Estimation apparatus, estimation system, learning apparatus, method of generating trained model, estimation method, and program
The estimation device uses machine learning to predict consumers' power saving acceptance, enhancing the accuracy of demand suppression by analyzing appliance operation data, thereby improving power demand management.
Patent Information
- Application Number
- JP2023077755
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-10
- Publication Date
- 2024-11-21
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing systems struggle to accurately estimate consumers' acceptance of power saving measures, leading to inadequate suppression of electricity demand during peak periods.
An estimation device that utilizes machine learning to analyze operating information of home appliances, such as air conditioners, to generate a learned model that predicts consumers' power saving acceptance based on historical data, enabling targeted power saving requests.
Accurately estimates consumers' power saving acceptance, allowing for more effective demand suppression and reducing the risk of power outages by identifying consumers likely to cooperate with power saving measures.
Smart Images

Figure 2024162351000001_ABST
Abstract
Description
[Technical field]
[0001] The present disclosure relates to an estimation device, an estimation system, a learning device, a method for generating a trained model, an estimation method, and a program. [Background technology]
[0002] In recent years, there has been an increasing demand to control power demand according to the amount of power supply, due to various circumstances such as the liberalization of the power market, difficulties in restarting nuclear power plants, and the use of renewable energy for carbon neutrality. In relation to this, for example, when a system monitoring the power supply and demand balance requests power saving, cases are expected in which non-users will automatically suppress power demand by changing the temperature setting of air conditioners or turning them off.
[0003] For example, Patent Document 1 discloses an information processing device that automatically controls devices to save energy. The information processing device disclosed in Patent Document 1 changes the priority, which indicates the order in which devices are controlled, to a lower value in response to a user's operation to refuse control. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] JP 2014-215728 A Summary of the Invention [Problem to be solved by the invention]
[0005] When reducing power demand, if consumers are not receptive to requests to save power, it may be possible to reduce power demand less than planned, and power demand reduction may not be realized. Therefore, in order to more reliably reduce power demand, it is important to identify consumers who are likely to cooperate in reducing power demand. Under these circumstances, there is a demand for technology that can estimate consumers' receptivity to saving power with a high degree of accuracy.
[0006] The present disclosure has been made to solve the above-mentioned problems, and aims to provide an estimation device etc. capable of estimating a consumer's acceptance of power saving with high accuracy. [Means for solving the problem]
[0007] In order to achieve the above object, an estimation device according to the present disclosure includes: An estimation device for estimating a power saving acceptance level of a consumer, An acquisition means for acquiring operation information of a home appliance used by the consumer to be estimated; The system further includes an estimation means for estimating the power saving acceptance of the target consumer from the operation information acquired by the acquisition means, based on a trained model generated by machine learning using learning data in which operation information of a plurality of home appliances used by the target consumer is associated with the power saving acceptance of the target consumer. Effect of the Invention
[0008] In the present disclosure, the power saving acceptance level of a consumer to be estimated is estimated from the operation information of a home appliance used by the consumer to be estimated based on a trained model generated by machine learning using learning data in which the operation information of a plurality of home appliances used by the plurality of consumers to be learned is associated with the power saving acceptance levels of the plurality of consumers to be learned. Therefore, according to the present disclosure, the power saving acceptance level of a consumer can be estimated with high accuracy. [Brief description of the drawings]
[0009] [Figure 1] FIG. 1 is a diagram showing an overall configuration of an estimation system according to a first embodiment. [Diagram 2] FIG. 1 is a block diagram showing a hardware configuration of an estimation device according to a first embodiment. [Diagram 3] FIG. 1 is a block diagram showing a functional configuration of an estimation device according to a first embodiment. [Figure 4] FIG. 1 is a diagram showing an example of an operation history according to the first embodiment; [Diagram 5]FIG. 1 is a diagram showing the transition of driving information on a normal day and on a warning issuance day in the first embodiment; [Figure 6] FIG. 1 is a diagram for explaining the characteristics of consumers with a high acceptance rate for power saving and consumers with a low acceptance rate for power saving in the first embodiment. [Figure 7] FIG. 1 is a diagram showing an example in which power saving acceptance levels of multiple consumers are derived in the first embodiment. [Figure 8] FIG. 1 is a diagram showing a comparison example of driving information on a normal day and on a warning issuance day in the first embodiment. [Figure 9] FIG. 1 is a diagram showing an example of learning data according to the first embodiment; [Figure 10] FIG. 1 is a diagram showing an example of input and output data in a trained model according to the first embodiment; [Figure 11] FIG. 13 is a diagram showing a first example of a notification screen displayed on the operation terminal according to the first embodiment; [Figure 12] FIG. 11 is a diagram showing a second example of a notification screen displayed on the operation terminal according to the first embodiment; [Figure 13] A flowchart showing the flow of a learning phase process according to the first embodiment. [Figure 14] A flowchart showing a flow of utilization phase processing according to the first embodiment. [Figure 15] FIG. 13 is a diagram showing an example of a notification screen displayed on an operation terminal according to the second embodiment; [Figure 16] FIG. 13 is a diagram showing an example of a questionnaire screen displayed on an operation terminal according to the third embodiment; [Figure 17] FIG. 13 is a diagram showing an example of a selection screen for a contract plan regarding power supply displayed on an operation terminal according to the third embodiment; [Figure 18] FIG. 13 is a block diagram showing a functional configuration of an estimation device according to a fourth embodiment. [Figure 19] FIG. 13 is a block diagram showing a configuration of an estimation system according to a fifth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0010] Hereinafter, embodiments will be described in detail with reference to the drawings, in which the same or corresponding parts are designated by the same reference numerals.
[0011] (Embodiment 1) 1 shows the overall configuration of an estimation system 1 according to embodiment 1. The estimation system 1 includes a plurality of air conditioners 3, a plurality of operation terminals 5, and an estimation device 10.
[0012] The estimation system 1 is a system that estimates the power saving acceptance level of a consumer by using a learning result by machine learning. Here, the power saving acceptance level is a value indicating the degree of consciousness of the consumer regarding power saving. Specifically, the power saving acceptance level is a value indicating the degree to which the consumer is willing to accept suppression of power consumption of home appliances used by the consumer when power saving is requested. A power saving request is issued by the government, a local government, etc. when the power supply and demand is tight or when the power supply and demand is expected to be tight. A power saving request is issued, for example, on a daily or time-zone basis by the day before the target day.
[0013] Each of the multiple air conditioners 3 is equipment that conditions an indoor space that is the target of air conditioning. Each air conditioner 3 is, for example, a room air conditioner, a packaged air conditioner, etc. The indoor space is, for example, a room in a house, an office, a store, a factory, etc. Each air conditioner 3 is an example of a home appliance that is the target of control in the estimation system 1.
[0014] In the example of Fig. 1, multiple air conditioners 3 are installed in the homes of consumer A, consumer B, .... Note that, although Fig. 1 shows an example in which one air conditioner 3 is installed in one consumer's home, multiple air conditioners 3 may be installed in one consumer's home. Here, each consumer is a party that demands power supplied from a commercial power source, and corresponds to a user that receives the supply of power. Each consumer uses home appliances using power supplied from the commercial power source.
[0015] Each air conditioner 3 is, for example, a heat pump type air conditioning facility that uses carbon dioxide, HFC (hydrofluorocarbon), etc. as a refrigerant. Although not shown in the figure, each air conditioner 3 includes an outdoor unit installed outdoors and an indoor unit installed indoors.
[0016] The outdoor unit and the indoor unit are connected by refrigerant piping through which a refrigerant flows. The outdoor unit includes a compressor that compresses the refrigerant and circulates it through the refrigerant piping, a four-way valve that switches the refrigerant circulation direction, an outdoor heat exchanger that exchanges heat between outdoor air and the refrigerant, an expansion valve that reduces the pressure of the refrigerant and expands it, and an outdoor fan that sends outdoor air to the outdoor heat exchanger. The indoor unit includes an indoor heat exchanger that exchanges heat between air in an indoor space and the refrigerant, and an indoor fan that sends the air after heat exchange to the indoor space. The compressor, the four-way valve, the outdoor heat exchanger, the expansion valve, and the indoor heat exchanger are connected in a ring shape by refrigerant piping. This forms a refrigeration cycle.
[0017] Each air conditioner 3 repeatedly operates and stops the compressor so that the temperature difference between the room temperature and the set temperature falls within an allowable range. For example, during cooling, the air conditioner 3 operates the compressor when the room temperature rises above the set temperature, and stops the compressor when the room temperature falls below the set temperature. Conversely, during heating, the air conditioner 3 operates the compressor when the room temperature falls below the set temperature, and stops the compressor when the room temperature rises above the set temperature.
[0018] Each of the multiple operation terminals 5 is a communication terminal such as a smartphone or tablet terminal, and functions as a user interface for the air conditioner 3. Consumers operate the operation terminal 5 to input operation instructions for the air conditioner 3. Specifically, the operation instructions include instructions to switch between operation and stop, instructions to switch operation modes, instructions to change the set temperature, set humidity, air volume or air direction, etc. Each operation terminal 5 communicates with the air conditioner 3 to be operated via wireless communication or wired communication, and transmits the input operation instructions to the air conditioner 3. The air conditioner 3 operates according to the operation instructions input to the operation terminal 5.
[0019] Although not shown, each air conditioner 3 and each operation terminal 5 includes a CPU (Central Processing Unit), ROM (Read Only Memory), RAM (Random Access Memory), a communication interface, and a readable / writable non-volatile semiconductor memory. In each device, the CPU uses the RAM as a working memory and executes a control program stored in the ROM, thereby controlling the operation of each device.
[0020] Each air conditioner 3 also includes various sensors, none of which are shown in the figures, such as a room temperature sensor that measures the room temperature in the indoor space, an outside air temperature sensor that measures the outside air temperature, etc. For example, the room temperature sensor is installed in the indoor unit, and the outside air temperature sensor is installed in the outdoor unit.
[0021] The estimation device 10 is a device that controls and manages multiple air conditioners 3. As an example, the estimation device 10 is a server that provides cloud computing resources. The estimation device 10 communicates with each of the multiple air conditioners 3 and multiple operation terminals 5 via a wide area communication network, and remotely manages each air conditioner 3.
[0022] As shown in FIG. 2, the estimation device 10 includes a control unit 11, a storage unit 12, and a communication unit 13.
[0023] The control unit 11 includes a CPU, a ROM, and a RAM. The CPU is also called a central processing unit, a processor, a microprocessor, a microcomputer, or the like, and functions as a central processing unit that executes processing and calculations related to the control of the estimation device 10. In the control unit 11, the CPU reads out programs and data stored in the ROM, and uses the RAM as a work area to perform overall control of the estimation device 10.
[0024] The storage unit 12 includes a non-volatile semiconductor memory such as a flash memory, an EPROM (Erasable Programmable ROM), or an EEPROM (Electrically Erasable Programmable ROM), and serves as a so-called secondary storage device or auxiliary storage device. The storage unit 12 stores programs and data used by the control unit 11 to perform various processes. The storage unit 12 also stores data generated or acquired by the control unit 11 performing various processes.
[0025] For example, the storage unit 12 stores driving history data 121, learning data 123, and a learned model 125. Details of each of these data will be described later.
[0026] The communication unit 13 includes a communication interface for the estimation device 10 to communicate with an external device. The communication unit 13 communicates with each of the multiple air conditioners 3 and the multiple operation terminals 5 via a communication network, which is a wide area communication network. The protocol used for communication is not particularly limited, but a protocol with a certain degree of versatility, such as the ECHONET Lite (registered trademark) standard, can be adopted.
[0027] Although not shown in the drawings, the estimation device 10 may include a user interface such as an operation unit that accepts user operations and a display unit that displays various images.
[0028] The control unit 11 functionally includes a learning phase processing unit 110 and an utilization phase processing unit 150. Fig. 3 shows details of the functions included in the learning phase processing unit 110 and the utilization phase processing unit 150. The learning phase processing unit 110 includes a learning acquisition unit 111 which is an example of a learning acquisition means, a learning data generation unit 113 which is an example of a learning data generation means, and a learning processing unit 115 which is an example of a learning processing means. The utilization phase processing unit 150 includes an estimation acquisition unit 151 which is an example of an estimation acquisition means, an estimation unit 153 which is an example of an estimation means, an alarm unit 155 which is an example of an alarm means, and an equipment control unit 157 which is an example of an equipment control means.
[0029] When there is no need to distinguish between the learning acquisition unit 111 and the estimation acquisition unit 151, they are collectively referred to as an “acquisition unit.” The acquisition unit is an example of an acquisition means.
[0030] Each of these functions is realized by software, firmware, or a combination of software and firmware. The software and firmware are written as programs and stored in the ROM or storage unit 12. The CPU then executes the programs stored in the ROM or storage unit 12 to realize each of these functions.
[0031] <<Learning Phase>> First, the learning phase will be described. The learning phase is a phase in which machine learning is performed using learning data 123 to generate a trained model 125. The processing of the learning phase is executed by the learning phase processing unit 110.
[0032] In the learning phase processing unit 110, the learning acquisition unit 111 acquires operation information of multiple air conditioners 3 used by multiple consumers to be learned. The operation information of the air conditioner 3 is information related to the air conditioning operation of the air conditioner 3. As an example, the operation information of the air conditioner 3 is information indicating the operation time of the air conditioner 3, the set temperature of the air conditioner 3, and the room temperature when the air conditioner 3 is operated.
[0033] Here, the operating time of the air conditioner 3 means the length of time that the air conditioner 3 is operating for air conditioning. The set temperature is the target temperature for air conditioning. The set temperature can be set by the consumer operating the operation terminal 5 or a remote control (not shown). The room temperature is the air temperature of the indoor space that is to be conditioned by the air conditioner 3. The room temperature is measured, for example, by a temperature sensor installed in the indoor unit.
[0034] The room temperature when the air conditioner 3 is operated is the room temperature when the consumer operates the air conditioner 3, for example, by stopping operation, starting operation, changing the set temperature, etc. In the following, the room temperature when the air conditioner 3 is operated is referred to as the "room temperature during operation", and an explanation will be given using the room temperature when the consumer stops operation of the air conditioner 3 as an example.
[0035] In order to acquire operation information of the air conditioners 3, the learning acquisition unit 111 repeatedly acquires the operation state from each of the multiple air conditioners 3 and generates operation history data 121. The operation history data 121 is data that indicates the operation state of each of the multiple air conditioners 3 in chronological order.
[0036] Specifically, the learning acquisition unit 111 communicates with each of the multiple air conditioners 3 via the communication unit 13, and transmits an acquisition request for the operating state to each air conditioner 3. When each air conditioner 3 receives an acquisition request from the estimation device 10, it transmits information on the operating state at that time to the estimation device 10 in response to the received acquisition request.
[0037] More specifically, each air conditioner 3 transmits information on its operating state, such as its operation ON / OFF, operation mode, set temperature, and room temperature at the time when the acquisition request is received, to the estimation device 10. Operation ON / OFF indicates whether the air conditioner 3 is performing air conditioning operation or is stopped. Operation mode is an air conditioning mode that can be performed by the air conditioner 3, such as cooling, heating, dehumidification, humidification, etc.
[0038] The learning acquisition unit 111 transmits an acquisition request to each air conditioner 3 at predetermined time intervals, and repeatedly acquires the operating state of each air conditioner 3. In this way, the learning acquisition unit 111 acquires time-series data of the operating state of each air conditioner 3.
[0039] When the learning acquisition unit 111 acquires the operating state from each air conditioner 3, the learning acquisition unit 111 stores the acquired operating state data in the storage unit 12 in association with the acquisition date and time. Specifically, the learning acquisition unit 111 stores the acquired operating state data in the operation history data 121 stored in the storage unit 12.
[0040] Fig. 4 shows, as an example, operation history data 121 of the air conditioner 3 used by consumer A. The operation history data 121 stores time, operation ON / OFF, operation mode, set temperature, and room temperature in association with one another. In the operation history data 121, the time is the time when the learning acquisition unit 111 acquired the operation state from the air conditioner 3. The example of Fig. 4 shows a case where the learning acquisition unit 111 acquires the operation state from the air conditioner 3 every 5 minutes, which is a predetermined time interval.
[0041] More specifically, the operation history data 121 stores, for each of "normal days" and "alarm days," the time, operation ON / OFF, operation mode, set temperature, and room temperature in association with each other. Here, an "alarm day" is a day on which an alert for tight power supply and demand was issued, and corresponds to a period during which power saving was requested. In contrast, a "normal day" is a day on which no alert was issued, and corresponds to a period during which power saving was not requested.
[0042] The storage unit 12 stores such operation history data 121 for each of the multiple air conditioners 3 that can communicate via the communication unit 13, on an air conditioner 3 basis or on a consumer basis. When the learning acquisition unit 111 acquires an operating state from an air conditioner 3, it updates the operation history data 121 by adding the acquired data on the operating state to the operation history data 121 of that air conditioner 3. The operation history data 121 is read out as necessary and subjected to calculation processing within the estimation device 10.
[0043] The learning acquisition unit 111 acquires operation information of each air conditioner 3 used by multiple consumers to be learned, based on such operation history data 121. Here, the multiple consumers to be learned are multiple consumers that are targets for generating the learning data 123, among consumers A, B, ... shown in Fig. 1. More specifically, the multiple consumers to be learned are multiple consumers that have experienced both "alarm issued days", which are periods when power saving was requested, and "normal days", which are periods when power saving was not requested, in the past.
[0044] The learning acquisition unit 111 refers to the operation history data 121 and acquires information on the operation time, the set temperature, and the room temperature at the time of operation in units of a predetermined time length as operation information for each air conditioner 3. The predetermined time length may be any length of time, but the following description will be given taking one day as an example.
[0045] Specifically, the learning acquisition unit 111 derives, as the operation time, the length of time that the operation ON / OFF in the operation history data 121 is ON on a daily basis. Furthermore, the learning acquisition unit 111 derives, as the set temperature, the average value of the set temperature stored in the operation history data 121 on a daily basis. Note that a representative value may be used instead of the average value. Furthermore, the learning acquisition unit 111 derives, as the room temperature at the time of operation, the room temperature at the time when the operation ON / OFF in the operation history data 121 changes from ON to OFF on a daily basis.
[0046] Returning to Fig. 3, the learning data generation unit 113 generates learning data 123 based on the operation information acquired by the learning acquisition unit 111. Here, the learning data 123 is also called teacher data, supervised data, training data, etc., and is data for generating a trained model 125. The learning data 123 is used for the purpose of learning the relationship between the operation information of the air conditioner 3 and the power saving acceptance level of the consumer of the air conditioner 3. Specifically, the learning data 123 is data in which the operation information of multiple air conditioners 3 used by multiple consumers to be learned is associated with the power saving acceptance levels of the multiple consumers.
[0047] To generate the learning data 123, information on the power saving acceptance levels of multiple consumers to be learned is required as a target variable of supervised learning. To this end, the learning data generating unit 113 derives the power saving acceptance levels of each of the multiple consumers to be learned, based on the operation history data 121 stored in the storage unit 12.
[0048] Specifically, the learning data generation unit 113 derives the power saving acceptance levels of multiple consumers to be learned, based on the operation information of the multiple air conditioners 3 during periods when power saving is requested and the operation information of the multiple air conditioners 3 during periods when power saving is not requested, which are included in the operation history data 121. Here, the period when power saving is requested refers to days or time periods when power saving is requested, and corresponds to, for example, "alert issued days." In contrast, the period when power saving is not requested refers to days or time periods when power saving is not requested, and corresponds to, for example, "normal days."
[0049] The learning data generation unit 113 compares the operation information for one air conditioner 3 during a period when power saving is requested and a period when power saving is not requested, and derives the power saving acceptance level of the consumer of that air conditioner 3 based on the difference between the compared operation information. The learning data generation unit 113 executes such derivation processing for each of the multiple consumers to be learned, thereby deriving the power saving acceptance level of each of the multiple consumers to be learned.
[0050] Fig. 5 shows an example of a graph depicting operation history data 121 of the same air conditioner 3 for the same time period for each of "normal days" and "alarm issued days" in winter. Fig. 5 shows the trends from 0 to 24 hours for operation ON / OFF, set temperature, and room temperature contained in the operation history data 121 for each of "normal days" and "alarm issued days." As it is winter, the operation mode is heating.
[0051] 5, focusing on the timing when the operation ON / OFF switch changed from ON to OFF, on a normal day, the room temperature when the consumer of this air conditioner 3 turned off the heating operation was 26.0°C. On the other hand, on the day an alert was issued, the room temperature when the consumer turned off the heating operation was 23.0°C. In other words, it can be seen that the consumer of this air conditioner 3 turned off the heating operation at a lower room temperature on the day an alert was issued than on a normal day.
[0052] In winter, warning days are days when the weather is very cold and the amount of electricity consumed for heating is expected to increase significantly, resulting in a tight electricity supply and demand balance. Nevertheless, consumers who turn off heating on warning days when the room temperature is lower than on normal days can be assumed to have a strong desire to save electricity in response to the warning. Therefore, it can be assumed that the consumers in the example of Figure 5 are highly receptive to requests to save electricity, i.e., have a high receptiveness to saving electricity.
[0053] In addition to the room temperature when heating operation is OFF, on normal days, the set temperature when heating operation is ON is 24.0°C. On the other hand, the consumer in Figure 5 changes the set temperature to 22.5°C on the day the warning is issued. Therefore, it is considered that the consumer in Figure 5 is even more likely to have a high acceptance of power saving.
[0054] In this way, the learning data generation unit 113 derives the power saving acceptance of the consumer from the operation history data 121 and sets it as the objective variable. At this time, by overlapping the results derived from a plurality of explanatory variables, it is possible to derive the power saving acceptance with higher accuracy. For example, in addition to the room temperature when the operation is OFF and the set temperature when the operation is ON, the room temperature when the set temperature is changed, whether the set temperature was increased or decreased, the length of time the air conditioner 3 is ON, etc. may be used as explanatory variables.
[0055] In the example of Figure 5, the explanation was given for the winter season when heating is expected to be used. However, in the summer season when air conditioning is expected to be used, the explanation can be given in the same way by reversing the high and low temperatures from the winter season.
[0056] Figure 6 summarizes examples of characteristics of consumers with high and low power saving acceptance in summer and winter, respectively. Consumers on the day an alert is issued will exhibit the characteristics shown in Figure 6 according to their power saving acceptance with regard to operation time, room temperature when operation is off, and set temperature. In summary, if a consumer uses the air conditioner 3 in a way that makes them uncomfortable and unbearable on the day an alert is issued, it can be said that the consumer is highly conscious about saving electricity and is highly likely to have a high power saving acceptance.
[0057] The learning data generation unit 113 executes a process of deriving such a power saving acceptance level of a consumer for each of a plurality of consumers to be learned. Fig. 7 shows an example of a case where the power saving acceptance level of each of a plurality of consumers to be learned is derived. In Fig. 7, "ID" is identification information of the air conditioner 3 and the consumer. The ID uniquely identifies each consumer to be learned, and information that does not overlap with other consumers is used.
[0058] 7 shows an example of winter when heating is expected to be used. However, since the usage of the air conditioner 3 varies greatly depending on the season, for example, when deriving the power saving acceptance level in summer, it is necessary to generate learning data 123 using summer data separately from winter data.
[0059] In Fig. 7, in order to derive the power saving acceptance degree with higher accuracy, the learning data generating unit 113 uses three variables, namely, "operation time / day", "set temperature", and "room temperature at time of operation". Here, "operation time / day" means the operation ON time per day. "Room temperature at time of operation" means the room temperature when the consumer performs the operation to turn off the operation.
[0060] The values of these three variables in Figure 7 indicate the average values on normal days and on days when a warning is issued. If the variables of a specific air conditioner 3 indicate extreme values, the median may be used instead of the average value. Also, taking into account the influence of the day of the week, values on the same day of the week for normal days and days when a warning is issued may be tallied.
[0061] If all three variables indicate a trend toward reducing power consumption on warning days compared to normal days, the consumer can be judged to have a high acceptance of power saving.To make it easier to understand, Figure 8 shows the results of plotting each variable on normal days and warning days for the consumers with IDs "ABCD1234" and "CDEF7654" in Figure 7.
[0062] As shown in Figure 8, for the customer with ID "CDEF7654," the operating time is shorter, the set temperature is lower, and the room temperature during operation is lower on the day when the warning was issued than on a normal day. Therefore, this customer acts to reduce power consumption on the day when the warning was issued, and it can be inferred that the customer's acceptance of power saving is high. Reflecting this result, the customer with ID "CDEF7654" in Figure 7 is marked with a "○" for the acceptance of power saving.
[0063] On the other hand, for the customer with ID "ABCD1234," the operating time is longer, the set temperature is higher, and the room temperature during operation is higher on the day the warning is issued than on a normal day. Therefore, it can be inferred that this customer does not take any action to reduce power consumption on the day the warning is issued, and that the power saving acceptance is not high. Reflecting this result, the power saving acceptance of the customer with ID "ABCD1234" in Figure 7 is marked with an "X."
[0064] When the power saving acceptance level of each consumer to be learned is derived in this manner, the learning data generation unit 113 generates learning data 123 in which the operation information of the multiple air conditioners 3 used by the consumer to be learned during the "normal day" during which power saving is requested is associated with the derived power saving acceptance level. For example, the learning data generation unit 113 generates the learning data 123 shown in FIG. 9.
[0065] 9, the learning data 123 has a plurality of data sets, each of which is a combination of "operating hours / day", "set temperature", "room temperature at time of operation", and "power saving acceptance" on a daily basis for each consumer to be learned. In this manner, the learning data generation unit 113 generates learning data 123 using the operation information acquired by the learning acquisition unit 111, that is, "operating hours / day", "set temperature", and "room temperature at time of operation", as explanatory variables, and using the derived "power saving acceptance" as a target variable.
[0066] 3, the learning processing unit 115 generates a trained model 125 based on the learning data 123 generated by the learning data generation unit 113. The trained model 125 is a model for estimating the power saving acceptance level of a consumer of an air conditioner 3 from the operation information of the air conditioner 3.
[0067] More specifically, the trained model 125 receives operation information of the air conditioner 3 used by the consumer to be estimated, and outputs an estimate of the consumer's power saving acceptance. The input of the trained model 125 is three variables, namely, "operation time," "set temperature," and "room temperature when operation is OFF," which are operation information of the air conditioner 3, and the output of the trained model 125 is a binary value indicating whether the power saving acceptance is high or low. Note that the "room temperature when operation is OFF" is an example of the "room temperature during operation."
[0068] The learning processing unit 115 performs machine learning using the learning data 123 generated by the learning data generation unit 113 to generate a trained model 125. The learning processing unit 115 performs machine learning using an appropriate supervised learning method. Examples of the supervised learning method include a support vector machine (SVM) and a neural network. However, the present invention is not limited to these, and other learning algorithms can be used as long as they are supervised learning methods capable of binary discrimination.
[0069] Using such a supervised learning technique, the learning processing unit 115 learns the relationship between the operation information and the power saving acceptance level that are associated with each other in the learning data 123 generated by the learning data generating unit 113. Specifically, when three variables of "operating time", "set temperature", and "room temperature when operation is OFF" that are input as operation information, the learning processing unit 115 learns the relationship between these three variables and whether the power saving acceptance level corresponding to these three variables is "high" or "low".
[0070] More specifically, the learning processing unit 115 learns the relationship between the operation information and the power saving acceptance separately for each of cooling and heating. Specifically, as shown in Fig. 10, the learned model 125 receives input of the operation time and set temperature of the air conditioner 3 during cooling and heating, and the room temperature when the operation is OFF, and outputs the power saving acceptance during each of cooling and heating.
[0071] As described above, the learning phase processing unit 110 learns the relationship between the operation information of the air conditioner 3 used by multiple consumers to be learned and the power saving acceptance of those multiple consumers, thereby generating a learned model 125 for estimating the power saving acceptance of the consumers from the operation information of the air conditioner 3.
[0072] <<Utilization Phase>> Next, returning to Fig. 3, the utilization phase will be described. The utilization phase is a phase in which the trained model 125 generated in the learning phase is utilized. The processing of the utilization phase is executed by the utilization phase processing unit 150. For example, when a new power saving request is made, the processing of the utilization phase is executed on the day before the target date of the power saving request.
[0073] In the utilization phase processing unit 150, the estimation acquisition unit 151 acquires operation information of the air conditioner 3 used by a consumer that is the subject of estimation. Here, the consumer that is the subject of estimation is a consumer that is the subject of estimation of the power-saving acceptance degree for a new power-saving request, among consumers A, B, ... shown in FIG. 1 .
[0074] The consumer to be estimated is, for example, a new consumer who has not yet experienced a period during which power saving was requested. However, the consumer to be estimated may be a consumer who has experienced a period during which power saving was requested in the past, in other words, any one of the multiple consumers to be learned. Even for a consumer whose power saving acceptance level has been determined in the past by the learning data generating unit 113, the power saving acceptance level may have changed over time, and therefore the consumer may be a target for estimating the power saving acceptance level.
[0075] The acquisition unit for estimation 151 acquires operation information of the air conditioners 3 used by the consumer that is the subject of estimation, by referring to the operation history data 121 stored in the memory unit 12. The operation history data 121 stores time series data of the operation states of the multiple air conditioners 3 acquired by the acquisition unit for learning 111. The acquisition unit for estimation 151 acquires time series data of the operation states of the air conditioners 3 used by the consumer that is the subject of estimation from the operation history data 121.
[0076] More specifically, the estimation acquisition unit 151 acquires operation information of home appliances used by the consumer to be estimated during cooling and heating during a period when power saving is not requested. Specifically, the estimation acquisition unit 151 acquires information on operation time, set temperature, and room temperature when operation is stopped in predetermined time units by referring to data on a "normal day" of the consumer to be estimated from the operation history data 121. The method of acquiring each piece of information is the same as the method of acquiring operation information by the learning acquisition unit 111 described above.
[0077] The estimation acquisition unit 151 may acquire the operating information by directly communicating with the air conditioner 3 used by the consumer that is the subject of estimation via the communication unit 13, without referring to the operating history data 121.
[0078] The estimation unit 153 estimates the power saving acceptance level of the consumer to be estimated from the operation information acquired by the estimation acquisition unit 151, based on the trained model 125 generated by the learning processing unit 115. To be more specific, the estimation unit 153 inputs the operation information acquired by the estimation acquisition unit 151 to the trained model 125.
[0079] As described above, the trained model 125 outputs an estimated value of the power saving acceptance level in two values, high / low, in response to the input of the operation information. The estimation unit 153 estimates the estimated value output from the trained model 125 in response to the input of the operation information as the power saving acceptance level of the consumer to be estimated.
[0080] 10, trained model 125 outputs the cooling power-saving acceptance degree and the heating power-saving acceptance degree as two values for each of the input six explanatory variables. Estimation unit 153 estimates the cooling power-saving acceptance degree and the heating power-saving acceptance degree output from trained model 125 as the power-saving acceptance degree of the consumer during cooling and heating, respectively.
[0081] As a result, it becomes possible to estimate whether the consumer's acceptance of power saving is high or low. The estimation result by the estimation unit 153 contributes to extracting consumers who are likely to accept a power saving request, and in turn to increasing the accuracy of estimating the actual amount of power saving during the period when power saving is requested. In addition, the estimation result by the estimation unit 153 can also be used when a retail electricity supplier, an aggregator, or the like notifies a consumer of an action recommended by the retail electricity supplier, an aggregator, or the like.
[0082] Returning to Fig. 3, the notification unit 155 notifies, based on the power saving acceptance level estimated by the estimation unit 153, of proposal information related to the operation of at least one appliance used by the consumer that is the estimation target. Here, the at least one appliance used by the consumer is, for example, the air conditioner 3, but is not limited to the air conditioner 3 and may be another home appliance used by the consumer. Examples of other home appliances include a lighting device, a water heater, etc.
[0083] The notification unit 155 communicates with the operation terminal 5 of the consumer via the communication unit 13, and transmits the proposal information to the operation terminal 5. Then, the notification unit 155 displays an image indicating the transmitted proposal information on the display unit of the operation terminal 5. In this way, the notification unit 155 notifies the consumer of the proposal information that encourages the consumer to operate a device that leads to power saving.
[0084] As an example, the notification unit 155 notifies the operation terminal 5 of a consumer estimated to have a low power saving acceptance level of the proposal information shown in Fig. 11 on a normal day, which is a period during which power saving is not requested. The proposal information shown in Fig. 11 proposes recommended air conditioning settings for power saving to the consumer with a low power saving acceptance level.
[0085] The notification unit 155 notifies such suggested information to consumers with low power saving acceptance, including for the purpose of enlightenment, regardless of whether the power supply and demand are tight. By notifying such suggested information, it is possible to eliminate inappropriate operating states of devices on normal days, such as excessively high or low temperature settings. This leads to minimizing the occurrence of situations in which the power supply and demand are tight.
[0086] 3, the appliance control unit 157 executes control to suppress power consumption of at least one appliance used by the consumer that is the estimation target, based on the power saving acceptance degree estimated by the estimation unit 153. As described above, the at least one appliance used by the consumer is the air conditioner 3, the lighting device, the water heater, etc.
[0087] The appliance control unit 157 transmits a control instruction to at least one appliance used by a consumer having a high power saving acceptance estimated by the estimation unit 153, via the communication unit 13. Then, the appliance control unit 157 controls at least one appliance in a direction to reduce power consumption. Here, examples of control to reduce power consumption include lowering the set temperature for heating or raising the set temperature for cooling when the appliance is an air conditioner 3, and lowering the indoor illuminance when the appliance is a lighting device. Moreover, power consumption may be suppressed by stopping the operation of the appliance.
[0088] The appliance control unit 157 performs such control to suppress power consumption during a period for which power saving is requested. At that time, in the case of high urgency, for example, when there is a risk of a power outage if the control is not performed, or when the consumer has given prior consent to the control, the appliance control unit 157 performs the control automatically without confirming the consumer's intention. In this case, the notification unit 155 displays notification information, for example, as shown in Fig. 12, on the operation terminal 5 of the consumer of the appliance for which the control was performed. In this way, the notification unit 155 notifies the consumer that the automatic control has been performed.
[0089] In such a case, if automatic control is performed randomly, some consumers may return to the original settings afterwards. If the settings are returned to the original settings, it will not be possible to suppress consumption demand. However, by giving priority to automatic control to consumers who are highly receptive to power saving, or by providing incentives for automatic control, it is possible to increase the likelihood of suppressing power consumption.
[0090] Next, the flow of processing executed by the estimation device 10 will be described with reference to the flowcharts shown in FIGS.
[0091] First, Fig. 13 shows a flow of the learning phase process executed by the learning phase processing unit 110. The learning phase processing unit 110 repeatedly executes the learning phase process at a predetermined time interval, such as every day, every week, every month, or every year. Alternatively, since the learning data 123 is updated taking into account the response of the consumer when a power saving request is made, the learning phase processing unit 110 may execute the learning phase process after a period in which power saving is requested. The learning phase process shown in Fig. 13 is an example of a method for generating a trained model 125.
[0092] When the learning phase process starts, the learning phase processing unit 110 functions as a learning acquisition unit 111 and acquires operation information of multiple air conditioners 3 used by multiple consumers to be learned (step S11). Specifically, the learning phase processing unit 110 periodically acquires operation states from the multiple air conditioners 3 used by the multiple consumers to be learned and stores the acquired operation status in operation history data 121. Then, the learning phase processing unit 110 acquires the operation time, set temperature, and room temperature during operation from the operation history data 121 as operation information of each air conditioner 3 on normal days and on alert issuance days.
[0093] Upon acquiring the operation information, the learning phase processing unit 110 functions as the learning data generating unit 113 and generates the learning data 123 (step S12). Specifically, the learning phase processing unit 110 derives a power saving acceptance level for each of the multiple consumers to be learned, based on the difference between the operation information on the normal day and the warning issuance day acquired in step S11. Then, the learning phase processing unit 110 generates the learning data 123 by associating the derived power saving acceptance level with the operation information on the normal day.
[0094] When the learning data 123 is generated, the learning phase processing unit 110 functions as the learning processing unit 115 and executes the learning process (step S13). Specifically, the learning phase processing unit 110 executes machine learning to learn the relationship between the operation information of the air conditioner 3 and the power saving acceptance level of the consumer of that air conditioner 3. In this way, the learning phase processing unit 110 generates a trained model 125 that outputs the power saving acceptance level in response to the input of the operation information. With this, the learning phase processing shown in FIG. 13 ends.
[0095] Secondly, Fig. 14 shows a flow of utilization phase processing executed by the utilization phase processing unit 150. The utilization phase processing unit 150 executes the utilization phase processing at a timing before the target day of the power saving request, for example, in the afternoon or evening of the previous day. The utilization phase processing shown in Fig. 14 is an example of an estimation method.
[0096] When the utilization phase process starts, the utilization phase processing unit 150 functions as the estimation acquisition unit 151 and acquires operation information of the air conditioner 3 used by the consumer that is the subject of estimation (step S21). Specifically, the utilization phase processing unit 150 acquires the operation time, the set temperature, and the room temperature at the time of operation from the operation history data 121 as operation information of the air conditioner 3 that is the subject of estimation on a normal day.
[0097] Upon acquiring the operation information, the utilization phase processing unit 150 functions as the estimation unit 153 and estimates the power saving acceptance of the consumer to be estimated (step S22). Specifically, the utilization phase processing unit 150 inputs the operation information acquired in step S21 to the trained model 125 generated in step S13. Then, the utilization phase processing unit 150 estimates the power saving acceptance output from the trained model 125 as the power saving acceptance of the consumer to be estimated.
[0098] When the power saving acceptance level is estimated, the utilization phase processing unit 150 executes a notification or device control process based on the estimated power saving acceptance level (step S23). Specifically, the utilization phase processing unit 150 functions as a notification unit 155 and notifies suggested information regarding the operation of at least one device used by a consumer having a low estimated power saving acceptance level. In addition, the utilization phase processing unit 150 functions as a device control unit 157 and suppresses power consumption of at least one device used by a consumer having a high estimated power saving acceptance level. With the above, the utilization phase processing shown in FIG. 14 ends.
[0099] As described above, the estimation device 10 according to the first embodiment generates the trained model 125 by machine learning using the learning data 123 in which operation information of the multiple air conditioners 3 used by the multiple consumers to be learned is associated with the power saving acceptance levels of the multiple consumers to be learned. Then, based on the generated trained model 125, the estimation device 10 estimates the power saving acceptance level of the consumer to be estimated from the operation information of the air conditioners 3 used by the consumer to be estimated.
[0100] In this way, the estimation device 10 according to the first embodiment estimates the power saving acceptance level of a consumer by using machine learning, and therefore, it is not necessary to wait until the operation information history is accumulated, and it is possible to estimate the power saving acceptance level in a shorter period of time. Therefore, it is possible to estimate the power saving acceptance level with high accuracy even for consumers who do not have sufficient operation information history, such as consumers who have just started using the service or consumers who do not react.
[0101] In particular, by estimating the power saving acceptance of consumers with high accuracy, it becomes possible to more reliably implement the suppression of power demand in response to a power saving request. To be more specific, although it is assumed that consumers will receive some kind of incentive in return for complying with a power saving request, the level of awareness regarding power saving varies from one consumer to another. Therefore, it is expected that some consumers will feel very dissatisfied and will return to their original operating state after a power saving request is made, making it impossible to suppress the power demand. On the other hand, if the power demand suppression is only suppressed to a smaller amount than planned, it will be necessary to increase the supply of power to cover the shortage. In that case, economic losses may occur to retailers, aggregators, etc. in the form of the procurement price of power. In addition, cases where there is no power to be procured in the first place are also assumed. In that case, it may lead to a power outage, and there is a possibility that damage will be suffered over a wider area.
[0102] According to the estimation device 10 of the first embodiment, the power saving acceptance level of the consumer can be estimated with high accuracy, and therefore the consumer who is likely to cooperate with the demand reduction can be identified. Therefore, the above-mentioned situation can be avoided, and the power demand reduction due to the power saving request can be more reliably implemented.
[0103] (Embodiment 2) Next, a description will be given of embodiment 2. Descriptions of configurations and functions similar to those of embodiment 1 will be omitted where appropriate.
[0104] In the above-mentioned first embodiment, the notification unit 155 notifies the consumer having a low power saving acceptance level estimated by the estimation unit 153 of the proposal information shown in Fig. 11. In contrast, in the second embodiment, the notification unit 155 notifies the consumer having a low power saving acceptance level estimated by the estimation unit 153 of the proposal information shown in Fig. 15 on the warning issuance date, which is the period during which power saving is requested.
[0105] The proposal information shown in Fig. 15 includes a button labeled "OK (set)" in addition to the proposal information shown in Fig. 11. When the consumer agrees to the proposal of the recommended settings, the consumer inputs his / her intention to agree by pressing this button.
[0106] The equipment control unit 157 suppresses the power consumption of at least one device used by the consumer to be estimated in response to the consent to the proposal information notified by the notification unit 155. Specifically, when consent to the proposal information is obtained from any one of the multiple consumers to which the proposal information is notified by the notification unit 155, the equipment control unit 157 transmits a control instruction to at least one device used by the consumer. As a result, the equipment control unit 157 controls the at least one device in a direction to reduce power consumption. Details of the control of the equipment are the same as those in the first embodiment.
[0107] As described above, in the second embodiment, a series of processes from notification of proposal information to control of appliances can be performed on one screen, thereby reducing the effort required for consumers to control appliances by themselves.
[0108] (Embodiment 3) Next, a description will be given of embodiment 3. Descriptions of configurations and functions similar to those of embodiments 1 and 2 will be omitted where appropriate.
[0109] In the third embodiment, the learning data generation unit 113 corrects the power saving acceptance level estimated by the estimation unit 153 in accordance with the behavior of the consumer to be estimated, and updates the learning data 123 based on the corrected power saving acceptance level. In other words, if the consumer takes an action related to power saving after the estimation unit 153 estimates the power saving acceptance level of the consumer to be estimated, the learning data generation unit 113 corrects the power saving acceptance level of the consumer based on the action.
[0110] Here, the consumer behavior is behavior that reflects the consumer's awareness of saving electricity. Three examples of consumer behavior are described below: (1) acceptance of the proposal information shown in FIG. 15, (2) answering a questionnaire, and (3) selection of a contract plan for electricity supply.
[0111] (1) Regarding consent to the proposed information, when a consumer consents to the proposed information shown in Fig. 15 by pressing the "OK (set)" button, it means that the consumer has at least some interest in saving electricity. In other words, a consumer who has acted by consenting to the proposed information is likely to have a high degree of acceptance of saving electricity.
[0112] (2) Regarding the response to the questionnaire, the notification unit 155, for example, displays a questionnaire screen shown in Fig. 16 on the operation terminal 5 of the consumer. When the consumer enters a tolerant response to a power saving request on such a questionnaire screen, there is a high possibility that the consumer has a high acceptance of power saving.
[0113] (3) Regarding the selection of a power supply contract plan, the notification unit 155, for example, causes the consumer's operation terminal 5 to display a power supply contract plan selection screen shown in Fig. 17. When a consumer selects a plan on such a contract plan selection screen that allows the consumer to be controlled from the outside at all times, not just in emergencies, there is a high possibility that the consumer has a high acceptance of power saving.
[0114] When a consumer who has been estimated by the estimation unit 153 to have a low power-saving acceptance level takes such an action, the learning data generation unit 113 attaches importance to the action and corrects the power-saving acceptance level of the consumer to high. Then, the learning data generation unit 113 updates the learning data 123 based on the corrected power-saving acceptance level.
[0115] Specifically, the learning data generation unit 113 generates a new data set by combining the correct answer of "power saving acceptance is high" with the operation information of the air conditioner 3 of that consumer on a normal day, and adds it to the existing learning data 123. Alternatively, when the existing learning data 123 includes a data set that combines the power saving acceptance and operation information of that consumer, the learning data generation unit 113 may correct the level of the power saving acceptance in that data set.
[0116] When the learning data generation unit 113 updates the learning data 123, the learning processing unit 115 updates the trained model 125 using the updated learning data 123. In other words, the learning processing unit 115 regenerates the trained model 125 using the new learning data 123. The process of the learning processing unit 115 at that time is the same as that in the first embodiment, and therefore a description thereof will be omitted.
[0117] In this way, in the third embodiment, the trained model 125 is updated using new learning data 123 in which the power-saving acceptance level estimated from the behavior of the consumer in response to the content notified to the consumer is set as the correct answer. This improves the estimation accuracy of the power-saving acceptance level. In addition, the power-saving awareness of the consumer may change over time. By introducing such a method and appropriately re-learning and regenerating the trained model 125, it becomes easier to make an estimation that follows the changes in the power-saving awareness of the consumer.
[0118] (Embodiment 4) Next, a description will be given of embodiment 4. Descriptions of configurations and functions similar to those of embodiments 1 to 3 will be omitted where appropriate.
[0119] 18 shows a functional configuration of the estimation device 10 according to embodiment 4. In embodiment 4, the learning acquisition unit 111 and the estimation acquisition unit 151 acquire driving information similarly to embodiment 1, and further acquire an influence factor 127.
[0120] Here, the influencing factors 127 are factors that can affect the power saving acceptance of the consumer other than the operation information of the air conditioner 3. The influencing factors 127 are, for example, information such as the outside temperature, the consumer's address, a weather forecast, the consumer's family composition, the power rate plan subscribed to by the consumer, and the presence or absence of solar power generation equipment in the consumer's home. The learning acquisition unit 111 and the estimation acquisition unit 151 acquire at least one of these pieces of information as the influencing factors 127.
[0121] To explain in more detail, in winter in Japan, the weather is often such that "heavy snow falls on the Japan Sea side and clear skies on the Pacific Ocean side." It is not uncommon for outdoor temperatures to be below freezing in northern Japan while temperatures are around 10°C in Shikoku and Kyushu. It is clear that the higher the outdoor temperature, the lower the demand for heating, and the higher the receptivity to requests to conserve electricity by reducing heating. Therefore, for example, outdoor temperatures, the address of the consumer, and weather forecasts can be influencing factors 127.
[0122] Similarly, the family structure of the consumer, the electricity rate plan subscribed to by the consumer, the presence or absence of a solar power generation facility in the consumer's home, and the like can also be influencing factors 127. For example, in a home where someone is always at home, it is difficult to reduce power consumption, so the consumer's acceptance of power saving is likely to be low. In contrast, if a person lives alone, he or she can easily go out and reduce power consumption in response to a power saving request, so the consumer's acceptance of power saving is likely to be high. In addition, if the electricity rate plan subscribed to by the consumer is a plan that allows external control as shown in FIG. 17, the consumer's acceptance of power saving is likely to be high. If a solar power generation facility is installed in the consumer's home, the consumer's acceptance of power saving is likely to be high.
[0123] The acquisition unit for learning 111 and the acquisition unit for estimation 151 acquire information on the outside air temperature at the location where the target air conditioner 3 is located, for example, from a temperature sensor installed in the outdoor unit of the air conditioner 3, or from a weather server via a wide area communication network. The acquisition unit for learning 111 and the acquisition unit for estimation 151 also acquire information on the weather forecast at the location where the target air conditioner 3 is located, for example, from a weather server via a wide area communication network. Furthermore, the acquisition unit for learning 111 and the acquisition unit for estimation 151 acquire information such as the address, family composition, power rate plan, and presence or absence of solar power generation equipment of each target consumer, for example, from information input by the consumer at the operation terminal 5. In this way, the acquisition unit for learning 111 and the acquisition unit for estimation 151 acquire the influence factors 127 for each consumer that is the learning target and the estimation target.
[0124] The learning data generation unit 113 takes into account the influence factors 127 acquired by the learning acquisition unit 111 when generating the learning data 123. Specifically, the learning data generation unit 113 generates learning data 123 in which operation information of the air conditioners 3 used by multiple consumers to be learned on normal days, the influence factors 127 acquired by the learning acquisition unit 111, and the power saving acceptance level of each consumer are associated with each other. The learning processing unit 115 performs machine learning using the learning data 123 generated by the learning data generation unit 113. As a result, the learning processing unit 115 receives input of the operation information and the influence factors 127 and generates a trained model 125 that outputs the power saving acceptance level.
[0125] The estimation unit 153 estimates the power saving acceptance level of the consumer to be estimated from the operation information of the air conditioner 3 used by the consumer to be estimated on a normal day acquired by the estimation acquisition unit 151 and the influence factors 127, based on the trained model 125 generated by the learning processing unit 115. Specifically, the estimation unit 153 inputs the operation information of the air conditioner 3 used by the consumer to be estimated and the influence factors 127 to the trained model 125. Then, the estimation unit 153 estimates the output value output from the trained model 125 as the power saving acceptance level of the consumer to be estimated.
[0126] In this way, the estimation device 10 according to the fifth embodiment generates the trained model 125 by taking into account the influencing factors 127 in addition to the operation information of the air conditioner 3, and estimates the power saving acceptance level using the generated trained model 125. By taking into account the influencing factors 127, it is possible to estimate the power saving acceptance level of the consumer with higher accuracy.
[0127] (Embodiment 5) Next, a description will be given of embodiment 5. Descriptions of configurations and functions similar to those of embodiments 1 to 4 will be omitted where appropriate.
[0128] In the first embodiment, the estimation device 10 includes both the learning phase processing unit 110 and the utilization phase processing unit 150. In contrast, in the fifth embodiment, the learning phase processing unit 110 and the utilization phase processing unit 150 are included in separate devices.
[0129] 19 shows a configuration of an estimation system 1 according to embodiment 5. The estimation system 1 according to embodiment 5 includes a learning device 20 and an estimation device 10. As shown in FIG. 19, the learning device 20 includes each function of a learning phase processing unit 110, and the estimation device 10 includes each function of a utilization phase processing unit 150.
[0130] The learning device 20 is a device independent of the estimation device 10, and is realized by an appropriate information processing device such as a computer, a server, etc. The learning device 20 has a hardware configuration similar to that of the control unit 11, the storage unit 12, and the communication unit 13 in the estimation device 10, for realizing each function of the learning phase processing unit 110.
[0131] The learning device 20 generates a trained model 125 by executing the processing of the learning phase processing unit 110. The estimation device 10 communicates with the learning device 20 via the communication unit 13, and acquires the trained model 125 generated by the learning device 20. Then, the estimation device 10 executes the processing of the utilization phase processing unit 150 using the acquired trained model 125. Details of the processing of each function of the learning phase processing unit 110 and the utilization phase processing unit 150 are the same as those in the first embodiment.
[0132] In the estimation device 10, the estimation acquisition unit 151 may communicate with the learning device 20 via the communication unit 13 and acquire operation information of the air conditioner 3 used by the consumer that is the subject of estimation from the operation history data 121 stored in the learning device 20. Alternatively, the estimation acquisition unit 151 may acquire the operation information by directly communicating with the air conditioner 3 used by the consumer that is the subject of estimation via the communication unit 13 without referring to the operation history data 121.
[0133] As a modification of the fifth embodiment, the learning data generation unit 113 and the learning processing unit 115 may be provided in separate devices. In other words, the device that generates the learning data 123 and the device that executes the learning process to generate the trained model 125 are not limited to being the same learning device 20, and may be separate devices. In addition, the estimation unit 153, and the notification unit 155 and the appliance control unit 157 that utilize the power saving acceptance level estimated by the estimation unit 153 are not limited to being provided in the same estimation device 10, and may be provided in separate devices.
[0134] (Modification) Although the embodiments have been described above, it is possible to combine the embodiments, or to modify or omit the embodiments as appropriate.
[0135] For example, in the above embodiment, the power saving acceptance level is expressed by two values, high and low, representing the level of the consumer's behavior to suppress power consumption during the period when power saving is requested. However, the power saving acceptance level may be expressed in stages with three or more values, representing the level of the consumer's behavior to suppress power consumption during the period when power saving is requested. For example, the power saving acceptance level may be expressed by three values, high, medium, and low, representing the level of the consumer's behavior to suppress power consumption during the period when power saving is requested. By expressing the power saving acceptance level in many stages in this way, it becomes possible to estimate the power saving acceptance level with higher accuracy.
[0136] In the above embodiment, the trained model 125 outputs two objective variables, namely, the cooling power-saving acceptance degree and the heating power-saving acceptance degree, in response to the input of six explanatory variables that are operation information, as shown in Fig. 10. However, the trained model 125 is not limited to this.
[0137] For example, the trained model 125 may include two models: a cooling model that receives input of operation information during cooling and outputs a power-saving acceptance level during cooling, and a heating model that receives input of operation information during heating and outputs a power-saving acceptance level during heating. In this case, the learning processing unit 115 performs machine learning separately for cooling and heating, respectively, to generate a cooling model and a heating model. Alternatively, the trained model 125 may output one power-saving acceptance level as a response variable in response to input of operation information during cooling or heating, without distinguishing between cooling and heating.
[0138] In the above embodiment, a case has been described in which the home appliance whose operation information is to be acquired by the estimation device 10 is the air conditioner 3. However, the home appliance is not limited to the air conditioner 3, and may be other appliances such as a lighting device, a water heater, etc.
[0139] For example, when the home appliance is a lighting device, if the indoor illuminance when turning on or off the light is darker on the day when the alarm is issued than on a normal day, it is estimated that the power saving acceptance of the consumer is high. Also, when the set illuminance is darker on the day when the alarm is issued than on a normal day, it is estimated that the power saving acceptance of the consumer is high. Or, when the home appliance is a water heater, if the amount of water filled with water on a day with a similar outdoor temperature in winter is less on the day when the alarm is issued than on a normal day, it is estimated that the power saving acceptance of the consumer is high. Also, if the water filling temperature and the water supply temperature are lower on a day with a similar outdoor temperature in winter on the day when the alarm is issued than on a normal day, it is estimated that the power saving acceptance of the consumer is high. In this way, even if the home appliance is other than the air conditioner 3, it can be described in the same manner as in the above embodiment by appropriately replacing the parameters of the operation information.
[0140] In the above embodiment, the control unit 11 of the estimation device 10 functions as each unit in the learning phase processing unit 110 and the utilization phase processing unit 150 by the CPU executing a program stored in the ROM or the storage unit 12. However, the control unit 11 may be dedicated hardware. The dedicated hardware is, for example, a single circuit, a composite circuit, a programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination of these. When the control unit 11 is dedicated hardware, the functions of each unit may be realized by individual hardware, or the functions of each unit may be realized together by a single piece of hardware.
[0141] In addition, some of the functions of each unit may be realized by dedicated hardware, and other parts may be realized by software or firmware. In this way, the control unit 11 can realize each of the above-mentioned functions by hardware, software, firmware, or a combination of these.
[0142] It is also possible to make an existing computer, such as a personal computer or an information terminal device, function as the estimation device 10 by applying a program that defines the operation of the estimation device 10 to the existing computer.
[0143] Furthermore, such programs may be distributed in any manner, and may be distributed by storing them on a computer-readable recording medium such as a CD-ROM (Compact Disk ROM), a DVD (Digital Versatile Disk), an MO (Magneto Optical Disk), or a memory card, or may be distributed via a communications network such as the Internet.
[0144] Various embodiments and modifications of the present disclosure are possible without departing from the broad spirit and scope of the present disclosure. The above-described embodiments are for explaining the present disclosure and do not limit the scope of the present disclosure. That is, the scope of the present disclosure is indicated by the claims, not the embodiments. Various modifications made within the scope of the claims and the scope of the disclosure equivalent thereto are considered to be within the scope of the present disclosure.
[0145] Various aspects of the present disclosure are summarized below as appendices.
[0146] (Appendix 1) An estimation device for estimating a power saving acceptance level of a consumer, An acquisition means for acquiring operation information of a home appliance used by the consumer to be estimated; and an estimation means for estimating the power saving acceptance level of the target consumer from the operation information acquired by the acquisition means, based on a trained model generated by machine learning using training data in which operation information of a plurality of home appliances used by the target consumer is associated with the power saving acceptance level of the target consumer. Estimation device. (Appendix 2) The power saving acceptance level estimation method further includes a notification means for notifying a user of a power saving acceptance level estimated by the estimation means of suggested information regarding an operation of at least one appliance used by the user. 2. An estimation apparatus as described in claim 1. (Appendix 3) and a device control unit that executes control to suppress power consumption of at least one device used by the consumer to be estimated in response to consent to the proposal information notified by the notification unit. 3. An estimation apparatus as described in claim 2. (Appendix 4) Further, the power saving acceptance level is estimated by the estimation means, and a device control means is configured to execute control to suppress power consumption of at least one device used by the consumer to be estimated. 2. An estimation apparatus as described in claim 1. (Appendix 5) The power saving acceptance level of each of the plurality of consumers to be learned is derived according to a difference in operation information of a home appliance between a period when power saving is requested and a period when power saving is not requested. 5. An estimation apparatus according to any one of claims 1 to 4. (Appendix 6) The power saving acceptance level of the consumer to be estimated and the power saving acceptance levels of the plurality of consumers to be learned are expressed in three or more stages based on the degree of the consumer's behavior of suppressing power consumption during a period in which power saving is requested. 6. An estimation apparatus according to any one of claims 1 to 5. (Appendix 7) The home appliance used by the consumer to be estimated and each of the plurality of home appliances used by the plurality of consumers to be learned are air conditioners, The operation information indicates at least one of an operation time of the air conditioner, a set temperature of the air conditioner, and a room temperature when the air conditioner is operated. 7. An estimation apparatus according to any one of claims 1 to 6. (Appendix 8) A learning data generating means for generating the learning data; A learning processing means for generating the trained model by the machine learning using the training data generated by the training data generation means, 8. An estimation apparatus according to any one of claims 1 to 7. (Appendix 9) The learning data generating means Deriving the power saving acceptance degree of the plurality of consumers to be learned based on operation information of the plurality of home appliances during a period when power saving is requested and operation information of the plurality of home appliances during a period when power saving is not requested; generating the learning data in which operation information of the plurality of home electric appliances during the period when power saving is not requested is associated with the derived power saving acceptance degree; 9. The estimation apparatus of claim 8. (Appendix 10) the learning data generation means corrects the power-saving acceptance level estimated by the estimation means in accordance with the behavior of the consumer being the estimation target, and updates the learning data based on the corrected power-saving acceptance level; The learning processing means updates the trained model using the learning data updated by the learning data generation means. 10. The estimation device according to claim 8 or 9. (Appendix 11) The estimation device according to any one of Supplementary Notes 1 to 10, and the home appliance used by the consumer to be estimated. Estimation system. (Appendix 12) An acquisition means for acquiring operation information of a plurality of home appliances used by a plurality of consumers to be learned; a learning data generating means for generating learning data in which the operation information acquired by the acquiring means is associated with the power saving acceptance levels of the plurality of consumers to be learned; and a learning processing means for performing machine learning using the learning data generated by the learning data generating means to generate a trained model that outputs a power saving acceptance level of the consumer to be estimated in response to input of operation information of a home appliance used by the consumer to be estimated. Learning device. (Appendix 13) Acquire operation information of a plurality of home appliances used by a plurality of consumers to be learned; generating learning data in which the acquired operation information is associated with the power saving acceptance levels of the plurality of consumers to be learned; By performing machine learning using the generated learning data, a trained model is generated that outputs a power saving acceptance level of the consumer to be estimated in response to input of operation information of a home appliance used by the consumer to be estimated. How to generate a trained model. (Appendix 14) A method for estimating a consumer's acceptance of power saving, comprising: Acquire operation information of a home appliance used by the consumer that is the subject of estimation; Based on a trained model generated by machine learning using learning data in which operation information of a plurality of home appliances used by a plurality of target consumers is associated with the power saving acceptance of the plurality of target consumers, the power saving acceptance of the plurality of target consumers is estimated from the acquired operation information. Estimation method. (Appendix 15) Computer, An acquisition means for acquiring operation information of a home appliance used by a consumer to be estimated; and functioning as an estimation means for estimating the power saving acceptance level of the target consumer from the operation information acquired by the acquisition means, based on a trained model generated by machine learning using learning data in which operation information of a plurality of home appliances used by the target consumer is associated with the power saving acceptance level of the target consumer. program. [Explanation of symbols]
[0147] 1 Estimation system, 3 Air conditioner, 5 Operation terminal, 10 Estimation device, 11 Control unit, 12 Memory unit, 13 Communication unit, 20 Learning device, 110 Learning phase processing unit, 111 Learning acquisition unit, 113 Learning data generation unit, 115 Learning processing unit, 121 Operation history data, 123 Learning data, 125 Learned model, 127 Influence factor, 150 Utilization phase processing unit, 151 Estimation acquisition unit, 153 Estimation unit, 155 Notification unit, 157 Equipment control unit
Claims
1. An estimation device for estimating a power saving acceptance level of a consumer, An acquisition means for acquiring operation information of a home appliance used by the consumer to be estimated; and an estimation means for estimating the power saving acceptance level of the target consumer from the operation information acquired by the acquisition means, based on a trained model generated by machine learning using training data in which operation information of a plurality of home appliances used by the target consumer is associated with the power saving acceptance level of the target consumer. Estimation device.
2. The power saving acceptance level is estimated by the estimation means, and a notification means is configured to notify the customer of the power saving acceptance level estimated by the estimation means of suggested information regarding the operation of at least one appliance used by the customer. The estimation device according to claim 1 .
3. and a device control unit that executes control to suppress power consumption of at least one device used by the consumer to be estimated in response to consent to the proposal information notified by the notification unit. The estimation device according to claim 2 .
4. The power saving acceptance level is estimated by the estimation means, and based on the estimated power saving acceptance level, the power saving acceptance level is estimated by the estimation means, and a device control means is further provided to execute control to suppress power consumption of at least one device used by the consumer to be estimated. The estimation device according to claim 1 .
5. The power saving acceptance level of each of the plurality of consumers to be learned is derived according to a difference in operation information of a home appliance between a period when power saving is requested and a period when power saving is not requested. The estimation device according to any one of claims 1 to 4.
6. The power saving acceptance level of the consumer to be estimated and the power saving acceptance levels of the plurality of consumers to be learned are expressed in stages using three or more values representing the degree of the consumer's behavior of suppressing power consumption during a period in which power saving is requested. The estimation device according to any one of claims 1 to 4.
7. The home appliance used by the consumer to be estimated and each of the plurality of home appliances used by the plurality of consumers to be learned are air conditioners, The operation information indicates at least one of an operation time of the air conditioner, a set temperature of the air conditioner, and a room temperature when the air conditioner is operated. The estimation device according to any one of claims 1 to 4.
8. A learning data generating means for generating the learning data; A learning processing means for generating the trained model by the machine learning using the training data generated by the training data generation means, The estimation device according to any one of claims 1 to 4.
9. The learning data generating means Deriving the power saving acceptance degree of the plurality of consumers to be learned based on operation information of the plurality of home appliances during a period when power saving is requested and operation information of the plurality of home appliances during a period when power saving is not requested; generating the learning data in which operation information of the plurality of home electric appliances during the period when power saving is not requested is associated with the derived power saving acceptance degree; The estimation device according to claim 8.
10. the learning data generation means corrects the power-saving acceptance level estimated by the estimation means in accordance with the behavior of the consumer being the estimation target, and updates the learning data based on the corrected power-saving acceptance level; The learning processing means updates the trained model using the learning data updated by the learning data generation means. The estimation device according to claim 8.
11. A power consumption estimation system comprising: the estimation device according to claim 1 ; and the home appliance used by a consumer to be estimated. Estimation system.
12. An acquisition means for acquiring operation information of a plurality of home appliances used by a plurality of consumers to be learned; a learning data generating means for generating learning data in which the operation information acquired by the acquiring means is associated with the power saving acceptance levels of the plurality of consumers to be learned; and a learning processing means for performing machine learning using the learning data generated by the learning data generating means to generate a trained model that outputs a power saving acceptance level of the consumer to be estimated in response to input of operation information of a home appliance used by the consumer to be estimated. Learning device.
13. Acquire operation information of a plurality of home appliances used by a plurality of consumers to be learned; generating learning data in which the acquired operation information is associated with the power saving acceptance levels of the plurality of consumers to be learned; By performing machine learning using the generated learning data, a trained model is generated that outputs a power saving acceptance level of the consumer to be estimated in response to input of operation information of a home appliance used by the consumer to be estimated. How to generate a trained model.
14. A method for estimating a consumer's acceptance of power saving, comprising: Acquire operation information of a home appliance used by the consumer that is the subject of estimation; Based on a trained model generated by machine learning using learning data in which operation information of a plurality of home appliances used by a plurality of target consumers is associated with the power saving acceptance of the plurality of target consumers, the power saving acceptance of the plurality of target consumers is estimated from the acquired operation information. Estimation method.
15. Computer, An acquisition means for acquiring operation information of a home appliance used by a consumer to be estimated; and functioning as an estimation means for estimating the power saving acceptance level of the target consumer from the operation information acquired by the acquisition means, based on a trained model generated by machine learning using learning data in which operation information of a plurality of home appliances used by the target consumer is associated with the power saving acceptance level of the target consumer. program.
Citation Information
Patent Citations
Information processor and information processing method
JP2014215728A
