there are many bad data. In this case, W can be set as a new arbitrary value, another minimal value can be found, bad data can be removed, and the learning can be resumed.

If W is determined so that Y '= Y without doing so, it can be judged that the learning of the system has been successfully completed. If the number of learned data is large enough, then the post-learning W will be able to successfully classify new inputs with high probability.

The peak computing units 111 and 211 in FIGS. 1 and 4 can learn a decision algorithm using a polynomial logistic regression method and a cross entropy technique. The monitoring apparatuses 120 and 220 of FIGS. 1 and 4 can learn a peak pattern determination system that compares peak pattern data (PEAK) with reference peak pattern data using a polynomial logistic regression method and a cross entropy method.

The embodiments have been disclosed in the drawings and specification as described above. Although specific terms have been employed herein, they are used for purposes of illustration only and are not intended to limit the scope of the invention as defined in the claims or the claims. Therefore, those skilled in the art will appreciate that various modifications and equivalent embodiments are possible without departing from the scope of the present invention. Accordingly, the true scope of the present invention should be determined by the technical idea of the appended claims.

10: Distribution board
11: Internal wiring
21 ~ 2n: Devices
100, 200: Home appliance monitoring device
110, 210: Frequency measuring device
111 and 211:
112, 212: first meter
113, 213: a second meter
120, 220: Monitoring device
121:
122:
130 and 230:
214: Third meter

Claims (14)

1. A frequency measuring device for measuring a voltage signal of a plurality of devices supplied with power from a distribution board,
Measuring a use voltage signal at a specific location on an internal wiring connected to the devices based on a first sampling frequency and converting a first voltage signal measured according to the first sampling frequency into a frequency domain to obtain a first peak frequency spectrum A first metering device for generating a first metering device;
The second voltage signal being measured in accordance with the first sampling frequency, and the second voltage signal being measured in accordance with the first sampling frequency, A second meter for converting the second peak frequency spectrum to generate a second peak frequency spectrum; And
And a peak computing unit for extracting an original peak frequency pattern included in the use voltage signal based on the first and second peak frequency spectra,
Wherein the first and second sampling frequencies are less than two times one of the peak frequencies included in the original peak frequency pattern,
Wherein the first and second meters normalize to an overall spectral intensity to produce the first and second peak frequency spectra,
Wherein the peak computing unit determines peak frequency components having relatively similar sizes in the first and second peak frequency spectra as corresponding peak frequency pairs,
Frequency measuring device. The method according to claim 1,
Wherein the first and second sampling frequencies are smaller than a minimum one of peak frequencies included in the original peak frequency pattern. The method according to claim 1,
Wherein the first and second sampling frequencies are prime numbers. delete delete The method according to claim 1,
If the first peak frequency of the first peak frequency spectrum and the second peak frequency of the second peak frequency spectrum are determined as the pair of peak frequencies, overlapping in the effective frequency spectrum according to the sampling frequency based on the Nyquist theorem Based on the sampling characteristic of the high frequency component, the peak calculating section calculates a first candidate frequency set of the original peak frequency included in the original peak frequency pattern matched with the first peak frequency, And calculates a second candidate frequency set of the original peak frequency. The method according to claim 6,
Wherein the peak computing unit selects the original peak frequency based on a frequency spreading pattern in an intersection of the first and second candidate frequency sets. The method according to claim 6,
Wherein the first and second meters filter a specific frequency range through an analog filter to measure the used voltage signal,
Wherein the peak computing unit removes a candidate frequency belonging to the specific frequency region from the intersection of the first and second candidate frequency sets to select the original peak frequency. The method according to claim 6,
Measuring the use voltage signal on the basis of a third sampling frequency different from the first and second sampling frequencies in synchronization with the first and second meters at the specific location, And a third meter for converting the voltage signal into a frequency domain to generate a third peak frequency spectrum. 10. The method of claim 9,
When it is determined that the third peak frequency of the third peak frequency spectrum corresponds to the peak frequency pair in accordance with the relative magnitude of the peak frequency, the peak computing section calculates the peak frequency of the original peak frequency matched with the third peak frequency 3 candidate frequency sets and determines the frequency included in common to the third candidate frequency set as the original peak frequency at the intersection of the first and second candidate frequency sets. The method according to claim 6,
When the first sampling frequency is m1, the second sampling frequency is m2, the first peak frequency is a1, and the second peak frequency is a2,
Wherein the first candidate frequency set includes frequencies less than a least common multiple of the first and second sampling frequencies among values of m1 * n + a1 (where n is a natural number)
Wherein the second candidate frequency set includes frequencies less than a least common multiple of the first and second sampling frequencies among values of m2 * n + a2 (where n is a natural number). A household appliance monitoring system for measuring a use voltage signal of a plurality of appliances receiving power from a distribution board to identify a use state of the appliances,
And a plurality of meters for measuring the operating voltage signal using different sampling frequencies at the same time in a specific location on an internal wiring connected to the devices, wherein the voltage signals measured by the meters are Fourier transformed to obtain peak frequency spectra A frequency measurement device for generating an original peak frequency pattern of the used voltage signal based on the peak frequency spectra; And
And a monitoring device for comparing the original peak frequency pattern with a reference frequency pattern to identify the use state of the devices,
Wherein the sampling frequencies are less than two times one of the peak frequencies included in the original peak frequency pattern,
Wherein the monitoring apparatus includes a peak frequency determination system for applying a weight to the original peak frequency pattern to compare the peak frequency pattern with the reference frequency pattern,
The peak frequency determination system may be configured to determine a weighted value by multiplying the weighted value by a multi-logistic regression method, a Recurrent Neural Network, or a Convolutional Neural Network,
Home appliance monitoring system. 13. The method of claim 12,
The frequency measuring apparatus includes:
A first meter for measuring the use voltage signal at the specific location based on a first sampling frequency and converting a first voltage signal measured according to the first sampling frequency into a frequency domain to generate a first peak frequency spectrum;
The second voltage signal being measured in accordance with the first sampling frequency, and the second voltage signal being measured in accordance with the first sampling frequency, A second meter for converting the second peak frequency spectrum to generate a second peak frequency spectrum; And
And a peak computing unit for extracting an original peak frequency pattern included in the use voltage signal based on the first and second peak frequency spectra. delete

Images