JP7265847B2 - Power data processing system and method of processing power data using power data processing system - Google Patents

Description

The present invention includes a smart meter having a communication route for measuring the amount of electric power distributed to electrical equipment and communicating the measured amount of electric power as electric power data to the outside, and processing means for processing the electric power data communicated from the smart meter. A power data processing system and a method of processing power data using the power data processing system.

In general, a smart meter has a function of detecting the power amount and power value of electrical equipment in power distribution equipment, and also has a function of externally communicating the detected power amount and the like as power data.

When the power consumption of electrical equipment detected by the smart meter is communicated to the outside as power data, the electric power company or another business operator detects the usage state of the electric equipment by the user based on this power data. , it is possible to get an overview of the user's activity.

Patent Literature 1 discloses a living activity pattern extraction device that extracts a user's living activity pattern based on the user's power consumption obtained by a smart meter.

According to the living activity pattern extraction device of this patent document 1, it is determined whether or not a living activity has occurred when the power consumption exceeds the threshold, and if it is determined that the living activity has occurred, the It is possible to determine that a life event has occurred by associating the corresponding life events, and to estimate the frequency, probability, etc. of occurrence of the life event in the target period based on the life event set in which the life events are accumulated.

Japanese Patent Application Laid-Open No. 2015-185040

However, the life activity pattern extraction device of Patent Document 1 grasps changes in the power consumption of the user and determines whether or not a life event has occurred. Therefore, there is concern that the life event, that is, the activity state of the user cannot be determined appropriately.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a power data processing system and a method of processing power data using the power data processing system that can appropriately estimate the user's activity state. This is an issue.

A power data processing system for achieving the above object is a smart meter having a communication route for measuring the amount of power to be distributed to electrical equipment and externally communicating the measured power amount as power data, and receiving communication from the smart meter. In a power data processing system comprising processing means for processing power data, the processing means adds to the power data power amount attribute data relating to the attribute of the power amount when the power amount constituting the power data for a certain period of time is measured. a preprocessing program that generates processed data by means of a preprocessing program, and an activity that acquires the user's activity state as activity state data during a specific period arbitrarily set within a fixed period corresponding to the fixed period of the processed data generated by the preprocessing program. A user's activity corresponding to the processed data by the learning data generated by the state data acquisition program, the activity state data acquired by the activity state data acquisition program, the specific period in which the activity state data was acquired, and the processed data in the corresponding specific period and an artificial intelligence program for estimating the user's activity state in a remaining period obtained by subtracting the specific period from the fixed period based on a learned model obtained by subjecting the state to machine learning.

According to this electric power data processing system, a preprocessing program is provided for generating processing data by adding electric energy attribute data to electric power data for a certain period configured based on the electric energy communicated via the smart meter. Therefore, the amount of information is compounded so that the processing data can be analyzed based on the power data and the power amount attribute data.

On the other hand, in the electric power data processing system, the learning data is generated from the activity state data acquired in the specific period and the processing data of the specific period corresponding to this specific period, so that the artificial intelligence program performs machine learning. Data can be generated with precision.

Since the artificial intelligence program performs machine learning and generates a learned model based on the precisely generated learning data, the artificial intelligence program processes the remaining period based on this learned model. The activity state of the user corresponding to the data can be estimated appropriately and in detail.

Furthermore, the artificial intelligence program of this electric power data processing system generates a trained model in each of a plurality of subsequent unit times set in advance by setting a time difference while overlapping the preset unit time and the unit time and the time zone. Then, the activity state of the user is estimated in a time zone overlapping with each unit time among new unit times set in advance within the remaining period subsequent to each unit time.

In this way, a plurality of unit times are set between a preset unit time and a new unit time, a trained model is generated for each of the preset unit time and the plurality of unit times, and each Based on the learned model, the activity state of the user is estimated in a time period overlapping with a preset unit time and a plurality of unit times in the new unit time. Therefore, the activity state of the user in the new unit time can be estimated early.

In particular, the artificial intelligence program of the electric power data processing system generates a plurality of trained models in each unit of time. or not, selects an estimation result of the user's activity state estimated based on the learned model evaluated as high, and estimates the user's activity state based on the selected estimation result It is characterized in that estimated data is generated.

Therefore, since the user's activity state is selected as the estimation result based on the learned model with high learning accuracy among the plurality of trained models generated in each unit time, the activity state estimation data as the estimation result It is expected that the accuracy will be improved.

Further, the user's activity state estimated by the artificial intelligence program of the power data processing system is communicated as data to the user's terminal. Therefore, the user can grasp the outline of the activity state.

Moreover, the activity data acquisition program of the power data processing system is characterized by being stored in the user's terminal. Activity data is thus easily obtained on-site by the user's terminal.

A method for processing power data using a power data processing system for achieving the above object has a communication route for measuring the amount of power distributed to electrical equipment and communicating the measured power amount to the outside as power data. In a method of processing power data using a smart meter and a power data processing system including a processing means for processing power data communicated from the smart meter, when the power amount constituting the power data for a certain period of time is measured A preprocessing step of adding electric energy attribute data related to electric energy attributes to electric power data to generate processing data; an activity data acquisition step of acquiring the user's activity state as activity data during a period; the activity data acquired in the activity data acquisition step; the activity data acquired in the activity data acquisition step; an activity state estimating step of estimating the user's activity state in a residual period obtained by subtracting a specific period from a certain period based on a learned model obtained by machine-learning the user's activity state corresponding to the processed data using the generated learning data; , is provided.

According to the method of processing power data using this power data processing system, it is possible to appropriately and in detail estimate the state of activity of the user corresponding to the processed data of the remaining period.

According to this invention, the user's activity state can be estimated appropriately and in detail.

1 is a block diagram outlining a deployment environment in which a power data processing system according to an embodiment of the present invention is deployed; FIG. Similarly, it is a block diagram for explaining the outline of the power data processing system according to the present embodiment. Similarly, it is a figure explaining the outline of the storage of the server of the electric power data processing system which concerns on this Embodiment. Similarly, it is a figure explaining the outline of the smart phone of the electric power data processing system which concerns on this Embodiment. Similarly, it is a figure explaining the outline of the data processing in the pre-processing program of the electric power data processing system which concerns on this Embodiment. Similarly, FIG. 4 is a diagram for explaining an outline of profile data input to the profile input interface of the power data processing system according to the present embodiment. Similarly, it is a figure explaining the outline of the activity state data acquired by the activity state data acquisition program of the electric power data processing system which concerns on this Embodiment. Similarly, it is a block diagram for explaining an outline of data processing in the artificial intelligence program of the electric power data processing system according to the present embodiment. Similarly, it is the figure which represented typically the learning data produced|generated by the artificial intelligence program. Similarly, it is a figure explaining the outline of the trained model produced|generated based on the learning data produced|generated by the artificial intelligence program. Similarly, it is a figure explaining the outline of the procedure which presumes a user's activity state based on the learned model produced|generated by the artificial intelligence program. Similarly, it is a figure explaining the outline of the activity state estimation data produced|generated by an artificial intelligence program. Similarly, it is a figure explaining the outline of the procedure which presumes a user's activity state based on the learned model produced|generated by the artificial intelligence program. Similarly, it is a figure explaining the outline of the procedure which presumes a user's activity state based on the learned model produced|generated by the artificial intelligence program. Similarly, it is a flowchart for explaining the outline of the operation of the power data processing system according to the present embodiment.

Next, embodiments of the present invention will be described with reference to FIGS. 1 to 15. FIG.

FIG. 1 is a block diagram illustrating an overview of a deployment environment in which a power data processing system according to this embodiment is deployed. As shown, a deployment environment 100 provides services to a home 10, an electric utility 20 that supplies power to the home 10, and a user U associated with the home 10 regarding the power that the electric utility 20 supplies to the home 10. It is configured by the service provider 30 .

The house 10 is provided with a measurement communication facility 11 for measuring and communicating the electric power distributed to the electrical equipment 12 in the house 10 . In this embodiment, the measurement communication equipment 11 includes a smart meter 11a that measures the power consumption and power value of the electrical equipment 12 for a certain period of time and communicates the measured power consumption and the like to the outside as power data.

Furthermore, the measurement communication equipment 11 collects the electric energy of all the electric devices 12 measured by the smart meter 11a as electric power data, and communicates the electric power data to the service provider 30 via the HEMS (Home Energy Management System) 11b. A router 11c is provided.

The electric power company 20 is provided with an MDM system 21 that communicates with the smart meter 11 a to receive power data and communicates the received power data with the service company 30 .

On the other hand, the service provider 30 receives electric power data by communicating with the smart meter 11a via the router 11c installed in the measurement communication equipment 11 of the house 10, and receives power data from the MDM system deployed in the electric power provider 20. A server 31 is provided that communicates with 21 to receive power data.

In the present embodiment, communication between the smart meter 11a and the MDM system 121 of the electric power company 20 is performed via the A route 101, and communication between the smart meter 11a and the server 31 of the service company 30 is performed via the router 11c. Communication between is performed via B route path 102 consisting of B route 102a and Internet 102b, and communication between MDM system 21 of electric utility company 20 and server 31 of service provider 30 is performed via C route 103. run through.

In this embodiment, the power data communicated through the A route 101, the B route 102, and the C route 103 are configured based on the amount of power (kWh) detected in units of 30 minutes.

In such a deployment environment 100 , the smartphone 40 , which is a terminal owned by the user U, is configured to be able to access the MDM system 21 of the electric power company 20 and the server 31 of the service company 30 .

The power data processing system is deployed in such a deployment environment 100. As shown in the block diagram of FIG. 31 and the smartphone 40 owned by the user U.

The smart meter 11a communicates power data to the server 31 via the B route 102 in this embodiment.

The server 31 includes a processor 32 , a memory 33 , a storage 34 , a transmission/reception section 35 and an input/output section 36 as main components, which are electrically connected to each other via a bus 37 .

The processor 32 is an arithmetic device that controls the operation of the server 31, controls transmission and reception of data between elements, and performs processes necessary for executing application programs.

The processor 32 is, for example, a CPU (Central Processing Unit) in this embodiment, and executes an application program stored in a storage 34 described later and developed in a memory 33 to perform each process.

The memory 33 includes a main storage device composed of a volatile memory device such as a DRAM (Dynamic Random Access Memory), and an auxiliary storage device composed of a non-volatile memory device such as a flash memory or a HDD (Hard Disc Drive). .

The memory 33 is used as a work area for the processor 32, and stores a BIOS (Basic Input/Output System) executed when the server 31 is started, various setting information, and the like.

The storage 34 stores application programs, data used for various processes, and the like.

The transmitter/receiver 35 connects the server 31 to the network. The transmission/reception unit 35 may include a short-range communication interface such as Bluetooth (registered trademark) or BLE (Bluetooth Low Energy).

Information input devices such as a keyboard and a mouse and output devices such as a display are connected to the input/output unit 36 as necessary.

The bus 37 transmits, for example, address signals, data signals, and various control signals among the connected processor 32 , memory 33 , storage 34 , transmission/reception unit 35 and input/output unit 36 .

FIG. 3 is a diagram illustrating an outline of the storage 34 of the server 31 of the power data processing system 1 according to this embodiment.

As shown in the figure, the storage 34 stores a preprocessing program 34a that processes electric power data input from the smart meter 11a to generate processed data, and the activity status of the user U based on the processed data and activity status data described later. An artificial intelligence program 34b to estimate is stored.

Further, the storage 34 stores a post-processing program 34c that generates user activity state data, which will be described later, based on the estimation of the artificial intelligence program 34b.

FIG. 4 is a diagram illustrating an outline of smartphone 40 of power data processing system 1 according to the present embodiment.

As shown in the figure, the smartphone 40 has a profile input interface 41 for inputting a profile related to the attributes of the user U as profile data described later, and activity state data by inputting the activity state of the user U as activity state data. An activity state data acquisition program 42 to acquire is stored.

The pre-processing program 34a, the artificial intelligence program 34b, the post-processing program 34c, the profile input interface 41 and the activity state data acquisition program 42 constitute processing means in this embodiment.

Next, a specific configuration of each part of the power data processing system 1 will be described.

For convenience of explanation, the pre-processing program 34a, the profile input interface 41, the activity data acquisition program 42, the artificial intelligence program 34b, and the post-processing program 34c will be described in this order.

FIG. 5 is a diagram for explaining an overview of data processing in the preprocessing program 34a of the power data processing system 1 according to the present embodiment.

As shown in FIG. 5A, the power data D1 for a certain period of time X processed by the preprocessing program 34a is, in the present embodiment, the date d1a when the power amount is measured, and is measured in units of 30 minutes. 30-minute power consumption d1b.

In this power data D1, the power data D1a for a specific period Y, which is arbitrarily set in the power data D1 for the fixed period X, is set as the power data D1a. The power data of the reduced residual period Z is set as the power data D1b.

In the present embodiment, the preprocessing program 34a adds electric energy attribute data D2 related to the attribute of the 30-minute electric energy d1b when the 30-minute electric energy d1b constituting the electric power data D1 is measured to the electric power data D1. , to generate processing data D3 as shown in FIG. 5(b).

The electric energy attribute data D2 is configured to have a day of the week d2 in this embodiment. Furthermore, it is assumed that the electric energy attribute data D2 includes various given conditions such as the weather and temperature when the 30-minute electric energy d1b is detected, for example.

In this embodiment, the processing data D3 to which the electric energy attribute data D2 is added corresponds to the fixed period X, the specific period Y, and the remaining period Z set in the electric power data D1. and the remaining period Z is set, and the processing data for a specific period Y set arbitrarily out of the processing data D3 for the fixed period X is set as the processing data D3a, and the processing data for the specific period Y is set from the processing data D3 for the fixed period X The processing data for the remaining period Z obtained by subtracting D3a is set as the processing data D3b.

The profile input interface 41 is used by the user U to input a profile relating to attributes of the user U as profile data.

FIG. 6 is a diagram illustrating an outline of profile data input to profile input interface 41 of power data processing system 1 according to the present embodiment.

As shown in the figure, the profile data D4 includes, in this embodiment, a user ID d4a that is the IDs of the user U and a plurality of other users U1 to Un, and the user U and the plurality of other users U1 such as the number of residents and housing configuration. ~Un profile d4b.

The activity state data acquisition program 42 acquires the activity state by inputting the activity state of the user U as activity state data to the user U during a specific period Y set arbitrarily within the fixed period X set in the processing data D3. It is a program that acquires data.

FIG. 7 is a diagram for explaining the outline of the activity state data acquired by the activity state data acquisition program 42 of the power data processing system 1 according to the present embodiment. As illustrated, the activity state data D5 is composed of a date d5a when the user U grasped the activity state, and an activity state d5b of the user U such as getting up, cooking, and going to bed.

In the present embodiment, the activity state data acquisition program 42 reads, for example, "On April 1, 2018, the user U got up at 6:00 am, cooked at 8:00 am, and went to bed at 11:30 pm. , etc. are acquired as the activity state data D5.

FIG. 8 is a block diagram outlining data processing in the artificial intelligence program 34b of the power data processing system 1 according to the present embodiment, and FIG. 9 schematically shows learning data generated by the artificial intelligence program 34b. It is a diagram of

As shown in FIG. 8, in the present embodiment, the artificial intelligence program 34b generates learning data D6 from the activity state data D5 acquired by the activity state data acquisition program 42 and the processing data D3a of the specific period Y. .

On the other hand, if the activity state data D5 does not exist, for example, if the user U did not enter the activity state d5b during the specific period Y, a plurality of Profile data D4 of other users U1 to Un are referenced.

In the present embodiment, the profile data D4 of another user U1 having a profile d4b similar to the profile d4b of the user U is referred to, and learning data D6 generated based on the activity state data D5 of the user U1 is used thereafter. is processed.

As schematically shown in FIG. 9(a), the learning data D6 is the power data D1a included in the processing data D3 detected in a predetermined unit time within the specific period Y, 24 hours in this embodiment. The activity state d5b of the activity state data D5 input in the specific period Y corresponding to this specific period Y is superimposed on the active state data D5.

In this embodiment, the activity state d5b input by the user U during the specific period Y, such as "sleep at midnight", "wake up at 7 am", "clean at 10 am", etc., is stored in the processing data D3 of the specific period Y. It is superimposed on the included power data D1a. As a result, learning data D6 as shown in FIG. 9B is generated.

Furthermore, the artificial intelligence program 34b generates a learned model D7 by performing machine learning with the generated learning data D6.

Various algorithms such as neural networks, random forests, and SVMs (support vector machines) are appropriately used as techniques for machine learning.

FIG. 10 is a diagram for explaining an outline of generating a trained model D7 based on the learning data D6 generated by the artificial intelligence program 34b. As shown in the figure, the learned model D7 is obtained by machine-learning the activity state of the user U, and is a machine learning model that receives the profile data D4 and the processing data D3a and outputs the activity state d5b. A trained model D7 is generated by autonomously adjusting so as to satisfy the relationship of

Following the generation of the learned model D7, the artificial intelligence program 34b estimates the activity state of the user U corresponding to the processing data D3b of the residual period Z based on the generated learned model D7.

FIG. 11 is a diagram outlining a procedure for estimating the activity state of the user U based on the learned model D7. As shown in the figure, for example, when 24 hours on April 27, 2018 is set in advance as the unit time T1, the trained model D7a corresponding to the unit time T1 estimates when the unit time T1 has passed. processing takes place.

On the other hand, in the artificial intelligence program 34b, 24 hours on April 28, 2018 within the remaining period Z is set in advance as a new unit time T2 following the unit time T1, and after the unit time T2 has passed, At a certain time t1, the activity state of the user U in the unit time T2 is estimated by the trained model D7b corresponding to the unit time T2.

When the activity state of the user U is estimated in the artificial intelligence program 34b, activity state estimation data D8 is generated.

As shown in FIG. 12, the activity state estimation data D8 is constructed by adding an activity state d8 as an estimation result to the processing data D3b of the residual period Z. In this embodiment, for example, the residual It is estimated that user U woke up at 7:00 am on April 28, 2018 within period Z.

By the way, in the present embodiment, a plurality of unit times are set between the unit time T1 and the unit time T2 in the artificial intelligence program 34b. Each activity state of the user U is estimated based on D7.

An outline of the procedure for estimating the activity state of the user U in this case will be described with reference to FIGS. 13 and 14. FIG. As illustrated, in the present embodiment, a plurality of unit times are set between the unit time T1 and the unit time T2 in the artificial intelligence program 34b.

For example, when 24 hours on April 27, 2018 is set as the unit time T1, the unit time T1a follows the unit time T1 with a time difference of 8 hours while overlapping the unit time T1. (April 27 to 28, 2018) is set, and the unit time T1b (April 27, 2018 to 28th) is set.

The unit time T1b overlaps with the unit time T2 and is set with a time difference of 8 hours.

After the unit times T1-T2 have passed, learned models D7a-D7d are generated.

In the present embodiment, based on the generated trained models D7a-D7c, the activity state of the user U is estimated in the time zone overlapping each of the unit times T1-T1b in the unit time T2.

When estimating the activity state of the user U in the unit time T2, the estimation result based on the learned model D7a generated in the unit time T1a or the estimation result based on the learned model D7b generated in the unit time T1b is used. be done.

For example, when estimating the activity state of the user U at an arbitrary time t2 (7:00 am) of the unit time T2, the learned model D7c corresponding to the unit time T1a is used for the unit time after the unit time T1a has passed. At time t2 of T2, the activity state of user U is estimated.

Therefore, compared to the case where a plurality of unit times T1a and T1b are not set between the unit time T1 and the unit time T2, the activity state of the user U at any time of the unit time T2 is estimated early.

When the activity state of the user U is estimated in the artificial intelligence program 34b, activity state estimation data D8 is generated.

On the other hand, in the artificial intelligence program 34b of the present embodiment, an arbitrary estimation result is selected from among the estimation results generated by the trained models D7a, D7c, and D7d corresponding to each unit time T1 to T1b. Activity state estimation data D8 is generated based on the estimated result obtained.

The selection of the estimation results is performed by firstly selecting the trained models D7a, D7c when the artificial intelligence program 34b generates the estimated results from the trained models D7a, D7c, and D7d corresponding to the unit times T1 to T1b. and D7d, and autonomously evaluates whether or not there is a high rate of matching the activity state d5b of the user U.

Thereafter, an estimation result estimated based on the learned model autonomously evaluated as having a high rate of matching with the activity state d5b is selected, and based on the selected estimation result, the activity state of the user U in the unit time T2 is determined. Activity state estimation data D8 is generated.

Therefore, since the activity state of the user U estimated based on the learned model D7 with high learning accuracy among the plurality of learned models D7a, D7c, and D7d is selected as the estimation result, the accuracy of the estimation result is also high. expected to be

When the activity state estimation data D8 is generated, user activity state data D9 is generated as shown in FIG. This user activity data D9 is generated by the post-processing program 34c in this embodiment.

In this embodiment, the user activity data D9 is represented by an interface using various charts (for example, bar graphs, line graphs, etc.) that are easy for the user U to see.

Next, operation of the power data processing system 1 according to this embodiment will be described.

In this embodiment, an example will be described in which a unit time T1a and a unit time T1b are set between the unit times T1 and T2.

FIG. 15 is a flow chart for explaining the outline of the operation of the power data processing system 1 according to this embodiment.

As shown in the figure, in step S 1 , the power data D 1 for a certain period of time X communicated to the server 31 via the B route 102 is input to the preprocessing program 34 a stored in the server 31 .

When the power data D1 for a certain period of time X is input to the preprocessing program 34a, in step S2, the preprocessing program 34a adds the power amount attribute data D2 based on the date d1a included in the power data D1. , to generate processing data D3 (preprocessing step).

The generated processing data D3 is input to the artificial intelligence program 34b in this embodiment.

On the other hand, in step S3, when the activity state data D5 is acquired by the activity state data acquisition program 42 (activity state data acquisition step), the acquired activity state data D5 is sent to the artificial intelligence data through the transmission/reception unit 35 of the server 31. input to program 34b.

Thus, the learning data D6 is generated in step S4 from the processing data D3 and the activity data D5 input to the artificial intelligence program 34b.

When the learning data D6 is generated, the artificial intelligence program 34b performs machine learning based on the learning data D6 by appropriately using various algorithms such as a neural network, random forest, and SVM in step S5. Generate a finished model D7.

In this embodiment, a learned model D7a corresponding to a preset unit time T1 is generated.

Subsequently, based on the generated learned model D7a, the artificial intelligence program 34b autonomously estimates the activity state of the user U in the unit time T1 in step S6 (activity state estimation step).

Next, in step S7, it is determined whether or not the unit time has passed three times, that is, whether or not the unit time T1b has passed.

In the present embodiment, the unit time T1a elapses eight hours after the unit time T1 elapses. After the unit time T1a has passed, steps S2 to S6 are repeated.

At this time, in step S5, the trained model D7c is generated when the unit time T1a has passed, and based on the generated trained model D7c, in step S6, the artificial intelligence program 34b calculates the user Assume U's activity state autonomously.

Similarly, eight hours after the unit time T1a has passed, the unit time T1b passes. After the unit time T1b has passed, steps S2 to S6 are repeated.

At this time, in step S5, the trained model D7d is generated when the unit time T1b has passed, and based on the generated trained model D7d, in step S6, the artificial intelligence program 34b generates the user Assume U's activity state autonomously.

When it is determined in step S7 that the estimation process corresponding to the unit time T1b has been completed, in step S8, the , an arbitrary estimation result (an estimation result of the activity state of the user U based on a learned model autonomously evaluated as having a high rate of matching with the activity state data D5) is selected, and based on the selected estimation result , activity state estimation data D8 are generated.

The generated activity state estimation data D8 is processed by the post-processing program 34c, and the user activity state data D9 is generated by the post-processing program 34c.

In this embodiment, this user activity data D9 is communicated to the smartphone 40 of the user U via the transmission/reception unit 35 of the server 31 in step S9.

As a result, the user U can obtain an outline of the activity state of the user U during the fixed period X because the estimation result of the activity state in the remaining period Z is communicated.

The service provider 30 provides the user U with the activity state of the user U during the fixed period X as the user activity state data D9, thereby improving the level of service provided to the user U by the service provider 30. FIG.

On the other hand, if the activity state data D5 is not acquired by the activity state data acquisition program 42 in step S3 (activity state data acquisition step), in step S10, another user U1 having a profile d4b similar to the profile d4b of user U active state data D5 is referred to.

After that, in step S11, a plurality of trained models D7a, D7c and D7d generated by learning data D6 generated based on the activity state data D5 of the user U1 are applied, and the activity of the user U in the unit time T1 to T1b is applied. Estimate the state autonomously.

As described above, the power data processing system 1 of the present embodiment uses power data D1 for a certain period of time X based on the 30-minute power amount d1b communicated via the B route 102 of the smart meter 11a. Since the preprocessing program 34a that generates the processing data D3 by adding the amount attribute data D2 is provided, the amount of information is complex so that the processing data D3 can be analyzed based on the power data D1 and the power amount attribute data D2. increased exponentially.

On the other hand, in the power data processing system 1, since the learning data D6 is generated by the activity state data D5 acquired in the specific period Y and the processing data D3a of the specific period Y corresponding to this specific period Y, artificial intelligence The program 34b can precisely generate learning data D6 for machine learning.

Thus, based on the precisely generated learning data D6, the artificial intelligence program 34b of the present embodiment performs machine learning to generate the trained model D7. Based on the model D7, the activity state of the user U corresponding to the processing data D3b of the remaining period Z can be estimated appropriately and in detail.

In particular, in the power data processing system 1 of the present embodiment, a plurality of unit times T1a and T1b are set between the unit time T1 and the unit time T2, and the learned models D7a to D7a to T1b are set in each of the unit times T1 to T2. D7d is generated.

Among these trained models D7a to D7d, based on each trained model D7a, D7c, and D7d generated in a time zone that overlaps each unit time T1 to T1b in unit time T2, at an arbitrary time in unit time T2 User U's activity state is estimated.

Therefore, the activity state of the user U is estimated early at an arbitrary time of the unit time T2 (in this embodiment, for example, the time t2 shown in FIG. 13).

Moreover, the electric power data processing system 1 of the present embodiment can provide the user U based on the learned model autonomously evaluated as having a high rate of matching with the activity state data D5 among the learned models D7a, D7c, and D7d. is selected, and activity state estimation data D8 is generated based on the selected estimation result.

Therefore, since the activity state of the user U is selected as the estimation result based on the trained model with high learning accuracy among the plurality of trained models D7a, D7c, and D7d, it is expected that the accuracy of the estimation result will be improved. be done.

Furthermore, in the present embodiment, since the activity data acquisition program 42 is stored in the smartphone 40 owned by the user U, the user U of the power distribution equipment 11 inputs the activity status at the residence 10 on-site. be able to. Therefore, the on-site activity data D5 is easily obtained.

It should be noted that the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the invention. In the above embodiment, a case has been described in which the learned model D7 is generated based on a plurality of activity states of the user U, such as "getting up", "cooking", or "going to bed". A trained model D7 may be generated for each individual activity, such as a trained model D7 for "cooking" is generated.

In the above embodiment, the unit times T1a and T1b are set with a time difference of 8 hours between the unit time T1 and the unit time T2. The number of hours can be increased or decreased to the number of units of time as desired.

In the above embodiment, a case has been described in which the user activity state data D9 is communicated to the smartphone 40 of the user U, but it may be configured to be communicated to the computer of the user U, or It may be configured to communicate with the controller of the HEMS 11b.

1 Electric power data processing system 10 House 11 Measurement communication equipment 11a Smart meter 20 Electricity provider 30 Service provider 31 Server 34 Storage 34a Preprocessing program 34b Artificial intelligence program 40 Smartphone (terminal)
42 Activity state data acquisition program D1 Power data d1b 30 minute power consumption (power consumption)
D3 Processed data D5 Activity state data D6 Learning data D7 Trained model D8 Activity state estimation data X Fixed period Y Specific period Z Remaining period

Claims (5)

A smart meter having a communication route for measuring the amount of electric power distributed to an electrical device and externally communicating the measured amount of electric power as electric power data, and processing means for processing the electric power data communicated from the smart meter. In a power data processing system,
The processing means
a preprocessing program for generating processing data by adding to the power data power amount attribute data relating to attributes of the power amount when the power amount constituting the power data for a certain period of time is measured;
an activity state data acquisition program for acquiring a user's activity state as activity state data during a specific period arbitrarily set within a certain period corresponding to the certain period of the processed data generated by the preprocessing program;
learning data generated by the activity state data acquired by the activity state data acquisition program, the specific period during which the activity data was acquired, and the processing data in the corresponding specific period; an artificial intelligence program for estimating the state of activity of the user in a remaining period obtained by subtracting the specific period from the certain period based on a learned model obtained by subjecting the activity state to machine learning ;
The artificial intelligence program is
generating the trained model in each of a preset unit time and a plurality of succeeding unit times preset by setting a time difference while overlapping the unit time and the time zone, and following each unit time; estimating the activity state of the user in a time zone that overlaps with each of the new unit times set in advance within the remaining period;
A power data processing system characterized by: The artificial intelligence program is
autonomously evaluating whether or not the user's activity state learned by the learned model has a high rate of matching with the activity state data when a plurality of the learned models are generated in each unit time; and an estimation result of the activity state of the user estimated based on the learned model evaluated as high is selected, and activity state estimation data obtained by estimating the activity state of the user based on the selected estimation result is generated. 2. The power data processing system of claim 1 , wherein the power data processing system is generated. 3. The power data processing system according to claim 1 , wherein the user's activity state estimated by the artificial intelligence program is communicated as data to the user's terminal. The activity data acquisition program includes:
4. The power data processing system of claim 3 , stored in the terminal of the user. A smart meter having a communication route for measuring the amount of electric power distributed to an electrical device and externally communicating the measured amount of electric power as electric power data, and processing means for processing the electric power data communicated from the smart meter. In a method of processing power data using a power data processing system, comprising:
a preprocessing step of generating processing data by adding to the power data power amount attribute data relating to attributes of the power amount when the power amount constituting the power data for a certain period of time is measured;
an activity state data acquisition step of acquiring the user's activity state as activity state data during a specific period arbitrarily set within a certain period corresponding to the certain period of the processed data generated in the preprocessing step;
learning data generated by the activity state data acquired in the activity state data acquiring step, the specific period during which the activity data was acquired, and the processing data in the corresponding specific period; an activity state estimating step of estimating the user's activity state in a remaining period obtained by subtracting the specific period from the certain period based on a learned model obtained by subjecting the activity state to machine learning ;
The activity state estimation step includes:
generating the trained model in each of a preset unit time and a plurality of succeeding unit times preset by setting a time difference while overlapping the unit time and the time zone, and following each unit time; estimating the activity state of the user in a time zone that overlaps with each of the new unit times set in advance within the remaining period;
A method of processing power data using a power data processing system characterized by:

Images