WO2012101552A2

WO2012101552A2 - Disaggregation apparatus

Info

Publication number: WO2012101552A2
Application number: PCT/IB2012/050257
Authority: WO
Inventors: Ying Wang; Alessio Filippi; Ronald Rietman
Original assignee: Koninklijke Philips Electronics NV
Current assignee: Koninklijke Philips NV
Priority date: 2011-01-28
Filing date: 2012-01-19
Publication date: 2012-08-02
Anticipated expiration: 2013-07-28
Also published as: WO2012101552A3

Abstract

The invention relates to a disaggregation apparatus for determining electrical statuses of electrical consumers (2, 3, 4) in an electrical network (13). An overall electrical parameter, which is measured at least two different times and which is represented by a first number of sampled electrical parameter values, is used for determining a compressed difference overall electrical parameter, which is represented by a second number of sampled electrical parameter values being smaller than the first number. Compressed electrical signatures of the electrical consumers are provided, and the statuses of the electrical consumers are determined depending on the compressed difference overall electrical parameter and the compressed electrical signatures. Since compressed parameters and signatures are used for the disaggregation, less data have to be processed and have to be stored, thereby reducing the computational and storing efforts needed for the disaggregation.

Description

DISAGGREGATION APPARATUS

FIELD OF THE INVENTION

The invention relates to a disaggregation apparatus, a disaggregation method and a disaggregation computer program for determining electrical statuses of electrical consumers in an electrical network.

BACKGROUND OF THE INVENTION

The article "Nonintrusive Appliance Load Monitoring Based on Integer Programming" by Kosuke Suzuki et al, SICE Annual Conference 2008, August 20 to 22, 2008, The University of Electro-Communications, Japan discloses an apparatus for nonintrusive appliance load monitoring in an electrical network. The apparatus measures the consumed overall electrical current of the electrical network and compares the measured consumed overall electrical current with electrical current signatures of electrical consumers of the electrical network, in order to determine which electrical consumer is switched on and which electrical consumer is switched off. However, this electrical current signature based disaggregation method relies on the high frequency components of the consumed overall electrical current. Accordingly, the consumed overall electrical current of the electrical network has to be measured with a relatively high sampling rate. This high sampling rate leads to a large amount of data, which have to be processed and stored.

SUMMARY OF THE INVENTION

It is an object of the present invention to provide a disaggregation apparatus, a disaggregation method and a disaggregation computer program for determining electrical statuses of electrical consumers in an electrical network, wherein the disaggregation can be performed with less data.

In a first aspect of the present invention a disaggregation apparatus for determining electrical statuses of electrical consumers in an electrical network is presented, wherein the disaggregation apparatus comprises: an electrical parameter measuring unit for measuring an overall electrical parameter of the electrical network at least two different times, wherein the overall electrical parameter is represented by a first number of sampled electrical parameter values, a compressed difference overall electrical parameter determination unit for determining a compressed difference overall electrical parameter from the overall electrical parameter measured at the at least two different times, wherein the compressed difference overall electrical parameter is represented by a second number of sampled electrical parameter values being smaller than the first number,

an electrical signature providing unit for providing compressed electrical signatures of the electrical consumers, wherein a compressed electrical signature is indicative of a status of an electrical consumer and the compressed electrical signatures are represented by a number of sampled electrical parameter values being equal to the second number,

a status determination unit for determining the statuses of the electrical consumers depending on the compressed difference overall electrical parameter and the compressed electrical signatures.

Since a compressed difference overall electrical parameter determination unit determines a compressed difference overall electrical parameter, which is used together with compressed electrical signatures for determining the statuses of the electrical consumers, less data have to be processed and have to be stored.

The status of an electrical consumer is, for example, a switched-on status or a switched-off status of the electrical consumer. An electrical consumer can also comprise more than two statuses, for example, it can comprise a switched-on status, a standby status and a switched-off status. Moreover, an electrical consumer can comprise several statuses which relate to different operation modes of the electrical consumer. For instance, an electrical consumer can be a washing machine, which comprises different statuses depending on different operational modes of the washing machine.

The electrical parameter measuring unit is preferentially adapted to measure the overall electrical parameter of the electrical network at a single point such that it is not necessary to measure an electrical parameter at each electrical consumer for determining the statuses of the electrical consumers. The overall electrical parameter of the electrical network can be measured at a single point and this overall electrical parameter can be used for determining the statuses of the electrical consumers of the electrical network. The overall electrical parameter is preferentially periodic and the overall electrical parameter measured at a certain time corresponds preferentially to a period of the overall electrical parameter measured at that certain time, wherein the period of the overall electrical parameter is sampled by the first number of sampled electrical parameter values. Correspondingly, also the compressed difference overall electrical parameter corresponds to a period. But, the period of the compressed difference overall electrical parameter is sampled by a second number of electrical parameter values being smaller than the first number.

Moreover, also the compressed electrical signature corresponds preferentially to a period, which is sampled by a number of electrical parameter values being equal to the second number.

It is preferred that the electrical parameter measuring unit is adapted to measure the overall electrical current of the electrical network as the overall electrical parameter. It is further preferred that the first number of sampled electrical parameter values is chosen such that the Nyquist theorem is fulfilled. It is also preferred that the second number is smaller than needed for fulfilling the Nyquist theorem. Thus, the difference overall electrical parameter and the electrical signatures can be compressed to a number of sampled electrical parameter values being smaller than the number required for fulfilling the Nyquist theorem, wherein the statuses of the electrical consumers can still be determined depending on the compressed difference overall electrical parameter and the compressed electrical signatures. In an embodiment, the second number is M = O(K₀hi N_d) .

The compressed difference overall electrical parameter determination unit is preferentially adapted to apply a compression procedure for determining the compressed difference overall electrical parameter. The compressed difference overall electrical parameter determination unit can be adapted a) to subtract the overall electrical parameters measured at the at least two different times from each other and to apply the compression procedure to the subtraction result, or b) to apply the compression procedure to the overall electrical parameters measured at the at least two different times and to subtract the resulting compressed overall electrical parameters from each other, in order to determine the compressed difference overall electrical parameter.

It is preferred that the compressed difference overall electrical parameter determination unit is adapted to apply the compression procedure by applying a compression matrix of size second number times first number. It is also preferred that the compression matrix is a random matrix. In a preferred embodiment, the compression matrix comprises matrix elements from a random Gaussian distribution with zero mean and a variance which depends on the inverse of the second number. In a further preferred embodiment, the compression matrix comprises a number of rows being equal to the second number, which are randomly taken from an identity matrix of size first number times first number. A random matrix comprises preferentially randomly distributed matrix values. Using a random matrix as the compression matrix improves the accuracy of determining the statuses of the electrical consumers based on the compressed overall electrical parameter and the compressed electrical signatures, although the amount of data used for determining the statuses of the electrical consumers have been reduced by the compression procedure.

It is preferred that the status determination unit is adapted to determine the statuses by iterative ly solving an equation describing the compressed difference overall electrical parameter as a combination of the compressed electrical signatures of the electrical consumers. It is further preferred that for iteratively solving the equation an -norm minimization method is applied. It is also preferred that the -norm minimization method is a greedy pursuit algorithm. In a preferred embodiment the greedy pursuit algorithm is one of an appliance matching pursuit algorithm and an appliance orthogonal matching pursuit algorithm. The use of these algorithms further improves the accuracy of determining the statuses of the electrical consumers in the electrical network.

It is also preferred that the status determination unit is adapted to apply at least one of an appliance matching pursuit method and an appliance orthogonal matching pursuit method for determining the statuses of the electrical consumers. These methods further improve the quality of determining the statuses of the electrical consumers in the electrical network.

It is further preferred that

the electrical parameter measuring unit is adapted to initially measure an initial overall electrical parameter at an initial time, wherein the initial overall electrical parameter is represented by a number of sampled electrical parameter values being equal to the first number,

the electrical signature providing unit is adapted to provide non- compressed electrical signatures, wherein the non-compressed electrical signatures are represented by a number of sampled electrical parameter values being equal to the first number,

the status determination unit is adapted to determine initial statuses of the electrical consumers depending on the initial overall electrical parameter and the non- compressed electrical signatures, the electrical parameter measuring unit is adapted to measure a further overall electrical parameter at a further time, wherein the further overall electrical parameter is represented by a number of sampled electrical parameter values being equal to the first number,

- the compressed difference overall electrical parameter determination unit is adapted to determine a compressed difference overall electrical parameter from the initial overall electrical parameter and the further overall electrical parameter, wherein the compressed difference overall electrical parameter is represented by a second number of sampled electrical parameter values being smaller than the first number,

- the status determination unit is adapted to determine the statuses of the electrical consumers depending on the compressed difference overall electrical parameter, the compressed electrical signatures and the initial statuses. In particular, if at a first time the statuses of the electrical consumers have been determined, the statuses of the electrical consumers at a second time can be determined based on the compressed difference overall electrical parameter which corresponds to the first time and the second time, the compressed electrical signatures and the statuses determined for the first time. The compressed difference overall electrical parameter and the compressed electrical signatures can be used to determine status changes between the first time and the second time and the statuses of the electrical consumers at the second time can be determined based on the statuses of the electrical consumers at the first time and the determined status changes. The statuses at the first time can be known from, for example, an initialization procedure, which does not use the compression procedure, or the statuses at the first time can be known from other methods, for example, a user can give the status of an electrical consumer at a certain time.

In a preferred embodiment, the disaggregation apparatus comprises a filtering unit for applying a filter to the measured overall electrical parameter, wherein the electrical signature providing unit is adapted to provide filtered compressed electrical signatures, which have been filtered by using the filter, wherein the filter is adapted for at least one of reducing correlations between the compressed electrical signatures and enhancing the dynamic range for a compression used for determining the compressed difference overall electrical parameter. Based on known electrical signatures, the filter can be configured such that correlations between the electrical signatures are reduced and/or the dynamic range for the sampling is enhanced. A lot of different filters, which reduce correlations between electrical signatures and/or enhance the dynamic range for the sampling by the sampling unit, are possible. In an embodiment, the filter is, for example, a notch filter removing utility frequency components, for example, ± 50 Hz components, of the respective electrical parameter. The reduction of the correlations between the compressed electrical signatures and/or enhancing the dynamic range for a compression used for determining the compressed difference overall electrical parameter lead to a further improved accuracy of determining the statuses of the electrical consumers in the electrical network.

In a further aspect of the present invention a disaggregation method for determining electrical statuses of electrical consumers in an electrical network is presented, wherein the disaggregation method comprises:

measuring an overall electrical parameter of the electrical network at at least two different times, wherein the overall electrical parameter is represented by a first number of sampled electrical parameter values,

determining a compressed difference overall electrical parameter from the overall electrical parameter measured at the at least two different times, wherein the compressed difference overall electrical parameter is represented by a second number of sampled electrical parameter values being smaller than the first number,

providing compressed electrical signatures of the electrical consumers, wherein a compressed electrical signature is indicative of a status of an electrical consumer and the compressed electrical signatures are represented by a number of sampled electrical parameter values being equal to the second number,

determining the statuses of the electrical consumers depending on the compressed difference overall electrical parameter and the compressed electrical signatures.

In a further aspect of the present invention a disaggregation computer program for determining electrical statuses of electrical consumers in an electrical network is presented, wherein the computer program comprises program code means for causing a disaggregation apparatus as defined in claim 1 to carry out the steps of the disaggregation method as defined in claim 14, when the computer program is run on a computer controlling the disaggregation apparatus.

It shall be understood that the disaggregation apparatus of claim 1, the disaggregation method of claim 14, and the disaggregation computer program of claim 15 have similar and/or identical preferred embodiments, in particular, as defined in the dependent claims.

It shall be understood that a preferred embodiment of the invention can also be any combination of the dependent claims with the respective independent claim. These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter.

BRIEF DESCRIPTION OF THE DRAWINGS

In the following drawings:

Fig. 1 shows schematically and exemplarily an embodiment of an electrical network comprising a disaggregation apparatus for determining electrical statuses of electrical consumers in the electrical network,

Fig. 2 shows an example of a period of a steady-state current drawn by an electrical consumer in the time domain,

Fig. 3 shows exemplarily the period of the steady-state current drawn by the electrical consumer in the frequency domain,

Figs. 4 to 17 show exemplarily current shapes of electrical signatures of electrical consumers,

Fig. 18 shows exemplarily the rms current strengths of the electrical signatures,

Figs. 19 to 32 show exemplarily the current shape of filtered electrical signatures of the electrical consumers,

Fig. 33 shows exemplarily the rms current strengths of the filtered electrical signatures of the electrical consumers,

Fig. 34 shows schematically and exemplarily an illustration of correlations of current shapes of the electrical signatures,

Fig. 35 shows schematically and exemplarily an illustration of a correlation of the current shapes of filtered electrical signatures of the electrical consumers,

Fig. 36 shows exemplarily an unfiltered measured current signal,

Fig. 37 shows schematically and exemplarily a filtered measured current signal,

Fig. 38 shows exemplarily the filtered current signal at a first time,

Fig. 39 shows exemplarily the filtered current signal at a second time,

Fig. 40 shows exemplarily the difference between the filtered current signals at the first and second times, and Fig. 41 shows a flowchart exemplarily illustrating an embodiment of a disaggregation method for determining electrical statuses of electrical consumers in an electrical network.

DETAILED DESCRIPTION OF EMBODIMENTS

Fig. 1 shows schematically and exemplarily an embodiment of a disaggregation apparatus 1 for determining electrical statuses of electrical consumers 2, 3, 4 in an electrical network 13. In particular, in Fig. 1 the electrical network 13 is represented by an electrical circuit model of a household, where multiple electrical consumers 2, 3, 4 are connected as parallel loads that can be switched on and switched off independently.

The disaggregation apparatus 1 comprises an electrical parameter measuring unit 5 for measuring an overall electrical parameter of the electrical network 13 at at least two different times. In this embodiment, the electrical parameter measuring unit 5 is a current meter for measuring the overall electrical current of the electrical network 13 as the overall electrical parameter. The current meter 5 and a voltage meter 1 1 are placed at a central monitoring point, for example, at a main electric panel of a home. The voltage meter 1 1 measures the source voltage delivered by a power source 12, for example, a power utility. In this embodiment, the source voltage comprises a sinusoidal wave of 50 Hz. When an electrical consumer is turned on and is in the d -th ( d = 1,2, ...,N_d ) operation mode, it draws a current signal denoted as i_d during steady-state. For example, if three electrical consumers are present, wherein each electrical consumer has only a single operation mode, there are three operation modes being the switched-on modes of the three electrical consumers, wherein, if the second electrical consumer is turned on, it is in the second operation mode. The steady- state current waveform has several properties: 1) repeatable with, in this embodiment, a fundamental period of 1/50 s, 2) unique so that it can be used as an electrical signature of the operation mode of the respective electrical consumer, and 3) additive if multiple electrical consumers are switched on. The current meter 5 measures the overall electrical current i^ , i.e. the current signal drawn by the total load, which is the sum of the current signals i_d of the electrical consumers 2, 3, 4 which are switched on.

The disaggregation apparatus 1 further comprises an electrical signature providing unit 7 for providing compressed electrical signatures of the electrical consumers 2, 3, 4. The compressed electrical signatures can be determined by firstly measuring the steady- state current waveforms of each electrical consumer and by then compressing the respective steady-state current waveform. The steady-state current waveform of an electrical consumer can be measured by the current meter 5, while only the respective electrical consumer is switched on. It is also possible that the steady-state current waveforms are measured individually off-line, for example, by using the current meter 5 or another current meter, which is connected to the respective electrical consumer such that only the current of the respective electrical consumer is measured. The compressed steady-state current waveforms form a signature data base for an on-line determination of statuses of the electrical consumers, in particular, for an on-line identification of the electrical consumers which are switched on.

Fig. 2 shows schematically and exemplarily a steady-state current in ampere drawn by a light emitting diode (LED) light bulb being an electrical consumer depending on the time t in seconds s . Fig. 2 corresponds to one period of the steady-state current in the time domain. Fig. 3 shows schematically and exemplarily the corresponding frequency domain spectrum, i.e. the Fourier transform of the time domain waveform shown in Fig. 2, wherein / denotes the frequency.

The Nyquist rate of the current signal, which is defined as the bandwidth of the spectral components satisfying a threshold, e.g., < 20 dB below the peak spectrum, is denoted by F_s . The spectrum in Fig. 3 suggests that the Nyquist rate of the current signal is F_s = 10 kHz. One period (T = l/50s) of the current signal contains N = TF_S Nyquist rate samples.

The current signal of the d -th operation mode is expressed by an Nx 1 vector (for one period), I = A_d\_d (1) where \_d is normalized s.t. \"\'_d = 1 and A_d is the amplitude. The Nyquist rate digital signal representation of (1) can conceptually be viewed as the raw analog current signal.

Although Fig. 1 shows exemplarily only three electrical consumers 2, 3, 4, the electrical network 13 can of course comprise less or more than three electrical consumers. In an embodiment, the electrical network 13 comprises different types of lamps, a vacuum cleaner, a television, a DVD-player, a hairdryer, and a kettle. This set-up is fed by the power source 12 being, in this embodiment, mains power that behaves as a service entry point delivering electricity to a house, in which, in this embodiment, the electrical network is installed. The current meter 5 measures the overall electrical current and the voltage meter 11 measures the mains voltage. The current meter 5 is, for example, an Agilent current probe placed around a single core of a life or neutral wire. The voltage meter 11 is preferentially a differential voltage probe, which in principle could be placed at any socket in the house. Usually, each electrical consumer has its own steady-state current waveform as its unique electrical signature. For some multi-mode electrical consumers like for the kettle, which may have a standby and a cooking mode or like for the vacuum cleaner, which may have a low- power and a high-power mode, each operation mode, i.e. each status of the respective electrical consumer, has a unique steady-state current waveform counted as an electrical signature. The number N _d denotes the number of electrical signatures, which, in this case, is different to the number of electrical consumers. For example, if the electrical network 13 comprises twelve electrical consumers including a kettle and a vacuum cleaner, N _d is fourteen, if the kettle and the vacuum cleaner have two electrical signatures and the other electrical consumers each have a single electrical signature.

In this embodiment, the signal from the current meter 5 is sampled by a 24-bit analog-two-digital converter (ADC) running at E_s =10 kHz, which is the above exemplarily mentioned Nyquist rate. In other embodiments, the current signal from the current meter 5 can also be sampled by another ADC running at another frequency. Since, in this

embodiment, the current signal from the current meter 5 is sampled with a sampling rate being the Nyquist rate, the digitized current signal can conceptually viewed as an analog signal, which can be further processed.

The electrical signature providing unit 7 comprises preferentially a signature data base, which is constructed by collecting the current waveform of each electrical consumer, or, if an electrical consumer comprises several operation modes, of each operation mode of the respective electrical consumer, individually beforehand, i.e. before performing the disaggregation. The respective current waveform can be measured by the current meter 5, while only the respective electrical consumer is switched on, in particular, and is operated in a known certain operation mode, and while the other electrical consumers are switched off. The electrical signature providing unit 7 can be adapted to provide a self-learning algorithm which can be designed to build the signature data base automatically when an unknown electrical consumer is plugged in. For example, in this case a user can be asked via the output unit 10, which is preferentially a display, to switch all other electrical consumers off and to switch only the new unknown electrical consumer on, wherein then the current meter 5 measures the current waveform of the new unknown electrical consumer, which is stored in the signature data base as the signature of this new unknown electrical consumer.

Figs. 4 to 18 show the electrical signatures of the raw current, i.e. without having applied the below described pre- filtering procedure, wherein Figs. 4 to 17 describe the respective current shape denoted by \_d and wherein Fig. 18 shows the rms current strengths

I_rms = A_d I N in ampere, wherein 230 A_d I N gives the rms power in Watts. In Figs. 4 to

18 reference number 14 relates to a compact fluorescent lamp (CFL) with 20 W, reference number 15 relates to a 5 W CFL, reference number 16 relates to a halogen lamp with 50 W, reference number 17 relates to another halogen lamp with 20 W, reference number 18 relates to an incandescent light bulb, reference number 19 relates to an LED, reference number 20 relates to a hair dryer, reference number 21 relates to a vacuum cleaner in the low-power mode, reference number 22 relates to a vacuum cleaner in the high-power mode, reference number 23 relates to a television, reference number 24 relates to a kettle in the standby mode, reference number 25 relates to the kettle in a cooking mode, reference number 26 relates to lcRH and reference number 27 relates to a DVD player.

As it can be seen in Figs. 4 to 17, different loads produce different current shapes, while certain loads with similar circuit characteristics produce very similar current shapes like the resistive loads including hairdryer, kettle in the cooking mode, halogen lamp, and incandescent lamp, which all have a sinusoidal current shape. Fig. 18 shows that, in this embodiment, the current strength ranges from 0.03 A to 9.29 A.

The disaggregation apparatus 1 further comprises a filtering unit 6 for applying a filter to the measured overall electrical parameter, i.e. to the electrical overall current measured by the current meter 5. The filter is adapted to reduce correlations between electrical signatures, in particular, between compressed electrical signatures, and/or to enhance the dynamic range for the compression used for determining the compressed difference overall electrical parameter. The electrical signatures \_d are filtered to improve the signal condition for later processing. The corresponding filtered version of equation (1) is

(2) where W{ ·} represents the filter, in particular without changing the length of

\_d , u_d is normalized s.t. u" u_d = 1 , and B_d is the amplitude. Generally, W{] can be any filter or transform designed to enhance the signal dynamic range for the compression, and/or to reduce the signal correlations for disaggregation. In the remainder of this description, equation (2) is used as a general expression for the current signal, which includes equation

(1) as a special case, i.e., when w \_d ]= \_d , u_d = \_d and B_d = A_d . l_N is an Nx N identity matrix.

A basis matrix of size Nx N_d is defined as *F = [_Ulu₂ u_¾ ] (3)

Ignoring the additive noise term, the overall current signal is

wherein B =

J is a diagonal matrix, and s is an N_d x 1 status signal vector, with elements s_d equal to 1 or 0 indicating whether the d -th status is present or not. Preferentially, equation (4) represents conceptually the analog current signal before ADC, with the bandwidth of F_s Hz.

In an example, the filter W{ ·} is a notch filter removing the ±50 Hz components of the raw signal, with the transfer function of

H(z) = (l - 2cosco₀z^~l + z^~2)/(l - 2rcosco₀z^~1 + r²z^~2) , where ω₀ = 2π50/^ and r = 0.8 is a parameter (0 < r < 1) controlling the width of the notch. The signatures of the filtered current are exemplarity shown in Figs. 19 to 33. Without the 50 Hz components, the higher order harmonics become more dominant and the current shapes become more fluctuating as shown in Figs. 19 to 32. The filtered current strengths ranges from 0.01 A to 0.77 A as shown in Fig. 33, which decreases a lot as compared to the un- filtered current strengths shown in Fig. 18. The similarity between two current shapes (indexed by d_l and d₂ ) can be described by calculating the correlation as K_i (d_l ,d₂) = i"i_d2 for the raw signal, and K_u (d_l ,d_s ) = u^u^ for the filtered signal. The results are exemplarily illustrated in Figs. 34 and 35, where the correlation is represented by gray values, the whiter the more correlated and the darker the less. The signature correlations decrease significantly after notch- filtering, which assists the disaggregation.

The status signal s is denoted as being K₀ -sparse if it contains only

K₀(K₀ « N_d) non-zero entries. If it assumed that at any steady state only a small number of electrical consumers can be simultaneously switched on, the status signal s is K₀ -sparse. However, this assumption may not be correct and the status signal s may be not sparse. Therefore, a new sparse signal is introduced, which considers that it can be assumed that in a short time interval only a small number of electrical consumers are expected to change their status. Denoting τ as the time instant, i.e. the time, when the overall current i is observed, it can be defined by: ΐ = ΨΒ_8τ (5) where s_T is the status signal at the time instant τ . Let τ_ι , τ₂ be two time instants respectively before and after certain status changes of appliances, and τ₂ -τ_ι be a small time interval within which only K₀(K₀ « N_d) appliances change their status. A difference signal can then be defined as Γ= ΐ₂ - ΐ_ι = ΨΒ8' (6) where s' = s_T2 - s contains the change-of- status information. The element of s' can take one of the three values, 0 (no status change), 1 (for example off to on), and -1 (for example on to off). Even if none of s or s_T2 is sparse, s' is very likely K₀ -sparse.

Fig. 36 shows a measured over all current, which has been measured while some electrical consumers are switched on sequentially and some are switched on

simultaneously. Fig. 36 shows the raw signal and Fig. 37 shows the notch- filtered signal. Figs. 36 and 37 show two highlighted short segments with a duration of 0.6 s each, denoted by Ti and τ₂, respectively. The segment τι contains the steady-state measurement where the electrical consumers indexed by [17 19 23 27] are switched on. In between τι and τ₂, some electrical consumers indexed by [14 15 21 26] are switched on simultaneously, and the segment τ₂ contains the steady-state measurement where the electrical consumers indexed by [14 15 17 19 21 23 26 27] are all switched on. The waveform of one period is determined by averaging over the periods in each segment and denoted as and for the segments τ_ι and τ₂ , respectively. For example, if one period is 0.02 s and each segment has a duration of 0.6 s, it is averaged over thirty periods for obtaining the waveforms for the segments τι and τ₂. The difference waveform i ' is given by — . The , and i ' of the notch-filtered signal are exemplarily shown in Figs. 38 to 40.

The disaggregation apparatus 1 further comprises a compressed difference overall electrical parameter determination unit 8 for determining a compressed difference overall electrical parameter from the overall electrical parameter measured at two different times τ_ι and τ₂ . For example, the current meter 5 measures the overall electrical current at the times τ_ι and τ₂ , wherein the measured overall electrical currents are pre-filtered by the filtering unit 6 and a difference signal is calculated in accordance with equation (6). The difference signal is then compressed by the compressed difference overall electrical parameter determination unit 8 for determining the compressed difference overall electrical parameter i'_m preferentially in accordance with following equation:

A new basis matrix ^v¥_m is given by

Ψ_Β = ΦΨ = [ιη₁ ιη₂ - -η_¾ ] (8) where = Φιι^ is the basis vector of the measured sampled signal.

Thus, the compressed difference overall electrical parameter determination unit 8 is adapted to apply a compression procedure for determining the compressed difference overall electrical parameter i'_m . In accordance with equations (6) and (7), the compressed difference overall electrical parameter determination unit subtracts preferentially the overall electrical parameters and , which have been measured at two different times τ_ι and τ₂ and which have been filtered, from each other and applies the compression procedure to the subtraction result. In another embodiment, the compression procedure can also be applied to the overall electrical parameters, which have been measured at the two different times and which have been filtered, and the compressed overall electrical parameters can be subtracted from each other, in order to determine the compressed difference overall electrical parameter.

The columns of Ψ_ΜΒ can be regarded as being the compressed electrical signatures, which are used for determining the difference status signal s' based on the compressed difference overall electrical parameter V_m .

The compressed difference overall electrical parameter determination unit 8 is adapted to apply the compression procedure by applying a compression matrix Φ having a size of M x JV (M < N) being a random matrix, i.e. having randomly distributed matrix values. The number N is a first number of sampled electrical parameter values of the measured overall electrical parameter, i.e., in this embodiment, of the overall electrical current measured by the current meter 5. Correspondingly, also the un-compressed difference overall electrical parameter i ' is sampled by N electrical parameter values. Since M < N , the compressed difference overall electrical parameter i'_m is sampled by a second number M being smaller than the first number N .

The compressed difference overall electrical parameter determination unit 8 also compresses the electrical signatures by using the compression matrix Φ , wherein the resulting compressed electrical signature are stored in the signature data base of the electrical signature providing unit 7. The compressed electrical signatures are each represented by M sampled electrical parameter values.

The compression matrix Φ comprises preferentially matrix elements from a random Gaussian distribution, in particular with zero mean and a variance which depends on the inverse of M . In another embodiment, the compression matrix can comprise a number of rows being equal to M , wherein the rows are randomly taken from the identity matrix with the size N x N . The compression matrix Φ represents a compressive sampling with an analog-to-information converter (AIC) running at (M I N F_s Hz. Since s' is K₀ -sparse, the compressive random matrix Φ , which is preferentially non-adaptive, can be used to compress i ' to only M = O(K₀hi N_d) samples, from which a reliable estimation of s' can be achieved.

The compression procedure defined by equation (7) can be implemented by using analog circuits, where the multiplication of Φ and i ' is performed using mixers and integrators as disclosed in, for example, the article "Random sampling for analog-to- information conversion of wideband signals", by J. Laska et al., IEEE Power and Energy Magazine, pp. 56-63, Mar/ Apr 2003, which is herewith incorporated by reference. The resulting compressed overall electrical parameter \'_m is used by a status determination unit 9 for determining the statuses of the electrical consumers depending on the compressed difference overall electrical parameter and the compressed electrical signatures. The status determination unit can comprise a local microprocessor attached to, for example, the current meter and/or the voltage meter, or the status determination unit can comprise a remote personal computer, to which the current meter 5 may send the measured currents, wherein in the latter case the current meter 5 is equipped with a wireless or wired communication module. The compression lowers preferentially both, the processing complexity and the communication burden.

The status determination unit 9 is preferentially adapted to determine the difference status signal s' based on the compressed difference overall electrical parameter \'_m , given ^v¥_m and B representing the known compressed electrical signatures of the electrical consumers. Equation (7) can be rewritten as m Ψ m z' (9) where z'= Bs' is the difference status signal weighted by the amplitudes. One straightforward solution of z' can be obtained via the least-square (LS) estimate,

where† denotes the Moore-Penrose pseudoinverse of a matrix. The LS minimizes the /₂ -norm. However, preferentially the l_x -norm minimization approach is employed, i.e., z'= arg min z' l s.t. i' = Ψ z (11)

The optimized problem of 11 can be solved by greedy pursuit algorithms such as orthogonal matching pursuit (OMP) as disclosed in, for example, the article "Signal recovery from random measurements via orthogonal matching pursuit" by J. Tropp and A. Gilbert, IEEE Trans, on Information Theory, pp. 4655-4666, December 2007 which is herewith incorporated by reference. The status determination unit 9 can, for example, be adapted to apply an OMP algorithm, which is adapted to the settings of the electrical consumer identification, wherein this algorithm is in the following denoted as A-OMP.

However, the status determination unit 9 can also be adapted to apply another algorithm for solving equation 11 like an appliance orthogonal matching (A-MP) algorithm.

The A-MP and the A-OMP algorithms will in the following be described in more detail. The set Ω = {ΐ,2,...,Λ^ } is the index set of all operation modes of the electrical consumers, wherein = { d_l , d₂ , ... , d_K } <= Ω is the index set of the active operation modes selected by the algorithms. An electrical consumer can comprise one or several operation modes. If an electrical consumer has only one operation mode, then this operation mode corresponds to an on-status of the respective electrical consumer. If an electrical consumer has several operation modes defining several possible statuses of the respective electrical consumer, the respective electrical consumer can be in one of the several statuses. The A-MP algorithm has as input i'_m ,Ψ_Μ ,Β, and the thresholds α,ε . The A-MP algorithm is initialized by setting the residual r₀ to the compressed difference overall electrical parameter (r₀ = \'_m ) and the index set Ω₀ to an empty set (Ω₀ = 0) . In the k -th iteration preferentially following steps are performed:

1. Select some candidate electrical consumers

P_d = r m_rf /|| m_rf | .

C = { d:p_d≥ a max p_d d e Ω \ Ω_ί_₁ .

2. From the candidate set C , select the one electrical consumers that minimizes the residual power

d_k = argmin|| r_i_₁ - B_{d d} \ ,

n_k =n_k_^ { d_k } .

3. Update the residual

The iteration is preferentially performed until a stop criterion is fulfilled. In an embodiment, the stop criterion is fulfilled if e_k < ε or k = N_d . The result is the set Ω₄ , wherein the index k denotes the number of performed iterations.

The A-OMP algorithm has as input i'_m , Ψ_ιη , and a threshold ε . The A-OMP algorithm is initialized by setting the residual r₀ to the compressed difference overall electrical parameter (r₀ = Y_m ) and the index set Ω₀ to an empty set (Ω₀ = 0) . Moreover, initially a matrix containing the orthogonalized selected basis vectors Γ₀ = [ ] is provided.

The k -th iteration of the A-OMP algorithm comprises preferentially the following steps:

1. Select one electrical consumer and add to the index set

d_k = arg max p Ω, = Ω^ u { d_k } .

d e l,-,N_d

2. Orthogonalize and normalize the selected basis vector, and add it to the orthogonalized basis matrix

g* =™¾ ^/||™¾ || ^»

3. Update the residual

rk ^{= r}k-l ^~ (^ri-l§i )§i: '

The iteration stops, if a stop criterion is fulfilled, for example, if e_k < ε or k = N_d . The output of the A-OMP algorithm is the index set Ω^ .

A-MP has less computational complexity than A-OMP, since A-OMP performs matrix multiplication in the orthogonalization step of each iteration and also needs more iterations to converge. A-OMP is more robust than A-MP when the electrical consumers signature waveforms are very correlated. Given Ω^ , a mutilated basis matrix

can be obtained such that equation (9) can be rewritten as \'_m = Ύ _{m K}z . Then z'_K is estimated by z ^Λ' K = *P '^† m,K i^l' , (12) Similarly, a mutilated difference status signal vector is defined as

S K ^• -s'_d ]^T . Knowing the amplitudes, we estimate s'^ and obtain a refined index set as

where a threshold is set to be 0.5 given that the difference status signal is either 0 or ±1 with equal probability assumed.

The status determination unit 9 is preferentially adapted to use the difference status signal together with a previously determined status signal, which defines the previous statuses of the electrical consumers, for determining a current status signal, which is indicative of the current statuses of the electrical consumers. For example, the status at a certain time τ _n can be obtained by summing the initial statuses, for example, at an initial time τ = 0 , and all determined difference status signals encountered between the time, at which the initial statuses have been determined, and the certain time τ .

The status determination unit 9 can be adapted to determine an initial status signal at an initial time by using the above described equations for determining the difference status signal, wherein the strokes at the variables are removed, for example, wherein \'_m , z',s' are replaced by i_m , z,s , respectively. The variable \_m denotes a compressed overall electrical parameter, which corresponds to a certain time, the variable z is defined by equation (10) without strokes, and the variable s denotes the status signal at the certain time.

For determining the status signal s equation (14) becomes:

where a threshold is set to be 0.5 given that the status signal is either 0 or 1 with equal probability assumed.

In an embodiment, the disaggregation apparatus 1 is adapted to employ a two- step approach. The time τ_ι is, in this embodiment, an initial time instant at which the statuses of the electrical consumers are unknown. Between this initial time instant τ_ι and a second time τ₂ , i.e. a second time instant shortly after the first time instant τ_ι , at most K₀ electrical consumers change their status, for example, their on/off status, and trigger a simple event detector which is, for example, based on a measured power variation of the total load. The difference between the initial time instant τ_ι and the second time instant τ₂ is, for example, some seconds, in particular, three seconds. The variable K₀ is preferentially smaller than ten and can be, for example, eight. In the initial detection step at the initial time τ_ι the electrical parameter measuring unit 5 initially measures preferentially an initial overall electrical current, wherein the initial overall electrical current is represented by a number of sampled electrical current values fulfilling the Nyquist criterion. Thus, initially a Nyquist rate sampling for measuring the overall electrical current is performed. The status determination unit 9 determines then in this initial step initial statuses of the electrical consumers depending on the initial overall electrical current and the electrical signatures, i.e. the initial status signal s is determined from , which is sampled with, for example, 10 kHz. After the initial step, a differential detection is performed at the time instant τ₂ and/or at any later time instants τ _n , wherein n is larger than two. The electrical parameter measurement unit 5 measures preferentially a further overall electrical current at the time instant τ₂ and/or at later time instants τ _n , wherein immediately after the measurement the overall electrical current is also represented by a number of sampled electrical parameter values, which fulfill the Nyquist criterion. This further overall electrical current is then compressed by the compressed difference overall electrical parameter determination unit 8 to determine a compressed difference overall electrical current from the initial overall electrical current at the time τ_ι and the further overall electrical current at the time τ₂ or, for further later time instants τ _n , from an overall electrical current measured at a previous time instant τ_η and the overall electrical current measured at the time τ _n . The resulting compressed difference overall electrical current V_m is represented by a number M of sampled electrical parameter values being smaller than the sampling number used in the initial step. For example, the number of sampled electrical parameter values of the overall electrical current \_m can correspond to a sampling rate of 1 kHz. The status determination unit 9 is then preferentially adapted to determine the statuses of the electrical consumers depending on the compressed difference overall electrical current Y_m , the compressed electrical signatures and the initial statuses. In particular, the statuses at the time τ₂ can be given by s_T2 = s + s' and the statuses at the time τ_η can be given by s_T = s + s' .

The initial step is preferentially performed once at the beginning of the monitoring process, and the further steps at the times τ₂ and at possible further times τ_η are performed during the remaining of the monitoring process. These further steps can be triggered by a simple event detector. For example, the disaggregation apparatus can be adapted such that the statuses are determined at the further time instants τ₂ and τ _n , if a change in a measured overall power is larger than a predefined threshold. The disaggregation apparatus can be adapted to re-set the entire monitoring process according to a programmed schedule or by user intervention.

The change-off-status signal s' is K₀ -sparse ensured by the known switch continuity principle (SCP), which allows compressing the measured overall current to only K₀\nN_d samples and still estimating s' correctly. If s_T2 is directly determined from the compressed overall current measured at τ₂ , then s_T2 may not be sparse enough to allow much compression.

The compressive sampling described by the compression matrix Φ is preferentially a linear operation. Thus, in a preferred embodiment it doesn't matter whether overall electrical currents measured at different times are first compressed and then subtracted from each other or whether these overall electrical currents are firstly subtractive from each other and then compressed.

In the following an embodiment of a disaggregation method for determining electrical statuses of electrical consumers in an electrical network will exemplarily be described with reference to a flowchart shown in Fig. 41.

In step 101 , an overall electrical current of the electrical network is measured at at least two different times τ_ι and τ₂ , wherein the overall electrical current measured at the at least two different times is represented by a first number N of sampled electrical parameter values. In step 102, a compressed difference overall electrical current V_m is determined from the overall electrical current measured at the at least two different times τ_ι and τ₂ , wherein the compressed difference overall electrical current \'_m is represented by a second number M of sampled electrical parameter values being smaller than the first number N . In step 103, compressed electrical signatures of the electrical consumers are provided, and, in step 104, the statuses of the electrical consumers are determined depending on the compressed difference overall electrical current and the compressed electrical signatures. In particular, status changes s' are determined from the compressed overall electrical current and the compressed electrical signatures and the determined status changes are used together with previously determined statuses, which have been determined for the first time τ_ι of the at least two times τ_ι and τ₂ , for determining the statuses of the electrical consumers at the time τ₂ as described above.

Known current signature based disaggregation methods rely on their high frequency components of the current to make the individual appliance, i.e. the individual electrical consumers, more unique/distinguishable, which typically requires higher sampling rates, for example, a sampling rate of 40 kHz, than power signature based methods, which may require a sampling rate of, for example, 2 to 10 Hz. Besides the corresponding sampling burden on the ADC, the resulting high data rate also imposes burden on the processing power, i.e. the computational complexity of the disaggregation algorithm, the data storage, i.e., for example, the signature data base, and the data communication, if the electrical parameter measuring unit, for example, a current meter, needs to send the measured electrical parameters to a remote personal computer or digital signal processor, which perform the disaggregation algorithm.

The disaggregation apparatus and disaggregation method described above with reference to Figs. 1 to 41 and can provide an on-line monitoring solution which performs compression on the measured current without degrading the disaggregation performance. As result, the data rate can be significantly reduced to enable more efficient computation, storage, and communication. Preferentially, the compressed sampling (CS) theory and the switched continuity principle (SCP), according to which in a short time interval only a small number of electrical consumers are expected to change their status, are exploited, in order to apply a sparse reconstruction approach on the compressed measurement and still reliably identify the individual electrical consumers' status, in particular, their on/off status, as if the non-compressed measurement were used.

The disaggregation method can be regarded as being a difference current waveform measurement method, which is based on the SCP and exploits the sparsity of the current of the total load in terms of the change-of-status information, for example, off-to-on or on-to-off, for the individual electrical consumers. The sparsity allows us to compress the measured current without destroying the essential information for disaggregation.

As explained above, the disaggregation apparatus and method may perform a two-step approach for an on-line monitoring process. A first step being the initial detection step is performed once at the beginning of the monitoring process, wherein the non- compressed current is used. A second step being the differential detection step can be repeatedly performed during the remaining of the monitoring process, wherein the compressed current is used. The second step is preferentially triggered by a simple/coarse event detector, for example, based on the power variation of the total load. The entire monitoring process can be reset according to a programmed schedule or by user intervention, wherein, in this case, the disaggregation apparatus comprises a user interface like a reset button for indicating that the entire monitoring process should be reset. The disaggregation apparatus and the disaggregation method can be adapted to perform the compression in the digital domain, wherein a conventional ADC is still used, which lowers the burden of computation, storage and communication. However, the disaggregation apparatus and disaggregation method can also be adapted to perform the compression in the analog domain, wherein the conventional ADC is replaced with the AIC, which also lowers the sampling burden.

The compression can reduce the number of the measurement samples and also the size of the signature database, which can lower the processing complexity, the storage and communication burden.

The disaggregation apparatus and disaggregation method can be used for, for example, smart energy monitoring, energy disaggregation, smart energy control, et cetera.

Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.

In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality.

A single unit or device may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Determinations like the determination of the compressed difference overall electrical parameter or the determination of the statuses of the electrical consumers performed by one or several units or devices can be performed by any other number of units or devices. For example, steps 102 to 104 can be performed by a single unit or by any other number of different units. The determinations and/or the control of the disaggregation apparatus in accordance with the disaggregation method can be implemented as program code means of a computer program and/or as dedicated hardware. A computer program may be stored/distributed on a suitable medium, such as an optical storage medium or a solid-state medium, supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems.

Any reference signs in the claims should not be construed as limiting the scope.

The invention relates to a disaggregation apparatus for determining electrical statuses of electrical consumers in an electrical network. An overall electrical parameter, which is measured at least two different times and which is represented by a first number of sampled electrical parameter values, is used for determining a compressed difference overall electrical parameter, which is represented by a second number of sampled electrical parameter values being smaller than the first number. Compressed electrical signatures of the electrical consumers are provided, and the statuses of the electrical consumers are determined depending on the compressed difference overall electrical parameter and the compressed electrical signatures. Since compressed parameters and signatures are used for the disaggregation, less data have to be processed and have to be stored, thereby reducing the computational and storing efforts needed for the disaggregation.

Claims

CLAIMS:

1. A disaggregation apparatus for determining electrical statuses of electrical consumers (2, 3, 4) in an electrical network (13), the disaggregation apparatus (1) comprising:

- an electrical parameter measuring unit (5) for measuring an overall electrical parameter of the electrical network (13) at least two different times, wherein the overall electrical parameter is represented by a first number of sampled electrical parameter values,

- a compressed difference overall electrical parameter determination unit (8) for determining a compressed difference overall electrical parameter from the overall electrical parameter measured at the at least two different times, wherein the compressed difference overall electrical parameter is represented by a second number of sampled electrical parameter values being smaller than the first number,

- an electrical signature providing unit (7) for providing compressed electrical signatures of the electrical consumers (2, 3, 4), wherein a compressed electrical signature is indicative of a status of an electrical consumer and the compressed electrical signatures are represented by a number of sampled electrical parameter values being equal to the second number,

- a status determination unit (9) for determining the statuses of the electrical consumers (2, 3, 4) depending on the compressed difference overall electrical parameter and the compressed electrical signatures.

2. The disaggregation apparatus as defined in claim 1, wherein the electrical parameter measuring unit (5) is adapted to measure the overall electrical current of the electrical network (13) as the overall electrical parameter.

3. The disaggregation apparatus as defined in claim 1, wherein the first number of sampled electrical parameter values is chosen such that the Nyquist theorem is fulfilled.

4. The disaggregation apparatus as defined in claim 1, wherein the second number is smaller than needed for fulfilling the Nyquist theorem.

5. The disaggregation apparatus as defined in claim 1, wherein the compressed difference overall electrical parameter determination unit (8) is adapted to apply a compression procedure for determining the compressed difference overall electrical parameter.

6. The disaggregation apparatus as defined in claim 5, wherein the compressed difference overall electrical parameter determination unit (8) is adapted a) to subtract the overall electrical parameters measured at the at least two different times from each other and to apply the compression procedure to the subtraction result, or b) to apply the compression procedure to the overall electrical parameters measured at the at least two different times and to subtract the resulting compressed overall electrical parameters from each other, in order to determine the compressed difference overall electrical parameter.

7. The disaggregation apparatus as defined in claim 5, wherein the compressed difference overall electrical parameter determination unit (8) is adapted to apply the compression procedure by applying a compression matrix of size second number times first number.

8. The disaggregation apparatus as defined in claim 7, wherein the compression matrix is a random matrix.

9. The disaggregation apparatus as defined in claim 8, wherein the compression matrix comprises matrix elements from a random Gaussian distribution with zero mean and a variance which depends on the inverse of the second number.

10. The disaggregation apparatus as defined in claim 8, wherein the compression matrix comprises a number of rows being equal to the second number, which are randomly taken from an identity matrix of size first number times first number.

11. The disaggregation apparatus as defined in claim 1 , wherein the status determination unit (9) is adapted to determine the statuses by iterative ly solving an equation describing the compressed difference overall electrical parameter as a combination of the compressed electrical signatures of the electrical consumers.

12. The disaggregation apparatus as defined in claim 1 , wherein

- the electrical parameter measuring unit (5) is adapted to initially measure an initial overall electrical parameter at an initial time, wherein the initial overall electrical parameter is represented by a number of sampled electrical parameter values being equal to the first number,

- the electrical signature providing unit (7) is adapted to provide non- compressed electrical signatures, wherein the non-compressed electrical signatures are represented by a number of sampled electrical parameter values being equal to the first number,

- the status determination unit (9) is adapted to determine initial statuses of the electrical consumers depending on the initial overall electrical parameter and the non- compressed electrical signatures,

- the electrical parameter measuring unit (5) is adapted to measure a further overall electrical parameter at a further time, wherein the further overall electrical parameter is represented by a number of sampled electrical parameter values being equal to the first number,

- the compressed difference overall electrical parameter determination unit (8) is adapted to determine a compressed difference overall electrical parameter from the initial overall electrical parameter and the further overall electrical parameter, wherein the compressed difference overall electrical parameter is represented by a second number of sampled electrical parameter values being smaller than the first number,

- the status determination unit (9) is adapted to determine the statuses of the electrical consumers depending on the compressed difference overall electrical parameter, the compressed electrical signatures and the initial statuses.

13. The disaggregation apparatus as defined in claim 1, wherein the

disaggregation apparatus (1) comprises a filtering unit (6) for applying a filter to the measured overall electrical parameter, wherein the electrical signature providing unit (7) is adapted to provide filtered compressed electrical signatures, which have been filtered by using the filter, wherein the filter is adapted for at least one of reducing correlations between the compressed electrical signatures and enhancing the dynamic range for a compression used for determining the compressed difference overall electrical parameter.

14. A disaggregation method for determining electrical statuses of electrical consumers in an electrical network, the disaggregation method comprising:

- measuring an overall electrical parameter of the electrical network at at least two different times, wherein the overall electrical parameter is represented by a first number of sampled electrical parameter values,

- determining a compressed difference overall electrical parameter from the overall electrical parameter measured at the at least two different times, wherein the compressed difference overall electrical parameter is represented by a second number of sampled electrical parameter values being smaller than the first number,

- providing compressed electrical signatures of the electrical consumers, wherein a compressed electrical signature is indicative of a status of an electrical consumer and the compressed electrical signatures are represented by a number of sampled electrical parameter values being equal to the second number,

- determining the statuses of the electrical consumers depending on the compressed difference overall electrical parameter and the compressed electrical signatures.

15. A disaggregation computer program for determining electrical statuses of electrical consumers in an electrical network, the computer program comprising program code means for causing a disaggregation apparatus as defined in claim 1 to carry out the steps of the disaggregation method as defined in claim 14, when the computer program is run on a computer controlling the disaggregation apparatus.