JP7483122B2 - Air conditioning system and learning device - Google Patents

Description

The present disclosure relates to an air conditioning system.

The facility management device described in Patent Document 1 creates conditions for changing the set temperature of the air conditioning equipment in accordance with preset rules if it determines whether there is any regularity in the contents of the candidate data for creating operation conditions when changing the set temperature of the air conditioning equipment. This facility management device creates operation instruction information for changing the set temperature of the air conditioning equipment and transmits the operation instruction information to the air conditioning equipment.

JP 2016-45567 A

However, while the facility management device described in Patent Document 1 reduces the workload of the operator or user, it does not improve the comfort of air conditioning.

Therefore, the object of the present disclosure is to provide an air conditioning system and a learning device that can improve air conditioning comfort.

The air conditioning system disclosed herein comprises an air conditioning device including an outdoor unit and at least one indoor unit, a first inference device that infers the earliest arrival time for the day based on the earliest arrival time for each day for the day of the week, the month, and a certain period of time in the past, a second inference device that infers the thermal load required to set the room temperature to a set temperature at the earliest arrival time for the day, and a control device that controls the air conditioning device based on the thermal load.

The air conditioning system disclosed herein infers the earliest arrival time for the day based on the earliest arrival time for each day for the current day of the week, the current month, and a certain period of time in the past, and infers the thermal load required to bring the room temperature to the set temperature at the earliest arrival time for the day.

The learning device disclosed herein includes a data acquisition unit that acquires first learning data including a combination of input data including the earliest arrival time at work for the current day's day of the week, the current month, and for each day over a certain period of time in the past, and teacher data including the earliest arrival time at work for the current day, and a model generation unit that uses the first learning data to generate a first trained model for inferring the earliest arrival time at work for the current day from the earliest arrival time at work for the current day's day of the week, the current month, and for each day over a certain period of time in the past.

The learning device disclosed herein includes a data acquisition unit that acquires second learning data including input data including the outdoor temperature at the operation start time of the indoor unit of each zone, the room temperature at the operation start time of each zone, the compressor frequency at the operation start time, and the set temperature of each zone, and teacher data including the thermal load for bringing the room temperature of each zone to the set temperature, and a model generation unit that uses the second learning data to generate a second learned model for each zone that estimates the thermal load for bringing the room temperature of the corresponding zone to the set temperature from the outdoor temperature at the operation start time of the indoor unit of the corresponding zone, the room temperature at the operation start time of the corresponding zone, the compressor frequency at the operation start time, and the set temperature of the corresponding zone.

This enables the air conditioning system and learning device disclosed herein to improve air conditioning comfort.

1 is a diagram showing the configuration of an air conditioning system according to a first embodiment. 1 is a diagram showing the configuration of an air conditioning device 1. FIG. A figure showing an example of input data (B1) and output data (teacher data) B2 of the first trained model in embodiment 1. FIG. 2 is a diagram showing a configuration of a first learning device 3. 4 is a flowchart showing a procedure of a learning process of a first learning device 3. 2 is a diagram showing the configuration of a first inference device 4. 4 is a flowchart showing the procedure of an inference process of a first inference device 4. 13 is a diagram showing the configuration of a second learning device 6. FIG. 10 is a flowchart showing a procedure of a learning process of a second learning device 6. 2 is a diagram showing the configuration of a second inference device 7. FIG. 11 is a flowchart showing the procedure of an inference process of a second inference device 7. FIG. 11 is a diagram showing the configuration of an air conditioning system according to a second embodiment. 1 is a diagram showing the configuration of an air conditioning device 1a. 13A is a diagram showing an example of input data (B1) and output data (teacher data) B2 of a first trained model for zone A in embodiment 2. FIG. 13B is a diagram showing an example of input data (B1) and output data (teacher data) B2 of a first trained model for zone B in embodiment 2. FIG. 2 is a diagram showing the configuration of a first learning device 3a. 10 is a flowchart showing a procedure of a learning process of a first learning device 3a. FIG. 2 is a diagram showing the configuration of a first inference device 4a. 11 is a flowchart showing the procedure of an inference process of a first inference device 4a. 13A is a diagram showing an example of input data (B1) and output data (teacher data) B2 of a second trained model for zone A in embodiment 2. FIG. 13B is a diagram showing an example of input data (B1) and output data (teacher data) B2 of a second trained model for zone B in embodiment 2. FIG. 2 is a diagram showing the configuration of a second learning device 6a. 10 is a flowchart showing a procedure for creating second learning data for zone A by a second learning device. 10 is a flowchart showing a procedure of a learning process of a second learning device 6a. FIG. 2 shows the configuration of a second inference device 7a. 11 is a flowchart showing the procedure of an inference process of a second inference device 7a. 13A is a diagram showing an example of input data (B1) and output data (teacher data) B2 of a first trained model for zone A in embodiment 3. FIG. 13B is a diagram showing an example of input data (B1) and output data (teacher data) B2 of a first trained model for zone B in embodiment 3. FIG. 13 is a diagram showing an example of management data. 13 is a flowchart showing a processing procedure when changing seats. FIG. 13 is a diagram showing the configuration of an air conditioning system according to a modified example. 29 is a block diagram showing an example of the configuration of the air conditioning control device and the information processing device shown in FIG. 28. FIG. 30 is a diagram for explaining the control performed by the air conditioning control means shown in FIG. 29 on the air conditioner. FIG. 30 is a diagram illustrating an example of an operation procedure of a thermal load learning means and a schedule determination means illustrated in FIG. 29 . FIG. 30 is a diagram illustrating an example of another operation procedure of the thermal load learning means and the schedule determination means illustrated in FIG. 29 . 32 is a flowchart showing a machine learning procedure performed by the thermal load learning means shown in FIG. 31 . 4 is a flowchart showing an operation procedure of the information processing device 10. FIG. 11 is a diagram illustrating an example in which an air conditioning control device controls a compressor and an expansion device according to a schedule. A diagram showing the hardware configurations of a first learning device, a second learning device, a first inference device, a second inference device, and a control device.

Hereinafter, embodiments will be described with reference to the drawings.
Embodiment 1.
FIG. 1 is a diagram showing the configuration of an air conditioning system according to a first embodiment.

The air conditioning system comprises an air conditioning unit 1, a control unit 2, a first learning device 3, a first learned model memory device 5, a first inference device 4, a second learning device 6, a second inference device 7, and a second learning result memory device 8.

The air conditioning device 1 includes an outdoor unit and an indoor unit.
The first learning device 3 creates a first learned model.

The first learned model memory device 5 stores the first learned model generated by the first learning device 3.

The first inference device 4 uses the first learned model to infer the earliest arrival time for the current day based on the earliest arrival time for each day for the current day of the week, the current month, and the past week.

The second learning device 6 creates a second learning result.
The second learning result storage device 8 stores the second learning result.

The second inference device 7 uses the second learning result to infer the thermal load required to bring the room temperature to the set temperature at the earliest arrival time at work on that day inferred by the first inference device 4.

FIG. 2 is a diagram showing the configuration of the air conditioning device 1.
The air-conditioning apparatus 1 includes a heat source unit 50 which is an outdoor unit, and a load unit 40a which is an indoor unit.

The heat source side unit 50 has a compressor 51 that compresses and discharges the refrigerant, a heat source side heat exchanger 52 that exchanges heat between the outside air and the refrigerant, and a four-way valve 53 that switches the flow direction of the refrigerant according to the operating mode.

The load side unit 40a has a load side heat exchanger 41a in which heat is exchanged between the refrigerant and the air in the room that is the space to be air-conditioned by the load side unit 40a, and an expansion device 42a that reduces the pressure of the high-pressure refrigerant and expands it.

The load unit 40a is provided with a space temperature sensor 43a that detects the room temperature. The heat source unit 50 is provided with an outside air temperature sensor 54 that detects the outside air temperature.

Compressor 51 is, for example, an inverter type compressor whose capacity can be changed by changing the operating frequency.

The expansion device 42a is, for example, an electronic expansion valve.
The heat source side heat exchanger 52 and the load side heat exchanger 41a are, for example, fin-and-tube type heat exchangers. In the heat source side unit 50 and the load side unit 40a, the compressor 51, the heat source side heat exchanger 52, the expansion device 42a, and the load side heat exchanger 41a are connected to form a refrigerant circuit 60a in which the refrigerant circulates.

The first learned model storage device 5 stores the first learned model.
FIG. 3 is a diagram showing an example of input data (B1) and output data (teacher data) B2 of the first trained model in embodiment 1.

The input data (B1) includes the day of the week, the month, and the earliest arrival time for each day in the past week. The output data (teacher data) B2 includes the earliest arrival time for the current day.

FIG. 4 is a diagram showing the configuration of the first learning device 3. As shown in FIG.
The first learning device 3 includes a first learning data creation unit 61 , a first learning data storage unit 62 , a data acquisition unit 63 , and a model generation unit 64 .

The first learning data creation unit 61 creates multiple first learning data including a combination of input data (B1) including the earliest arrival time at work for the current day, the current month, and each day of the past week, and teacher data (B2) including the earliest arrival time at work for the current day, according to data obtained from the control device 2.

The first learning data storage unit 62 stores multiple first learning data created by the first learning data creation unit 61.

The data acquisition unit 63 acquires multiple first learning data from the first learning data storage unit 62.

The model generation unit 64 uses multiple first learning data to generate a first learned model for inferring the earliest arrival time at work on the current day from the earliest arrival time for each day on the current day of the week, the current month, and the past week.

FIG. 5 is a flowchart showing the procedure of the learning process of the first learning device 3.
In step S101, the data acquisition unit 63 acquires a plurality of first learning data from the first learning data storage unit 62. The first learning data includes a combination of the input data (B1) and the teacher data (B2) shown in Fig. 3. The data acquisition unit 63 acquires the input data (B1) and the teacher data (B2) simultaneously, but it is sufficient that the input data (B1) and the teacher data (B2) are input in association with each other, and the data acquisition unit 63 may acquire the input data (B1) and the teacher data (B2) at different times.

In step S102, the model generation unit 64 generates a first trained model in which teacher data (B2) is output when input data (B1) is input, by so-called supervised learning, in accordance with a plurality of first training data including a combination of input data (B1) and teacher data (B2) acquired by the data acquisition unit 63.

In step S103, the first learned model storage device 5 stores the first learned model generated by the model generation unit 64.

FIG. 6 is a diagram showing the configuration of the first inference device 4. As shown in FIG.
The first inference device 4 includes a data acquisition unit 65 and an estimation unit 66 .

The data acquisition unit 65 acquires from the control device 2 the day of the week, the month, and the earliest arrival time for each day of the past week.

The estimation unit 66 uses a first learned model for inferring the earliest work arrival time of the day from the earliest work arrival time of the day for the day, the month for the day, and each day for the past week stored in the first learned model storage device 5, and outputs the earliest work arrival time of the day (B2) from the earliest work arrival time of the day for the day, the month for the day, and each day for the past week (B1) input from the data acquisition unit 65. That is, the estimation unit 66 can output output data (B2) inferred from the input data (B1) by inputting the input data (B1) acquired by the data acquisition unit 65 to the first learned model.

FIG. 7 is a flow chart showing the procedure of the inference process of the first inference device 4.
In step S201, the data acquisition unit 65 acquires the input data (B1) shown in FIG.

In step S202, the estimation unit 66 inputs input data (B1) to the first learned model stored in the first learned model memory device 5 and obtains output data (B2).

In step S203, the estimation unit 66 outputs the output data (B2) obtained by the first learned model to the control device 2.

FIG. 8 is a diagram showing the configuration of the second learning device 6. As shown in FIG.
The second learning device 6 includes a setting unit 67 , a data acquisition unit 68 , and a coefficient calculation unit 69 .

The setting unit 67 sets the start time t0, the room temperature setting temperature Ts, and the frequency f0 of the compressor 51.

The data acquisition unit 68 acquires the outside air temperature T0, the time t1, and the like.
The coefficient calculation unit 69 calculates a time coefficient K.

FIG. 9 is a flow chart showing the procedure of the learning process of the second learning device 6.
In step S301, the setting unit 67 sets the current time to the start time t0, sets the set temperature Ts of the room, and sets the frequency f0 of the compressor 51.

In step S302, the data acquisition unit 68 acquires the outside air temperature T0 at time t0 from the control device 2.

In step S303, the setting unit 67 instructs the control device 2 to start operation of the air conditioning device 1 at time t0. The control device 2 starts operation of the air conditioning device 1.

In step S304, when the current room temperature Tr becomes equal to the set temperature Ts, the process proceeds to step S305.
In step S305, the data acquisition unit 68 acquires the current time t1.

In step S306, the coefficient calculation unit 69 calculates the time coefficient K according to the following formula. The coefficient calculation unit 69 stores the time coefficient K in the second learning result storage device 8.

K = (t1-t0)/(Ts-T0)…(A1)
The time coefficient K obtained by performing the above steps S301 to S306 a plurality of times may be averaged and stored in the second learning result storage device 8.

FIG. 10 is a diagram showing the configuration of the second inference device 7. As shown in FIG.
The second inference device 7 includes a data acquisition unit 71 and an optimal operation start time estimation unit 72 .

The data acquisition unit 71 acquires from the control device 2 the earliest arrival time of the day, the room temperature setting at the earliest arrival time of the day, and the outside air temperature of the air conditioning unit.

The optimal operation start time estimation unit 72 estimates the optimal operation start time of the air conditioning unit 1 based on the earliest arrival time at work on that day and the room temperature setting at the earliest arrival time at work on that day.

FIG. 11 is a flow chart showing the procedure of the inference process of the second inference device 7.
In step S401, the data acquisition unit 71 acquires the earliest work arrival time tx of the day from the control device 2. The earliest work arrival time tx is data inferred by the first inference device 4. The data acquisition unit 71 acquires the room temperature setting Ts at the time tx from the control device 2. The room temperature setting Ts may be fixed, or may be set by the user via a remote control and held by the control device 2. Alternatively, the room temperature setting Ts may be changed automatically depending on the season.

In step S402, the data acquisition unit 71 acquires the time coefficient K from the second learning result memory device 8.

In step S403, a time that is a certain time α earlier than time tx is set as the tentative operation start time ty of the air conditioning unit 1.

In step S404, the data acquisition unit 71 acquires the outside air temperature Ty at the tentative operation start time ty from the control device 2.

In step S405, the optimal operation start time estimation unit 72 calculates the required time Ton according to the following formula.

Ton = (Ts-Ty) × k…(A2)
In step S406, the optimum operation start time estimation unit 72 calculates a time that is later than the tentative operation start time ty by the necessary time Ton as the set temperature reaching time tz.

In step S407, if the difference between the earliest work arrival time tx on that day and the set temperature arrival time tz is less than the threshold TH1, processing proceeds to step S409. If the difference between the earliest work arrival time tx on that day and the set temperature arrival time tz is equal to or greater than the threshold TH1, processing proceeds to step S408.

In step S408, the optimal operation start time estimation unit 72 delays the tentative operation start time ty by a certain time Δt. Then, the process returns to step S404.

In step S409, the optimal operation start time estimation unit 72 sets the currently set tentative operation start time as the optimal operation start time of the air conditioning device 1. The control device 2 starts operation of the air conditioning device 1 at the set optimal operation start time.

The air conditioning system of this embodiment infers the earliest arrival time for the day based on the earliest arrival time for each day for the day of the week, the month, and a certain period of time in the past, and infers the heat load for bringing the room temperature to the set temperature at the earliest arrival time for the day. This allows the air conditioning system of this embodiment to improve the comfort of air conditioning.

Embodiment 2.
FIG. 12 is a diagram showing the configuration of an air conditioning system according to the second embodiment.

The air conditioning system comprises an air conditioning unit 1a, a control unit 2a, a first learning device 3a, a first inference device 4a, a first learned model memory device 5a, a second learning device 6a, a second inference device 7a, and a second learned model memory device 8a.

The air conditioning unit 1a includes an outdoor unit and a number of indoor units, each of which is provided for a different zone.

The first learning device 3a creates a first learned model.
The first learned model storage device 5a stores the first learned model generated by the first learning device 3a.

The first inference device 4a uses the first learned model to infer the earliest arrival time for each zone on that day based on the day of the week, the month, and the earliest arrival time for each day in the past week for each zone.

The second learning device 6a generates a second learned model.
The second learned model storage device 8a stores the second learned model.

The second inference device 7a uses the second learned model to set the operation start time of the indoor unit of each zone to a time a certain time before the earliest arrival time of each zone inferred by the first inference device 4, and infers the thermal load for bringing the room temperature of each zone to the set temperature based on the outside air temperature at the operation start time of the indoor unit of each zone, the room temperature at the operation start time of the indoor unit of each zone, the compressor frequency at the operation start time of the indoor unit of each zone, and the set temperature of each zone.

FIG. 13 is a diagram showing the configuration of an air conditioning device 1a.
The air conditioner 1a further includes a load-side unit 40b, which is an outdoor unit, in addition to the components of the air conditioner 1 of embodiment 1. The load-side unit 40a is installed in zone A, and the load-side unit 40b is installed in zone B.

The load side unit 40b has a load side heat exchanger 41b in which the refrigerant exchanges heat with the air in the room that is the target space for air conditioning of the load side unit 40b, and an expansion device 42b that reduces the pressure of the high-pressure refrigerant and expands it. The load side unit 40b is provided with a space temperature sensor 43b that detects the room temperature.

The expansion device 42b is, for example, an electronic expansion valve.
The load-side heat exchanger 41b is, for example, a fin-and-tube heat exchanger. In the heat source-side unit 50 and the load-side unit 40b, the compressor 51, the heat source-side heat exchanger 52, the expansion device 42b, and the load-side heat exchanger 41b are connected to each other to form a refrigerant circuit 60b in which the refrigerant circulates.

The first learned model memory device 5a stores a first learned model for zone A and a second learned model for zone B.

Figure 14 (a) is a diagram showing an example of input data (B1) and output data (teacher data) B2 of the first trained model for zone A in embodiment 2. The input data (B1) includes the day of the week, the month of the day, and the earliest start time of work for each day in the past week in zone A. The output data (B2) includes the earliest start time of work for the day in zone A.

Figure 14 (b) is a diagram showing an example of input data (B1) and output data (teacher data) B2 of the first trained model for zone B in embodiment 2. The input data (B1) includes the day of the week, the month of the day, and the earliest start time of work for each day in the past week in zone B. The output data (B2) includes the earliest start time of work for zone B on the day.

FIG. 15 is a diagram showing the configuration of a first learning device 3a.
The first learning device 3a includes a first learning data creation unit 61a, a first learning data storage unit 62a, a data acquisition unit 63a, and a model generation unit 64a.

The first learning data creation unit 61a creates multiple first learning data for zone A, including a combination of input data (B1) including the current day of the week, the current month, and the earliest arrival time for each day in the past week in zone A, and teacher data (B2) including the earliest arrival time for the current day in zone A, according to data obtained from the control device 2.

The first learning data creation unit 61a creates multiple first learning data for zone B, including a combination of input data (B1) including the current day of the week, the current month, and the earliest arrival time for each day in the past week in zone B, and teacher data (B2) including the earliest arrival time for the current day in zone B, according to data obtained from the control device 2.

The first learning data storage unit 62a stores a plurality of first learning data for zone A and a plurality of first learning data for zone B created by the first learning data creation unit 61a.

The data acquisition unit 63a acquires from the first learning data storage unit 62a a plurality of first learning data including a combination of input data (B1) including the day of the week, the month, and the earliest arrival time for each day in the past week for each zone, and teacher data (B2) including the earliest arrival time for each zone for the day. That is, the data acquisition unit 63a acquires a plurality of first learning data for zone A and a plurality of first learning data for zone B.

The model generation unit 64a uses the first learning data to generate a first learned model for each zone for inferring the earliest work arrival time of the day in the corresponding zone from the earliest work arrival time of each day in the past week for the day, the month, and the corresponding zone. That is, the model generation unit 64a uses a plurality of first learning data for zone A to generate a first learned model for zone A for inferring the earliest work arrival time of the day in zone A from the earliest work arrival time of the day in the day, the month, and the corresponding zone. The model generation unit 64a uses a plurality of first learning data for zone B to generate a first learned model for zone B for inferring the earliest work arrival time of the day in zone B from the earliest work arrival time of the day in the day, the month, and the corresponding zone B.

FIG. 16 is a flowchart showing the procedure of the learning process of the first learning device 3a.
In step S601, the data acquiring unit 63a acquires a plurality of first learning data for zone A from the first learning data storage unit 62a. The first learning data for zone A includes a combination of input data (B1) and teacher data (B2) shown in Fig. 14(a). The data acquiring unit 63a is configured to acquire the input data (B1) and the teacher data (B2) simultaneously, but it is sufficient that the input data (B1) and the teacher data (B2) are input in association with each other, and the data acquiring unit 63a may acquire the input data (B1) and the teacher data (B2) at different times.

In step S602, the model generation unit 64a generates a first learned model for zone A in which teacher data (B2) is output when input data (B1) is input, by so-called supervised learning, in accordance with a plurality of first learning data for zone A including a combination of input data (B1) and teacher data (B2) acquired by the data acquisition unit 63a.

In step S603, the first learned model storage device 5a stores the first learned model for zone A generated by the model generation unit 64a.

In step S604, the data acquisition unit 63a acquires a plurality of first learning data for zone B from the first learning data storage unit 62a. The first learning data for zone B includes a combination of input data (B1) and teacher data (B2) shown in FIG. 14(b). The data acquisition unit 63a acquires the input data (B1) and the teacher data (B2) simultaneously, but it is sufficient that the input data (B1) and the teacher data (B2) are input in association with each other, and the data acquisition unit 63a may acquire the input data (B1) and the teacher data (B2) at different times.

In step S605, the model generation unit 64a generates a first learned model for zone B in which teacher data (B2) is output when input data (B1) is input, by so-called supervised learning, according to a plurality of first learning data for zone B including a combination of input data (B1) and teacher data (B2) acquired by the data acquisition unit 63a.

In step S606, the first learned model memory device 5a stores the first learned model for zone B generated by the model generation unit 64a.

FIG. 17 is a diagram showing the configuration of a first inference device 4a.
The first inference device 4a includes a data acquisition unit 65a and an estimation unit 66a.

The data acquisition unit 65a acquires the earliest arrival time for each day of the current week for the current day, the current month, and each zone for the past week. That is, the data acquisition unit 65a acquires input data (B1) for zone A shown in FIG. 14(a) and input data (B1) for zone B shown in FIG. 14(b) from the control device 2.

The estimation unit 66a uses a first learned model for each zone for inferring the earliest arrival time of the corresponding zone on the day from the earliest arrival time of each day of the past week for the day, the month of the day, and the corresponding zone stored in the first learned model storage device 5a, and outputs the earliest arrival time of the day for each zone from the earliest arrival time of the day for the day, the month of the day, and each day of the past week for each zone input from the data acquisition unit 65a. The estimation unit 66a infers output data (B2) obtained by using the first learned model for zone A. That is, the estimation unit 66a can output output data (B2) for zone A inferred from the input data (B1) for zone A by inputting the input data (B1) for zone A acquired by the data acquisition unit 65a into the first learned model for zone A. The estimation unit 66a infers output data (B2) obtained by using the first learned model for zone B. That is, the estimation unit 66a can output output data (B2) for zone B inferred from the input data (B1) for zone B by inputting the input data (B1) for zone B acquired by the data acquisition unit 65a into the first learned model for zone B.

FIG. 18 is a flow chart showing the procedure of the inference process of the first inference device 4a.
In step S701, the data acquisition unit 65a acquires the input data (B1) for zone A shown in FIG. 14(a).

In step S702, the estimation unit 66a inputs input data (B1) for zone A to a first learned model for zone A stored in the first learned model memory device 5a, and obtains output data (B2) for zone A.

In step S703, the estimation unit 66a outputs output data (B2) for zone A obtained by the first learned model for zone A to the control device 2.

In step S704, the data acquisition unit 65a acquires input data (B1) for zone B shown in FIG. 14 (b).

In step S705, the estimation unit 66a inputs input data (B1) for zone B to a first learned model for zone B stored in the first learned model memory device 5a, and obtains output data (B2) for zone B.

In step S706, the estimation unit 66a outputs output data (B2) for zone B obtained by the first learned model for zone B to the control device 2.

Figure 19 (a) is a diagram showing an example of input data (B1) and output data (teacher data) B2 of the second trained model for zone A in embodiment 2. The input data (B1) includes the outdoor air temperature at the operation start time of the indoor unit of zone A, the room temperature at the operation start time of the indoor unit of zone A, the compressor frequency at the operation start time of the indoor unit of zone A, and the set temperature of zone A. The output data (B2) includes the thermal load for setting the room temperature of zone A to the set temperature.

Figure 19 (b) is a diagram showing an example of input data (B1) and output data (teacher data) B2 of the second trained model for zone B in embodiment 2. The input data (B1) includes the outdoor air temperature at the operation start time of the indoor unit of zone B, the room temperature at the operation start time of the indoor unit of zone B, the compressor frequency at the operation start time of the indoor unit of zone B, and the set temperature of zone B. The output data (B2) includes the thermal load for setting the room temperature of zone B to the set temperature.

FIG. 20 is a diagram showing the configuration of the second learning device 6a.
The second learning device 6 a includes a second learning data creation unit 73 , a second learning data storage unit 74 , a data acquisition unit 75 , and a model generation unit 76 .

The second learning data creation unit 73 creates second learning data for zone A and second learning data for zone B according to the data obtained from the control device 2.

The second learning data storage unit 74 stores a plurality of second learning data for zone A including a combination of input data (B1) and teacher data (B2) for zone A shown in FIG. 19(a), and a plurality of second learning data for zone B including a combination of input data (B1) and teacher data (B2) for zone B shown in FIG. 19(b).

The data acquisition unit 75 acquires second learning data including input data including the outdoor temperature at the operation start time of the indoor unit of each zone, the room temperature at the operation start time of the indoor unit of each zone, the compressor frequency at the operation start time of the indoor unit of each zone, and the set temperature of each zone, and teacher data including the heat load for setting the room temperature of each zone to the set temperature. That is, the data acquisition unit 75 acquires multiple second learning data for zone A and multiple second learning data for zone B from the second learning data storage unit 74.

The model generation unit 76 uses the second learning data to generate a second trained model for each zone that estimates the heat load for setting the room temperature of the corresponding zone to the set temperature from the outdoor air temperature at the operation start time of the indoor unit of the corresponding zone, the room temperature at the operation start time of the indoor unit of the corresponding zone, the compressor frequency at the operation start time of the indoor unit of the corresponding zone, and the set temperature of the corresponding zone. That is, the model generation unit 76 generates a second trained model for zone A according to the second learning data for zone A, and generates a second trained model for zone B according to the second learning data for zone B.

Figure 21 is a flowchart showing the procedure for creating second learning data for zone A by the second learning device. The procedure for creating second learning data for zone B by the second learning device is similar.

In step S500, the second learning data creation unit 73 sets the control variable i = 0.

In step S501, the second learning data creation unit 73 obtains from the control device 2 the outdoor air temperature Tout at the start of operation of the indoor unit in zone A, the room temperature Tr(0) of zone A at the start of operation of the indoor unit in zone A, the compressor frequency f0 at the start of operation of the indoor unit in zone A, and the set temperature of zone A. The control device 2 starts operation of the heat source side unit 50, which is the outdoor unit, and the load side unit 40a, which is the indoor unit in zone A. The frequency f0 of the compressor 51 of the heat source side unit 50 is fixed.

In step S502, when time Δt has elapsed, processing proceeds to step S503.

In step S503, the second learning data creation unit 73 increments i by 1.

In step S504, the second learning data creation unit 73 obtains the room temperature Tr(i) of zone A.

In step S505, the second learning data creation unit 73 calculates the air conditioning heat quantity QHVAC(i) at time Δt according to the following formula.

aQHVAC(i)=bQr+(Tout-Tr(i))/Rwin-C(Tr(i)-Tr(i-1))/Δt+α…(B1)
where a and b are coefficients. C is the indoor heat capacity [kJ/K], and α is the amount of heat due to other heat load influencing factors. The air conditioning heat quantity QHVAC(i) is the amount of heat removed in cooling operation, and the amount of heat supplied in heating operation. Rwin is the window thermal conductivity [kw/K]. The indoor heat generation quantity Qr can be set to a fixed value, for example 0, since no one is present in the room. Alternatively, since the air conditioner is operating, the indoor heat generation quantity Qr is an amount that depends on the compressor frequency f0.

In step S506, if the room temperature of zone A is equal to the set temperature of zone A, the process proceeds to step S507. If the room temperature Tr(i) of zone A is not equal to the set temperature of zone A, the process returns to step S502.

In step S507, the second learning data creation unit 73 calculates the thermal load Q required to set the room temperature of zone A to the set temperature according to the following formula:

Q = QHVAC(1) + QHVAC(2) + ... + QHVAC(i) ... (B2)
In step S508, the second learning data creation unit 73 sets the outdoor air temperature Tout at the start of operation of the indoor unit of zone A, the room temperature Tr(0) of zone A at the start of operation of the indoor unit of zone A, the compressor frequency at the start of operation of the indoor unit of zone A, and the set temperature of zone A acquired in step S501 as the outdoor air temperature at the start of operation of the indoor unit of zone A, the room temperature at the start of operation of the indoor unit of zone A, the compressor frequency at the start of operation of the indoor unit of zone A, and the set temperature of zone A of the input data (B1) of the second learning data for zone A. The second learning data creation unit 73 sets the heat load Q calculated in step S508 as the heat load for setting the room temperature of zone A to the set temperature of the teacher data (B2) of the second learning data for zone A. The second learning data creation unit 73 stores the second learning data in the second learning data storage unit 74.

FIG. 22 is a flowchart showing the procedure of the learning process of the second learning device 6a.
In step S801, the data acquiring unit 75 acquires second learning data for zone A from the second learning data storage unit 74. The second learning data for zone A includes a combination of input data (B1) and teacher data (B2) shown in Fig. 19(a). The data acquiring unit 75 acquires the input data (B1) and the teacher data (B2) simultaneously, but it is sufficient that the input data (B1) and the teacher data (B2) are input in association with each other, and the data acquiring unit 75 may acquire the input data (B1) and the teacher data (B2) at different times.

In step S802, the model generation unit 76 generates a second trained model for zone A in which teacher data (B2) is output when input data (B1) is input, by so-called supervised learning, according to second learning data for zone A including a combination of input data (B1) and teacher data (B2) acquired by the data acquisition unit 75.

In step S803, the second learned model storage device 8a stores the second learned model for zone A generated by the model generation unit 76.

In step S804, the data acquisition unit 75 acquires second learning data for zone B from the second learning data storage unit 74. The second learning data for zone B includes a combination of input data (B1) and teacher data (B2) shown in FIG. 19(b). The data acquisition unit 75 acquires the input data (B1) and the teacher data (B2) simultaneously, but it is sufficient that the input data (B1) and the teacher data (B2) are input in association with each other, and the data acquisition unit 75 may acquire the input data (B1) and the teacher data (B2) at different times.

In step S805, the model generation unit 76 generates a second trained model for zone B, in which teacher data (B2) is output when input data (B1) is input, by so-called supervised learning according to second learning data for zone B including a combination of input data (B1) and teacher data (B2) acquired by the data acquisition unit 75.

In step S806, the second learned model storage device 8a stores the second learned model for zone B generated by the model generation unit 76.

FIG. 23 shows the configuration of the second inference device 7a.
The second inference device 7 a includes a data acquisition unit 77 and an estimation unit 78 .

The data acquisition unit 77 acquires the outdoor air temperature at the start time of operation of the indoor unit of each zone, the room temperature at the start time of operation of the indoor unit of each zone, the compressor frequency at the start time of operation of the indoor unit of each zone, and the set temperature of each zone.

The estimation unit 78 uses a second learned model for each zone that estimates the thermal load for bringing the room temperature of the corresponding zone to the set temperature from the outdoor air temperature at the operation start time of the indoor unit of the corresponding zone, the room temperature at the operation start time of the indoor unit of the corresponding zone, the compressor frequency at the operation start time of the indoor unit of the corresponding zone, and the set temperature of the corresponding zone, and outputs the thermal load for bringing the room temperature of each zone to the set temperature from the outdoor air temperature at the operation start time of the indoor unit of each zone, the room temperature at the operation start time of the indoor unit of each zone, the compressor frequency at the operation start time of the indoor unit of each zone, and the set temperature of each zone input from the data acquisition unit 77.

The estimation unit 78 infers output data (B2) obtained by utilizing the second trained model for zone A. That is, the estimation unit 78 can output output data (B2) for zone A inferred from the input data (B1) for zone A by inputting the input data (B1) for zone A acquired by the data acquisition unit 77 into the second trained model for zone A. The estimation unit 78 infers output data (B2) obtained by utilizing the second trained model for zone B. That is, the estimation unit 78 can output output data (B2) for zone B inferred from the input data (B1) for zone B by inputting the input data (B1) for zone B acquired by the data acquisition unit 77 into the second trained model for zone B.

The control device 2 creates an operation schedule for the air conditioning device 1a based on the heat load for setting the room temperature of each zone to the set temperature, which is the inference result. The operation schedule includes, for example, a setting schedule for the operation frequency of the compressor 51, and a setting schedule for the opening degree of the expansion devices 42a and 42b. The control device 2 controls the air conditioning device 1a based on the operation schedule. The heat load, which is the inference result, is the heat load when operating at the frequency of the compressor 51 at the operation start time of the indoor unit of each zone. Therefore, if the frequency of the compressor 51 is changed from that at the operation start time, the heat load also changes, so the operation schedule is created taking into account the changing heat load.

The operation schedule may be adjusted according to an evaluation function specified by the user. The evaluation function can be changed according to what the user values. The evaluation function is, for example, the comfort of the user, the amount of power consumed by the air conditioning device 1, and the electricity charge for the air conditioning device 1. The operation schedule that is output changes according to the evaluation function specified by the user.

For example, if the evaluation function is user comfort, the control device 2 calculates a schedule for each zone to reach the designated set temperature at the earliest arrival time at work. The control device 2 determines the start time and operating frequency of the compressor 51 by prioritizing the room temperature reaching the designated set temperature at the earliest arrival time at work over the power consumption of the compressor 51.

When the evaluation function is power consumption, the control device 2 determines the start time and operating frequency of the compressor 51 so that the set temperature specified for each zone is reached at the earliest arrival time at work and the power consumption of the compressor 51 is minimized.

The evaluation function may be the electricity rate. For example, if the electricity rate for the time period from 9:00 p.m. to 5:00 a.m. is lower than the electricity rate for other time periods, the control device 2 may determine the start time of the compressor 51 to be between 9:00 p.m. and 5:00 a.m.

FIG. 24 is a flow chart showing the procedure of the inference process of the second inference device 7a.
If the earliest arrival time of the day in zone A is earlier than the earliest arrival time of the day in any zone, the process proceeds to step S902. If the earliest arrival time of the day in zone A is not earlier than the earliest arrival time of the day in any zone, the process proceeds to step S903.

In step S902, the data acquisition unit 77 sets the operation start time of the indoor unit in zone A to a time ta that is a certain time ΔT earlier than the earliest arrival time of work in zone A on that day.

In step S903, the data acquisition unit 77 acquires input data (B1) for zone A shown in FIG. 19 (a).

In step S904, the estimation unit 78 inputs input data (B1) for zone A to a second learned model for zone A stored in the second learned model storage device 8a, and obtains output data (B2) for zone A.

In step S905, the estimation unit 78 outputs output data (B2) for zone A obtained by the second learned model for zone A to the control device 2.

In step S906, the control device 2 sets the operation schedule for the heat source unit 50, which is the outdoor unit, and the load unit 40a, which is the indoor unit of zone A, based on the thermal load for setting the room temperature of zone A to the set temperature, which is the inference result.

In step S907, the data acquisition unit 77 sets the operation start time of the indoor unit in zone B to a time tb that is a certain time ΔT earlier than the earliest arrival time of the day in zone B.

In step S908, the data acquisition unit 77 acquires input data (B1) for zone B shown in FIG. 19 (b).

In step S909, the estimation unit 78 inputs input data (B1) for zone B to a second learned model for zone B stored in the second learned model storage device 8a, and obtains output data (B2) for zone B.

In step S910, the estimation unit 78 outputs output data (B2) for zone B obtained by the second learned model for zone B to the control device 2.

In step S911, the control device 2 sets operation schedules for the heat source unit 50, which is the outdoor unit, the load unit 40a, which is the indoor unit of zone A, and the load unit 40b, which is the indoor unit of zone B, based on the heat load for bringing the room temperature of zone B to the set temperature, which is the inference result, and the remaining heat load for bringing the room temperature of zone A to the set temperature (the value obtained by subtracting the heat load already supplied from the heat load of the inference result).

In step S912, the data acquisition unit 77 sets the operation start time of the indoor unit in zone B to a time tb that is a certain time ΔT earlier than the earliest arrival time of the day in zone B.

In step S913, the data acquisition unit 77 acquires input data (B1) for zone B shown in FIG. 19 (b).

In step S914, the estimation unit 78 inputs input data (B1) for zone B to a second learned model for zone B stored in the second learned model storage device 8a, and obtains output data (B2) for zone B.

In step S915, the estimation unit 78 outputs output data (B2) for zone B obtained by the second learned model for zone B to the control device 2.

In step S916, the control device 2 sets the operation schedule for the heat source unit 50, which is the outdoor unit, and the load unit 40b, which is the indoor unit of zone B, based on the thermal load for setting the room temperature of zone B to the set temperature, which is the inference result.

In step S917, the data acquisition unit 77 sets the operation start time of the indoor unit in zone A to a time ta that is a certain time ΔT earlier than the earliest arrival time of work in zone A on that day.

In step S918, the data acquisition unit 77 acquires input data (B1) for zone A shown in FIG. 19 (a).

In step S919, the estimation unit 78 inputs input data (B1) for zone A to a second learned model for zone A stored in the second learned model storage device 8a, and obtains output data (B2) for zone A.

In step S920, the estimation unit 78 outputs output data (B2) for zone A obtained by the second learned model for zone A to the control device 2.

In step S921, the control device 2 sets operation schedules for the heat source unit 50, which is the outdoor unit, the load unit 40a, which is the indoor unit of zone A, and the load unit 40b, which is the indoor unit of zone B, based on the heat load for bringing the room temperature of zone A to the set temperature, which is the inference result, and the remaining heat load for bringing the room temperature of zone B to the set temperature (the value obtained by subtracting the heat load already supplied from the heat load of the inference result).

Note that, in the process of FIG. 24, the indoor unit in the zone with the earliest arrival time is started first, and then the indoor unit in the zone with the latest arrival time is started, but this is not limited to this. The indoor unit in zone A and the indoor unit in zone B may be started simultaneously at a certain time ΔT earlier than the earlier of the two earliest arrival times. In such a case, the operation schedules of the load side unit 40a, which is the indoor unit in zone A, and the load side unit 40b, which is the indoor unit in zone B, can be set at the same time.

Embodiment 3.
25(a) is a diagram showing an example of input data (B1) and output data (teacher data) B2 of the first trained model for zone A in embodiment 3. The input data (B1) includes the day of the week, the month of the day, the earliest arrival time at work for each day in the past week in zone A, and the person who arrived at work first for each day in the past week in zone A. The output data (B2) includes the earliest arrival time at work for each day in zone A, and the person who arrives at work first for each day in zone A.

Figure 25 (b) is a diagram showing an example of input data (B1) and output data (teacher data) B2 of the first trained model for zone B in embodiment 3. The input data (B1) includes the day of the week, the month of the day, the earliest arrival time at work for each day in the past week in zone B, and the person who arrived at work first for each day in zone B in the past week. The output data (B2) includes the earliest arrival time at work for each day in zone B, and the person who arrives at work first for each day in zone B.

The first learning device 3a operates as follows.
The data acquisition unit 63a acquires from the first learning data storage unit 62a first learning data including a combination of input data (B1) including the day of the week, the month of the day, the earliest arrival time of work for each day in the past week for each zone, and the person who arrived at work first for each day in the past week for each zone, and teacher data (B2) including the earliest arrival time of work for each zone for that day and the person who will arrive at work first for each zone for that day.

The model generation unit 64a uses a plurality of first learning data to generate a first trained model for each zone for inferring the earliest arrival time of the corresponding zone on that day and the person who will arrive first in the corresponding zone from the day of the week of the current day, the month of the current day, the earliest arrival time of each day of the corresponding zone in the past week, and the person who arrived first in the corresponding zone on that day. That is, the model generation unit 64a uses the first learning data for zone A to generate a first trained model for zone A for inferring the earliest arrival time of the corresponding zone on that day and the person who will arrive first in the corresponding zone on that day from the day of the week of the current day, the month of the current day, the earliest arrival time of each day of the corresponding zone in the past week, and the person who arrived first in the corresponding zone on that day. The model generation unit 64a uses the first learning data for zone B to generate a first trained model for zone B for inferring the earliest arrival time at work on that day in zone B and the person who will be the earliest to arrive at work on that day in zone B from the current day of the week, the current month, the earliest arrival time at work on each day in zone B in the past week, and the person who arrived at work first on each day in zone B in the past week.

The first inference device 4a operates as follows.
The data acquiring unit 65a acquires the day of the week, the month of the month, the earliest arrival time for each day of the past week for each zone, and the earliest person who arrived at work for each day of the past week for each zone. That is, the data acquiring unit 65a acquires input data (B1) for zone A shown in Fig. 14(a) and input data (B1) for zone B shown in Fig. 14(b) from the control device 2.

The estimation unit 66a uses a first learned model for each zone to infer the earliest arrival time of the day in the corresponding zone and the person who will arrive at work earliest on that day in the corresponding zone from the day of the week of the day, the month of the day, the earliest arrival time of the day in each day of the past week in the corresponding zone, and the person who arrived at work earliest on that day in the corresponding zone, input from the data acquisition unit 65a, and outputs the earliest arrival time of the day in the corresponding zone and the person who will arrive at work earliest on that day in the corresponding zone from the day of the week of the day, the month of the day, the earliest arrival time of the day in each day of the past week in each zone, and the person who arrived at work earliest on that day in each zone.

The second learning device 6a operates in the same manner as in the second embodiment.
The second inference device 7a operates as follows.

The data acquisition unit 77 acquires the outdoor temperature at the start time of the indoor unit of each zone, the room temperature at the start time of the indoor unit of each zone, the compressor frequency at the start time of the indoor unit of each zone, and the set temperature according to the person who will be the first to arrive at work in each zone. The person who will be the first to arrive at work can use the inference result of the first inference device 4a. The set temperature according to the person who will be the first to arrive at work can be identified by referring to a table that defines the correspondence between the set temperature of the remote control created based on past remote control operations and the person who operated the remote control. Alternatively, a table created by each person (user) inputting their preferred set temperature may be referenced.

The estimation unit 78 uses a second learned model for each zone that estimates the thermal load for bringing the room temperature of the corresponding zone to the set temperature from the outdoor air temperature at the operation start time of the indoor unit of the corresponding zone, the room temperature at the operation start time of the indoor unit of the corresponding zone, the compressor frequency at the operation start time of the indoor unit of the corresponding zone, and the set temperature of the corresponding zone, and outputs the thermal load for bringing the room temperature of each zone to the set temperature from the outdoor air temperature at the operation start time of the indoor unit of each zone, the room temperature at the operation start time of the indoor unit of each zone, the compressor frequency at the operation start time of the indoor unit of each zone, and the set temperature according to the first person to arrive at work in each zone, which are input from the data acquisition unit.

Embodiment 4.
FIG. 26 is a diagram showing an example of management data.

The management data includes, for each user (person), the zone in which the user's seat is located and the time the user arrives at work each day. The first learning data creation unit 61a of the first learning device 3a can create the first learning data by referring to this management data. The model generation unit 64a of the first learning device 3a uses the first learning data to generate a first trained model.

When a user changes seats, the zone may change. When the user's zone is changed, the management data is updated. The first learning data creation unit 61a of the first learning device 3a can update the first learning data by referring to the updated management data. The model generation unit 64a of the first learning device 3a updates the first learned model using the updated first learning data.

FIG. 27 is a flowchart showing a processing procedure when changing seats.
In step S1001, if the user moves to a different room and the management table is updated, the process proceeds to step S1002.

In step S1002, the first learning data creation unit 61a of the first learning device 3a updates the first learning data by referring to the updated management data.

In step S1003, the model generation unit 64a of the first learning device 3a updates the first learned model using the updated first learning data.

Variation example 1.
FIG. 28 is a diagram showing the configuration of a modified air conditioning system.

The air conditioning system comprises an air conditioning unit 1a, a control unit 2a, a first learning device 3a, a first inference device 4a, a first learned model memory device 5a, an air conditioning management device 9, and an information processing device 10.

The air conditioning device 1a, the first learning device 3a, the first learned model storage device 5a, and the first inference device 4a are the same as in embodiment 2, so the description will not be repeated.

Figure 29 is a block diagram showing an example configuration of the air conditioning management device and information processing device shown in Figure 28. Figure 30 is a diagram for explaining the control performed by the air conditioning control means shown in Figure 29 on the air conditioner.

The memory unit 12 of the air conditioning management device 9 has an air conditioning data storage means 101. The control unit 11 has a communication means 102, an air conditioning control means 103, and an air conditioning device communication means 104. The communication means 102, the air conditioning control means 103, and the air conditioning device communication means 104 are configured by the CPU 14 executing a program.

The air conditioning data storage means 101 stores air conditioning data. The air conditioning data includes, for example, the operating frequency of the compressor 51, the openings of the expansion devices 42a and 42b, and the detection values of various sensors. The communication means 102 transmits and receives data to and from the information processing device 10 via the communication line 95, and manages the communication with the information processing device 10. Specifically, the communication means 102 reads air conditioning data from the air conditioning data storage means 101 and transmits it to the information processing device 10. In addition, when the communication means 102 receives an operating state schedule including the start time of the compressor 51 and the operating frequency set in the compressor 51 at the time of start from the information processing device 10, it passes the schedule to the air conditioning control means 103. The operating state may include information on the openings of the expansion devices 42a and 42b. The schedule may include set values associated with time changes in the operating state including the operating frequency of the compressor 51 and the openings of the expansion devices 42a and 42b.

The air-conditioning device communication means 104 relays communication between the air-conditioning device 1 and the air-conditioning control means 103 and memory unit 12, and manages communication with the air-conditioning device 1. The air-conditioning device communication means 104 acquires data from the compressor 51, the four-way valve 53, the expansion devices 42a and 42b, the outside air temperature sensor 54, and the space temperature sensors 43a and 43b via the communication line 96, and stores the acquired data in the air-conditioning data storage means 101. The data acquired from the compressor 51, the four-way valve 53, and the expansion devices 42a and 42b indicate the operating state. The data acquired from the outside air temperature sensor 54 and the space temperature sensors 43a and 43b are the detection values of each sensor.

When the load side unit 40a is operating, the air conditioning control means 103 controls the operating frequency of the compressor 51 and the opening degree of the expansion device 42a so that the room temperature Tra detected by the space temperature sensor 43a becomes the set temperature Tsa. When the load side unit 40b is operating, the air conditioning control means 103 controls the operating frequency of the compressor 51 and the opening degree of the expansion device 42b so that the room temperature Tb detected by the space temperature sensor 43b becomes the set temperature Tsb. When the load side units 40a and 40b are operating, the air conditioning control means 103 sets the operating frequency of the compressor 51 and the opening degree of the expansion devices 42a and 42b so that the room temperature Tra becomes the set temperature Tsa and the room temperature Trb becomes the set temperature Tsb. Hereinafter, the set temperature when both the set temperatures Tsa and Tsb are included is referred to as Tset.

The set temperature Tsa is stored in memory, for example, by a user using a room that is the target space for air conditioning of the load side unit 40a operating a remote controller (not shown). The set temperature Tsb is stored in memory, for example, by a user using a room that is the target space for air conditioning of the load side unit 40b operating a remote controller. In addition, the air conditioning control means 103 may take into account the influence of the outside air temperature Tout when determining the operating frequency of the compressor 51.

Furthermore, when the air conditioning control means 103 receives a schedule of the operating state from the information processing device 10 via the communication means 102, it controls the air conditioner 1 in accordance with the schedule. For example, if the schedule includes information on the start time and operating frequency of the compressor 51, the air conditioning control means 103 sets the operating frequency to the compressor 51 in accordance with the schedule and starts the compressor 51 at the start time. If the schedule includes time-series setting values for the opening degrees of the expansion devices 42a and 42b, the air conditioning control means 103 controls the opening degrees of the expansion devices 42a and 42b in accordance with the schedule.

The memory unit 22 of the information processing device 10 has a learning data holding means 121. The control unit 21 has a communication means 122, a heat load learning means 123, and a schedule determination means 124. The communication means 122, the heat load learning means 123, and the schedule determination means 124 are configured by the CPU executing a program.

The learning data storage means 121 stores learning data used for machine learning among the data included in the air conditioning data stored by the air conditioning data storage means 101 of the air conditioning management device 9. The learning data is composed of a set of input data and output data. In the case of the load side unit 40a, the input data is the outside air temperature Tout, the detection value of the space temperature sensor 43a, the set temperature Tsa, and the operating frequency of the compressor 51. The output data is the heat load processed by the load side unit 40a. The heat load is proportional to the temperature difference between the room temperature Tr and the set temperature Tset. In addition, the heat load is affected by the outside air temperature Tout and the amount of heat generated indoors. The amount of heat generated indoors is a different amount for each air-conditioned space of the load side units 40a and 40b. Therefore, the learning data needs to be accumulated for each unit of the load side units 40a and 40b.

The communication means 122 transmits and receives data to and from the air conditioning management device 9, and manages communication with the air conditioning management device 9. Specifically, the communication means 122 acquires learning data from the air conditioning data stored in the air conditioning data storage means 101 via the communication means 102, and stores the acquired learning data in the learning data storage means 121. The heat load learning means 123 executes machine learning according to a machine learning program using the learning data stored in the learning data storage means 121, and estimates the heat loads of the load side units 40a and 40b relatively from the overall heat load. Then, the heat load learning means 123 obtains a learning model that indicates the input-output relationship between the input data and the output data for each of the load side units 40a and 40b. The heat load learning means 123 stores the obtained learning model in memory. The storage destination of the learning model may be the storage unit 22 instead of the memory.

The schedule determination means 124 determines a schedule of the operating state for each load unit 40a and 40b, based on the learning model obtained by the heat load learning means 123 for each load unit 40a and 40b, such that the room temperature reaches a designated set temperature at a designated set time. For example, the schedule is such that the operating frequency of the compressor 51 is set to F1 [Hz] at 6:00 a.m., and is set to F2 [Hz], which is an increase of Δf from F1, at 8:00 a.m.

Here, the operation of the thermal load learning means 123 and the schedule determination means 124 shown in Fig. 29 will be described with reference to Fig. 31. Fig. 31 is a diagram showing a schematic example of the operation procedure of the thermal load learning means and the schedule determination means shown in Fig. 29.

The input data for the learning data includes, for example, the outside air temperature Tout, the room temperature Tr, the set temperature Tset, and the operating state. The operating state will be described in the case of the operating frequency of the compressor 51, but it may also be the opening degree of the expansion devices 42a and 42b. Weather forecast information may be used for the outside air temperature Tout. The case where weather forecast information is used will be described later. Figure 31 shows four cases where the input data is the outside air temperature Tout, the room temperature Tr, the set temperature Tset, and the operating state.

As a pre-processing step for machine learning, as shown in FIG. 31, the thermal load learning means 123 may perform one or both of the optimization of input data and the reduction of input dimensions. The reduction of input dimensions is, for example, taking the temperature difference between the room temperature Tr and the set temperature Tset. The optimization of input data is, for example, normalizing the set temperatures Tsa and Tsb when the set temperatures Tset of the air-conditioned spaces of the load side units 40a and 40b are different.

In the machine learning shown in FIG. 31, the heat load learning means 123 performs, for example, supervised learning. Supervised learning is machine learning that obtains a learning model that shows an input/output relationship that estimates an unknown situation with high accuracy from multiple learning data. The machine learning performed by the heat load learning means 123 is not limited to supervised learning, and may be reinforcement learning. Deep learning may also be applied to supervised learning and reinforcement learning. Furthermore, for the input/output relationship, depending on the required accuracy and calculation efficiency, the heat load learning means 123 may select any one of the machine learning methods from supervised learning, reinforcement learning, and neural network.

Here, we will explain an example of how to determine the learning model. The thermal load influence factors of the thermal characteristic model include the outside air temperature Tout, the room temperature Tr, the amount of heat generated indoors Qr, the amount of air conditioning heat QHVAC [kW], the amount of solar radiation, and the temperature in the adjacent room. One thermal characteristic model using representative influence factors among these influence factors is expressed, for example, by the following formula (C1).

C*dTr/dt=aQHVAC+bQr+(Tout-Tr)/Rwin+α…(C1)
In formula (C1), a and b are coefficients, and C is the indoor heat capacity [kJ/K]. Rwin is the window thermal conductivity [kW/K], and α is the amount of heat caused by other heat load influencing factors. The air conditioning heat quantity QHVAC is the amount of heat removed in the case of cooling operation, and is the amount of heat supplied in the case of heating operation. The heat load learning means 123 calculates the air conditioning heat quantity QHVAC from a theoretical value based on the operating state including the operating frequency of the compressor 51.

In formula (C1), if we move the left side to the right side and move the air conditioning heat quantity QHVAC to the left side, we get the following formula (C2). Note that in formula (C2), the coefficient a of the air conditioning heat quantity QHVAC includes a plus or minus sign.

aQHVAC=bQr+(Tout-Tr)/Rwin-C*dTr/dt+α…(C2)
For the amount of heat shown on the right side of formula (C2), the integral value of the change over time until the room temperature Tr reaches the set temperature Tset can be considered as the thermal load. The thermal load learning means 123 substitutes input data that changes over time for each of the load side units 40a and 40b into formula (C2) and accumulates learning data that pairs the input data with the output data. Then, based on the accumulated learning data, the thermal load learning means 123 obtains a learning model that indicates the input/output relationship between the input data and the output data.

Note that the thermal characteristic model used to collect the learning data is not limited to formula (C1). There may be multiple thermal characteristic models used to collect the learning data. Furthermore, the thermal load learning means 123 may accumulate learning data for each of the multiple thermal characteristic models, extract parameters that are influencing factors that have a significant effect on the thermal load in the air-conditioned environment of the space to be air-conditioned, and select the optimal thermal characteristic model from the multiple thermal characteristic models.

The schedule determination means 124 determines a schedule for the operating state of the air conditioning device 1 based on the learning model for the thermal load of each load side unit obtained by the thermal load learning means 123. In this case, the schedule determination means 124 may adjust the schedule according to an evaluation function specified by the user. The evaluation function can be changed by the user by operating a remote controller (not shown) according to the contents that the user values. The evaluation function is, for example, the comfort of the user, the amount of power consumed by the air conditioning device 1, and the electricity fee for the air conditioning device 1.

The operating state of the output schedule changes according to the evaluation function specified by the user. When the evaluation function is user comfort, the schedule determination means 124 calculates a schedule for reaching the specified set temperature at the set time specified for each of the load side units 40a and 40b. In this case, the schedule determination means 124 determines the start time and operating frequency of the compressor 51 by prioritizing the room temperature reaching the specified set temperature at the specified set time over the power consumption of the compressor 51.

Figure 32 is a diagram showing a schematic example of another operating procedure of the thermal load learning means and schedule determination means shown in Figure 29. Figure 32 shows a case where there are four load side units. Figure 32 shows a case where the evaluation function is power consumption. The schedule determination means 124 determines the start time and operating frequency of the compressor 51 so that the room temperature reaches the specified set temperature at the set time specified by the user for each of the four load side units and the power consumption of the compressor 51 is minimized. The schedule determination means 124 outputs the start time and operating frequency of the compressor 51 for the four load side units collectively.

In addition, in FIG. 32, the evaluation function may be the electricity rate. When the evaluation function is the electricity rate, it is assumed that the memory stores information on the electricity rate corresponding to the time period. For example, if the electricity rate for the time period from 9:00 p.m. to 5:00 a.m. is lower than the electricity rate for other time periods, the schedule determination means 124 may determine the start time of the compressor 51 to be between 9:00 p.m. and 5:00 a.m. Furthermore, in FIG. 32, four air conditioners 1 may be used instead of the four load side units. In this case, the schedule determination means 124 may output the start times of the compressors 51 of all the air conditioners 1 at once. As described with reference to FIG. 31 and FIG. 32, the start time of the compressor 51 changes depending on the evaluation function.

Figure 33 is a flowchart showing the steps of machine learning performed by the heat load learning means shown in Figure 31. When the start time of the determined cycle is reached (step S1101), the communication means 122 reads out data including the room temperatures Tra and Trb, the outside air temperature Tout, and the set temperatures Tsa and Tsb from the air conditioning data stored in the air conditioning management device 9 (step S1102). Then, the communication means 122 stores the read data in the learning data storage means 121. The communication means 122 also acquires data indicating the operating state of the air conditioning device 1 from the air conditioning data stored in the air conditioning management device 9 (step S1103), and stores the acquired data in the learning data storage means 121. Figure 33 shows a case where the operating state is the operating frequency of the compressor 51.

The thermal load learning means 123 determines whether it is time to learn (step S1104), and if it is not time to learn, proceeds to step S1106. If the result of the determination in step S1104 is that it is time to learn, the thermal load learning means 123 executes the machine learning described with reference to FIG. 31 (step S1105) and stores the learning model in memory as the learning result. If the memory already stores the learning model, the thermal load learning means 123 updates the learning model stored in the memory. The cycle for executing the machine learning can be freely set by the user. The storage means for storing the learning model is not limited to a memory. The storage unit 22 may store the learning model.

Next, the operation of the information processing device 10 of the modified example will be described. FIG. 34 is a flowchart showing the operation procedure of the information processing device 10. Here, it is assumed that each user of the load side units 40a and 40b operates a remote controller (not shown) to specify the set time and set temperature so that the room temperature Tr becomes the set temperature at the set time. The communication means 102 of the air conditioning management device 9 stores the set time and set temperature of each load side unit in the memory unit 12. For the load side unit 40a, the specified set time is ta, and the specified set temperature is Tsa1. For the load side unit 40b, the specified set time is tb, and the specified set temperature is Tsb1. The dashed frame shown in FIG. 34 indicates processing based on the learning model obtained by the thermal load learning means 123.

The communication means 122 acquires the set time and set temperature specified for each load unit from the air conditioning management device 9 (step S1201). The communication means 122 acquires air conditioning data including the current room temperature Tr, outside air temperature Tout, and operating state from the air conditioning management device 9 a certain time before the earliest specified set time among the specified set times (step S1202). The certain time is, for example, one hour. Here, the set time ta is assumed to be earlier than the set time tb.

The schedule determination means 124 receives the set times ta and tb, the set temperatures Tsa1 and Tsb1, and the air conditioning data from the communication means 122. The schedule determination means 124 uses the information received from the communication means 122 as input data and calculates the relatively required heat load using the learning models of the load side units 40a and 40b (step S1203). The schedule determination means 124 determines the start time and operating frequency of the compressor 51 taking into account the overall heat load (step S1204). Then, the schedule determination means 124 determines a schedule of the operating state in which the room temperature Tra reaches the set temperature Tsa1 at the set time ta and the room temperature Trb reaches the set temperature Tsb1 at the set time tb. The schedule determination means 124 transmits the determined schedule to the air conditioning management device 9 (step S1205).

In the procedure shown in Figure 34, the control unit 21 may periodically repeat the processing of steps S1202 to S1205 until the time reaches the set time ta or tb, and update the schedule to be notified to the air conditioning management device 9. Also, in the procedure shown in Figure 34, in step S1205, the schedule determination means 124 transmits the schedule to the air conditioning management device 9, but the operating state may be instructed to the air conditioning management device 9 in accordance with the schedule. In this case, the air conditioning management device 9 does not need to hold the schedule. Also, if the schedule is updated over time, the air conditioning management device 9 does not need to receive a new schedule from the information processing device 10 each time the schedule is updated.

Figure 35 is a diagram showing an example of a case where the air conditioning management device controls the compressor and expansion device according to a schedule. The vertical axis of Figure 35 is room temperature, and the horizontal axis is time. Figure 35 shows a case where the load side units 40a and 40b perform heating operation.

For example, the schedule sets the operating frequency of the compressor 51 to F1 [Hz] at time t1, sets the opening of the expansion device 42a to Cva, and sets the operating frequency of the compressor 51 to F2 [Hz] and sets the opening of the expansion device 42b to Cvb at time t2. The relationship is F2>F1. The compressor 51 starts at time t1, which is earlier than the set time ta, so that the room temperature Tra reaches the set temperature Tsa1 at the set time ta. Also, the operating frequency of the compressor 51 increases to F2 at time t2, which is earlier than the set time tb, so that the room temperature Trb reaches the set temperature Tsb1 at the set time tb. In this way, each of the load side units 40a and 40b is controlled so that each air-conditioned space reaches the set temperature at the set time.

The communication line 95 may be a network. The network is, for example, the Internet. In this case, the information processing device 10 may obtain information on the outside air temperature in the area in which the air conditioning device 1 is installed from a server that provides weather forecasts via the network. The outside air temperature sensor 54 does not need to be provided. Furthermore, the number of load side units is not limited to two. The number of load side units may be three or more.

The information processing device 10 of the modified example has a schedule determination means 124 that determines a schedule of an operating state for making the temperature of the air-conditioned space of each load side unit reach a set temperature at a set time based on a learning model of each load side unit. The learning model indicates the input/output relationship between input data that is an influencing factor of the heat load of each load side unit and output data that indicates the heat load.

According to this modified example, the information processing device 10 estimates the heat load of each load unit using a learning model for multiple load side units 40a and 40b, and a schedule indicating an operating state suitable for the air conditioning environment of the air conditioned space of each load side unit is obtained. As a result, if the time at which the operating load of each load side unit becomes large shifts, the maximum value of the power consumption of the air conditioning device 1 can be suppressed from increasing. Therefore, the room temperature of each air conditioned space reaches the set temperature at the set time, ensuring user comfort and suppressing the maximum value of the overall power consumption of the air conditioning device 1. If the power fee increases in proportion to the maximum power consumption, the power fee can be suppressed. In addition, by using machine learning to calculate the learning model of the heat load, it is easy to change the parameters of the input data used to estimate the heat load.

The information processing device 10 of this modified example may have a learning data holding means 121 that stores learning data consisting of a pair of input data and output data of each load side unit, and a thermal load learning means 123. The thermal load learning means 123 estimates the thermal load processed by each load side unit based on the learning data of each load side unit, and obtains a learning model that indicates the input/output relationship between the input data and the output data.

To calculate the heat load more accurately, it is necessary to use a physical model and a statistical model that take into account influencing factors including the outside air temperature, which complicates the calculation process and increases the information processing load. In contrast, in this modified example, even if the air-conditioning environments of multiple air-conditioned spaces are different, the heat load learning means 123 obtains a learning model of the heat load of each air-conditioned space according to machine learning, thereby improving the accuracy of the heat load calculation while reducing the information processing load.

In this modified example, the user may specify an evaluation function. The schedule determination means 124 adjusts the schedule according to the specified evaluation function. For example, when the evaluation function is power consumption, the schedule determination means 124 adjusts the schedule for each air-conditioned space to reach the set temperature at the set time so that the power consumption of the air-conditioning device 1 is minimized. When the evaluation function is power rate, the schedule determination means 124 adjusts the schedule for each air-conditioned space to reach the set temperature at the set time so that the start time of the compressor 51 is shifted to a time period when the power rate is low. In this case, the schedule determination means 124 changes the operating frequency at the start of the compressor 51 to an operating frequency lower than that of the schedule before adjustment.

The schedule determination means 124 may obtain time-series information on the power consumption of the entire building from the centralized controller 200, and determine a schedule for the operating state of the air conditioning unit 1 so that the power consumption of the entire building is minimized. In this case, the maximum power consumption of the entire building in which the air conditioning unit 1 is installed is suppressed, and the power consumption of the entire building can be reduced. Furthermore, the schedule determination means 124 may adjust the start time of the compressor 51 for a schedule for each air-conditioned space to reach the set temperature at the set time, so as to shift it to a time period when the electricity rate is low. In this case, the electricity rate for the entire building can be suppressed.

Variation example 2.
The first learning device, the second learning device, the first inference device, and the second inference device are provided inside the air conditioning system, but may be devices connected to the air conditioning system via a network and separate from the air conditioning system. Furthermore, the first learning device, the second learning device, the first inference device, and the second inference device may exist on a cloud server.

Variation example 3.
In the embodiment, a case where supervised learning is applied to the learning algorithm used by the model generation unit has been described, but the present invention is not limited to this. As for the learning algorithm, reinforcement learning, unsupervised learning, semi-supervised learning, etc. can also be applied in addition to supervised learning.

Variation example 4.
The model generation unit may use learning data created in multiple air conditioning systems. The model generation unit may acquire learning data from multiple air conditioning systems used in the same area, or may use learning data collected from multiple air conditioning systems operating independently in different areas. It is also possible to add or remove air conditioning systems from which learning data is collected midway. Furthermore, a learning device that has learned about a certain air conditioning system may be applied to another air conditioning system, and re-learn about this other air conditioning system.

Variation example 5.
The learning algorithm used by the model generation unit may be deep learning, which learns to extract the features themselves, or machine learning may be performed according to other known methods, such as genetic programming, functional logic programming, or support vector machines.

Variation example 6.
FIG. 36 is a diagram showing the hardware configuration of the first learning device, the second learning device, the first inference device, the second inference device, and the control device.

The first learning device, the second learning device, the first inference device, the second inference device, and the control device can be configured with corresponding operations in hardware or software of a digital circuit. When the functions of the first learning device, the second learning device, the first inference device, the second inference device, and the control device are realized using software, the first learning device, the second learning device, the first inference device, the second inference device, and the control device can be, for example, as shown in FIG. 36, equipped with a processor 1001 and a memory 1002 connected by a bus 1003, and the processor 1001 can execute a program stored in the memory 1002.

Variation example 7.
In the above embodiment, the first inference device and the second inference device perform inference using a trained model, but this is not limited to this. For example, the first inference device and the second inference device may perform inference from the input data acquired by the data acquisition unit based on rule-based inference or case-based inference.

The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present disclosure is indicated by the claims, not the above description, and is intended to include all modifications within the meaning and scope of the claims.

1, 1a air conditioning device, 2, 2a control device, 3, 3a first learning device, 4, 4a first inference device, 5, 5a first learned model storage device, 6, 6a second learning device, 7, 7a second inference device, 8 second learning result storage device, 8a second learned model storage device, 9 air conditioning management device, 10 information processing device, 11, 21 control unit, 12, 22 storage unit, 40a, 40b load side unit, 41a, 41b load side heat exchanger, 42a, 42b expansion device, 43a, 43b space temperature sensor, 50 heat source side unit, 51 compressor, 52 heat source side heat exchanger, 53 four-way valve, 54 outdoor air temperature sensor, 60a, 60b refrigerant circuit, 61, 61a first learning data creation unit, 62, 62a First learning data storage unit, 63, 63a, 65, 65a, 68, 71, 75, 77 Data acquisition unit, 64, 64a, 76 Model generation unit, 66, 66a, 78 Estimation unit, 67 Setting unit, 69 Coefficient calculation unit, 72 Optimum operation start time estimation unit, 73 Second learning data creation unit, 74 Second learning data storage unit, 95, 96 Communication line, 101 Air conditioning data storage means, 102, 122 Communication means, 103 Air conditioning control means, 104 Air conditioning device communication means, 121 Learning data storage means, 123 Thermal load learning means, 124 Schedule determination means, 200 Centralized controller, 1001 Processor, 1002 Memory, 1003 Bus.

Claims (12)

An air conditioning device including an outdoor unit and at least one indoor unit;
a first inference device that infers an earliest office arrival time of the current day based on the latest day of the week, the latest month of the current day, and the latest office arrival time of each day of a certain period of time in the past;
A second inference device that infers a heat load for setting the room temperature to a set temperature at the earliest arrival time at work on the day;
A control device that controls the air conditioning device based on the thermal load. The first inference device:
A data acquisition unit that acquires the earliest arrival time for each day of the current day of the week, the current month, and a certain period of time in the past;
The air conditioning system of claim 1, further comprising: an estimation unit that uses a first learned model for inferring the earliest work arrival time on that day from the earliest work arrival time for the day of the week, the month of the day, and each day for a certain period of time in the past, and outputs the earliest work arrival time on that day from the earliest work arrival time for the day of the week, the month of the day, and each day for a certain period of time in the past input from the data acquisition unit. The indoor units are provided for each zone,
the first inference device infers the earliest work start time for each zone on the day based on the day of the week, the month of the day, and the earliest work start time for each day in a certain period of time in the past for each zone;
2. The air conditioning system according to claim 1, wherein the second inference device infers a heat load for bringing the room temperature of each zone to a set temperature at the earliest arrival time at work in each zone on the day. The first inference device:
A data acquisition unit that acquires the earliest arrival time of each day for a certain period of time in the past for the day of the week, the month, and each zone;
The air conditioning system of claim 3, further comprising: an estimation unit that uses a first learned model for each zone for inferring the earliest work arrival time for the corresponding zone on the current day from the earliest work arrival time for each day for the current day, the current month, and the corresponding zone during a certain period in the past, and outputs the earliest work arrival time for each zone on the current day from the earliest work arrival time for each day for the current day, the current month, and the corresponding zone during a certain period in the past, which are input from the data acquisition unit. The air conditioning system of claim 3, wherein the second inference device sets the operation start time of the indoor unit of each zone to a time a certain time before the earliest arrival time of each zone, and infers the heat load for bringing the room temperature of each zone to the set temperature based on the outdoor air temperature at the operation start time of the indoor unit of each zone, the room temperature of each zone at the operation start time, the compressor frequency at the operation start time, and the set temperature of each zone. The second inference device:
A data acquisition unit that acquires an outdoor air temperature at an operation start time of an indoor unit of each zone, an indoor room temperature at the operation start time of each zone, a compressor frequency at the operation start time, and a set temperature of each zone;
The air conditioning system of claim 5 further comprises an estimation unit that uses a second learned model for each zone that estimates the thermal load for bringing the room temperature of the corresponding zone to a set temperature from the outdoor air temperature at the operation start time of the indoor unit of the corresponding zone, the room temperature of the corresponding zone at the operation start time, the compressor frequency at the operation start time, and the set temperature of the corresponding zone, which are input from the data acquisition unit, and outputs the thermal load for bringing the room temperature of each zone to a set temperature from the outdoor air temperature at the operation start time of the indoor unit of the corresponding zone, the room temperature of the corresponding zone at the operation start time, the compressor frequency at the operation start time, and the set temperature of the corresponding zone. the first inference device infers the earliest arrival time of each zone on that day and the person who will arrive first at work on that day in each zone, based on the day of the week on that day, the month on that day, the earliest arrival time of each zone on each day in a fixed period of time in the past, and the person who arrived first at work on each day in each zone on that day in a fixed period of time in the past;
The air conditioning system of claim 3, wherein the second inference device infers a thermal load for setting the room temperature of each zone to a set temperature corresponding to the person who arrives at work earliest in each zone on that day at the earliest arrival time of the person in each zone on that day. The first inference device:
a data acquisition unit that acquires the day of the week, the month, the earliest arrival time for each day in a certain period of time in the past for each zone, and the earliest person who arrived at work for each day in a certain period of time in the past for each zone;
The air conditioning system of claim 7, further comprising: an estimation unit that uses a first learned model for each zone to infer the earliest arrival time of work on that day in the corresponding zone and the person who will be the first to arrive at work on that day in the corresponding zone from the day of the week of the current day, the month of the current day, the earliest arrival time of work on each day in the corresponding zone during a certain period in the past, and the person who was the first to arrive at work on each day in the corresponding zone during the certain period in the past, which are input from the data acquisition unit; and an estimation unit that outputs the earliest arrival time of work on that day in each zone and the person who will be the first to arrive at work on that day in each zone from the day of the week of the current day, the month of the current day, the earliest arrival time of work on each day in the corresponding zone during a certain period in the past, and the person who was the first to arrive at work on each day in the corresponding zone during the certain period in the past, which are input from the data acquisition unit. The second inference device:
a data acquisition unit that acquires the outdoor air temperature at the start time of operation of the indoor unit of each zone, the room temperature at the start time of the operation of each zone, the compressor frequency at the start time of the operation, and the set temperature according to the person who arrives at work first in each zone;
The air conditioning system of claim 7 further comprises: an estimation unit that uses a second learned model for each zone that estimates the thermal load for setting the room temperature of the corresponding zone to a set temperature from the outdoor air temperature at the operation start time of the indoor unit of the corresponding zone, the room temperature of the corresponding zone at the operation start time, the compressor frequency at the operation start time, and the set temperature of the corresponding zone, and outputs the thermal load for setting the room temperature of each zone to a set temperature corresponding to the person who is to arrive at work first in each zone from the outdoor air temperature at the operation start time of the indoor unit of the corresponding zone, the room temperature of the corresponding zone at the operation start time, the compressor frequency at the operation start time, and the set temperature of the corresponding zone input from the data acquisition unit. a first data acquisition unit that acquires first learning data including a combination of input data including the earliest arrival time at work for each day of the current day, the current month, and a certain period of time in the past, and teacher data including the earliest arrival time at work for the current day;
a first model generation unit that generates a first trained model for inferring the earliest arrival time of a day from the earliest arrival time of each day for the day of the week, the month of the day, and a certain period in the past, using the first learning data; and
a second data acquisition unit that acquires second learning data including input data including the outdoor air temperature at the operation start time of the indoor unit of each zone, the room temperature of each zone at the operation start time, the compressor frequency at the operation start time, and the set temperature of each zone, and teacher data including a heat load for setting the room temperature of each zone to the set temperature;
a second model generation unit that uses the second learning data to generate a second learned model for each zone that estimates the thermal load for bringing the room temperature of the corresponding zone to a set temperature, from the outside air temperature at the operation start time of the indoor unit of the corresponding zone, the room temperature of the corresponding zone at the operation start time, the compressor frequency at the operation start time, and the set temperature of the corresponding zone. the first data acquisition unit acquires input data including the day of the week of the current day, the month of the current day, and the earliest start time of work for each day in a certain period of time in the past for each zone, and the first learning data including the earliest start time of work for each zone on the current day;
11. The learning device according to claim 10, wherein the first model generation unit uses the first learning data to generate a first trained model for each zone for inferring the earliest work arrival time on the current day in the corresponding zone from the earliest work arrival time for each day for the current day of the week, the current month, and a certain period in the past for the corresponding zone. the first data acquisition unit acquires the first learning data including input data including the day of the week of the current day, the month of the current day, the earliest arrival time of each day for each zone during a certain period in the past, and the person who arrived at work earliest on each day for each zone during a certain period in the past, and teacher data including the earliest arrival time of each day for each zone, and the person who arrives at work earliest on each day for each zone;
11. The learning device according to claim 10, wherein the first model generation unit uses the first learning data to generate a first trained model for each zone for inferring the earliest arrival time at work on that day in the corresponding zone and the person who will arrive at work earliest on that day in the corresponding zone from the day of the week of the current day, the month of the current day, the earliest arrival time at work on each day in the corresponding zone during a certain period in the past, and the person who arrived at work earliest on each day in the corresponding zone during a certain period in the past.

Images