Device status determination system, electricity meter and device status determination method

Abstract

The device state determination system comprises an acquisition unit (31), a calculation processing unit (33), and a device state determination unit (34). The acquisition unit (31) acquires the detection result of a current sensor that detects the instantaneous value of the current flowing through the power line connected to a plurality of electrical devices. The calculation processing unit (33) calculates the effective value and frequency spectrum of the current flowing through the power line at a plurality of moments based on the detection result acquired by the acquisition unit (31). The device state determination unit (34) inputs the calculation result of the calculation processing unit (33) and the information at a plurality of moments into a trained calculation model generated by machine learning, and determines the operation state of a plurality of electrical devices, wherein in the machine learning, the information of the effective value and frequency spectrum of the current flowing through the power line at a plurality of moments and the information at a plurality of moments are used as input, and the operation state of a plurality of electrical devices is used as output. The determination of the operation state can play a role in the management of electrical devices, etc.

Description

Technical Field

[0001] The present invention relates to an equipment state determination system, an electric power meter and an equipment state determination method for determining the operation states of a plurality of electric equipment connected to a power line. Background Art

[0002] Currently, there is a known technique for determining the operating state of multiple electrical devices connected to the same power line based on the current flowing through the power line. For example, Patent Document 1 proposes a technique for determining the operating state of each electrical device based on the current waveform obtained by a sensor installed on the power line.

[0003] In the technology described in Patent Document 1, information on the current waveform of the power line when each electrical device operates independently is obtained as information on the reference waveform, and the electrical device corresponding to the reference waveform that is most similar to the difference waveform between the current waveform before and after the change when the current waveform has changed, that is, the difference waveform, is determined to be the electrical device whose operating state has changed.

[0004] Patent Document 1: Japanese Patent Application Publication No. 2013-150508 Summary of the invention

[0005] However, the technology described in Patent Document 1 determines the operating state of the electrical device by pattern matching of the current waveform and the reference waveform. Therefore, in the technology described in Patent Document 1, when there are multiple electrical devices having current waveforms matching the same reference waveform, it is difficult to determine which of the multiple electrical devices is in the operating state.

[0006] The present invention has been made in view of the above-mentioned problems, and an object of the present invention is to provide an equipment state determination system capable of accurately determining the operating states of a plurality of electric equipment connected to the same power line.

[0007] In order to solve the above-mentioned problems and achieve the purpose, the device state determination system of the present invention has an acquisition unit, an operation processing unit, and a device state determination unit. The acquisition unit acquires the detection result of a current sensor that detects the instantaneous value of the current flowing through the power line connected to multiple electrical devices. The operation processing unit calculates the effective value and frequency spectrum of the current flowing through the power line at multiple times based on the detection result obtained by the acquisition unit. The device state determination unit inputs the calculation result of the operation processing unit and the information at multiple times into a trained calculation model generated by machine learning, and determines the operation status of multiple electrical devices, wherein in the machine learning, the information of the effective value and frequency spectrum of the current flowing through the power line at multiple times and the information at multiple times are used as input, and the operation status of multiple electrical devices is used as output.

[0008] Effects of the Invention

[0009] According to the present invention, it is possible to accurately determine the operating states of a plurality of electrical devices connected to the same power line. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a diagram showing an example of the configuration of the device state determination system according to Embodiment 1 of the present invention.

[0011] Figure 2 This is a diagram for explaining data collected by the HEMS according to the first embodiment.

[0012] Figure 3 This is a diagram for explaining the frequency spectrum of the current calculated in the device state determination system according to the first embodiment.

[0013] Figure 4 This is a diagram for explaining an example of information on effective values ​​according to the first embodiment.

[0014] Figure 5 This is a diagram for explaining an example of processing in the device state determination system according to the first embodiment.

[0015] Figure 6 This is a diagram showing an example of the structure of the device state determination device involved in the first embodiment.

[0016] Figure 7 This is a diagram showing an example of a deep neural network involved in Implementation Example 1.

[0017] Figure 8 This is a diagram for explaining another example of the information of the effective value of the current input to the deep neural network involved in the first embodiment.

[0018] Fig. 9 This is a diagram for explaining another example of the spectrum information input to the deep neural network involved in Implementation Example 1.

[0019] Fig.10 This is a flowchart showing an example of processing performed by the processing unit of the device state determination apparatus according to the first embodiment.

[0020] Fig.11 This is a diagram showing an example of the hardware configuration of the device state determination apparatus according to the first embodiment.

[0021] Fig.12 This is a diagram showing an example of the configuration of a device state determination system according to Embodiment 2 of the present invention.

[0022] Fig.13This is a diagram for explaining data collected by the HEMS according to the second embodiment.

[0023] Fig.14 This is a diagram showing an example of the structure of the device state determination device involved in the second embodiment.

[0024] Fig.15 This is a diagram showing an example of the configuration of a device state determination system according to Embodiment 3 of the present invention.

[0025] Fig.16 This is a diagram showing an example of the structure of a cloud server involved in Implementation Example 3.

[0026] Fig.17 This is a diagram showing an example of the structure of an electricity meter according to the third embodiment.

[0027] Fig.18 This is a diagram showing another example of the configuration of the device state determination system according to the third embodiment. DETAILED DESCRIPTION

[0028] Hereinafter, the device state determination system, the electricity meter, and the device state determination method according to the embodiment of the present invention will be described in detail based on the drawings. In addition, the present invention is not limited to the embodiment.

[0029] Implementation Method 1

[0030] Figure 1 This is a diagram showing an example of the configuration of the device state determination system according to Embodiment 1 of the present invention. Figure 1 The device state determination system 100 shown in the figure has a trained calculation model generated by machine learning, and uses the trained calculation model to determine the status of multiple electrical devices 41, 42, 43, ..., 4 connected to the power line 8. m The operation state of the plurality of electrical devices 41, 42, 43, ..., 4 m The electrical devices 41, 42, 43, ..., 4 m When each is distinguished and individually indicated, it may be described as the electric device 4. The plurality of electric devices 4 is an example of the plurality of first electric devices. The power supply line 8 is an example of the first power supply line.

[0031] The power from the power supply device 2 is supplied to a plurality of electrical devices 41, 42, 43, ..., 4 through an electric meter 3 and a circuit breaker 6. m The electricity meter 3 supplies electricity to a plurality of electrical devices 41, 42, 43, ..., 4 via the power line 8. mThe amount of power supplied by the loads included in the electrical device 4 is measured. The electrical device 4 is, for example, a lighting device, an air conditioner, a microwave oven, a refrigerator, a washing machine, a fan heater, or a television receiver.

[0032] The device state determination system 100 includes a device state determination device 1, a measuring device 5, and a HEMS (Home Energy Management System) 71 to 7 k The measuring device 5 repeatedly measures the instantaneous value of the current flowing through the power line 8 at a predetermined sampling period T1. The measuring device 5 transmits measurement data including the measured instantaneous current value and information on the measurement time of the instantaneous current value to the device state determination apparatus 1 via the network 80.

[0033] Furthermore, when the device state determination device 1 has a clock function, the measuring device 5 may not have a clock function. In this case, the measurement data sent from the measuring device 5 does not include information about the time. When the device state determination device 1 receives measurement data including the instantaneous value of the current from the measuring device 5 when the measuring device 5 does not have a clock function, the device state determination device 1 generates measurement data that associates information about the time counted by the device state determination device 1 with the instantaneous value of the current included in the measurement data.

[0034] The device state determination device 1 receives the measurement data repeatedly transmitted from the measuring device 5 via the network 80. The device state determination device 1 is, for example, a cloud server composed of one or more servers and one or more memories, and the one or more servers and one or more memories constitute a cloud system constructed in a data center (not shown). In addition, the device state determination device 1 receives the measurement data repeatedly transmitted from the measuring device 5 via the network 80. k Receive various data from HEMS 71 to 7 k The transmitted data includes learning data for learning the above-mentioned calculation model. The learning data will be described in detail later. k When they are distinguished and individually indicated, they may be described as HEMS 7.

[0035] Figure 2 1 is a diagram for explaining data collected by the HEMS according to the first embodiment. Figure 2 As shown, the HEMS 7 can communicate with the measuring device 41 and a plurality of electrical devices 401, 402, 403, ..., 40 m A plurality of electrical devices 401, 402, 403, ..., 40 m The power supply device 2 is connected to the same power line 43 and is supplied with electric power via the power line 43 .

[0036] A plurality of electrical devices 401, 402, 403, ..., 40 m The plurality of electrical devices 401, 402, 403, ..., 40 m For example, a plurality of electrical devices 41, 42, 43, ..., 4 m The same, but may be partially different. m When each is distinguished and shown individually, it may be described as the electric device 40. The power line 43 is an example of a second power line, and the plurality of electric devices 40 are a plurality of second electric devices. The current flowing through the power line 43 is an example of the second current.

[0037] The measuring device 41 repeatedly measures the instantaneous value of the current flowing through the power line 43 at a predetermined sampling period T1 and transmits measurement data including the measured instantaneous current value and information on the measurement time of the instantaneous current value to the HEMS 7 .

[0038] The HEMS 7 acquires a plurality of current instantaneous values ​​from the measuring device 41. The HEMS 7 calculates the effective current value, which is the effective value of the current flowing through the power supply line 43, based on the acquired plurality of current instantaneous values. For example, each time the HEMS 7 acquires the current instantaneous value from the measuring device 41, it takes a moving average of the square values ​​of the plurality of current instantaneous values ​​and calculates the square root of the moving average result, thereby being able to obtain the current effective value for each sampling period T1.

[0039] HEMS 7 calculates the spectrum of the current flowing through power line 43 at a predetermined period T2 based on the acquired multiple current instantaneous values. The spectrum of the current is the power distribution of the frequency components of the current flowing through power line 43. Period T2 is longer than period T1.

[0040] Figure 3 1 is a diagram for explaining the frequency spectrum of the current calculated in the device state determination system according to the first embodiment. Figure 3 As shown, the HEMS 7 calculates the frequency spectrum of the current flowing through the power supply line 43 by FFT (Fast Fourier Transform) processing based on a plurality of current instantaneous values ​​corresponding to the period TA acquired from the measuring device 41 .

[0041] The period TA is the period of one cycle of the AC voltage applied from the power supply device 2 to the electric device 40. For example, the period TA is 1 / 60 [s] when the frequency of the AC voltage is 60 Hz, and is 1 / 50 = 0.02 [s] when the frequency of the AC voltage is 50 Hz. TA = T2.

[0042] In addition, the HEMS 7 determines the operating state of the plurality of electrical devices 40 at a predetermined period T2. The operating state of the electrical device 40 determined by the HEMS 7 is, for example, whether the electrical device 40 is in the on state or the electrical device 40 is in the off state. The HEMS 7 determines whether each electrical device 40 is in the on state or the off state based on information output from the plurality of electrical devices 40, for example.

[0043] HEMS 7 generates learning data and transmits the generated learning data to device state determination apparatus 1. The learning data includes information on the effective value of the current flowing through power line 43, information on the spectrum of the current flowing through power line 43, information on the operating states of the plurality of electrical devices 40, and time information for each time.

[0044] Figure 4 1 is a diagram for explaining an example of information on effective values ​​involved in Implementation Example 1. Figure 4 As shown, the effective value information includes a plurality of current effective values ​​corresponding to the period TA. For example, the effective value information includes the current effective value of each sampling period T1 corresponding to the period TA.

[0045] The time information included in the learning data is the time of each predetermined period T2, which is information expressed by hours, minutes, and seconds. The learning data includes information that associates information of each of the multiple time points in the period TB with information of effective values, information of frequency spectrum, and information of the operating states of the multiple electrical devices 40. The period TB is, for example, a period of 1 day.

[0046] Set the time t i The effective value of the current flowing through the power line 43 at i ), the time t i The spectrum of the current flowing through the power line 43 at is set to Is2(t i ), the time t i The operation states of the plurality of electrical devices 40 at the location are set as Do2(t i ). In addition, the time of each cycle T2 in the period TB is set to time t1~t m m is TB / T2. In this case, the learning data is time series data including current data, time data, and operation data. The current data includes Ir2(t1) to Ir2(t m ) and Is2(t1)~Is2(t m ) information, the time data includes t1~t m The action data includes Do2(t1)~Do2(t m) information. The current data and time data in the learning data are input data, and the action data are result data. In addition, the result data is also called answer data.

[0047] Do2(t i ) is represented by, for example, "1" or "0". i , the operating state of the electrical device 401 is the on state, the operating state of the electrical device 402 is the off state, the operating state of the electrical device 403 is the on state, and the electrical device 40 m The action state is the disconnected state. In this case, it is expressed as Do2(t i )={1,0,1,…,0}.

[0048] The operating state of each electrical device 40 is not limited to the on state and the off state, and may include, for example, a standby state. In this case, the operating state of each electrical device 40 is represented by a plurality of bits. For example, the on state is defined as "11", the off state is defined as "00", and the standby state is defined as "01". i , the state of the electrical device 401 is the standby state, the electrical device 402 is the disconnected state, the electrical device 403 is the connected state, and the electrical device 40 m In this case, it is expressed as Do2(t i )={

[01] ,

[00] ,

[11] , ...,

[00] }. The standby state is a state in which the electric device 40 consumes less power than the on state and can be switched to the on state in a shorter time than the off state.

[0049] Figure 5 1 is a diagram for explaining an example of processing in the device state determination system according to the first embodiment. Figure 5 As shown, the device state determination system 100 performs a process in a learning phase and a process in a determination phase. In the learning phase, the device state determination system 100 generates a trained calculation model by machine learning using the above-mentioned learning data.

[0050] In addition, in the determination phase, the device state determination system 100 uses the measurement data sent from the measuring device 5 as determination data and pre-processes the determination data. Pre-processing is the processing of the determination data before inputting the determination data into the trained calculation model. Specifically, the device state determination system 100 calculates the effective value of the current flowing through the power line 8 and the frequency spectrum of the current flowing through the power line 8 based on the multiple current instantaneous values ​​included in the determination data.

[0051] Furthermore, the device state determination system 100 generates the following data as preprocessed data, which includes information on the effective value of the current flowing through the power line 8, information on the frequency spectrum of the current flowing through the power line 8, and information on the time in units of time. The device state determination system 100 inputs the preprocessed data to the trained calculation model generated by the processing in the learning stage, and performs the preprocessed data on the plurality of electrical devices 41, 42, 43, ..., 4 m The action status is determined.

[0052] The effective value and spectrum of the current flowing through the power line 8 differ according to the operating state of each of the plurality of electrical devices 4. In addition, the effective value and spectrum of the current supplied to the electrical devices 4 differ between different types of electrical devices 4. In addition, the operating states of the plurality of electrical devices 4 differ according to time.

[0053] The device state determination system 100 determines the operating states of the plurality of electrical devices 4 by a calculation model generated by machine learning using the above-mentioned learning data, and thus can determine the operating states of the plurality of electrical devices 4 connected to the same power line 8 with high accuracy. For example, in the case where there are two or more electrical devices 4 of the same type, the operating states of the two or more electrical devices 4 are different depending on the time. In the device state determination system 100, since the input data includes information on the time, it is possible to determine the operating states of the two or more electrical devices 4 of the same type with high accuracy.

[0054] In addition, since the device state determination system 100 collects learning data from each HEMS 7, it is possible to generate a calculation model corresponding to various patterns of the operating states of the plurality of electrical devices 40 in the period TA, and it is possible to accurately determine the operating states of the plurality of electrical devices 4. In addition, the device state determination system 100 can also collect learning data from sources other than the HEMS 7.

[0055] In addition, time t i The effective value of the current at Ir2(t i ) includes the effective value of the current at each period T1 in the period TA. Therefore, the device state determination system 100 and Ir2(t i ) contains only one effective value, the operating states of the plurality of electrical devices 4 can be determined with high accuracy. i ) may also contain only one valid value during period TA.

[0056] Figure 6 FIG. 1 is a diagram showing an example of the structure of the device state determination device involved in Embodiment 1. Figure 6As shown, the device state determination apparatus 1 includes a communication unit 10, a storage unit 11, and a processing unit 12. The communication unit 10 is connected to a network 80 for communication, and receives learning data transmitted from the HEMS 7 and measurement data transmitted from the measurement device 5 via the network 80.

[0057] The storage unit 11 includes learning data 21 and determination data 22. The learning data 21 includes, for example, the data from the HEMS 71 to 7 k Each of the plurality of learning data included in the learning data 21 is data for each period TB, and is a data set including current data, time data, and operation data. The determination data 22 is measurement data sent from the measuring device 5 to the device state determination apparatus 1 via the network 80.

[0058] The processing unit 12 includes an acquisition unit 31, a learning processing unit 32, a calculation processing unit 33, and a device state determination unit 34. The acquisition unit 31 acquires the learning data received from the HEMS 7 by the communication unit 10, and stores the acquired learning data in the storage unit 11. As a result, the learning data 21 stored in the storage unit 11 is updated.

[0059] The acquisition unit 31 acquires the measurement data received from the measuring device 5 by the communication unit 10, and stores the acquired measurement data as determination data in the storage unit 11. Thus, the determination data 22 stored in the storage unit 11 is updated.

[0060] The learning processing unit 32 obtains the learning data 21 from the storage unit 11, and generates a trained calculation model by machine learning based on the obtained learning data 21. The learning processing unit 32 sets the generated trained calculation model to the device state determination unit 34. The trained calculation model is a calculation model generated by machine learning based on the data set of input data and result data included in the learning data 21. The input data is time series data including current data and time data. The result data is time series data including action data.

[0061] The learning processing unit 32 performs learning processing according to the deep neural network through so-called tutored learning, for example, to generate a computational model composed of the deep neural network. Here, tutored learning is a method of providing a large amount of data sets of certain inputs and results to the learning device, learning the characteristics of these data sets, and generating a computational model for inferring the results based on the inputs through machine learning. The deep neural network is composed of the following layers, namely, an input layer composed of multiple neurons, an intermediate layer composed of multiple neurons, and an output layer composed of multiple neurons. The intermediate layer is also called a hidden layer.

[0062] Figure 7 is a diagram showing an example of a deep neural network according to Implementation Example 1. Figure 7 In the example shown, the input layer of the deep neural network used by the learning processing unit 32 includes 18 neurons. Each neuron has a resolution equivalent to 8 bits.

[0063] exist Figure 7 In the example shown, among the 6 neurons in the input layer, 8 neurons are input with information about the effective value of the current, the other 8 neurons are input with information about the spectrum, and the remaining two neurons are input with information about the time. The information about the effective value of the current in each period T2 is as follows: Figure 4 As shown in FIG. 1 , the spectrum of each period T2 is divided into 8 parts in the time axis direction, and the effective current values ​​of the divided time periods T21 to T28 are input to the corresponding neurons in the 8 neurons. Figure 3 As shown, it is divided into 8 parts in the frequency axis direction, and the spectrum of the divided frequency bands fa1 to fa8 is input to the corresponding neurons in the 8 neurons. When the frequency of the AC voltage is 60Hz, the amount of information on the time contained in the learning model is 60 [cycles / second] × 360 [seconds] × 24 [hours] = 518400, log2 (518400) = 15.5. Therefore, the time of each cycle T2 can be represented by 16 bits, as shown in Figure 7 As shown, the information of the time is input into two neurons.

[0064] In addition, the above example shows a method in which the information of the effective value of the current is divided into 8 parts in the time axis direction in units of period T2 and input into 8 neurons, and the information of the spectrum is divided into 8 parts in the frequency axis direction in units of period T2 and input into 8 neurons, but it is not limited to this example. Figure 8 This is a diagram for explaining another example of the information of the effective value of the current input to the deep neural network involved in the first embodiment. Fig. 9 This is a diagram for explaining another example of the spectrum information input to the deep neural network involved in Implementation Example 1.

[0065] like Figure 8 As shown in FIG. 1 , the waveform of the effective value of the current in each period T2 can also be divided into a plurality of regions AR as image data. 11 ~AR 19 In this case, each area AR 11 ~AR 19 The image data is represented by 8-bit grayscale, for example, and is input to the corresponding neuron among the 9 neurons. Similarly, the waveform of the spectrum of each period T2 can also be divided into a plurality of regions AR as image data. 21 ~AR 29 In this case, each area AR21 ~AR 29 The image data is represented, for example, in 8-bit grayscale and input to the corresponding neurons among the 9 neurons.

[0066] in addition, Figure 7 The deep neural network shown is, for example, a deep neural network in the case where there are four electrical devices 4, and the number of neurons in the output layer is, for example, four. In this case, each neuron in the output layer outputs a score of the action state of the corresponding electrical device 4 among the multiple electrical devices 4, that is, an action state score. The action state score is, for example, a value from 0 to 1, and when the action state score is greater than or equal to a threshold, it is determined that the electrical device 4 is in the on state. Figure 7 In the example shown, the number of neurons in the middle layer is 5, but it can also be less than or equal to 4 or greater than or equal to 6. Similarly, the number of neurons in the input layer and the number of neurons in the output layer are also not limited to Figure 7 Example shown.

[0067] For example, in the device state determination system 100, when the standby state of the electrical device 4 is determined as the operation state of the electrical device 4, the output layer may include various types of neurons for the operation state of the electrical device 4. For example, the output layer may include neurons for determining the on state, neurons for determining the off state, and neurons for determining the standby state for each electrical device 4. In this case, the operation state corresponding to the neuron with the highest value among the values ​​output from the neurons for determining the on state, the neurons for determining the off state, and the neurons for determining the standby state is determined as the operation state of the electrical device 4.

[0068] In addition, as a learning algorithm used in the calculation model, in a deep neural network, in particular, by using a deep neural network such as a recurrent neural network or LSTM (Long Short Term Memory) that can learn and judge timing information, when the causal relationship between the actions of each electrical device 4 has a certain tendency, the action state of the electrical device 4 can be judged by learning this information.

[0069] In addition, in the above example, the case where the learning algorithm used in the calculation model is applied to the learning algorithm used in the calculation model is described, but the learning algorithm used in the calculation model is not limited to the learning algorithm used in the calculation model. For example, unsupervised learning can also be applied to the learning algorithm used in the calculation model. In this case, the device state determination system 100 performs clustering through unsupervised learning and then assigns labels to the results, thereby generating a calculation model.

[0070] In addition, the learning algorithm used as the computational model may replace the neural network and perform machine learning according to other well-known methods, such as genetic programming, functional logic programming, support vector machines, etc.

[0071] Return to Figure 6 , the description of the processing unit 12 is continued. The calculation processing unit 33 of the processing unit 12 obtains the determination data 22 from the storage unit 11, and generates preprocessed data by preprocessing the obtained determination data 22.

[0072] Specifically, the calculation processing unit 33 calculates the effective value of the current flowing through the power line 8 and the spectrum of the current flowing through the power line 8 based on the plurality of current instantaneous values ​​included in the determination data 22. The calculation processing of the effective value and the spectrum performed by the calculation processing unit 33 is the same as the calculation processing of the effective value and the spectrum performed by the HEMS 7.

[0073] The processing unit 33 generates time series data including information on the effective value of the current flowing through the power line 8, information on the frequency spectrum of the current flowing through the power line 8, and information on the time in units of time as preprocessing completion data, and outputs the generated preprocessing completion data to the device state determination unit 34.

[0074] In addition, the calculation processing unit 33 can also calculate the effective value of the current based on the instantaneous current value from the latest time to the period TA before, whenever the instantaneous current value measured at the latest time is obtained by the acquisition unit 31. In addition, the calculation processing unit 33 can also calculate the spectrum of the current based on the instantaneous current value from the latest time to the period TA before, with the period T2 as a unit.

[0075] The device state determination unit 34 inputs the preprocessed data to the trained calculation model set by the learning processing unit 32, and performs a plurality of electrical devices 41, 42, 43, ..., 44 based on the output of the trained calculation model. m Thus, the device state determination apparatus 1 can determine, for example, how many electrical devices 4 there are and at what timing each of the plurality of electrical devices 4 operates.

[0076] For example, when the predetermined time comes, the device state determination unit 34 inputs the preprocessing data calculated based on the instantaneous current value obtained in the period TB into the trained calculation model, thereby being able to determine the operation status of the plurality of electrical devices 4 in the period TB. In addition, the device state determination unit 34 repeatedly inputs the preprocessing data calculated based on the instantaneous current value from the latest time to before the period TA into the trained calculation model, for example, in units of period T2, thereby being able to determine the operation status of the plurality of electrical devices 4 at the latest time.

[0077] Here, the time t i The effective value of the current flowing through the power line 8 at i ), the time t i The spectrum of the current flowing through the power line 8 at is set to Is1(t i ). In addition, the time intervals of the period T2 in the period TB are set as time t1 to t m m is TB / T2. In this case, the preprocessed data is time series data including current data and time data. The current data includes Ir1(t1) to Ir1(t m ) and Is1(t1)~Is1(t m ) information, the time data includes t1~t m information.

[0078] The device state determination unit 34 inputs the preprocessed data into the trained calculation model to determine the time intervals t1 to t m Then, the acquisition unit 31 acquires the new time t m+1 Ir1(t m+1 ) information, Is1(t m+1 ) information and t m+1 In this case, the device state determination unit 34 converts the measurement data including Ir1(t m+1 ) information, Is1(t m+1 ) information and t m+1 The preprocessing including the information of the data is completed and input into the trained calculation model, so as to calculate the new time t m+1 The device state determination unit 34 repeats this process every time the acquisition unit 31 acquires measurement data at a new time, and can determine the operation states of the plurality of electrical devices 4 at a new time.

[0079] The device state determination unit 34 is Figure 7 In the case of the deep neural network shown, for example, when the action state score output from the output layer is greater than or equal to the threshold, the electrical device 4 is judged to be in the on state, and when the action state score is less than the threshold, the electrical device 4 is judged to be in the off state.

[0080] The device state determination unit 34 displays the plurality of electrical devices 41, 42, 43, ..., 4 m The information on the operation status is sent from the communication unit 10 to the external device via the network 80.

[0081] Next, the processing performed by the processing unit 12 of the device state determination apparatus 1 will be described using a flowchart. Fig.10 This is a flowchart showing an example of processing performed by the processing unit of the device state determination apparatus according to the first embodiment. Fig.10 The illustrated processing is repeatedly executed by the processing unit 12 of the device state determination apparatus 1 .

[0082] like Fig.10 As shown, the processing unit 12 of the device state determination device 1 determines whether the learning data is received through the communication unit 10 (step S10). When the processing unit 12 determines that the learning data is received through the communication unit 10 (step S10: Yes), the learning data received through the communication unit 10 is stored in the storage unit 11 (step S11).

[0083] When the processing of step S11 is completed or when it is determined that the learning data is not received through the communication unit 10 (step S10: No), the processing unit 12 determines whether the determination data is received through the communication unit 10 (step S12). When it is determined that the determination data is received through the communication unit 10 (step S12: Yes), the processing unit 12 stores the determination data received through the communication unit 10 in the storage unit 11 (step S13).

[0084] When the processing of step S13 is completed, or when it is determined that the determination data has not been received by the communication unit 10 (step S12: No), the processing unit 12 determines whether the determination timing has arrived (step S14). In step S14, the processing unit 12 determines that the determination timing has arrived when it has arrived at a predetermined time or when the current instantaneous value measured at the latest time has been acquired by the acquisition unit 31.

[0085] When the processing unit 12 determines that the determination timing has arrived (step S14: Yes), it obtains the determination data 22 from the storage unit 11 and generates pre-processed data based on the obtained determination data 22 (step S15). Then, the processing unit 12 inputs the pre-processed data into the trained calculation model and determines the operation states of the plurality of electrical devices 4 based on the output of the trained calculation model (step S16). When the processing of step S16 is completed or when it is determined that the determination timing has not arrived (step S14: No), the processing unit 12 Fig.10 The shown processing ends.

[0086] Fig.11 FIG. 1 is a diagram showing an example of the hardware configuration of the device state determination device according to Embodiment 1. Fig.11As shown, the device state determination apparatus 1 includes a computer having a processor 101 , a memory 102 , a communication device 103 , and an interface circuit 104 .

[0087] The processor 101, the memory 102, the communication device 103 and the interface circuit 104 can transmit and receive information to each other through the bus 105, for example. The storage unit 11 is implemented by the memory 102. The communication unit 10 is implemented by the communication device 103. The processor 101 reads and executes the program stored in the memory 102, thereby executing the functions of the acquisition unit 31, the learning processing unit 32, the calculation processing unit 33 and the device state determination unit 34. The processor 101 is, for example, an example of a processing circuit, including more than or equal to one of a CPU (Central Processing Unit), a DSP (Digital Signal Processor), a GPU (Graphical Processing Unit) and a system LSI (Large Scale Integration).

[0088] The memory 102 includes more than or equal to one of RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory) and EEPROM (registered trademark) (Electrically Erasable Programmable Read Only Memory). In addition, the memory 102 includes a computer-readable recording medium having a program recorded thereon. The recording medium includes more than or equal to one of a non-volatile or volatile semiconductor memory, a magnetic disk, a floppy disk, an optical disk, a high-density disk and a DVD (Digital Versatile Disc). In addition, the device state determination device 1 may also include an integrated circuit such as an ASIC (Application Specific Integrated Circuit) and an FPGA (Field Programmable Gate Array).

[0089] As described above, the device state determination system 100 according to the first embodiment includes an acquisition unit 31, a calculation processing unit 33, and a device state determination unit 34. The acquisition unit 31 acquires the detection result of the current sensor that detects the instantaneous value of the current flowing through the power line 8 to which the plurality of electrical devices 4 are connected. The calculation processing unit 33 calculates the effective value and spectrum of the current flowing through the power line 8 at a plurality of times based on the detection result acquired by the acquisition unit 31. The device state determination unit 34 inputs the calculation result of the calculation processing unit 33 and the information at a plurality of times into a trained calculation model generated by machine learning, and determines the operation states of the plurality of electrical devices 4, wherein in the machine learning, the information of the effective value and spectrum of the current flowing through the power line 8 at a plurality of times and the information at a plurality of times are input, and the operation states of the plurality of electrical devices 4 are output. Thus, the device state determination system 100 can determine the operation states of the plurality of electrical devices 4 with high accuracy. The determination of the operation state by the device state determination system 100 can play a role in, for example, the management of the electrical devices 4.

[0090] In addition, the device state determination system 100 has a storage unit 11 and a learning processing unit 32. The storage unit 11 stores learning data, which includes the following data, namely, data including information indicating the effective value and frequency spectrum of the current flowing through the power line 43 connected to the plurality of electrical devices 40 at each of the plurality of moments in the predetermined period TB, data including information indicating each of the plurality of moments in the predetermined period TB, and data including information indicating the operation state of the plurality of electrical devices 40 at each of the plurality of moments. The learning processing unit 32 generates a trained calculation model through machine learning based on the learning data stored in the storage unit 11. Thus, the device state determination system 100 can generate a trained calculation model, and the trained calculation model can determine the operation state of the plurality of electrical devices 4 with high accuracy.

[0091] Implementation Method 2

[0092] The device state determination system according to the second embodiment differs from the device state determination system 100 according to the first embodiment in that it uses information on the temperature and vibration of a plurality of electrical devices in addition to information on the effective value of the current, information on the spectrum of the current, and information on the time. Hereinafter, the same reference numerals are given to the structural elements having the same functions as those in the first embodiment, and the description thereof is omitted, and the description will be centered on the differences from the device state determination system 100 according to the first embodiment.

[0093] Fig.12 This is a diagram showing an example of the configuration of a device state determination system according to Embodiment 2 of the present invention. Fig.13This is a diagram for explaining data collected by the HEMS according to the second embodiment. Fig.14 This is a diagram showing an example of the structure of the device state determination device involved in the second embodiment.

[0094] like Fig.12 As shown, the device state determination system 100A according to the second embodiment includes a device state determination apparatus 1A, a measuring device 5A, and HEMS 7A1 to 7A2. k Next, if you do not use HEMS 7A1~7A k When they are distinguished and individually indicated, they may be described as HEMS 7A.

[0095] In addition to the functions of the measuring device 5, the measuring device 5A measures the temperature and vibration of the plurality of electrical devices 4 at a predetermined period T2, and sends the measurement data including the information of the measured temperature and vibration of the plurality of electrical devices 4 to the device state determination device 1A via the network 80. The temperature measured by the measuring device 5A is, for example, the internal temperature of the electrical device 4, and the vibration measured by the measuring device 5A is the vibration of the frame of the electrical device 4. The measuring device 5A may also be composed of a plurality of measuring devices.

[0096] in addition, Fig.13 The HEMS 7A shown measures the temperature and vibration of the plurality of electrical devices 40 at a predetermined period T2 in addition to the functions of the HEMS 7. The HEMS 7A generates learning data including information on the temperature and vibration of the plurality of electrical devices 40 in input data, and transmits the generated learning data to the device state determination device 1A. The temperature measured by the HEMS 7 is, for example, the internal temperature of the electrical device 40, and the vibration measured by the HEMS 7 is the vibration of the housing of the electrical device 40.

[0097] The learning data generated by the HEMS 7A includes information on the temperature of the plurality of electrical devices 4 and information on the vibration of the plurality of electrical devices 4 in units of time, in addition to the information included in the learning data generated by the HEMS 7 .

[0098] like Fig.14As shown, the device state determination device 1A has a communication unit 10, a storage unit 11A, and a processing unit 12A. The storage unit 11A stores the learning data 21A and the determination data 22A. The multiple learning data included in the learning data 21A are each data of each period TB, and are data sets including temperature data and vibration data in addition to the current data, time data, and action data included in the learning data 21. The temperature data includes information on the temperature of multiple electrical devices 4 in each period T2 in the period TB. The vibration data includes information on the vibration of multiple electrical devices 4 in each period T2 in the period TB.

[0099] The processing unit 12A includes an acquisition unit 31, a learning processing unit 32A, a calculation processing unit 33A, and a device state determination unit 34A. The acquisition unit 31 stores the learning data received by the communication unit 10 in the storage unit 11A, and updates the learning data 21A stored in the storage unit 11A. In addition, the acquisition unit 31 stores the measurement data received by the communication unit 10 as determination data in the storage unit 11A. As a result, the determination data 22A stored in the storage unit 11A is updated.

[0100] The learning processing unit 32A obtains the learning data 21A from the storage unit 11A, and generates a trained calculation model by machine learning using the data set of input data and result data included in the obtained learning data 21A. The learning processing unit 32A sets the generated trained calculation model to the device state determination unit 34A. The input data is time series data including current data, time data, operation data, temperature data, and vibration data. The result data is time series data including information indicating the operation state of multiple electrical devices 40 in units of time. Temperature data and vibration data are examples of additional information.

[0101] The calculation processing unit 33A calculates the effective value of the current flowing through the power line 8 and the spectrum of the current flowing through the power line 8 at a predetermined period T2 based on the multiple current instantaneous values ​​included in the determination data 22A. In addition, the calculation processing unit 33A generates time series data including information on the effective value of the current flowing through the power line 8, information on the spectrum of the current flowing through the power line 8, information on the temperature of the multiple electrical devices 4, information on the vibration of the multiple electrical devices 4, and information on the time as pre-processing completion data in units of time, and outputs the generated pre-processing completion data to the device state determination unit 34A.

[0102] The device state determination unit 34A inputs the preprocessed data to the trained calculation model set by the learning processing unit 32A, and determines the operating states of the plurality of electrical devices 4 based on the output of the trained calculation model.

[0103] The hardware configuration example of the device state determination device 1A according to the second embodiment is the same as Fig.11 The hardware configuration of the device state determination apparatus 1 shown is the same. The processor 101 reads and executes the program stored in the memory 102 to execute the functions of the acquisition unit 31, the learning processing unit 32A, the calculation processing unit 33A, and the device state determination unit 34A.

[0104] As described above, in the device state determination system 100A according to the second embodiment, the trained calculation model takes information indicating the temperature and vibration of the plurality of electrical devices 4 at each of the plurality of moments as input. The acquisition unit 31 acquires information indicating the temperature and vibration of the plurality of electrical devices 4 at each of the plurality of moments, i.e., additional information. The device state determination unit 34A inputs the calculation result of the operation processing unit 33A, the additional information, and the information at the plurality of moments into the trained calculation model to determine the operation state of the plurality of electrical devices 40. Thus, the device state determination system 100A can determine the operation state of the plurality of electrical devices 4 with higher accuracy.

[0105] Implementation 3

[0106] The device state determination system according to the third embodiment differs from the device state determination system 100 according to the first embodiment in that the operation states of a plurality of electrical devices are determined by an electricity meter. Hereinafter, the same reference numerals are given to the components having the same functions as those in the first embodiment and the description thereof is omitted, and the description will be centered on the differences from the device state determination system 100 in the first embodiment.

[0107] Fig.15 This is a diagram showing an example of the configuration of a device state determination system according to Embodiment 3 of the present invention. Fig.16 This is a diagram showing an example of the structure of a cloud server involved in Implementation Example 3. Fig.17 This is a diagram showing an example of the structure of an electricity meter according to the third embodiment.

[0108] like Fig.15 As shown, the device state determination system 100B according to the third embodiment includes a cloud server 81, an electric power meter 3B, a measuring device 5, and HEMS 71 to 7 k The cloud server 81 has the same functions as part of the functions of the device state determination device 1 , and the same reference numerals are given to the structural elements having the same functions as the device state determination device 1 , and the description of the structural elements having the same functions as the device state determination device 1 is sometimes omitted.

[0109] The cloud server 81 is composed of, for example, one or more servers and one or more memories, which constitute a cloud system built in a data center (not shown). Fig.16 As shown, the cloud server 81 is different from the device state determination device 1 in that it has a processing unit 12B instead of the processing unit 12. The processing unit 12B is different from the processing unit 12 of the device state determination device 1 in that it has an output processing unit 35 instead of the calculation processing unit 33 and the device state determination unit 34.

[0110] The output processing unit 35 causes the communication unit 10 to transmit the calculation model setting data for setting the trained calculation model generated by the learning processing unit 32 in the electricity meter 3B via the network 80 .

[0111] like Fig.17 As shown, the electricity meter 3B includes a communication unit 50, a processing unit 51, and a storage unit 52. The communication unit 50 is connected to a network 80 and can receive calculation model setting data from a cloud server 81 via the network 80. The processing unit 51 includes an acquisition unit 71, a calculation processing unit 72, and an equipment state determination unit 73.

[0112] The acquisition unit 71 acquires the measurement data received from the measuring device 5 by the communication unit 50 similarly to the acquisition unit 31, and stores the acquired measurement data as determination data in the storage unit 52. Thus, the determination data 61 stored in the storage unit 52 is updated.

[0113] The acquisition unit 71 acquires the calculation model setting data received from the cloud server 81 by the communication unit 50, and stores the acquired calculation model setting data in the storage unit 52. Thus, the calculation model setting data 62 stored in the storage unit 52 is updated.

[0114] The calculation processing unit 72 calculates the effective value of the current flowing through the power line 8 and the frequency spectrum of the current flowing through the power line 8 based on the multiple current instantaneous values ​​included in the determination data 61, similarly to the calculation processing unit 33. Then, the calculation processing unit 72 generates preprocessing completion data similarly to the calculation processing unit 33, and outputs the generated preprocessing completion data to the device state determination unit 73.

[0115] The device state determination unit 73 obtains the calculation model setting data 62 stored in the storage unit 52, and updates the trained calculation model based on the obtained calculation model setting data 62. The device state determination unit 73 inputs the preprocessed data to the trained calculation model in the same manner as the device state determination unit 34, and determines the operation states of the plurality of electrical devices 4 based on the output of the trained calculation model. The device state determination unit 73 transmits information indicating the determined operation states of the plurality of electrical devices 4 from the communication unit 50 to an external device via the network 80.

[0116] In the above example, the electricity meter 3B determines the operating states of the plurality of electrical devices 4 , but the operating states of the plurality of electrical devices 4 may be determined by a device different from the electricity meter 3B. Fig.18 FIG. 2 is a diagram showing another example of the structure of the device state determination system according to Embodiment 3. Fig.18 The device state determination system 100B shown includes an edge computer 9 that determines the operating states of a plurality of electric devices 4 .

[0117] Fig.18 The edge computer 9 shown is Fig.17 The configuration of the electricity meter 3B shown is the same. The edge computer 9 is connected to the electric device 4 by wireless or wired means to obtain the measurement data from the electric device 4 , but it is also possible to obtain the measurement data from the electric device 4 via the network 80 .

[0118] In the device state determination system 100B, the learning function requiring huge computer resources is installed on the cloud server 81, and the determination function requiring higher-speed execution is installed on the edge computer 9. Thus, the device state determination system 100B can quickly determine the operating states of multiple electrical devices 4.

[0119] In addition, the device state determination system 100B may be configured to use information on the temperature and vibration of multiple electrical devices 4 in addition to information on the effective value of the current, information on the spectrum of the current, and information on the time, similarly to the device state determination system 100A. In this case, the cloud server 81 includes a learning processing unit 32A instead of the learning processing unit 32. The output processing unit 35 causes the communication unit 10 to send calculation model setting data to the electricity meter 3B via the network 80, and the calculation model setting data is used to set the trained calculation model generated by the learning processing unit 32A to the electricity meter 3B. In this case, the device state determination system 100B includes a measuring device 5A instead of the measuring device 5. The cloud server 81 obtains measurement data from the measuring device 5A. The device state determination unit 73 of the electricity meter 3B determines the operating state of multiple electrical devices 4 based on the measurement data sent from the measuring device 5A, similarly to the device state determination unit 34. In addition, the electricity meter 3B may also be configured to have the measuring device 5 or the measuring device 5A built-in.

[0120] Hardware configuration examples of the cloud server 81, the electricity meter 3B, and the edge computer 9 according to the third embodiment Fig.11 The hardware structure of the device state determination device 1 shown is the same. The processor 101 reads and executes the program stored in the memory 102, thereby executing the functions of the acquisition unit 31, the learning processing unit 32, and the output processing unit 35. In addition, the processor 101 reads and executes the program stored in the memory 102, thereby executing the functions of the acquisition unit 71, the calculation processing unit 72, and the device state determination unit 73.

[0121] As described above, the electricity meter 3B involved in the third embodiment includes an acquisition unit 71, a calculation processing unit 72, and a device state determination unit 73. The acquisition unit 71 acquires the detection result of the current sensor that detects the instantaneous value of the current flowing through the power line 8 to which the plurality of electrical devices 4 are connected. The calculation processing unit 72 calculates the effective value and frequency spectrum of the current flowing through the power line 8 at a plurality of times based on the detection result acquired by the acquisition unit 31. The device state determination unit 73 inputs the calculation result of the calculation processing unit 72 and the information at a plurality of times into a trained calculation model generated by machine learning, and determines the operation states of the plurality of electrical devices 4, wherein in the machine learning, the information of the effective value and frequency spectrum of the current flowing through the power line 8 at a plurality of times and the information at a plurality of times are input, and the operation states of the plurality of electrical devices 4 are output. As a result, the electricity meter 3B can determine the operation states of the plurality of electrical devices 4 with high accuracy.

[0122] In the above-mentioned embodiments 1 to 3, an example with HEMS 7 is shown, but depending on the use environment, it is also possible to have an EMS (Energy Management System) other than HEMS 7, that is, FEMS (Factory Energy Management System) or BEMS (Building Energy Management System). In addition, the device state determination system 100, 100A, 100B can also collect learning data from other than HEMS 7 and EMS. In this case, the device state determination system 100, 100A, 100B may also be a structure without HEMS 7 and EMS.

[0123] In addition, in Embodiments 1 to 3, the device state determination systems 100, 100A, and 100B are described as examples each consisting of a plurality of devices, but the present invention is not limited thereto. The device state determination systems 100, 100A, and 100B may each have a structure required for accurately determining the operating states of a plurality of electrical devices 4, and may be a single device if the single device has the required structure.

[0124] The structure shown in the above embodiment is an example of the content of the present invention, and may be combined with other well-known technologies. Without departing from the scope of the present invention, part of the structure may be omitted or changed.

[0125] Description of the label

[0126] 1. 1A equipment status determination device, 2 power supply device, 3. 3B electricity meter, 4. 41, 42, 43, ..., 4 m 、40、401、402、403、…、40 m Electrical equipment, 5, 5A, 41 measuring equipment, 6 circuit breakers, 7, 71~7 k ,7A,7A1~7A k HEMS, 8, 43 power line, 9 edge computer, 10, 50 communication unit, 11, 11A, 52 storage unit, 12, 12A, 12B, 51 processing unit, 21, 21A learning data, 22, 22A, 61 judgment data, 31, 71 acquisition unit, 32, 32A learning processing unit, 33, 33A, 72 calculation processing unit, 34, 34A, 73 device state judgment unit, 35 output processing unit, 62 calculation model setting data, 80 network, 81 cloud server, 100, 100A, 100B device state judgment system.

Claims

1. A device status determination system, characterized in that: have: an acquisition unit that acquires a detection result of a current sensor that detects an instantaneous value of a current flowing through a power line to which a plurality of electrical devices are connected, the plurality of electrical devices including a plurality of identical electrical devices; an arithmetic processing unit that calculates effective values ​​and frequency spectra of currents flowing through the power line at respective times based on the detection results acquired by the acquisition unit; and The device state determination unit inputs the calculation results of the operation processing unit and the information at the multiple moments into a trained calculation model generated by machine learning, and determines the operation states of the multiple electrical devices, wherein in the machine learning, the effective value and spectrum information of the current flowing through the power line at each of the multiple moments and the information at the multiple moments are used as inputs, and the operation states of the multiple electrical devices are used as outputs. Each cycle of the AC voltage in the power line includes a plurality of the instants.

2. The device status determination system according to claim 1, characterized in that: The calculation model takes as input at least any one of the temperature and vibration of the plurality of electrical devices at each of the plurality of time periods, The acquisition unit acquires additional information which is information indicating at least any one of the temperature and the vibration of the plurality of electrical devices at each of the plurality of times. The device state determination unit inputs the calculation result of the calculation processing unit, the additional information, and the information at the plurality of times into the calculation model to determine the operation states of the plurality of electrical devices.

3. The device status determination system according to claim 1 or 2, characterized in that: have: a storage unit for storing, when the plurality of electrical devices are set as a plurality of first electrical devices and the power line is set as a first power line, learning data including information indicating effective values ​​and frequency spectra of currents flowing through a second power line different from the first power line and connected to a plurality of second electrical devices different from the plurality of first electrical devices at each of a plurality of moments in a predetermined period, data including information indicating each of the plurality of moments in the predetermined period, and data including information indicating operating states of the plurality of second electrical devices at each of the plurality of moments; as well as A learning processing unit generates the calculation model through the machine learning based on the learning data.

4. An electricity meter, characterized in that: have: an acquisition unit that acquires a detection result of a current sensor that detects an instantaneous value of a current flowing through a power line to which a plurality of electrical devices are connected, the plurality of electrical devices including a plurality of identical electrical devices; an arithmetic processing unit that calculates effective values ​​and frequency spectra of currents flowing through the power line at respective times based on the detection results acquired by the acquisition unit; and The device state determination unit inputs the calculation results of the operation processing unit and the information at the multiple moments into the trained calculation model generated by machine learning, and determines the operation states of the multiple electrical devices, wherein in the machine learning, the effective value and spectrum information of the current flowing through the power line at each of the multiple moments and the information at the multiple moments are used as input, and the operation states of the multiple electrical devices are used as output, Each cycle of the AC voltage in the power line includes a plurality of the instants.

5. A method for determining a device state, which is executed by a computer, and the method for determining a device state is characterized in that it comprises: The first step is to obtain a detection result of a current sensor that detects an instantaneous value of a current flowing through a power line to which a plurality of electrical devices are connected, wherein the plurality of electrical devices include a plurality of identical electrical devices; A second step of calculating an effective value and a frequency spectrum of a current flowing through the power line at each of a plurality of times based on the detection result obtained in the first step; and In the third step, the calculation result of the first step and the information of the plurality of moments are input into a trained calculation model generated by machine learning to determine the operation status of the plurality of electrical devices, wherein: In this machine learning, information on the effective value and spectrum of the current flowing through the power line at each of the plurality of times and information at the plurality of times are used as input, and the operating states of the plurality of electrical devices are used as output. Each cycle of the AC voltage in the power line includes a plurality of the instants.

Images