Presence / absence estimation system, power control system, and presence / absence estimation method

Abstract

To provide a presence / absence estimation system capable of estimating the presence / absence of a user even if there is no information indicating the presence / absence of the user whose presence / absence is to be estimated. [Solution] The server device 1 includes a memory unit 121 that stores a learning model 121 generated by machine learning using as training data first data indicating the amount of electricity used in multiple buildings B2 to B4 and second data indicating the presence or absence of people in multiple buildings B2 to B4, and a presence / absence estimation unit 132 that estimates the presence or absence of a user by inputting third data indicating the amount of electricity used in the user's building B1 into the learning model 121.

Description

Technical Field

[0004] ,

[0006] , , , , , ,

[0005] , , , , , ,

[0001] The present disclosure relates to a presence / absence estimation system, a power control system, and a presence / absence estimation method.

Background Art

[0002] Conventionally, a presence / absence prediction system that predicts the presence / absence of a person (user) based on data on power consumption at a predetermined location is known. In the presence / absence prediction system of Patent Document 1, learning data (learning model) is generated based on the power data of a building prior to a first time and information indicating whether the person was actually in the building (information indicating presence / absence). Then, based on the power data of the building at the first time and the learning data, it is predicted whether the person is in the building.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] As described above, in a presence / absence prediction system such as that of Patent Document No. 1, in order to generate a learning model, it is necessary to obtain information indicating the presence / absence of a user who is the target of presence / absence estimation. Therefore, if information indicating the presence / absence of the user cannot be obtained, the presence / absence of the user cannot be predicted.

[0005] Therefore, an object of the present invention is to provide a presence / absence estimation system that can estimate the presence / absence of a user even without information indicating the presence / absence of the user who is the target of presence / absence estimation.

Means for Solving the Problems

[0006] To solve the aforementioned problems, an embodiment of the present / absence estimation system is a present / absence estimation system for estimating the presence or absence of a user, comprising: a storage unit that stores a learning model generated by machine learning using first data indicating the amount of electricity used in a plurality of buildings and second data indicating the presence or absence of people in the plurality of buildings as training data; and an estimation unit that estimates the presence or absence of the user by inputting third data indicating the amount of electricity used in the user's building to the learning model.

[0007] According to this embodiment of the invention, a learning model is generated using first data showing the amount of electricity used in multiple buildings and second data showing the presence or absence of people in multiple buildings. Then, by inputting third data showing the amount of electricity used in the user's building into the learning model, the presence or absence of the user is estimated. As a result, when generating the learning model, the data showing the amount of electricity used in multiple buildings, including buildings other than the user's building, and the presence or absence of people in those buildings are used as training data, eliminating the need to acquire data showing the presence or absence of the user whose presence or absence is to be estimated as training data. Therefore, the presence or absence of a user can be estimated even without information showing the presence or absence of the user whose presence or absence is to be estimated.

[0008] Alternatively, the user's building may not be included in the plurality of buildings, the third data may be data indicating the amount of electricity used by the user's building during a first period including the current time, and the estimation unit may estimate the presence or absence of the user during a second period following the first period.

[0009] According to this embodiment of the invention, since data indicating the presence or absence of people in multiple buildings, excluding the user's building, is used as training data, the presence or absence of a user can be estimated without obtaining data indicating the presence or absence of the user to be estimated as training data. Furthermore, the estimation unit estimates the presence or absence of the user in a second period following the first period by inputting third data indicating the power consumption of the user's building in a first period including the current time to the learning model. In other words, since the presence or absence of the user is estimated not only for the period immediately following the current time, but also for a period after the current time (the second period), the future presence or absence of the user can be estimated.

[0010] Furthermore, the first data may be data showing the amount of electricity used by the multiple buildings during the third period, and the second data may be data showing the presence or absence of people in the multiple buildings during a fourth period different from the third period.

[0011] According to this embodiment of the invention, since the acquisition periods of the first and second data used as training data for the learning model are the third and fourth periods, there is no need to match the acquisition periods of the first and second data. This makes it easier to generate the learning model.

[0012] Furthermore, the first data and the second data may each be classified based on the calendar in which the data were acquired, and the learning model may use the classified first data and the second data as the training data.

[0013] According to this embodiment of the invention, first and second data classified based on the calendar are used as training data for a learning model. Generally, common patterns of presence or absence of people and changes in electricity usage can occur between buildings depending on the calendar, such as holidays, weekdays, days of the week, and seasons. For example, if a building is a residence, electricity usage decreases during the day on weekdays because people are generally absent during that time. Also, electricity usage increases during a certain period from evening to night when people return to the building and use equipment connected to the building. On the other hand, on holidays, electricity usage changes in a different pattern than the weekday patterns described above because the times when people are present or absent generally differ from those on weekdays. Therefore, by classifying the first and second data based on the calendar, it is possible to more accurately estimate the presence or absence of users based on the patterns of changes in electricity usage of buildings according to the calendar. Furthermore, while the calendar-based classification may be performed on the first data for all buildings, including the user's building, it may also be performed on the first data for each building group after the multiple buildings have been pre-classified into one or more building groups based on one or more common characteristics such as the attributes and number of people in the building, the type of building, and the equipment connected to the building. Examples of such building group classifications include classifications according to the type of building, such as residences, shops, apartment buildings, offices, factories, and warehouses, as well as further subdivisions for residences based on the number of household members and the age group of the head of household, and the presence or absence of vehicles with charging devices. In this embodiment, by using only data showing the power consumption of buildings classified into the same building group as the user's building as the first data, the user's presence or absence can be estimated more accurately based on the power consumption and calendar of building groups that share common characteristics with the user's building.

[0014] Furthermore, another embodiment of the present invention is a power control system for managing the amount of power supplied to at least one of a vehicle and equipment connected to a user's building, comprising a server device and a power control device communicated with the server device, wherein the server device has a storage unit that stores a learning model generated by machine learning using first data indicating the amount of power used in a plurality of buildings and second data indicating the presence or absence of people in the plurality of buildings as training data, and an estimation unit that estimates the presence or absence of the user by inputting third data indicating the amount of power used in the user's building to the learning model, and the power control device controls the amount of power supplied to at least one of the vehicle and equipment according to the estimation result of the estimation unit.

[0015] According to this embodiment of the invention, the power control device controls the amount of power supplied to at least one of the vehicle and equipment according to the estimation result of the estimation unit. Since at least one of charging the vehicle and supplying power to the equipment is controlled according to the estimation result of the estimation unit, for example, power generated by solar panels can be used efficiently within a building.

[0016] Furthermore, another embodiment of the present invention provides a method for estimating the presence or absence of a user, comprising the step of inputting third data indicating the power consumption of the user's building to a learning model generated by machine learning using first data indicating the power consumption in a plurality of buildings and second data indicating the presence or absence of people in the plurality of buildings as training data.

[0017] According to an embodiment of the present invention, a learning model is generated based on first data indicating power consumption in a plurality of buildings and second data indicating the presence or absence of people in the plurality of buildings. Then, by inputting the power consumption of a user's building into the learning model, the presence or absence of the user is estimated. As a result, when generating the learning model, it is not necessary to obtain the presence or absence of the user who is the target of presence or absence estimation. Therefore, even without information indicating the presence or absence of the user who is the target of presence or absence estimation, the presence or absence of the user can be estimated.

Advantages of the Invention

[0018] According to an embodiment of the present invention, in a presence or absence estimation system for estimating the presence or absence of a user, the presence or absence of the user can be estimated even without information indicating the presence or absence of the user.

Brief Description of the Drawings

[0019] [Figure 1] A block diagram showing the configuration of the power control system according to this embodiment. [Figure 2] A conceptual diagram of the first data used as teacher data for the learning model according to this embodiment. [Figure 3] A conceptual diagram of the second data used as teacher data for the learning model according to this embodiment. [Figure 4] A conceptual diagram of the third data input to the learning model according to this embodiment. [Figure 5] A conceptual diagram of the presence probability data estimated in the presence or absence estimation system according to this embodiment. [Figure 6] A flowchart showing the operation of the presence or absence estimation system according to this embodiment. [Figure 7] A flowchart showing the operation of the power control device according to this embodiment. [Figure 8] A diagram showing the experimental record of the presence or absence estimation system according to this embodiment.

Modes for Carrying Out the Invention

[0020] Embodiments of the present invention will be described in detail below with reference to the drawings. The following description of preferred embodiments is essentially illustrative and is not intended to limit the present invention, its applications, or its uses.

[0021] Figure 1 is a block diagram showing the configuration of the power control system according to this embodiment. As shown in Figure 1, the power control system according to this embodiment comprises a server device 1 and a power control device 21. The server device 1 is configured to communicate with the power control device 21 via a communication network N.

[0022] The presence / absence estimation system according to this embodiment is configured as part of a power control system. Specifically, the presence / absence estimation system 100 includes a server device 1.

[0023] The power control device 21 is for controlling the power to the user's building B1. Specifically, the power control device 21 is connected to the solar panels 22, a number of devices 23, and a vehicle 24 via power lines. The power control device 21 is also connected to the server device 1 via a communication network N. The solar panels 22 are solar panels used for solar power generation. The devices 23 are electrical equipment connected to building B1, and the amount of power supplied to them is controlled by the power control device 21. For example, each device 23 may be directly or indirectly connected to the power control device 21 by at least one wired and wireless communication means. The amount of power supplied to the devices 23 may be controlled by the power control device 21 via the communication means, such as switching the power input or increasing or decreasing the power consumption. Examples of devices 23 include those used especially when people are in the building, such as air conditioners and lighting fixtures; those used especially when people are not in the building, such as security devices such as surveillance cameras and cleaning robots; and those used regardless of whether people are present or absent, such as refrigerators.

[0024] Furthermore, vehicle 24 is a vehicle equipped with a charging device, such as an electric vehicle, hybrid vehicle, or fuel cell vehicle.

[0025] The power control device 21 controls the power supplied to each device connected to building B1. Specifically, it converts the DC power generated by the solar panels 22 into AC power and supplies power to the equipment 23 and vehicle 24. Furthermore, if the amount of power generated by the solar panels 22 is greater than the amount of power used by the equipment 23 and vehicle 24, the power control device 21 outputs the surplus power to the commercial power grid (not shown) and sells it. In addition, if the amount of power generated by the solar panels 22 is less than the amount of power used by the equipment 23 and vehicle 24, the power control device 21 receives the necessary power from the power grid.

[0026] Here, the power control device 21 may charge the vehicle 24 based on the occupancy probability data. As will be explained in more detail later, the occupancy probability data is data that indicates the probability that a user is in building B1 for each time period (in the following explanation, the probability of a person being in a building is referred to as the occupancy probability). For example, the power control device 21 performs a supply process in which it increases the amount of power supplied to the equipment 23 and decreases the amount of power supplied to the vehicle 24 during time periods when the occupancy probability is above a predetermined value. On the other hand, the power control device 21 performs a charging process in which it increases the amount of power supplied to the vehicle 24 and decreases the amount of power supplied to the equipment 23 during time periods when the occupancy probability is below a predetermined value. As an example of controlling the amount of power supplied to equipment 23, for example, the power control device 21 may increase the amount of power supplied to equipment 23 such as air conditioners and lighting fixtures during time periods when the occupancy probability is above a predetermined value, while decreasing the amount of power supplied to the equipment 23 during time periods when the occupancy probability is below a predetermined value. This function of the power control device 21 allows the power obtained from the solar panels 22 to be used efficiently within building B1.

[0027] The power control device 21 also includes a smart meter 211. The smart meter 211 is a power meter that measures the amount of electricity used in building B1. In this example, the smart meter 211 measures at least the amount of electricity generated by the solar panels 22 and the amount of electricity used by the equipment 23 and the vehicle 24 as the amount of electricity used in building B1. The smart meter 211 measures the amount of electricity used in building B1 at predetermined intervals (generally every 30 minutes) and transmits the measurement results of the amount of electricity used to the server device 1.

[0028] (Regarding the configuration of Server Device 1) As shown in Figure 1, the server device 1 comprises a communication unit 11, a storage unit 12, and a control unit 13.

[0029] The communication unit 11 is composed of, for example, an electrical circuit, and communicates with the power control system 2 via the communication network N.

[0030] The memory unit 12 is a storage medium composed of ROM (Read Only Memory), RAM (Random Access Memory), HDD (Hard Disk Drive), and SSD (Solid State Drive), etc. Various programs executed by the control unit 13 are stored in the memory unit 12.

[0031] Furthermore, the memory unit 12 stores the learning model 121. The learning model 121 is a learning model generated by machine learning using first data showing the amount of electricity used in multiple buildings and second data showing the presence or absence of people in multiple buildings (whether or not a person is in a building) as training data. The machine learning model or algorithm solves the classification problem and includes, for example, logistic regression (including ridge regularization and lasso regularization), random forest, support vector machine, neural network (including multilayer perceptron). In this embodiment, the learning model generated by machine learning may be one of the above-mentioned machine learning models or algorithms that is evaluated as having high accuracy by the evaluation of the accuracy of classification prediction. The evaluation of the accuracy of classification prediction may be performed by one or more of the following: the mean squared error of the estimated conditional probability, AUC (Area Under the Curve) quantified as the area under the ROC (Receiver Operating Characteristic) curve, logarithmic loss in the Kullback-Leibler divergence minimization problem, and a calibration plot that plots the observed value and the predicted value. Here, the accuracy of classification predictions by machine learning algorithms or models varies depending on the amount of training data and the features (for example, the number of days for which power consumption data used for training is acquired and the attribute information added). Therefore, the selection of a learning model may be made by comparing the evaluation results of the classification accuracy of each machine learning algorithm or model under predetermined conditions. Furthermore, when selecting a learning model, priority may be given to those with low computational costs required for training and classification.

[0032] Figure 2 is a conceptual diagram of the first data used as training data for the learning model according to this embodiment, and Figure 3 is a conceptual diagram of the second data used as training data for the learning model according to this embodiment.

[0033] As shown in Figures 2 and 3, the first data set includes data showing the electricity usage of buildings B2 to B4 every 30 minutes, and the second data set includes data showing the presence or absence of people in buildings B2 to B4 every 30 minutes. For example, the first data set is generated based on electricity usage data that is automatically transmitted at predetermined intervals (e.g., every 30 minutes) from smart meters installed in buildings B2 to B4. The second data set is generated based on the results of a presence or absence questionnaire conducted with people in buildings B2 to B4.

[0034] The learning model 121 is generated by machine learning using the first and second data as training data. Upon receiving input of the user's electricity usage in building B1, it outputs the probability of that user being present (details will be described later).

[0035] The control unit 13 is composed of a microcomputer, for example, a CPU (Central Processing Unit) and semiconductor memory. The control unit 13 controls each part of the server device 1 by executing programs stored in the storage unit 12.

[0036] Furthermore, the control unit 13 includes a power usage acquisition unit 131, a presence / absence estimation unit 132, and a presence / probability data distribution unit 133.

[0037] The power usage acquisition unit 131 acquires third data indicating the power usage of the user's building B1. Specifically, the power usage acquisition unit 131 acquires data indicating the power usage of building B1, which is transmitted every 30 minutes from the smart meter 211 installed in building B1, as third data.

[0038] The presence / absence estimation unit 132 inputs third data, which represents the electricity usage of building B1 and was acquired by the electricity usage acquisition unit 131, into the learning model 121 to estimate the probability of the user being present.

[0039] Figure 4 is a conceptual diagram of the third data input to the learning model according to this embodiment, and Figure 5 is a conceptual diagram of the presence probability data estimated in the power control system according to this embodiment.

[0040] For example, the presence / absence estimation unit 132 inputs data showing the electricity usage of building B1 during the first period (in this example, the period from 7 days ago (1 / 1) to the current day (1 / 8)) as third data to the learning model 121. The learning model 121 then outputs the user's presence probability for the second period (in this example, the current day (1 / 8) and the following day (1 / 9)). The presence / absence estimation unit 132 uses the user's presence probability output from the learning model 121 as the estimated result of the user's presence or absence.

[0041] The presence probability distribution unit 133 transmits (distributes) the estimation result of the presence / absence estimation unit 132 as presence probability data to the power control device 21. Based on the received presence probability data, the power control device 21 performs power control of building B1 (for example, charging the vehicle 24).

[0042] For example, the power control device 21 performs a supply process in which, during periods when the probability of presence is above a predetermined value (here, 50% or more), the amount of power supplied to the equipment 23 is increased and the amount of power supplied to the vehicle 24 is reduced. On the other hand, the power control device 21 performs a charging process in which, during periods when the probability of presence is below a predetermined value (here, less than 50%), the amount of power supplied to the vehicle 24 is increased and the amount of power supplied to the equipment 23 is reduced.

[0043] (Regarding the operation of the power control system) Figures 6 and 7 are flowcharts showing the operation of the power control system according to this embodiment. Specifically, Figure 6 shows the operation of the presence / absence estimation system 100 (server device 1), and Figure 7 shows the operation of the power control device 21.

[0044] First, let's explain the operation of server device 1.

[0045] The power usage acquisition unit 131 of the server device 1 acquires (receives) data from the smart meter 211 indicating the power usage of the user's building B1 (step S1).

[0046] The presence / absence estimation unit 132 inputs third data, which shows the amount of electricity used by building B1 during the first period, into the learning model 121 (step S2).

[0047] The learning model 121 outputs (calculates) the user's presence probability for the second period (step S3). The presence / absence estimation unit 132 uses the user's presence probability output from the learning model 121 as the estimated result of the user's presence probability.

[0048] The presence probability distribution unit 133 transmits (distributes) the estimation result of the presence / absence estimation unit 132 as presence probability data to the power control device 21.

[0049] Next, the operation of the power control device 21 will be explained.

[0050] When the power control device 21 receives presence probability data from the server device 1 (step S11), it detects the presence or absence of the user at the current time from the presence probability data. Then, the power control device 21 determines whether the user's presence probability at the current time is greater than or equal to a predetermined value (step S12).

[0051] If the probability of a user being present at the current time is above a predetermined level (Yes in step S12), the power control device 21 increases the amount of power supplied to the equipment 23 and reduces the amount of power supplied to the vehicle 24 (step S13). On the other hand, if the probability of a user being present at the current time is below a predetermined level (No in step S12), the power control device 21 increases the amount of power supplied to the vehicle 24 and reduces the amount of power supplied to the equipment 23 (step S14).

[0052] (Regarding the results of the estimated presence / absence) Figure 8 shows the experimental records of the occupancy estimation system according to this embodiment. In Figure 8, electricity usage and occupancy status were surveyed every 30 minutes for 880 households on a predetermined day on both weekdays and holidays. Of these, data from 616 households was used as training data for the learning model 121, and data from 264 households was used for testing the learning model 121. In Figure 8, "Number of Records" refers to the number of times electricity usage and occupancy status were recorded every 30 minutes.

[0053] As shown in Figure 8, the positive response rate for all records was 78.9% (0.789). This indicates that 78.9% of the total records used in this experiment were "present" and 21.1% were "absent". Although not shown in the figure, this experiment used logistic regression (ridge regularization), logistic regression (lasso regularization), random forest, support vector machine, and neural network as learning models 121. In this experiment, the five learning models 121 were compared using power consumption data for the current day and the past 7 days, as well as power consumption data for the current day and the past 14 days, as features. According to this experiment, the test results for each learning model 121 were as follows.

[0054] The mean squared error (for power consumption data for the current day and the past 7 days) was 0.145 for logistic regression (ridge regularization), 0.143 for logistic regression (Lasso regularization), 0.130 for random forest, 0.138 for support vector machine, and 0.137 for neural network. According to these results, the most suitable example of learning model 121 is the random forest, which had the smallest mean squared error. Furthermore, the mean squared error (for power consumption data for the current day and the past 14 days) was 0.140 for logistic regression (ridge regularization), 0.148 for logistic regression (Lasso regularization), 0.133 for random forest, 0.145 for support vector machine, and 0.133 for neural network. According to these results, the most suitable examples of learning model 121 are the random forest and neural network, which had the smallest mean squared error.

[0055] The AUC (for power consumption data for the current day and the past 7 days) was 0.770 for logistic regression (ridge regularization), 0.772 for logistic regression (lasso regularization), 0.802 for random forest, 0.786 for support vector machine, and 0.790 for neural network. According to these results, the most suitable example of learning model 121 is the random forest, which had the highest AUC. Furthermore, the AUC (for power consumption data for the current day and the past 14 days) was 0.749 for logistic regression (ridge regularization), 0.758 for logistic regression (lasso regularization), 0.773 for random forest, 0.760 for support vector machine, and 0.790 for neural network. According to these results, the most suitable example of learning model 121 is the neural network, which had the highest AUC.

[0056] The logarithmic loss (for power consumption data for the current day and the past 7 days) was 0.447 for logistic regression (ridge regularization), 0.449 for logistic regression (lasso regularization), 0.417 for random forest, 0.433 for support vector machine, and 0.431 for neural network. According to these results, the most suitable example of learning model 121 is the random forest, which had the smallest logarithmic loss. Furthermore, the logarithmic loss (for power consumption data for the current day and the past 14 days) was 0.446 for logistic regression (ridge regularization), 0.459 for logistic regression (lasso regularization), 0.442 for random forest, 0.459 for support vector machine, and 0.423 for neural network. According to these results, the most suitable example of learning model 121 is the neural network, which had the smallest logarithmic loss.

[0057] Based on the experimental results described above, a random forest or a neural network may be selected as a suitable example of the learning model 121 in this embodiment.

[0058] As described above, the power control system according to this embodiment includes an presence / absence estimation system 100 that estimates the presence or absence of a user. This power control system includes a storage unit 12 that stores a learning model 121 generated by machine learning using first data showing the amount of electricity used in multiple buildings B2 to B4 and second data showing the presence or absence of people in multiple buildings B2 to B4 as training data, and a presence / absence estimation unit 132 that estimates the presence or absence of a user by inputting third data showing the amount of electricity used in the user's building B1 to the learning model 121.

[0059] In this configuration, a learning model 121 is generated using first data showing the power consumption in multiple buildings B2-B4 and second data showing the presence or absence of people in multiple buildings B2-B4. Then, by inputting third data showing the power consumption in the user's building B1 into the learning model 121, the presence or absence of that user is estimated. As a result, when generating the learning model 121, the power consumption in multiple buildings including buildings B2-B4 other than the user's building B1, and the data showing the presence or absence of people in those buildings B2-B4, are used as training data. Therefore, it is not necessary to obtain data showing the presence or absence of the user whose presence or absence is to be estimated as training data. Consequently, even without information showing the presence or absence of the user whose presence or absence is to be estimated, the probability of the user being at home or not can be estimated.

[0060] Furthermore, since the learning model 121 is generated using first data showing the amount of electricity used in multiple buildings B2-B4 and second data showing the presence or absence of people in multiple buildings B2-B4, there is no need to collect other data, such as data showing the presence or absence of people in building B1. This also reduces the burden when generating the learning model.

[0061] Furthermore, the third data set is data showing the user's electricity usage in building B1 during the first period, which includes the current day (present time). The presence / absence estimation unit 132 estimates the user's presence or absence during the second period following the first period.

[0062] In this configuration, since the power consumption data for multiple buildings B2-B4 (excluding the user's building B1) and data indicating the presence or absence of people in those buildings are used as training data, the presence or absence of a user can be estimated without having to obtain data indicating the presence or absence of the user to be estimated as training data. Furthermore, the presence / absence estimation unit 132 estimates the presence or absence of a user in the second period following the first period by inputting third data, which indicates the power consumption of the user's building B1 in the first period, to the learning model 121. In other words, the presence or absence of a user is estimated not only for the period immediately following the current time, but also for the period after the current time (the second period), so the future presence or absence of a user can be estimated.

[0063] Furthermore, the server device 1 includes an occupancy / absence estimation unit 132. The server device 1 is communicatively connected to a power control system 2 that controls the charging status of a vehicle 24 connected to the user's building B1. The power control device 21 controls the amount of power supplied to the vehicle 24 according to the estimation result of the occupancy / absence estimation unit 132.

[0064] In this configuration, the power control device 21 controls the amount of power supplied to the vehicle 24 according to the estimation result of the presence / absence estimation unit 132. Because the charging status of the vehicle 24 is controlled according to the estimation result of the presence / absence estimation unit 132, for example, electricity generated by solar panels can be used efficiently within the building.

[0065] (Other embodiments) As described above, embodiments have been explained as examples of the technology disclosed in this application. However, the technology in this disclosure is not limited thereto and can be applied to embodiments that are modified, replaced, added, or omitted as appropriate.

[0066] In the above embodiment, as shown in Figures 2 and 3, the acquisition periods for the first and second data are the same, but they may be different. That is, the first data may be data showing the amount of electricity used in multiple buildings during the third period, and the second data may be data showing the presence or absence of people in multiple buildings during a fourth period different from the third period. This eliminates the need to match the acquisition periods of the first and second data, making it easier to generate the learning model. In this case, for example, a reference time (such as the same time on the same day of the week) can be determined, and the first and second data can be associated with the reference time in the first and second data to generate the learning model 121.

[0067] Furthermore, in the above embodiment, the first and second data are acquired at the same time intervals (every 30 minutes in Figure 2), but this is not limited to this. For example, the time at which the first and second data are acquired does not have to be every 30 minutes. Also, the first and second data may be acquired at different time intervals.

[0068] Furthermore, in the above embodiment, the third data input to the learning model 211 is data showing the electricity usage of the user's building B1 from 7 days prior to the current day, but it may also be data showing the electricity usage of the user's building B1 from more than 7 days prior to the current day.

[0069] Furthermore, the first and second data may be classified based on the calendar (month, day, day of the week, weekday, holiday, season, etc.) from which they were acquired, and the classified first and second data may be used as training data for the learning model 121. Generally, common patterns may emerge between buildings in the presence or absence of people in the building and changes in electricity consumption, depending on whether it is a holiday or a weekday, and the day of the week. For example, if the building is a residence, on weekdays, people are generally absent during the day, so the building's electricity consumption decreases during that time. Also, during a certain period from evening to night when people return to the building, the building's electricity consumption increases because people use equipment connected to the building. On the other hand, on holidays, the times when people are present or absent generally differ from those on weekdays, so electricity consumption changes in a different pattern than the pattern of changes in building electricity consumption on weekdays as described above. For this reason, by classifying (distinguishing) the first and second data based on the calendar, it is possible to more accurately estimate the presence or absence of users based on the pattern of changes in building electricity consumption according to the calendar. The calendar-based classification may be performed on the first data for all buildings, including the user's building B1. Alternatively, the multiple buildings may be pre-classified into one or more building groups based on one or more common characteristics such as the attributes and number of people in the buildings, the type of building, and the equipment connected to the buildings, and then the classification may be performed on the first data for each building group. Examples of such building group classifications include classifications based on building type such as residences, shops, apartment buildings, offices, factories, and warehouses, as well as further subdivisions for residences based on household size and the age group of the head of household, and the presence or absence of vehicles with charging devices. In this embodiment, by using only data showing the power consumption of buildings classified into the same building group as the user's building for which presence or absence is to be estimated as the first data, the user's presence or absence can be estimated more accurately based on the power consumption and calendar of building groups that share common characteristics with the user's building. Furthermore, the learning model 121 may be generated without classifying the first and second data based on the calendar. Alternatively, the first and second data, which have been pre-classified based on the calendar, may be used as training data for the learning model 121.

[0070] Furthermore, although the above embodiment was described as an example in which the power control device 21 controls the amount of power supplied to the vehicle 24 and the equipment 23 using presence probability data, the power control device 21 may also control the amount of power supplied to either the vehicle 24 or the equipment 23 using presence probability data.

[0071] In the above embodiment, the server device 1 constituting the presence / absence estimation system 100 was described as using the user's presence / absence status in cooperation with the power control device 21 to control the charging of the vehicle 24. However, the device (or system) that cooperates with the server device 1 is not limited to the power control device 21. For example, the server device 1 may be cooperated with a delivery management system. In this case, the delivery management system may determine the delivery route using the user's presence / absence status. The server device 1 may also be cooperated with a user monitoring system. In this case, the monitoring system may issue a warning when the user's presence / absence status is high or when the user's presence cannot be confirmed. The server device 1 may also be cooperated with an advertising distribution system. In this case, the advertising distribution system may determine which advertisements to deliver depending on the user's presence / absence status.

[0072] Furthermore, the server device 1 may be equipped with a function to sequentially receive the first and second data and update the learning model 121 based on the received data.

[0073] Alternatively, data indicating the power consumption and occupancy status of the user's building B1 may be used as training data for the learning model 121.

[0074] Furthermore, the power control system according to this embodiment may be linked with a credit card payment system or a bank transfer system. In this case, for example, payment information such as the user's credit card number or bank account number may be received via the user's communication terminal (such as a mobile phone or PC), and the payment information may be registered in the credit card payment system or bank transfer system. This allows the presence / absence estimation system to manage the user's payment information.

[0075] Furthermore, the server device 1 of the power control system according to this embodiment may be equipped with a function that can detect information related to a credit card from an image of the credit card. For example, the server device 1 may be configured to communicate with the user's communication terminal (such as a mobile phone or PC) and the credit card payment system. The server device 1 may receive an image of the user's credit card from the user's communication terminal, and from the received image, read information such as the user's credit card number using, for example, a character recognition (OCR) technology, and notify the credit card payment system of this information. This eliminates the need to manually input payment information.

[0076] Furthermore, the power control system according to this embodiment may be linked with a customer management system that manages various information used to bill users for their electricity charges, such as the user's name, address, telephone number, and electricity usage. This would allow, for example, the system to bill users for electricity charges based on their electricity usage.

[0077] Furthermore, server device 1 may consist of multiple devices or it may be a single device. One server device may be provided for each function of server device 1. Alternatively, multiple server devices may work together to realize each function of server device 1. [Industrial applicability]

[0078] The presence / absence estimation system according to this embodiment can estimate whether a user is present or absent. Therefore, by linking with a power control system that controls vehicle charging, for example, electricity can be used efficiently within the user's building. [Explanation of Symbols]

[0079] 100 Presence / Absence Estimation System 1 Server device 12 Storage section 121 Learning Models 13 Control Unit 131 Power usage acquisition section 132 Presence / Absence Estimation Department 133 Probability Distribution Department 21 Power control device 211 Smart Meter 24 vehicles

Claims

[Claim 1] A power control system that manages the amount of electricity supplied to at least one of the vehicles and equipment connected to the user's building, Server device, A power control device that is communicatively connected to the server device and Equipped with, The server device is A storage unit that stores a learning model generated by machine learning using first data showing the amount of electricity used in multiple buildings excluding the user's building, and second data showing the presence or absence of people in the multiple buildings excluding the user's building as training data. The learning model is provided with a presence / absence estimation unit that, by inputting only third data indicating the amount of electricity used in the user's building during a first period including the current time, outputs the probability of the user's presence during a second period following the first period as an estimated absence result. It has, The first and second data sets are classified based on the calendar in which each data set was acquired. The learning model is generated by using the classified first data and second data as training data. The power control device, in accordance with the estimation result of the presence / absence estimation unit, controls the amount of power supplied to the equipment and the amount of power supplied to the vehicle during the second period to increase the amount of power supplied to the equipment and decrease the amount of power supplied to the vehicle. Power control system.

Images