JP2016017917A - Apparatus power consumption estimation device and computer program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an apparatus power consumption estimation device and a computer program capable of creating a database used in estimating the operating state of an electrical apparatus.SOLUTION: An apparatus power consumption estimation device of the embodiment comprises: an apparatus feature learning unit; an apparatus feature database; an operating state estimation unit; and a power consumption estimation unit. The apparatus feature learning unit acquires a feature quantity of the operating state of an apparatus from the higher harmonic wave of a current or electric power obtained from time-series data of a voltage and current measured in a power feeding path. The apparatus feature database stores the acquired feature quantity of the operating state of the apparatus. The operating state estimation unit estimates the operating state of the apparatus on the basis of the feature quantity of the higher harmonic wave acquired from the higher harmonic wave of the current or electric power and the feature quantity of the operating state of the apparatus stored in the apparatus feature database. The power consumption estimation unit estimates power consumption of the apparatus on the basis of the estimated operating state.SELECTED DRAWING: Figure 1

Description

  Embodiments described herein relate generally to an apparatus power consumption estimation apparatus and a computer program.

  As a technology to measure the operating status of electrical equipment including inverter equipment individually, instead of installing a measuring instrument for each electrical equipment, each electrical equipment is non-intrusive based on measurements on the lead-in line from the utility pole to the home. Techniques for estimating the operating state have been proposed. As one of the techniques, there is known a method of using harmonics of a current waveform when estimating an operation state from a current waveform of an electric device. In such a method, the operation state is estimated by comparing the characteristics of the harmonics of the electric device with a database in which the characteristics of the harmonics in the operation state of each electric device are stored in advance. However, it may be time-consuming to store the harmonic characteristics of each electrical device in the database.

JP 2012-189526 A JP 2013-79853 A

  The problem to be solved by the present invention is to provide a device power consumption estimation device and a computer program that can easily create a database used for estimating an operating state of an electrical device.

  The device power consumption estimation apparatus of the embodiment includes a device feature learning unit, a device feature database, an operation state estimation unit, and a power consumption estimation unit. The device feature learning unit acquires a feature amount of the operation state of the device from the harmonics of the current or power obtained from the time-series data of the voltage and current measured in the power feeding path. The device feature database stores the acquired feature amount of the operating state of the device. The operation state estimation unit estimates the operation state of the device based on the feature quantity of the harmonic obtained from the harmonic of the current or power and the feature amount of the operation state of the device stored in the device feature database. To do. The power consumption estimation unit estimates the power consumption of the device based on the estimated operation state.

The schematic block diagram showing the functional structure of the apparatus power consumption estimation apparatus 100 of embodiment. The schematic block diagram showing the structure of the apparatus characteristic learning part 30 concerning 1st Embodiment. The figure for demonstrating the process which the clustering part 32 performs. The flowchart which shows the flow of a process of the apparatus characteristic learning part 30 concerning 1st Embodiment. The schematic block diagram showing the structure of the apparatus characteristic learning part 30a concerning 2nd Embodiment. The figure for demonstrating cluster number control by an information criterion. The flowchart which shows the flow of a process of the apparatus characteristic learning part 30a concerning 2nd Embodiment. The schematic block diagram showing the structure of the apparatus characteristic learning part 30b concerning 3rd Embodiment. The flowchart which shows the flow of a process of the apparatus characteristic learning part 30b concerning 3rd Embodiment. The schematic block diagram showing the structure of the apparatus power consumption estimation apparatus 100a concerning a modification. FIG. 3 is a schematic block diagram illustrating a configuration of a position information acquisition unit 70.

Hereinafter, the apparatus power consumption estimation apparatus 100 of embodiment is demonstrated with reference to drawings.
FIG. 1 is a schematic block diagram illustrating a functional configuration of a device power consumption estimation apparatus 100 according to the embodiment.
The device power consumption estimation apparatus 100 estimates the operating state of each home appliance based on voltage and current measurement data of a household electrical appliance (hereinafter simply referred to as “home appliance”), and the estimation result of the power consumption of each home appliance. Is a device that outputs.

  The apparatus power consumption estimation apparatus 100 includes a CPU (Central Processing Unit), a memory, an auxiliary storage device, and the like connected by a bus, and executes a power consumption output program. By executing the power consumption output program, the device power consumption estimation apparatus 100 includes a voltage / current measurement unit 10, a harmonic calculation unit 20, a device feature learning unit 30, a device feature DB 40, an operation state estimation unit 50, and a power consumption estimation unit 60. It functions as a device provided with. Note that all or part of each function of the device power consumption estimation apparatus 100 is realized by using hardware such as an application specific integrated circuit (ASIC), a programmable logic device (PLD), and a field programmable gate array (FPGA). Also good. The power consumption output program may be recorded on a computer-readable recording medium. The computer-readable recording medium is a storage device such as a flexible disk, a magneto-optical disk, a portable medium such as a ROM (Read Only Memory) and a CD-ROM, and a hard disk built in the computer system. The power consumption output program may be transmitted / received via a telecommunication line.

The voltage / current measurement unit 10 records time-series data of voltage and current measured in the power feeding path at a constant period. The power supply path is a path through which electric power is supplied to the home appliance, for example, a power supply line or a distribution board.
The harmonic calculation unit 20 calculates the feature quantity of current and power harmonics based on the time-series data. Specifically, first, the harmonic calculation unit 20 receives time-series data and detects the change time of the operating state from the time-series data. The change time of the operation state means that the home appliance is changed from ON to OFF or OFF to ON, or the operation mode is changed from “washing” to “rinsing” of the washing machine, for example. This represents the time when the operating state has changed. There may be a plurality of operating state change times instead of one. Next, the harmonic calculation unit 20 extracts characteristic waveforms before and after the change time from the detected change time, voltage, and current measurement data. The harmonic calculation unit 20 performs FFT (Fast Fourier Transform) on the extracted feature waveform to calculate the FFT coefficient (harmonic feature amount) of the harmonic. The harmonic calculation unit 20 may calculate a harmonic feature value for each waveform, or may calculate a harmonic feature value at a predetermined interval.

The device feature learning unit 30 calculates a harmonic feature amount for each operation state of the home appliance from the harmonic feature amount calculated by the harmonic calculation unit 20.
The device feature DB 40 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The device feature DB 40 stores harmonic feature values for each operating state of the home appliance calculated by the device feature learning unit 30.

The operation state estimation unit 50 is based on the harmonic feature amount calculated by the harmonic calculation unit 20 and the harmonic feature amount for each operation state of the home appliance stored in the device feature DB 40. Is estimated.
The power consumption estimation unit 60 stores a regression model created in advance from the relationship between the harmonic feature quantity and the power consumption for each operating state of each home appliance. The power consumption estimation unit 60 estimates the power consumption of each home appliance based on the operation state estimated by the operation state estimation unit 50 and the regression model.

  The apparatus power consumption estimation apparatus 100 creates a database used for estimating the operation state of the home appliance by storing the harmonic feature quantity for each operation state of the home appliance calculated by the device feature learning unit 30 in the device feature DB 40. To do. Therefore, in the following description, the embodiment (first embodiment to third embodiment) is specifically described as an example of the method for calculating the harmonic feature quantity for each operation state of the home appliance by the device feature learning unit 30. explain. In the first embodiment and the second embodiment, the feature quantity of home appliance harmonics is not stored in the device feature DB 40, and the case of calculating the home appliance harmonic feature quantities for the first time will be described as an example. In the third embodiment, a case will be described as an example where harmonics of home appliances are stored in the device feature DB 40 and additional harmonics of home appliances are stored.

(First embodiment)
FIG. 2 is a schematic block diagram illustrating the configuration of the device feature learning unit 30 according to the first embodiment.
As illustrated in FIG. 2, the device feature learning unit 30 includes a harmonic usage history unit 31, a clustering unit 32, a comparison determination unit 33, and a feature amount calculation unit 34.

  The harmonic usage history unit 31 is a household appliance (hereinafter referred to as “a predetermined period”) from when a learning start signal is input until a learning end signal is input (hereinafter referred to as “during a predetermined period”). The information used for calculating the feature quantity of the harmonic for each operation state of “target household appliance”) is recorded. For example, the harmonic usage history unit 31 includes the harmonic feature amount input to the device feature learning unit 30 each time a harmonic feature amount is input, the device name of the target home appliance, and the harmonic feature amount. The power consumption value of the target home appliance at the input time is recorded in association with each other. The harmonics of the current are matched in phase with the voltage waveform as a reference.

  The learning start signal is a signal that is input when the appliance power consumption estimation apparatus 100 starts learning the harmonic feature quantity for each operation state of the target home appliance. The learning end signal is a signal input when the learning of the harmonic feature quantity for each operation state of the target home appliance is ended. The learning start signal and the learning end signal may be input by the user operating an input device (not shown), or may be input remotely by a support center administrator who has received an instruction from the user. The device name of the target home appliance is input together with the input of the learning start signal by the user or the administrator of the support center.

  When the learning end signal is input, the clustering unit 32 clusters the harmonic feature values recorded in the harmonic usage history unit 31 during a predetermined period. The clustering method performed by the clustering unit 32 may be any method as long as the number of clusters can be specified. For example, examples of clustering methods that can specify the number of clusters include a K-means method, a K-means ++ method, and an EM (Expextation Maximization) method. The clustering unit 32 creates a cluster with the minimum number of clusters as 1 in the first clustering after the learning end signal is input.

The comparison determination unit 33 determines whether it is necessary to cluster the clustered harmonic feature values again. For example, the comparison / determination unit 33 satisfies a predetermined condition based on power consumption of harmonic feature amounts in a cluster (hereinafter simply referred to as “cluster”) including harmonic feature amounts classified by clustering. It is determined whether or not the harmonic feature values need to be clustered again according to whether or not they have been.
The feature amount calculation unit 34 calculates the feature amount of the cluster as the feature amount of the operation state of the target home appliance when it is determined in the determination process by the comparison determination unit 33 that it is not necessary to perform clustering again. A device name is associated with the calculated feature amount of the cluster.

FIG. 3 is a diagram for explaining processing performed by the clustering unit 32.
In FIG. 3, the vertical axis represents the imaginary axis, and the horizontal axis represents the real axis. FIG. 3A shows an example in which the feature quantities of individual harmonics recorded in the harmonic usage history unit 31 during a predetermined period are plotted on the complex plane as a plurality of points. As shown in FIG. 3A, the harmonic feature quantities can be plotted as points on the complex plane for each order. The harmonic feature quantity is distributed in the vicinity of the same coordinates when the home appliance is in the same operating state. Therefore, the clustering unit 32 classifies the harmonic feature values for each operation state by clustering the harmonic feature values. FIG. 3B shows an example in which the clustering unit 32 clusters the harmonic feature values as one cluster 200.

  Information of power consumption is associated with the feature quantity of each harmonic. The comparison determination unit 33 determines whether it is necessary to perform clustering again based on the power consumption information associated with the feature quantity of each harmonic. When it is determined that it is not necessary to perform clustering again, the feature amount calculation unit 34 calculates the feature amount of the cluster as the feature amount of the operation state of the target home appliance, and outputs the calculated cluster feature amount to the device feature DB 40. To do. On the other hand, if it is determined that it is necessary to perform clustering again, the clustering unit 32 increases the number of clusters to be classified by clustering and clusters the harmonic feature values again. For example, the clustering unit 32 performs clustering again for all harmonic feature quantities. Each time the clustering unit 32 determines that it is necessary to perform clustering again, the clustering unit 32 increases the number of clusters by one and clusters all the harmonic feature quantities again.

  FIG. 3C shows an example when it is determined that clustering needs to be performed again. As shown in FIG. 3C, the clustering unit 32 performs clustering again to classify the harmonic feature quantity into two clusters 201 and 202. For clustering, the Euclidean distance may be used, or the Mahalanobis distance may be used.

Hereinafter, the process of the comparison determination unit 33 will be specifically described with reference to FIG.
As described above, power consumption information is associated with the harmonic feature values plotted on the complex plane. Therefore, the comparison / determination unit 33 determines whether or not the following Expression 1 is satisfied based on the power consumption information corresponding to the feature quantity of each harmonic in the cluster.

Pa in Equation 1 represents average power consumption, Px represents maximum power consumption, Pn represents minimum power consumption, and α1 and α2 represent variables. α1 and α2 may be appropriately changed by the user. α1 and α2 may be common to all clusters or may be different for each cluster. When all the clusters satisfy the above equation 1, the comparison determination unit 33 determines that it is not necessary to perform clustering again. On the other hand, if there is even one cluster that does not satisfy Equation 1, the comparison / determination unit 33 determines that it is necessary to perform clustering again.
In the following description, when all the clusters satisfy the above formula 1, it is referred to as “cluster condition is satisfied”.

FIG. 4 is a flowchart illustrating a process flow of the device feature learning unit 30 according to the first embodiment. In the description of FIG. 4, processing after the learning start signal is input to the device feature learning unit 30 will be described.
The harmonic usage history unit 31 records the harmonic feature quantity of the home appliance calculated by the harmonic calculation unit 20 for a predetermined period (step S101). Moreover, the harmonic usage log | history part 31 records the apparatus name and power consumption value of object household appliance with the feature-value of a harmonic. When the learning end signal is input to the device feature learning unit 30, the harmonic use history unit 31 ends the recording of the harmonic feature quantity. Thereafter, the clustering unit 32 clusters the harmonic feature quantities recorded in the harmonic usage history unit 31 during a predetermined period (step S102). For example, the clustering unit 32 clusters the harmonic feature quantities into one cluster. The comparison determination unit 33 determines whether or not the cluster condition is satisfied (step S103).

When the cluster condition is satisfied (step S103—YES), the feature amount calculation unit 34 calculates a feature amount for each cluster (step S104). For example, the feature amount calculation unit 34 calculates a statistical value as a feature amount for each cluster. Statistical values include the center (center of gravity) of the cluster, the radius of the cluster, the deviation of the distribution of the harmonic feature quantity, the variance, the covariance matrix, and the like. The feature amount calculation unit 34 outputs the calculated feature amount of the cluster to the device feature DB 40 as the feature amount of the operation state of the target home appliance (step S105). The device feature DB 40 stores the output cluster feature value as the feature value of the operation state of the target home appliance.
On the other hand, if the cluster condition is not satisfied in the process of step S103 (step S103-NO), the clustering unit 32 increases the number of clusters and clusters the harmonic feature values again (step S106).

  According to the device feature learning unit 30 configured as described above, clustering is automatically performed based on harmonics input during a predetermined period. And the harmonic feature-value for every operation state of an object household appliance is computed. Therefore, the user only needs to input a learning start signal, a learning end signal, and a device name. Therefore, the database used for estimation of the operation state of the target home appliance can be easily created.

  The cluster corresponds to the operating state and power consumption of the target home appliance. Therefore, if the difference in power consumption in the cluster is large, the modeling error becomes large. As a result, it is conceivable that the estimation accuracy of the final power consumption is reduced. Therefore, in the present embodiment, in order to reduce the modeling error during clustering, when the above equation 1 is not satisfied, the number of clusters is increased and clustering is executed again. Thereby, the accuracy of classification for each operation state is further improved. As a result, it is possible to increase the accuracy of the feature quantity of the harmonic for each operation state of the target home appliance that is calculated.

(Second Embodiment)
FIG. 5 is a schematic block diagram showing the configuration of the device feature learning unit 30a according to the second embodiment.
As illustrated in FIG. 5, the device feature learning unit 30 a includes a harmonic usage history unit 31, a clustering unit 32 a, a comparison determination unit 33, a feature amount calculation unit 34, and a likelihood evaluation unit 35.
The device feature learning unit 30a according to the second embodiment is different in configuration from the device feature learning unit 30 in that a clustering unit 32a is provided instead of the clustering unit 32 and a likelihood evaluation unit 35 is newly provided. The device feature learning unit 30a is the same as the device feature learning unit 30 in other configurations. Therefore, the description of the entire device feature learning unit 30a is omitted, and the clustering unit 32a and the likelihood evaluation unit 35 will be described.

  When the learning end signal is input, the clustering unit 32a clusters the harmonic feature values recorded in the harmonic use history unit 31 during a predetermined period. Further, the clustering unit 32a clusters again the harmonic feature quantities recorded in the harmonic usage history unit 31 during a predetermined period based on the determination result of the likelihood evaluating unit 35.

  The likelihood evaluator 35 is based on the calculated cluster feature quantity and the harmonic feature quantity recorded in the harmonic use history part 31, and the likelihood of the harmonic feature quantity for all clusters. Is calculated. In addition, the likelihood evaluation unit 35 calculates an information amount reference value. For the calculation of the information amount reference value, for example, an information amount reference such as AIC (Akaike's Information Criterion) or BIC (Bayesian Information Criterion) is used.

  The likelihood evaluation unit 35 compares the likelihood sum or the information amount reference value with the previous value of the number of clusters (the likelihood sum or the information amount reference value), and the change amount is less than a predetermined threshold value. It is determined that the likelihood condition is satisfied. On the other hand, when the amount of change is equal to or greater than a predetermined threshold, it is determined that the likelihood condition is not satisfied. In this case, the likelihood evaluation unit 35 instructs the clustering unit 32a to perform clustering again.

FIG. 6 is a diagram for explaining the cluster number control based on the information amount criterion.
In FIG. 6, the vertical axis represents the information amount reference value, and the horizontal axis represents the number of clusters. As shown in FIG. 6, the information amount reference value tends to have a downward convex shape as the number of clusters increases. In the information amount standard, the smaller the information amount reference value, the better the model (here, the number of clusters) is applied. Therefore, as a result, the information is classified into the number of clusters (for example, the number of clusters indicated by point 203 in FIG. 6) in which the amount of change in the information amount reference value before and after the increase is less than a predetermined threshold. The estimation accuracy of the operating state can be improved. Further, it is determined that there is still room for improvement in clustering accuracy for the number of clusters whose change amount is equal to or greater than a predetermined threshold, and the number of clusters is increased, and clustering is executed until the change amount is less than the predetermined threshold. As a result, the estimation accuracy of the operation state can be improved as a result.

FIG. 7 is a flowchart illustrating a process flow of the device feature learning unit 30a according to the second embodiment. In addition, about the process similar to FIG. 4, the code | symbol similar to FIG. 4 is attached | subjected in FIG. 7, and description is abbreviate | omitted. In the description of FIG. 7, processing after a learning start signal is input to the device feature learning unit 30a will be described.
When the processing up to step S104 is completed, the likelihood evaluation unit 35 calculates the likelihood of the harmonic feature value for all the clusters based on the calculated cluster feature value and the harmonic feature value. (Step S201). Among the likelihoods calculated for all clusters, the highest likelihood is set as the likelihood of the feature quantity of the harmonic. The likelihood evaluating unit 35 calculates the likelihood of all the harmonic feature amounts by performing this process on all the harmonic feature amounts, and calculates the total sum of the likelihoods. Thereafter, the likelihood evaluation unit 35 calculates an information amount reference value based on the current number of clusters, the sum of likelihoods, and the number of harmonic feature values (step S202).

Thereafter, the likelihood evaluation unit 35 determines whether or not the likelihood condition is satisfied (step S203). When the likelihood condition is satisfied (step S203—YES), the likelihood evaluating unit 35 outputs the cluster feature amount calculated in the process of step S104 as the feature amount of the operation state of the target home appliance (step S204).
On the other hand, when the likelihood condition is not satisfied (step S203—NO), the processing after step S106 is executed.

According to the device feature learning unit 30a configured as described above, it is determined whether or not to perform clustering again based on the likelihood of all the harmonic feature values with respect to the cluster. Therefore, clustering with higher accuracy than in the first embodiment is possible. As a result, it is possible to improve the estimation accuracy of the operating state and the estimation accuracy of the power consumption as compared with the first embodiment.
Further, the second embodiment can obtain the same effects as those of the first embodiment.

Hereinafter, modified examples will be described.
In the present embodiment, the number of clusters until the likelihood condition is satisfied, that is, until the improved amount becomes less than a predetermined threshold value as compared with the previous value of the number of clusters (total likelihood or information amount reference value). However, it may be configured to provide an upper limit value for the number of clusters. In this case, the following processing is performed. After the feature amount is calculated by the feature amount calculation unit 34, the likelihood evaluation unit 35 determines whether or not the likelihood condition is satisfied. When the likelihood condition is not satisfied, the likelihood evaluating unit 35 determines whether or not the number of clusters is a set upper limit value. When it is not the set upper limit value, the likelihood evaluating unit 35 instructs the clustering unit 32a to perform clustering again. On the other hand, when it is the set upper limit value, the likelihood evaluation unit 35 determines that the likelihood condition is satisfied even if the improved amount is equal to or greater than a predetermined threshold. Then, the likelihood evaluating unit 35 outputs the cluster feature amount calculated in the process of step S104 as the feature amount of the operation state.

  By being configured in this way, it is possible to reduce the possibility of a long processing time. Specifically, it is as follows. As the number of clusters is increased, the accuracy of clustering is improved, and as a result, the accuracy of estimating the operating state and the accuracy of estimating the power consumption can be improved. Therefore, it is effective to increase the number of clusters until the improved amount is less than a predetermined threshold. However, on the other hand, there is a problem that the time required for processing increases. Therefore, as described above, by setting the upper limit value of the number of clusters, it is possible to improve the accuracy of clustering and reduce the possibility that the time required for processing will become long.

(Third embodiment)
FIG. 8 is a schematic block diagram showing the configuration of the device feature learning unit 30b according to the third embodiment.
As illustrated in FIG. 8, the device feature learning unit 30b includes a harmonic use history unit 31b, a clustering unit 32a, a comparison determination unit 33, a feature amount calculation unit 34, a likelihood evaluation unit 35, and a cluster prediction unit 36.

  The device feature learning unit 30b according to the third embodiment is provided with a harmonic use history unit 31b instead of the harmonic use history unit 31 and a device feature learning unit 30a with a new cluster prediction unit 36. And the configuration is different. The device feature learning unit 30b is the same as the device feature learning unit 30a in other configurations. Therefore, description of the entire device feature learning unit 30b is omitted, and the harmonic usage history unit 31b and the cluster prediction unit 36 will be described.

Based on the newly input harmonic feature quantity and the harmonic feature quantity for each operating state of the home appliance, that is, the device feature DB 40 in which the cluster feature quantity is stored, the cluster prediction unit 36 newly It is determined whether the input harmonic feature quantity is classified as one of the cluster feature quantities stored in the device feature DB 40.
The harmonic usage history unit 31b records, for a predetermined period, the harmonic feature value determined not to be classified as any of the cluster feature values stored in the device feature DB 40.

FIG. 9 is a flowchart showing a flow of processing of the device feature learning unit 30b according to the third embodiment. In addition, about the process similar to FIG. 7, the code | symbol similar to FIG. 7 is attached | subjected in FIG. 9, and description is abbreviate | omitted. In the description of FIG. 7, processing after a learning start signal is input to the device feature learning unit 30 will be described.
When a new harmonic feature value is input to the device feature learning unit 30, the cluster prediction unit 36 uses the newly input harmonic feature value as a cluster feature value stored in the device feature DB 40. It is determined whether it is classified into any of the above (step S301). Specifically, first, the cluster prediction unit 36 calculates the likelihood of the input harmonic feature quantity with respect to the cluster feature quantity stored in the device feature DB 40. Next, the cluster prediction unit 36 selects a likelihood having the smallest value from the calculated likelihoods. Then, the cluster prediction unit 36 determines that the smallest likelihood is not classified as any of the cluster feature values stored in the device feature DB 40 when the likelihood is less than a predetermined value. On the other hand, when the likelihood with the smallest value is equal to or greater than a predetermined value, it is determined that the class is classified into one of the cluster feature quantities stored in the device feature DB 40.

As a result of the determination, when it is determined that the newly input harmonic feature value is classified into one of the cluster feature values stored in the device feature DB 40 (step S302—YES), the cluster prediction unit 36 discards the newly input harmonic feature.
On the other hand, when it is determined that the newly input harmonic feature value is not classified into any of the cluster feature values stored in the device feature DB 40 (step S302—NO), the processing after step S101 is performed. Executed.

According to the device feature learning unit 30b configured as described above, the newly input harmonic feature value is a cluster feature value already stored in the device feature DB 40, that is, a home appliance harmonic feature value. Only when it is determined that it is not related, a new harmonic feature quantity of the home appliance is calculated for each operation state and added to the device feature DB 40. In other words, only the feature quantity of the harmonic for each operation state of the home appliance that is not stored in the device feature DB 40 can be stored in the device feature DB 40. Therefore, it is not necessary to calculate the feature quantity of the home appliance from the harmonic feature quantity input to the device feature learning unit 30. Therefore, it is possible to reduce the processing load when creating the database. Furthermore, since it is not necessary to newly store the harmonic feature values already stored in the device feature DB 40, the memory amount of the device feature DB 40 can be reduced.
Further, the third embodiment can obtain the same effects as those of the second embodiment.

Hereinafter, modified examples will be described.
The third embodiment may be modified in the same manner as the second embodiment.

Hereinafter, modified examples common to the respective embodiments (first embodiment to third embodiment) will be described. In the following description, the device feature learning unit 30, the device feature learning unit 30a, and the device feature learning unit 30b are referred to as the device feature learning unit 30 unless particularly distinguished.
FIG. 10 is a schematic block diagram illustrating a configuration of a device power consumption estimation apparatus 100a according to a modification.
The device power consumption estimation apparatus 100a includes a CPU, a memory, an auxiliary storage device, and the like connected by a bus, and executes a power consumption output program. By executing the power consumption output program, the device power consumption estimation apparatus 100a includes a voltage / current measurement unit 10, a harmonic calculation unit 20, a device feature learning unit 30, a device feature DB 40a, an operation state estimation unit 50, and a power consumption estimation unit 60. , Functions as an apparatus including the position information acquisition unit 70. All or some of the functions of the device power consumption estimation apparatus 100a may be realized using hardware such as an ASIC, PLD, or FPGA. The power consumption output program may be recorded on a computer-readable recording medium. The computer-readable recording medium is, for example, a portable medium such as a flexible disk, a magneto-optical disk, a ROM, a CD-ROM, or a storage device such as a hard disk built in the computer system. The power consumption output program may be transmitted / received via a telecommunication line.

  The device power consumption estimation apparatus 100a according to the modified example is different in configuration from the device power consumption estimation apparatus 100 in that a device feature DB 40a is provided instead of the device feature DB 40, and a position information acquisition unit 70 is newly provided. The apparatus power consumption estimation apparatus 100a is the same as the apparatus power consumption estimation apparatus 100 in other configurations. Therefore, description of the entire apparatus power consumption estimation apparatus 100a is omitted, and the apparatus feature DB 40a and the position information acquisition unit 70 will be described.

The device feature DB 40a is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The appliance feature DB 40 stores the harmonic feature quantity of the home appliance calculated by the appliance feature learning unit 30, that is, the feature quantity of the cluster and the location information of the target home appliance measured by the location information acquisition unit 70 in association with each other. ing.
The position information acquisition unit 70 acquires position information of the target home appliance.

FIG. 11 is a schematic block diagram illustrating the configuration of the position information acquisition unit 70.
As shown in FIG. 11, the position information acquisition unit 70 includes a plane position acquisition unit 71, an altitude acquisition unit 72, and a data integration unit 73.
The plane position acquisition unit 71 acquires information on the plane position of the target home appliance (hereinafter referred to as “plane position information”). The plane position information is, for example, latitude and longitude. The plane position acquisition unit 71 may acquire the plane position information of the target home appliance by communicating with, for example, a GPS (Global Positioning System) satellite. In addition, the plane position acquisition unit 71 may acquire the relative position of the target home appliance from the radio field intensity with the surrounding wireless base station devices. Moreover, the plane position acquisition part 71 may acquire the plane position information of the target home appliance by an input from the user.

The altitude acquisition unit 72 acquires information about the height at which the target home appliance is installed (hereinafter referred to as “altitude information”). For example, the altitude acquisition unit 72 acquires altitude information of the target home appliance using a surveying instrument such as a barometer. In addition, the altitude acquisition unit 72 may acquire altitude information of the target home appliance by an input from the user.
The data integration unit 73 integrates the harmonic feature quantity, the plane position information, and the altitude information into one data. The data integration unit 73 outputs the integrated data to the device feature DB 40a.

According to the apparatus power consumption estimation apparatus 100a configured as described above, it is possible to estimate in which location, in what state, and how much power the target home appliance consumes. Therefore, by displaying the estimation result on an image display device or the like, the user can grasp at a glance which home appliance is operating in which room. Therefore, fine HEMS (Home Energy Management System) control becomes possible.
Moreover, the apparatus power consumption estimation apparatus 100a can acquire the effect similar to 1st Embodiment-3rd Embodiment.

The feature amount calculation unit 34 in each embodiment is an example of a feature amount acquisition unit that acquires the feature amount of the cluster as the feature amount of the operation state of the target home appliance.
The harmonic feature values used in clustering may be the harmonic feature values of all orders, or the harmonics of some orders such as the first and third harmonics. A feature amount may be used. Further, the clustering unit 32 may perform clustering on a (2n + 1) -dimensional data set in which n orders of harmonics and power consumption are set. In this case, since it is necessary to input the number of clusters, the clustering unit 32 creates a cluster with the minimum number of clusters being 1 at the start of learning.

  Further, the determination process of whether or not the cluster condition is satisfied may be performed based on a variance value calculated from the power consumption distribution. When configured in this way, the comparison / determination unit 33 determines whether or not the following Expression 2 is satisfied.

P _{k in} Equation 2 represents power consumption corresponding to the feature quantity of each harmonic in the cluster, and α3 represents a variable. α3 may be appropriately changed by the user. α3 may be common to all clusters or may be different for each cluster. When all the clusters satisfy the above formula 2, the comparison determination unit 33 determines that it is not necessary to perform clustering again. On the other hand, if there is even one cluster that does not satisfy Equation 2, the comparison determination unit 33 determines that it is necessary to perform clustering again.

  In each embodiment, the configuration in which the number of clusters is increased until the cluster condition is satisfied, but the upper limit value of the number of clusters may be provided. In this case, the following processing is performed. After clustering is performed by the clustering unit 32, the comparison determination unit 33 determines whether or not the cluster condition is satisfied. When the cluster condition is not satisfied, the comparison determination unit 33 determines whether or not the number of clusters is a set upper limit value. If it is not the set upper limit value, the comparison / determination unit 33 instructs the clustering unit 32 to perform clustering again. On the other hand, when it is the set upper limit value, the comparison determination unit 33 determines that the cluster condition is satisfied by the number of clusters at that time. Thereafter, the processing after step S104 is executed.

  By being configured in this way, it is possible to reduce the possibility of a long processing time. Specifically, it is as follows. As the number of clusters is increased, the accuracy of clustering is improved, and as a result, the accuracy of estimating the operating state and the accuracy of estimating the power consumption can be improved. Therefore, it is effective to increase the number of clusters until the improved amount is less than a predetermined threshold. However, on the other hand, there is a problem that the time required for processing increases. Therefore, as described above, by setting the upper limit value of the number of clusters, it is possible to improve the accuracy of clustering and reduce the possibility that the time required for processing will become long.

  In each embodiment, the clustering units 32 and 32a (referred to as the clustering unit 32 unless otherwise distinguished) have been described as an example in which the number of clusters is increased by one when performing clustering again. The number of clusters may be two, or may be changed as appropriate according to the situation. In addition, when performing clustering again, the clustering unit 32 may not only increase the number of clusters but also set an initial value for clustering to set a specific cluster as a target for division. Specifically, it is as follows.

When one of the two clusters does not satisfy the cluster condition or the likelihood condition, the clustering unit 32 performs initial clustering again to determine the initial value in the vicinity of the harmonic feature amount included in the cluster that does not satisfy the condition. Set two. Thereafter, the clustering unit 32 performs clustering again. The clustering unit 32 repeatedly executes this process until the condition is satisfied.
By being configured in this way, it is possible to cluster the feature quantities of harmonics included in a cluster that does not satisfy the condition with high accuracy. That is, the accuracy of clustering can be improved. As a result, the estimation accuracy of the operation state of the target home appliance can be improved.

  According to at least one embodiment described above, the clustering unit 32 that clusters the harmonic feature quantities of the target home appliances recorded during a predetermined period, and the feature amount for each cluster is the feature amount of the operation state of the target home appliances. By having the device feature learning unit 30 including the feature amount calculation unit 34 that outputs as a database, it is possible to easily create a database used for estimating the operating state of the home appliance.

  The apparatus power consumption estimation apparatus 100 and the apparatus power consumption estimation apparatus 100a of each embodiment described above may be realized by a program. A program that realizes the above functions of the device power consumption estimation apparatus 100 and the device power consumption estimation apparatus 100a is, for example, a server, smart meter, smart tap, and the like used in a HEMS that is a system for managing power consumption by devices in the home. It may operate on a cloud computer or other server for managing power consumption. Further, the above-described program may be recorded on a non-transitory computer-readable recording medium, and the control process may be performed by causing the computer system to read and execute the program recorded on the recording medium. The “computer system” here includes an OS (Operating System) and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system.

  Moreover, the apparatus power consumption estimation apparatus 100 and the apparatus power consumption estimation apparatus 100a may be computer program products including instructions that can be read by a computer when loaded into a computer processor and executed. In this case, the apparatus power consumption estimation apparatus 100 and the apparatus power consumption estimation apparatus 100a may be realized by software by installing a downloaded program via an external network such as the Internet.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and the equivalents thereof.

DESCRIPTION OF SYMBOLS 10 ... Voltage / current measurement part, 20 ... Harmonic calculation part, 30, 30a, 30b ... Equipment feature learning part, 31, 31b ... Harmonic use history part, 32, 32a ... Clustering part, 33 ... Comparison determination part, 34 ... feature amount calculation unit, 35 ... likelihood evaluation unit, 36 ... cluster prediction unit, 40, 40a ... device feature DB (database), 50 ... operating state estimation unit, 60 ... power consumption estimation unit, 70 ... position information acquisition unit , 71 ... Planar position acquisition unit, 72 ... Altitude acquisition unit, 73 ... Data integration unit, 100, 100a ... Device power consumption estimation device

Claims (8)

A device feature learning unit that acquires a feature value of an operation state of the device from harmonics of current or power obtained from time-series data of voltage and current measured in the power supply path;
A device feature database for storing feature quantities of the acquired device operating states;
An operation state estimation unit that estimates an operation state of the device based on a harmonic feature amount acquired from a harmonic of current or power and a feature amount of the operation state of the device stored in the device feature database; ,
A power consumption estimation unit that estimates power consumption of the device based on the estimated operating state;
An apparatus power consumption estimation apparatus comprising: The device feature learning unit
A clustering unit for clustering harmonic feature values obtained during a predetermined period;
A comparison / determination unit that determines whether the clustered harmonic feature values need to be clustered again;
When it is determined that it is not necessary to perform clustering again, a feature amount acquisition unit that acquires a feature amount of a cluster including harmonic feature amounts classified by the clustering as a feature amount of the operation state of the device;
With
The device power consumption estimation apparatus according to claim 1, wherein the device feature database stores the acquired feature amount of the cluster as a feature amount of an operation state of the device.   The apparatus power consumption estimation apparatus according to claim 2, wherein the clustering unit increases the number of clusters and clusters the harmonic feature amount again each time it is determined that clustering is necessary again. The harmonic feature quantity is associated with information on power consumption,
The comparison determination unit determines that it is not necessary to perform clustering again when all the clusters satisfy a predetermined condition based on power consumption of the harmonic feature amount in the cluster, and a cluster that does not satisfy the predetermined condition The apparatus power consumption estimation apparatus according to claim 2, wherein when it exists, it is determined that clustering is necessary again. The device feature learning unit
A likelihood evaluation unit for calculating a likelihood and an information criterion for the cluster of the harmonic feature quantity based on the acquired cluster feature quantity and the harmonic feature quantity in the cluster; Prepared,
The likelihood evaluation unit determines that it is not necessary to perform clustering again when the calculated likelihood or information amount reference value changes by less than a predetermined threshold before and after the increase in the number of clusters, and the predetermined threshold The apparatus power consumption estimation apparatus of any one of Claim 2 to 4 which determines with the need to cluster again when it changes above. The device feature learning unit
A cluster prediction unit for determining whether or not the newly obtained harmonic feature quantity is classified into any of the cluster feature quantities stored in the device feature database;
The cluster predicting unit calculates the likelihood of the newly obtained harmonic feature quantity with respect to the cluster feature quantity stored in the device feature database, and the smallest likelihood value is less than a predetermined value Is determined not to be classified into any of the cluster feature values stored in the device feature database, and is stored in the device feature database when the smallest likelihood value is equal to or greater than a predetermined value. The apparatus power consumption estimation apparatus according to claim 2, wherein the apparatus power consumption estimation apparatus determines that it is classified into one of the feature quantities of the cluster. A position information acquisition unit for acquiring position information of the device;
The device power consumption according to any one of claims 2 to 6, wherein the device feature database stores a cluster feature amount, which is a feature amount of the operation state of the device, and position information of the device in association with each other. Estimating device. A device feature learning step for acquiring a feature value of the operating state of the device from harmonics of current or power obtained from time-series data of voltage and current measured in the power supply path;
A recording step of storing the acquired feature quantity of the operation state of the device in the device feature database;
An operation state estimation step for estimating an operation state of the device based on a feature amount of a harmonic acquired from a harmonic of current or power and a feature amount of the operation state of the device stored in the device feature database; ,
A power consumption estimation step of estimating the power consumption of the device based on the estimated operating state;
A computer program for causing a computer to execute.

Images