Control system and control method

Abstract

To appropriately grasp a change in performance of an air conditioner.SOLUTION: The control system comprises: a model creation unit configured to create a trained machine learning model by performing machine learning using, as training data, a time series of indoor temperature in an installation location of an air conditioner, a time series of outdoor temperature or a prescribed outdoor temperature, and a time series of information indicating an operation state of the air conditioner, wherein the trained model receives, as input data, an initial value of the indoor temperature, a time series of the outdoor temperature or the prescribed outdoor temperature, and a time series of the information, and outputs a change in the indoor temperature caused by operation of the air conditioner or a time based on the change; an accumulation unit configured to accumulate, as an actual result time series, a time series of actual result values of the indoor temperature, a time series of actual result values of the outdoor temperature or actual result values of the prescribed outdoor temperature, and a time series of actual result values of the information; and a determination unit configured to determine whether to perform prescribed notification, by using a first trained machine learning model created using an actual result time series accumulated in a first period as training data, and part or all of an actual result time series accumulated in a second period subsequent to the first period.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a control system and a control method. [Background technology]

[0002] The data provision method described in Patent Document 1 displays a trend in the time it takes for the room temperature to reach the set temperature from the time the air conditioner starts operating, when the relationship between the initial room temperature and the outdoor temperature in the room where the air conditioner is installed is constant, using sample data from at least three time points indicating the time it takes for the room temperature to reach the set temperature. Specifically, the constant relationship between the initial room temperature and the outdoor temperature refers to a relationship where the outdoor temperature is 35°C and the initial room temperature is 33°C. The trend in the time it takes for the room temperature to drop from an outdoor temperature of 35°C and an initial room temperature of 33°C to 28°C is displayed, for example, as 5 minutes in 2013, 10 minutes in 2018, and 15 minutes in 2023. According to the data provision method described in Patent Document 1, by using log information measured under the same environmental conditions, it is possible to clearly display trends in the air conditioner's performance, such as a decline in cooling function. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2014 / 171118 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the data provision method described in Patent Document 1, changes in air conditioner performance are displayed based on multiple pieces of data acquired when the relationship between the initial room temperature and the outside air temperature is constant. Therefore, for example, when it is difficult to acquire data when the relationship between the initial room temperature and the outside air temperature is constant, there is a problem in that it can be difficult to properly grasp changes in air conditioner performance.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a control system and a control method that can appropriately grasp changes in the performance of an air conditioner. [Means for solving the problem]

[0006] In order to solve the above problem, the control system of the present disclosure includes a model creation unit that receives as input data an initial value of the room temperature at the installation location of the air conditioner, a time series of the outside air temperature or a predetermined outside air temperature, and a time series of information indicating the operating status of the air conditioner, and creates a trained machine learning model that calculates a change in room temperature due to operation of the air conditioner or a time based on that change by machine learning using the time series of the room temperature and the time series of the outside air temperature or the predetermined outside air temperature and the time series of the information as training data; a storage unit that accumulates the time series of actual values ​​of the room temperature and the time series of actual values ​​of the outside air temperature or the time series of actual values ​​of the predetermined outside air temperature and the information as actual time series; and a determination unit that determines whether to issue a predetermined notification using a first trained machine learning model created using the actual time series accumulated in a first period as training data, and part or all of the actual time series accumulated in a second period after the first period.

[0007] The control method disclosed herein includes the steps of: inputting, as input data, an initial value of the room temperature at the installation location of the air conditioner, a time series of the outside temperature or a predetermined outside temperature, and a time series of information indicating the operating status of the air conditioner; and creating a trained machine learning model that calculates a change in room temperature due to operation of the air conditioner or a time based on that change by machine learning using the time series of the room temperature and the time series of the outside temperature or the predetermined outside temperature and the time series of the information as training data; accumulating the time series of actual values ​​of the room temperature and the time series of actual values ​​of the outside temperature or the time series of actual values ​​of the predetermined outside temperature and the information as actual time series; and determining whether to issue a predetermined notification using a first trained machine learning model created using the actual time series accumulated in a first period as training data, and part or all of the actual time series accumulated in a second period after the first period. [Effects of the Invention]

[0008] According to the control system and control method of the present disclosure, it is possible to appropriately grasp changes in the performance of the air conditioner. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a block diagram showing an example of the configuration of an air conditioner according to a first embodiment of the present disclosure. [Figure 2] FIG. 1 is a schematic diagram illustrating a configuration example of a machine learning model according to a first embodiment of the present disclosure. [Figure 3] 5 is a flowchart showing an example of the operation of the control device according to the first embodiment of the present disclosure. [Figure 4] 3 is a schematic diagram for explaining an example of operation of the control device according to the first embodiment of the present disclosure. FIG. [Figure 5] 3A and 3B are schematic diagrams for explaining an example of operation of the control device according to the first embodiment of the present disclosure. [Figure 6] FIG. 10 is a block diagram showing a configuration example of an air conditioner according to a second embodiment of the present disclosure. [Figure 7] 10 is a flowchart showing an example of the operation of the control device according to the second embodiment of the present disclosure. [Figure 8] FIG. 10 is a schematic diagram for explaining an example of operation of the control device according to the second embodiment of the present disclosure. [Figure 9] FIG. 10 is a schematic diagram for explaining an example of operation of the control device according to the second embodiment of the present disclosure. [Figure 10] 10 is a flowchart illustrating an example of the operation of the control device according to the third embodiment of the present disclosure. [Figure 11] FIG. 10 is a schematic diagram for explaining an example of operation of the control device according to the third embodiment of the present disclosure. [Figure 12] FIG. 10 is a schematic diagram for explaining an example of operation of the control device according to the third embodiment of the present disclosure. [Figure 13] FIG. 1 is a schematic block diagram illustrating the configuration of a computer according to at least one embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, a control system and a control method according to an embodiment of the present disclosure will be described with reference to the drawings. Note that the same or corresponding components in the drawings are designated by the same reference numerals and descriptions thereof will be omitted as appropriate.

[0011] First Embodiment FIG. 1 is a block diagram showing an example configuration of an air conditioner according to a first embodiment of the present disclosure. As shown in FIG. 1, the air conditioner 1 includes a control device 10, an outdoor air temperature sensor 21, a room temperature sensor 22, a heat exchanger temperature sensor 23, a compressor 24, and an expansion valve 25. For example, the outdoor air temperature sensor 21 is provided in an outdoor unit (not shown) and detects the outdoor air temperature. The room temperature sensor 22 is provided in an indoor unit (not shown) and detects the room temperature of the location where the air conditioner 1 is installed. The heat exchanger temperature sensor 23 is provided in a heat exchanger (not shown) and detects the temperature of the heat exchanger. However, the locations where the outdoor air temperature sensor 21, room temperature sensor 22, etc. are installed are not limited to these.

[0012] The control device 10 is connected to the outside air temperature sensor 21, the room temperature sensor 22, and the heat exchanger temperature sensor 23, and the control device 10 can acquire and record the temperatures measured by the outside air temperature sensor 21, the room temperature sensor 22, and the heat exchanger temperature sensor 23. The control device 10 is connected to the compressor 24 and the expansion valve 25, and the control device 10 can control the compressor 24 and the expansion valve 25 and acquire and record information indicating their operating states. For example, the control device 10 can operate the compressor 24 at a desired rotation speed and record the rotation speed. The control device 10 can also drive the expansion valve 25 at a desired valve opening and record the valve opening.

[0013] The control device 10 can be configured using a computer such as a microcontroller, and includes the following functional blocks configured by a combination of hardware such as a computer and software such as a program executed by the computer. That is, the control device 10 includes, as functional blocks, a sensor information acquisition unit 11, a setting acceptance unit 12, a storage unit 13, a model creation unit 14, a determination unit 15, a notification unit 16, and a control unit 17. The control device 10 can be connected to, for example, an external server (not shown) via a predetermined communication network, and some or all of the functional blocks, such as the storage unit 13, the model creation unit 14, the determination unit 15, and the notification unit 16, may be included in the external server. The control device 10 is also an example configuration of a control system according to the present disclosure.

[0014] The sensor information acquisition unit 11 acquires the outside air temperature measured by the outside air temperature sensor 21, the room temperature measured by the room temperature sensor 22, the temperature of the heat exchanger measured by the heat exchanger temperature sensor 23, and the like.

[0015] The setting reception unit 12 receives settings such as a target temperature (set temperature) for the space to be air-conditioned and a designated time at which the target temperature should be achieved. For example, the target temperature and the designated time are set using a user terminal such as a remote control or a smartphone (not shown).

[0016] The accumulation unit 13 accumulates, as actual time series, for example, a time series of actual room temperature values ​​acquired by the sensor information acquisition unit 11, a time series of actual outside air temperature values, and a time series of information indicating the operating state of the air conditioner 1 acquired (or generated) as command values ​​or the like by the control unit 17 (described later). Alternatively, the accumulation unit 13 accumulates, as actual time series, for example, a time series of actual room temperature values ​​acquired by the sensor information acquisition unit 11, a predetermined outside air temperature, and a time series of information indicating the operating state of the air conditioner 1 acquired or generated as command values, sensor values, or the like by the control unit 17 (described later). Here, the predetermined outside air temperature is, for example, the outside air temperature at the start of operation of the air conditioner 1 (initial value of the outside air temperature). However, the predetermined outside air temperature is not limited to the outside air temperature at the start of operation, and may be, for example, the outside air temperature after a predetermined time has elapsed since the start of operation, an initial value, or an average value of the outside air temperatures after one or more predetermined times. The operating state of the air conditioner 1 refers to, for example, the operating state of the compressor 24 of the air conditioner 1, the operating state of the expansion valve 25, the operating mode (heating, cooling, dehumidification, etc.), etc. For example, the information indicating the operating state of the compressor 24 is data indicating values ​​such as the rotation speed and drive current of the drive motor of the compressor 24. For example, the information indicating the operating state of the expansion valve 25 is data indicating values ​​such as the valve opening of the expansion valve 25. In this embodiment, a time series is a series of values ​​that continuously (i.e., at predetermined intervals) represent changes over time, and the terms time series data and time series are synonymous. The accumulation unit 13 accumulates, for example, a first-period actual time series 131 that is an actual time series accumulated during a first period, which is a certain period after installation of the air conditioner 1, and a second-period actual time series 132 that is an actual time series accumulated during a second period after the first period. The second period is a fixed period of time immediately preceding the first period, and the accumulation unit 13 sequentially updates the actual time series 132 for the second period. The first period is not limited to a fixed period after installation, and can be, for example, a fixed period after maintenance such as cleaning of a filter (not shown) provided in the air conditioner 1, and the actual time series 131 for the first period can be accumulated again, for example, multiple times. In this embodiment, the actual time series 131 for the first period and the actual time series 132 for the second period are assumed to contain sufficient data for machine learning of a machine learning model, which will be described later.

[0017] The model creation unit 14 creates a first trained machine learning model 141 and a second trained machine learning model 142. The first trained machine learning model 141 and the second trained machine learning model 142 are trained machine learning models that receive as input data an initial value of the room temperature at the installation location of the air conditioner 1 (such as the room temperature value at each start-up of the air conditioner 1) and a time series of the outside air temperature, or a time series of a predetermined outside air temperature and information indicating the operating state of the air conditioner 1, and calculate, as output data, a change in the room temperature associated with the operation of the air conditioner 1 or the time based on that change. That is, the first trained machine learning model 141 and the second trained machine learning model 142 receive as input data an initial value of the room temperature and a time series of the outside air temperature, or a time series of a predetermined outside air temperature and information indicating the operating state of the air conditioner 1, and calculate the change over time from the initial value of the room temperature or the time it takes for the room temperature to reach a predetermined temperature from the initial value (for example, a temperature with a predetermined temperature difference from the initial value or a temperature that is a predetermined percentage (%) of the initial value), and output the calculated value as output data. Note that, when the time based on a change in room temperature is obtained as output data, it can be calculated by performing machine learning using training data that includes information representing the actual value of the time based on a change in room temperature, or by adding a process for obtaining the time based on the change from the calculated value of the change in room temperature to the output stage of the machine learning model. Furthermore, the input data may include, as information indicating the operating state of the air conditioner 1, information indicating the operating state of the compressor 24 (e.g., rotation speed), information indicating the operating state of the expansion valve 25 (e.g., valve opening), information indicating the temperature of the heat exchanger, information indicating the air volume setting, operating mode, and target temperature (set temperature). Furthermore, the input data may include data indicating the environmental conditions (e.g., room conditions) of the installation location of the air conditioner 1. Furthermore, the first trained machine learning model 141 may be created at a timing specified by a user.

[0018] The model creation unit 14 creates the first trained machine learning model 141 and the second trained machine learning model 142 by machine learning using as training data a time series of room temperature and a time series of outside air temperature or a predetermined outside air temperature and a time series of information indicating the operating state of the air conditioner 1. In this case, the model creation unit 14 creates the first trained machine learning model 141 by machine learning using as training data the actual time series 131 for a first period, and creates the second trained machine learning model 142 by machine learning using as training data the actual time series 132 for a second period.

[0019] In this embodiment, there is no limitation on how the machine learning model is configured. The machine learning model may or may not use a neural network, for example. FIG. 2 shows an example in which the first trained machine learning model 141 and the second trained machine learning model 142 are configured as an ARX model (Auto-Regressive with eXogenous model) (linear multiple regression model) as a model for predicting changes in room temperature over time. Note that while FIG. 2 shows an example in which the operation mode is heating, in the case of cooling, "heating" can be read as "cooling."

[0020] In the model shown in FIG. 2 , Tin(k) is the room temperature [K] one step later, Tin(k-1) is the room temperature [K] at the current step, φh(k-1) is the heating capacity [W], To-in(k-1) is the outdoor / indoor temperature difference [K], a1 is the weight [-] related to the current room temperature, b11 is the weight [K / W] related to the heating capacity, b21 is the weight [-] related to the outdoor / indoor temperature difference, and w0 is the constant bias term [K]. The model shown in FIG. 2 predicts the room temperature at the next step based on the room temperature, heating capacity, and outdoor air temperature at the current step. By repeating this step, the change in room temperature over time is calculated. In this case, the time series of information indicating the operating state of the air conditioner 1 is a time series representing the change in heating capacity over time. Note that the heating capacity or cooling capacity is the thermal energy added to or removed from the room per unit time and can be calculated based on, for example, the rotation speed of the compressor 24. Note that in this embodiment, the heating capacity and cooling capacity are collectively referred to as air conditioning capacity. The outdoor / indoor temperature difference [K] may be calculated by fixing the outdoor air temperature to a predetermined value (for example, the outdoor air temperature to a predetermined initial value).

[0021] The determination unit 15 determines whether to issue a predetermined notification using a first trained machine learning model 141 created using a first period actual time series 131, which is an actual time series accumulated during a first period, as training data, and a part or all of a second period actual time series 132, which is an actual time series accumulated during a second period that follows the first period. Examples of the predetermined notification include a notification regarding cleaning of the filter of the air conditioner 1 indicating that the filter needs to be cleaned, a notification that a change (or a trend of change) in the performance of the air conditioner 1 has been detected, etc.

[0022] When the determining unit 15 determines that a predetermined notification should be made, the notifying unit 16 sends the notification to a user terminal such as a remote control or a smartphone (not shown). The content of the notification may be changed depending on, for example, the magnitude of a change in performance. Furthermore, when it is estimated that a significant change in performance has occurred, the notification may also include the estimated cause.

[0023] The control unit 17 controls the compressor 24, the expansion valve 25, etc. to perform air conditioning.

[0024] Next, an example of operation of the control device 10 according to the first embodiment will be described with reference to Figs. 3 to 5. Fig. 3 is a flowchart showing an example of operation of the control device 10 according to the first embodiment of the present disclosure. Note that the processing shown in Fig. 3 is an example of processing up to the time when the notification unit 16 issues a first notification. Figs. 4 and 5 are schematic diagrams for explaining an example of operation of the control device 10 according to the first embodiment of the present disclosure. Note that each flowchart illustrates an example of processing in the case where the time series of the outside air temperature is used out of the time series of the outside air temperature and a predetermined outside air temperature.

[0025] The process shown in FIG. 3 is initiated when the air conditioner 1 starts operating (or when a trial run ends) after installation. In the process shown in FIG. 3, first, the accumulation unit 13 accumulates, for a predetermined period (e.g., for multiple days), actual values ​​of the room temperature, the outside air temperature, and information indicating the operating state of the air conditioner as an actual time series 131 for a first period (step S11). Next, the model creation unit 14 uses the actual time series 131 for the first period accumulated during the first period as training data to create a first trained machine learning model 141 (hereinafter also referred to as a normal model) (step S12). Next, for example, the accumulation unit 13 determines whether a certain period has elapsed after step S12 (step S13). This certain period can be a period during which the performance of the air conditioner may change due to, for example, filter contamination.

[0026] If a certain period has passed (step S13: YES), the accumulation unit 13 accumulates the actual values ​​of the room temperature, the outside temperature, and the information indicating the operating state of the air conditioner for the certain period as an actual time series 132 for a second period (step S14). Next, the model creation unit 14 creates a second trained machine learning model 142 (hereinafter also referred to as a model after a certain period) using the actual time series 132 for the second period accumulated during the second period as training data (step S15). Next, the determination unit 15 determines whether to issue a predetermined notification based on a result obtained by inputting predetermined input data into the first trained machine learning model 141 and a result obtained by inputting the same input data into the second trained machine learning model 142 (step S16).

[0027] Fig. 4 schematically shows the flow of data in the processing of step S16 and the like. Fig. 5, with the horizontal axis representing time, shows an example of the change over time of room temperature obtained by the first trained machine learning model 141 (predicted value using the normal model) and the change over time of room temperature obtained by the second trained machine learning model 142 (predicted value using the model after a certain period of time). In step S16, as shown in Fig. 4, the same input data D11 is input to the first trained machine learning model 141 (normal model) trained by machine learning using the actual time series 131 for a first period as training data D21, and the second trained machine learning model 142 (model after a certain period of time) trained by machine learning using the actual time series 132 for a second period as training data D22. The input data D11 includes, for example, an initial value of the room temperature (e.g., 35°C), a time series of changes in the outdoor temperature over time for model comparison (or a constant outdoor temperature such as the initial value of the outdoor temperature), and a time series of changes in information indicating the standard operating state of the air conditioner 1 (e.g., a time series of a standard pattern during cooling operation). When the input data D11 is input, the calculation results of the first trained machine learning model 141 (normal model) and the calculation results of the second trained machine learning model 142 (model after a certain period of time) are, for example, curves such as those shown in FIG. 5. In the example shown in FIG. 5, compared to the normal model, the model after a certain period of time is a model in which the performance of the air conditioner 1 has deteriorated to a certain extent due to filter contamination. In this case, by inputting a time series of information indicating the standard operating state (e.g., a time series of cooling capacity in a standard operating pattern) into the normal model, the room temperature drops more quickly than the model after a certain period of time. Furthermore, if the temperature at which the room temperature stabilizes in the predicted value of the model after a certain period of time is set as the predetermined temperature, the time T1 required to reach the predetermined temperature predicted by the normal model is shorter than the time T2 required to reach the predetermined temperature predicted by the model after a certain period of time. The determination unit 15 can determine whether to issue a predetermined notification by, for example, comparing the deviation of the time required to reach the predetermined temperature (the difference between time T1 and time T2) with a predetermined threshold. Alternatively, the determination unit 15 can determine whether to issue a predetermined notification by, for example, comparing the deviation of the temperature after a certain period of time has elapsed (the difference between the temperature according to the normal model after a certain period of time has elapsed and the temperature according to the model after a certain period of time has elapsed) with a predetermined threshold.

[0028] If the determination unit 15 determines that a notification should be made (step S17: YES), the notification unit 16 makes a predetermined notification (step S18), and ends the processing shown in Fig. 10. On the other hand, if the determination unit 15 determines that a notification should not be made (step S17: NO), the accumulation unit 13 updates the performance time series 132 for the second period in step S14 after a certain period has elapsed, and the processing from step S15 onwards is executed again.

[0029] As described above, in this embodiment, the determination unit 15 determines whether to issue a predetermined notification using the first trained machine learning model 141 created using the actual time series accumulated in the first period (actual time series 131 for the first period) as training data D21, and some or all of the actual time series accumulated in the second period that follows the first period (actual time series 132 for the second period). With this configuration, even if the actual time series 131 for the first period and the actual time series 132 for the second period do not include data corresponding to a case where the relationship between the initial room temperature and the outside air temperature is constant, for example, it is possible to appropriately grasp changes in the performance of the air conditioner.

[0030] Furthermore, the model creation unit 14 of this embodiment creates a second trained machine learning model 142 using the actual time series accumulated in the second period (the actual time series 132 for the second period) as training data D22, and the determination unit 15 determines whether to issue a notification based on a result obtained by inputting predetermined input data D11 into the first trained machine learning model 141 and a result obtained by inputting the same input data D11 into the second trained machine learning model 142. This configuration makes it possible to compare the difference in performance between the first period and the second period under the same desired conditions.

[0031] Second Embodiment An air conditioner 1a according to a second embodiment of the present disclosure will be described with reference to Figs. 6 to 9. Fig. 6 is a block diagram showing a configuration example of the air conditioner 1a according to the second embodiment of the present disclosure. Fig. 7 is a flowchart showing an operation example of the control device 10a according to the second embodiment of the present disclosure. Figs. 8 and 9 are schematic diagrams for explaining an operation example of the control device according to the second embodiment of the present disclosure.

[0032] In the air conditioner 1a of the second embodiment shown in Fig. 6, the configuration of the model creation unit 14a and the operation of the determination unit 15a provided in the control device 10a are partially different from the model creation unit 14 and the determination unit 15, which are corresponding configurations to those of the first embodiment shown in Fig. 1. Furthermore, compared to the flowchart shown in Fig. 3, the flowchart shown in Fig. 7 omits step S15 shown in Fig. 3 and the content of step 16a corresponding to step S16 shown in Fig. 3 is partially different.

[0033] The model creation unit 14a of the second embodiment shown in FIG. 6 creates a first trained machine learning model 141, but does not create the second trained machine learning model 142 shown in FIG.

[0034] In addition, in step S16a of Figure 7, the judgment unit 15a of the second embodiment shown in Figure 6 determines whether to issue a notification by comparing the result (change in room temperature or the time corresponding to the change) obtained by inputting the initial value of room temperature, the time series of outside air temperature, and the time series of information indicating the operating state of the air conditioner contained in the actual time series accumulated over the second period (actual time series 132 for the second period) as input data D12 into the first trained machine learning model 141, as shown in Figure 8, with the change (D31) in the actual value of room temperature based on the actual time series accumulated over the second period (actual time series 132 for the second period) or the time corresponding to the change.

[0035] In the second embodiment, for example, as shown in FIG. 9 , the first trained machine learning model 141 (normal model) can compare the change in room temperature over time (predicted value by the normal model) obtained by inputting the actual values ​​of the outdoor air temperature and the operating state of the air conditioner after a certain period of time with the change in room temperature over time included in the actual value after the certain period of time. If the performance of the air conditioner 1 is reduced to some extent due to filter contamination after the certain period of time, the actual value of the operating state of the air conditioner after the certain period of time will change in the direction of increasing the cooling capacity. Therefore, if this actual value of the operating state is input into the normal model, the room temperature will decrease more quickly than the model after the certain period of time. Furthermore, if the temperature at which the room temperature stabilizes in the actual value after the certain period of time is set as the predetermined temperature, the time T1 required to reach the predetermined temperature predicted by the normal model is shorter than the time T2 required to reach the predetermined temperature in the actual value after the certain period of time. The determination unit 15a can determine whether to issue a predetermined notification by, for example, comparing the deviation of the time required to reach the predetermined temperature (the difference between time T1 and time T2) with a predetermined threshold.

[0036] As described above, the judgment unit 15a may make a judgment based on the degree of discrepancy in the time it takes to reach the specified temperature, or may make a judgment by comparing the error between the predicted room temperature value and the actual room temperature value (RMSE (Root Mean Squared Error) or maximum error over a certain period of time) with a specified threshold value.

[0037] As described above, in this embodiment, the determination unit 15a determines whether to issue a predetermined notification using the first trained machine learning model 141 created using the actual time series accumulated in the first period (actual time series 131 for the first period) as training data D21, and some or all of the actual time series accumulated in the second period that follows the first period (actual time series 132 for the second period). With this configuration, even if the actual time series 131 for the first period and the actual time series 132 for the second period do not include data corresponding to a case where the relationship between the initial room temperature and the outside air temperature is constant, for example, it is possible to appropriately grasp changes in the performance of the air conditioner.

[0038] Furthermore, the determination unit 15a of this embodiment determines whether to issue a notification by comparing a result obtained by inputting, as input data D12, the time series of the initial value of room temperature and the outside air temperature (or a predetermined outside air temperature such as an initial value) and a time series of information indicating the operating state of the air conditioner contained in the actual time series accumulated for the second period (actual time series for the second period 132) into the first trained machine learning model 141 with the change in the actual room temperature value or the time corresponding to the change based on the actual time series accumulated for the second period (actual time series for the second period 132). With this configuration, there is no need to create a second trained machine learning model.

[0039] <Third embodiment> An air conditioner 1 according to a third embodiment of the present disclosure will be described with reference to FIGS. 10 to 12. FIG. 10 is a flowchart illustrating an example of the operation of a control device according to a third embodiment of the present disclosure. FIGS. 11 and 12 are schematic diagrams for explaining an example of the operation of a control device according to a third embodiment of the present disclosure. The air conditioner according to the third embodiment has the same basic configuration as the air conditioner 1 shown in FIG. 1. Note that the first and third embodiments differ partially in the operation of the model creation unit 14 and the determination unit 15 shown in FIG. 1. Furthermore, the flowchart shown in FIG. 10 differs from the flowchart shown in FIG. 3 in the content of step 16b, which corresponds to step S16 shown in FIG. 3.

[0040] Furthermore, in step S16b of FIG. 10, the determination unit 15 (see FIG. 1) of the third embodiment determines whether to issue a notification based on the value of a predetermined model parameter (parameter) 1411 included in the first trained machine learning model 141 and the value of a predetermined model parameter (parameter) 1412 included in the second trained machine learning model 142, as shown in FIG. 11. Note that, when creating the second trained machine learning model 142, the model creation unit 14 of the third embodiment may, as shown in FIG. 12, fix one or more other parameters (a1, b21, and w0) other than the predetermined parameter (b11) included in the machine learning model in the machine learning model shown in FIG. 2 to the same values ​​as the one or more other parameters (a1, b21, and w0) included in the first trained machine learning model 141, and create the second trained machine learning model 142 using the actual time series accumulated over the second period as training data D22. In this case, only the parameter (b11) changes due to machine learning.

[0041] The determination unit 15 can determine the degree of deviation between the parameter b11 included in the model parameter 1411 and the parameter b11 included in the model parameter 1412, for example, by determining whether the difference between the parameters is equal to or greater than a predetermined threshold value, or whether the sum of multiple normalized parameters is equal to or greater than a predetermined threshold value. The parameter b11 is a weight related to the heating capacity, and a decrease in b11 indicates a decrease in the capacity, which can be used to determine that dust has accumulated in the filter. Furthermore, if all parameters are learned, the effect of the reduced performance of the air conditioner would be dispersed across various parameters, but limiting the parameters to be learned can prevent this dispersion.

[0042] As described above, in this embodiment, the determination unit 15 determines whether to issue a predetermined notification using the first trained machine learning model 141 created using the actual time series accumulated in the first period (actual time series 131 for the first period) as training data D21, and some or all of the actual time series accumulated in the second period that follows the first period (actual time series 132 for the second period). With this configuration, even if the actual time series 131 for the first period and the actual time series 132 for the second period do not include data corresponding to a case where the relationship between the initial room temperature and the outside air temperature is constant, for example, it is possible to appropriately grasp changes in the performance of the air conditioner.

[0043] Furthermore, the model creation unit 14 of this embodiment creates a second trained machine learning model 142 using the actual time series accumulated in the second period (the actual time series 132 for the second period) as training data D22, and the determination unit 15 determines whether to issue a notification based on the values ​​of predetermined parameters included in the first trained machine learning model 141 and the values ​​of predetermined parameters included in the second trained machine learning model 142. According to this configuration, it is possible to determine whether to issue a notification at the stage when the machine learning model is created.

[0044] Furthermore, the model creation unit 14 of this embodiment can fix one or more parameters other than a predetermined parameter included in the machine learning model (e.g., weight parameter b11 related to air conditioning capacity) to the same values ​​as one or more other parameters included in the first trained machine learning model 141, and create the second trained machine learning model 142 using the actual time series accumulated over the second period as training data D22. With this configuration, for example, changes in air conditioning capacity can be reliably captured as changes in parameters.

[0045] (Action and effect) In the control system and control method configured as described above, the determination unit 15 or 15a determines whether to issue a predetermined notification using a first trained machine learning model 141 created using an actual time series accumulated in a first period (actual time series 131 for the first period) as training data D21, and some or all of an actual time series accumulated in a second period that follows the first period (actual time series 132 for the second period). With this configuration, even if the actual time series 131 for the first period and the actual time series 132 for the second period do not include data corresponding to a case where the relationship between the initial room temperature and the outside air temperature is constant, for example, it is possible to appropriately grasp changes in the performance of the air conditioner.

[0046] (Other embodiments) The above describes in detail the embodiments of the present disclosure with reference to the drawings, but the specific configuration is not limited to this embodiment, and design changes and the like are also included within the scope that does not deviate from the gist of the present disclosure.

[0047] <Computer Configuration> FIG. 13 is a schematic block diagram illustrating the configuration of a computer according to at least one embodiment. The computer 90 includes a processor 91 , a main memory 92 , a storage 93 , and an interface 94 . The above-described control devices 10 and 10a are implemented in a computer 90. The operations of the above-described processing units are stored in the form of a program in a storage 93. A processor 91 reads the program from the storage 93, loads it into a main memory 92, and executes the above-described processing in accordance with the program. The processor 91 also allocates storage areas in the main memory 92 corresponding to the above-described storage units in accordance with the program.

[0048] The program may be for realizing some of the functions to be performed by the computer 90. For example, the program may be combined with other programs already stored in storage or other programs implemented in other devices to perform the functions. In other embodiments, the computer may include a custom LSI (Large Scale Integrated Circuit) such as a PLD (Programmable Logic Device) in addition to or instead of the above configuration. Examples of PLDs include PAL (Programmable Array Logic), GAL (Generic Array Logic), CPLD (Complex Programmable Logic Device), and FPGA (Field Programmable Gate Array). In this case, some or all of the functions realized by the processor may be realized by the integrated circuit.

[0049] Examples of storage 93 include a hard disk drive (HDD), a solid state drive (SSD), a magnetic disk, a magneto-optical disk, a compact disc read-only memory (CD-ROM), a digital versatile disc read-only memory (DVD-ROM), and a semiconductor memory. Storage 93 may be an internal medium directly connected to the bus of computer 90, or an external medium connected to computer 90 via interface 94 or a communication line. Furthermore, when this program is distributed to computer 90 via a communication line, computer 90 that receives the program may load the program into main memory 92 and execute the above-described processing. In at least one embodiment, storage 93 is a non-transitory tangible storage medium.

[0050] <Additional Notes> The control system (control device 10, 10a) described in each embodiment can be understood, for example, as follows.

[0051] (1) A control system according to a first aspect includes a model creation unit 14, 14a that receives as input data an initial value of the room temperature at the installation location of the air conditioner 1, a time series of the outside air temperature or a predetermined outside air temperature, and a time series of information indicating the operating state of the air conditioner, and creates a trained machine learning model that calculates a change in the room temperature due to the operation of the air conditioner or a time based on that change by machine learning using the time series of the room temperature and the time series of the outside air temperature or the predetermined outside air temperature and the time series of the information as training data; and a storage unit 13 that stores, as an actual time series, a time series of the actual values ​​of the outside air temperature or a time series of the actual values ​​of the predetermined outside air temperature and the information, and a first trained machine learning model 141 created using the actual time series stored in a first period (actual time series 131 for the first period) as the teacher data, and a determination unit 15, 15a that determines whether to issue a predetermined notification using part or all of the actual time series stored in a second period after the first period (actual time series 132 for the second period). According to this aspect and each of the following aspects, it is possible to appropriately grasp changes in the performance of the air conditioner.

[0052] (2) A control system according to a second aspect is the control system according to (1), in which the predetermined notification is a notification related to cleaning of a filter of the air conditioner.

[0053] (3) A control system according to a second aspect is the control system of (1) or (2), wherein the operating state includes an operating state of a compressor of the air conditioner.

[0054] (4) A control system according to a fourth aspect is a control system according to (1) to (3), wherein the model creation unit creates a second trained machine learning model using the actual time series accumulated over the second period as the training data, and the determination unit determines whether to issue the notification based on the result obtained by inputting a specific input data into the first trained machine learning model and the result obtained by inputting the same input data into the second trained machine learning model.

[0055] (5) A control system according to a fifth aspect is a control system according to any one of (1) to (3), wherein the determination unit determines whether to issue the notification by comparing the result obtained by inputting the initial value of the room temperature and the time series of the outside temperature or the time series of the specified outside temperature and the information contained in the actual time series accumulated over the second period as the input data into the first trained machine learning model with the change in the actual value of the room temperature based on the actual time series accumulated over the second period or the time corresponding to the change.

[0056] (6) A control system according to a sixth aspect is a control system according to any one of (1) to (3), wherein the model creation unit creates a second trained machine learning model using the actual time series accumulated over the second period as the training data, and the determination unit determines whether to issue the notification based on the value of a predetermined parameter included in the first trained machine learning model and the value of the predetermined parameter included in the second trained machine learning model.

[0057] (7) A control system according to a seventh aspect is a control system according to (6), wherein the model creation unit fixes one or more other parameters other than the predetermined parameters included in the machine learning model to the same values ​​as one or more other parameters included in the first trained machine learning model, and creates the second trained machine learning model using the actual time series accumulated over the second period as the training data.

[0058] (8) A control system according to an eighth aspect is the control system of (1) to (6), wherein the input data includes data indicating an environmental condition of a location where the air conditioner is installed. [Explanation of symbols]

[0059] 1, 1a…Air conditioner 10, 10a...Control device (control system) 13...Storage section 131...First period performance time series 132...2nd period performance time series 14, 14a...Model creation section 141…First trained machine learning model 142…Second trained machine learning model 15, 15a...judgment section 16…Notification section

Claims

1. Inputting an initial value of the room temperature at the installation location of the air conditioner, a time series of the outdoor temperature or a predetermined outdoor temperature, and a time series of information indicating the operating state of the air conditioner as input data; A trained machine learning model that determines a change in room temperature due to operation of the air conditioner or a time based on that change, a model creation unit that creates a model by machine learning using the time series of the room temperature and the time series of the outside air temperature or the predetermined outside air temperature and the time series of the information as training data; a storage unit that stores a time series of the actual value of the room temperature and a time series of the actual value of the outside air temperature, or a time series of the actual value of the predetermined outside air temperature and an actual value of the information, as an actual time series; a determination unit that determines whether to issue a predetermined notification using a first trained machine learning model created using the performance time series accumulated in a first period as the training data and a part or all of the performance time series accumulated in a second period that is later than the first period; A control system comprising:

2. The predetermined notification is a notification regarding cleaning of the filter of the air conditioner. The control system of claim 1 .

3. The operating state includes an operating state of a compressor of the air conditioner. The control system of claim 2 .

4. the model creation unit creates the second trained machine learning model using the performance time series accumulated in the second period as the training data; The determination unit determines whether to issue the notification based on a result obtained by inputting predetermined input data into the first trained machine learning model and a result obtained by inputting the same input data into the second trained machine learning model. A control system according to any one of claims 1 to 3.

5. The determination unit determines whether to issue the notification by comparing a result obtained by inputting, as the input data, the initial value of the room temperature and the time series of the outside air temperature or the time series of the predetermined outside air temperature and the information included in the actual time series accumulated over the second period into the first trained machine learning model with a change in the actual value of the room temperature based on the actual time series accumulated over the second period or a time corresponding to the change. A control system according to any one of claims 1 to 3.

6. the model creation unit creates the second trained machine learning model using the performance time series accumulated in the second period as the training data; The determination unit determines whether to issue the notification based on a value of a predetermined parameter included in the first trained machine learning model and a value of the predetermined parameter included in the second trained machine learning model. A control system according to any one of claims 1 to 3.

7. The model creation unit fixes one or more other parameters other than the predetermined parameter included in the machine learning model to the same values ​​as the one or more other parameters included in the first trained machine learning model, and creates the second trained machine learning model using the performance time series accumulated in the second period as the training data. The control system of claim 6.

8. The input data includes data indicating the environmental conditions of the installation location of the air conditioner. A control system according to any one of claims 1 to 3.

9. Inputting an initial value of the room temperature at the installation location of the air conditioner, a time series of the outdoor temperature or a predetermined outdoor temperature, and a time series of information indicating the operating state of the air conditioner as input data; A trained machine learning model that determines a change in room temperature due to operation of the air conditioner or a time based on that change, creating the information by machine learning using the time series of the room temperature and the time series of the outside air temperature or the predetermined outside air temperature and the time series of the information as training data; accumulating a time series of the actual values ​​of the room temperature and the time series of the actual values ​​of the outside air temperature or a time series of the actual values ​​of the predetermined outside air temperature and the actual values ​​of the information as an actual time series; a step of determining whether to issue a predetermined notification using a first trained machine learning model created using the performance time series accumulated in a first period as the training data and a part or all of the performance time series accumulated in a second period that is later than the first period; A control method comprising:

Images