阅读说明：本技术 决定空调机的预冷运转或预热运转的运转条件的机械学习装置 (Machine learning device for determining operating conditions for precooling operation or preheating operation of air conditioner ) 是由 吉见学 西村忠史 于 2020-03-13 设计创作，主要内容包括：机械学习装置(100)决定空调机(10)的预冷运转或预热运转的运转条件。机械学习装置(100)具有取得部(451)以及学习部(452)。取得部(451)取得预冷运转或预热运转时的室温数据、设定温度数据(Ts)以及外部气温数据作为状态变量。学习部(452)基于状态变量、预冷运转或预热运转开始后的室温以及设定温度数据(Ts),学习预冷运转或预热运转的运转条件。(The machine learning device (100) determines the operating conditions for the pre-cooling operation or the pre-heating operation of the air conditioner (10). The machine learning device (100) is provided with an acquisition unit (451) and a learning unit (452). An acquisition unit (451) acquires room temperature data, set temperature data (Ts), and outside air temperature data during pre-cooling operation or pre-heating operation as state variables. A learning unit (452) learns the operating conditions of the pre-cooling operation or the pre-heating operation based on the state variables, the room temperature after the start of the pre-cooling operation or the pre-heating operation, and the set temperature data (Ts).)

1. A machine learning device (100) for determining an operation condition for a pre-cooling operation or a pre-heating operation of an air conditioner (10),

the machine learning device (100) is provided with:

an acquisition unit (451) that acquires room temperature data, set temperature data, and outside air temperature data during the pre-cooling operation or the pre-heating operation as state variables; and

and a learning unit (452) that learns the operating conditions for the pre-cooling operation or the warm-up operation based on the state variables, the room temperature after the start of the pre-cooling operation or the warm-up operation, and a set temperature.

2. The machine learning apparatus of claim 1,

the operation condition of the pre-cooling operation or the warm-up operation includes an operation start time.

3. The machine learning apparatus according to claim 1 or 2,

the air conditioner includes:

-using a heat exchanger (16);

using a fan (17); and

a compressor (11) for compressing the refrigerant,

the operating conditions of the pre-cooling operation or the warm-up operation include at least one of:

a temperature of the utilization heat exchanger;

the rotation speed of the fan is utilized; and

the rotational speed of the compressor.

4. The machine learning apparatus of claim 3,

the acquisition unit further acquires at least one of the following items as the state variable:

data relating to the amount of heat treatment of the air conditioner;

data of a main body surrounding a space adjusted by a utilization unit of the air conditioner;

outside air temperature data before the pre-cooling operation or the pre-heating operation;

sunshine data before the pre-cooling operation or the pre-heating operation;

meteorological data before the pre-cooling operation or the pre-heating operation; and

season or month in the pre-cooling operation or the pre-heating operation.

5. The machine learning apparatus of claim 4,

the data relating to the heat treatment amount of the air conditioner includes at least one of:

the rotational speed of the compressor;

said temperature of said utilizing heat exchanger; and

a suction temperature measured at an inlet side of the utilization heat exchanger.

6. The mechanical learning apparatus according to claim 4 or 5,

the data of the subject includes at least one of:

the number of years elapsed for the subject;

the thermal insulation of the body;

a specification of a window of the body;

the size of the room formed by the body;

ventilation of the subject;

an orientation of the body; and

and operation information of an air conditioner in a room adjacent to the room constituted by the main body.

7. The mechanical learning apparatus according to any one of claims 1 to 6,

the learning unit further includes a reward setting unit (453), the reward setting unit (453) determining a reward based on the room temperature and the set temperature after the start of the pre-cooling operation or the warm-up operation,

the learning unit learns the operation condition of the pre-cooling operation or the pre-heating operation based on the state variable and the return.

8. The machine learning apparatus of claim 7,

the return setting unit may determine the return based on an amount of electric power required until a certain time after the start of the pre-cooling operation or the warm-up operation.

9. The machine learning apparatus of claim 8,

the certain time after the start of the pre-cooling operation or the pre-heating operation is the time when the pre-cooling operation or the pre-heating operation is finished.

10. The machine learning apparatus according to claim 8 or 9,

the return setting unit increases the return when a difference between the room temperature and the set temperature at the end of the pre-cooling operation or the warm-up operation is small, or increases the return when the amount of electric power is small.

11. The mechanical learning apparatus according to any one of claims 7 to 10,

the learning unit further includes:

a behavior cost function holding unit (455) for holding a behavior cost function; and

a behavior cost function update unit (454) that updates the behavior cost function,

the behavior cost function represents an expected value of the return expected to be accepted for the operating condition, and the behavior cost function updating unit updates the behavior cost function based on the actually obtained return.

Technical Field

The present invention relates to a machine learning device that determines an operation condition for a pre-cooling operation or a pre-heating operation of an air conditioner.

Background

Patent document 1 (japanese patent No. 6270996) discloses that the absence control of an air conditioner is controlled during a time when there is no user in a space to be air-conditioned. This non-control is performed for the purpose of improving comfort when the user returns to the space to be air-conditioned. In the off-control, in order to determine the length of time for operating the air conditioner and the load handling capacity of the air conditioner, the length of time, the main structure performance of the building having the space to be air-conditioned, and the air-conditioning load at the start of the room need to be referred to.

Disclosure of Invention

Problems to be solved by the invention

In the above patent documents, no mechanical learning is mentioned.

Means for solving the problems

The machine learning device according to the first aspect determines the operation conditions of the pre-cooling operation or the pre-heating operation of the air conditioner. The machine learning device includes an acquisition unit and a learning unit. The acquisition unit acquires room temperature data, set temperature data, and outside air temperature data during pre-cooling operation or pre-heating operation as state variables. The learning unit learns the operating conditions of the pre-cooling operation or the pre-heating operation based on the state variables, the room temperature after the start of the pre-cooling operation or the pre-heating operation, and the set temperature.

According to this configuration, the operation conditions of the pre-cooling operation or the pre-heating operation are derived by machine learning. Therefore, user comfort or energy saving effect is improved.

In the machine learning device according to the second aspect, the operation condition of the pre-cooling operation or the warm-up operation includes an operation start time.

A third aspect of the present invention provides the machine learning device according to the first or second aspect, wherein the air conditioner includes a heat exchanger, a fan, and a compressor. The operating conditions for the pre-cooling operation or the pre-heating operation include at least one of the temperature of the heat exchanger, the rotational speed of the fan, and the rotational speed of the compressor.

A machine learning device according to a fourth aspect is the machine learning device according to the third aspect, wherein the acquisition unit further acquires at least one of the following items as the state variable: data relating to the amount of heat treatment of the air conditioner; data of a main body surrounding a space adjusted by a utilization unit of an air conditioner; outside air temperature data before pre-cooling operation or pre-heating operation; sunshine data before precooling operation or preheating operation; meteorological data before pre-cooling operation or pre-heating operation; and the season or month in the pre-cooling operation or the pre-heating operation.

A machine learning device according to a fifth aspect is the machine learning device according to the fourth aspect, wherein the data on the heat treatment amount of the air conditioner includes at least one of: the rotational speed of the compressor; utilizing the temperature of a heat exchanger; and a suction temperature measured at the inlet side using the heat exchanger.

A mechanical learning device according to a sixth aspect is the mechanical learning device according to the fourth or fifth aspect, wherein the data of the subject includes at least one of: the number of years elapsed for the subject; the thermal insulation of the main body; a specification of a window of the body; the size of the room formed by the body; ventilation volume of the subject; the orientation of the body; and operation information of the air conditioner in a room adjacent to the room constituted by the main body.

A machine learning device according to a seventh aspect is the machine learning device according to any one of the first through sixth aspects, wherein the learning unit further includes a return setting unit. The return setting unit determines a return based on the room temperature and the set temperature after the start of the pre-cooling operation or the pre-warming operation. The learning unit learns the operating conditions of the pre-cooling operation or the pre-heating operation based on the state variables and the returns.

The machine learning device according to an eighth aspect is the machine learning device according to the seventh aspect, wherein the return setting unit further determines the return based on an amount of electric power required until a certain time after the start of the precooling operation or the preheating operation.

A machine learning device according to a ninth aspect is the machine learning device according to the eighth aspect, wherein the time after the start of the pre-cooling operation or the warm-up operation is the time when the pre-cooling operation or the warm-up operation is finished.

A machine learning device according to a tenth aspect is the machine learning device according to the eighth or ninth aspect, wherein the return setting unit increases the return when a difference between the room temperature and the set temperature at the end of the pre-cooling operation or the pre-warming operation is small, or increases the return when the amount of electric power is small.

The mechanical learning device according to an eleventh aspect is the mechanical learning device according to any one of the seventh through tenth aspects, wherein the learning unit further includes: a behavior cost function holding unit that holds a behavior cost function; and a behavior cost function updating unit that updates the behavior cost function. The behavioral merit function represents an expected value of return for which the operating condition is expected to be accepted. A behavior cost function updating unit updates the behavior cost function based on the actually obtained return.

Drawings

Fig. 1 is a schematic diagram showing the configuration of a machine learning apparatus 100.

Fig. 2 is a schematic diagram showing the structure of the air conditioner 10.

Fig. 3 is a block diagram of the control unit 40.

Fig. 4 is a block diagram of the mechanical learning apparatus 100.

Fig. 5 is a flowchart showing the processing of the machine learning apparatus 100.

Detailed Description

(1) Integral structure

Fig. 1 shows a machine learning apparatus 100. The machine learning device 100 includes an air conditioner 10. The air conditioner 10 includes a heat source unit 10a and a use unit 10 b. The air conditioner 10 is installed in a room R to be air-conditioned.

The machine learning apparatus 100 can perform the off-control. The out-of-control refers to control of the air conditioner 10 performed during a time when no user is present in the space (room R) to be air-conditioned. The cooling operation not performed during the control is referred to as a pre-cooling operation. The heating operation that is not being controlled is referred to as a warm-up operation.

(2) Detailed structure

(2-1) air conditioner 10

The air conditioner 10 conditions air in the room R. Fig. 2 shows a structure of the air conditioner 10. The air conditioner 10 includes a compressor 11, a four-way switching valve 12, a heat source heat exchanger 13, a heat source fan 14, an expansion valve 15, a use heat exchanger 16, a use fan 17, and a controller 40. The air conditioner 10 further includes an outside air temperature sensor 19a, a room temperature sensor 19b, a heat source heat exchanger temperature sensor 19c, a utilization heat exchanger temperature sensor 19d, a compressor intake temperature sensor 19e, a compressor discharge temperature sensor 19f, a pressure sensor not shown, and the like. In the case of the cooling operation, the refrigerant circulates in the direction of the solid arrow. In the heating operation, the refrigerant circulates in the direction of the broken-line arrow.

In the configuration of fig. 2, 1 heat source unit 10a and 1 usage unit 10b are connected. Alternatively, a configuration may be adopted in which 1 heat source unit 10a is connected to a plurality of usage units 10 b. The control unit 40 may be mounted on either the heat source unit 10a or the use unit 10 b. Alternatively, the control unit 40 may be mounted on each of the heat source unit 10a and the usage unit 10 b.

(2-2) control section 40

Fig. 3 is a block diagram of the control unit 40. The control unit 40 is, for example, a microcomputer. The control unit 40 functions as an air conditioner control unit 41, a room information acquisition unit 44, and an operation content determination unit 45 by executing a dedicated program.

The air conditioner control unit 41 controls the air conditioner 10. Specifically, as shown in fig. 2, the air conditioner control unit 41 controls the operations of the compressor 11, the four-way switching valve 12, the heat source fan 14, the expansion valve 15, and the utilization fan 17. The air conditioner control unit 41 acquires temperature information from the temperature sensors 19a to 19 f. The air conditioner control unit 41 holds the set temperature data Ts input by the user. The air conditioner control unit 41 then recognizes the power consumption of the air conditioner 10.

The room information acquiring unit 44 shown in fig. 3 acquires information of the time when the user goes out of the room R and the time when the user returns to the room R, for example, based on the statistical result output from the human detection sensor. Alternatively, the room information acquiring unit 44 may receive information on the time when the user has left the room R and the time when the user has returned to the room R by manual input.

The operation content determination unit 45 comprehensively determines how to operate the air conditioner 10 in the pre-cooling operation or the warm-up operation to be executed.

(3) Details of the operation content determining unit 45

(3-1) Structure

Fig. 4 is a block diagram of the mechanical learning apparatus 100. The operation content determination unit 45 includes an acquisition unit 451 and a learning unit 452.

The acquisition unit 451 acquires, as the state variables, the outputs of the temperature sensors 19a to 19f of the air conditioner 10, the output of the room information acquisition unit 44, and other signals. For example, the acquiring unit 451 acquires the room temperature data and the outside air temperature data as the state variables from the outputs of the room temperature sensor 19b and the outside air temperature sensor 19a, respectively. The acquiring unit 451 also acquires the set temperature data Ts held by the air conditioner control unit 41 as a state variable. Further, the acquisition unit 451 acquires, as the state variable, the time until the user returns to the room R based on the output from the room information acquisition unit 44. As the state variables, the acquisition unit 451 acquires at least one of the following items:

data relating to the amount of heat treatment of the air conditioner 10 (e.g., the rotational speed of the compressor 11, the temperature using the heat exchanger 16, the suction temperature measured on the suction side using the heat exchanger 16 (i.e., room temperature));

data of a main body surrounding a space (room R) to be air-conditioned by the utilization unit 10b of the air conditioner 10 (for example, the number of years elapsed for the main body, the heat insulation of the main body, the specification of a window of the main body, the size of the room constituted by the main body, the ventilation amount of the main body, the orientation of the main body, and operation information of the air conditioner of a room adjacent to the room R constituted by the main body);

-outside air temperature data before pre-cooling operation or pre-heating operation;

-sunshine data before pre-cooling operation or pre-heating operation;

-meteorological data before pre-cooling or pre-heating operation; and

season or month in pre-cooling or warming-up operation.

The learning unit 452 learns the operating conditions of the pre-cooling operation or the pre-heating operation based on a training data set including state variables, room temperature data at a certain time after the start of the pre-cooling operation or the pre-heating operation, and the set temperature data Ts. The operation condition referred to herein may include an operation start time of the pre-cooling operation or the warm-up operation. Alternatively, the operating condition may include at least one of the temperature using the heat exchanger 16, the rotation speed using the fan 17, and the rotation speed of the compressor 11.

The learning unit 452 includes a reward setting unit 453, a behavior-value-function updating unit 454, and a behavior-value-function holding unit 455. The reward setting unit 453 calculates the reward based on the next value when the pre-cooling operation or the warm-up operation is performed.

- (a) difference between room temperature and set temperature at the time when the user returns to the room R.

- (B) the amount of electric power required until a certain time after the start of the pre-cooling operation or the warm-up operation (for example, at the end of the pre-cooling operation or the warm-up operation).

Specifically, the reward setting unit 453 assigns a larger reward when the "(a) difference" is small, and assigns a smaller reward when the "(a) difference" is large. The reward setting unit 453 gives a large reward when the "(B) amount of electric power" is small, and gives a small reward when the "(B) amount of electric power" is large.

The behavior cost function holding unit 455 holds a function (behavior cost function) for calculating a control method for the pre-cooling operation or the pre-heating operation. The function referred to herein also includes a numerical value (action value table) expressed in a tabular form. The function may also represent an expected value of the reward expected to be accepted for the operating conditions.

The behavior cost function update unit 454 updates the behavior cost function held in the behavior cost function holding unit 455, based on the state variables acquired by the acquisition unit 451 and the returns calculated by the return setting unit 453.

In this way, the learning unit 452 learns the operation conditions for the pre-cooling operation or the pre-heating operation based on the state variables and the returns. The learning section 452 may update the behavior cost function in real time. Further, for this purpose, the learning unit 452 may calculate the state variable acquired from the acquisition unit 451 by a multilayer neural network. In this case, the learning unit 452 preferably performs reinforcement learning.

(3-2) operation Contents

The operation content determination unit 45 determines the operation content of the air conditioner 10. Here, the operation content of the air conditioner 10 may include an operation start time of the air conditioner 10. Alternatively, the operating condition may include at least one of the temperature using the heat exchanger 16, the rotation speed using the fan 17, and the rotation speed of the compressor 11.

(4) Treatment of

Fig. 5 is a flowchart showing the processing of the machine learning apparatus 100. The initial value of the "action" of reinforcement learning is sometimes selected randomly. In step S1, the control parameters of the air conditioner 10 are randomly selected.

In step S2, the acquisition unit 451 acquires a state variable. Here, the state variables include the outputs of at least some of the temperature sensors 19a to 19f of the air conditioner 10, the output of the room information acquisition unit 44, and other signals. For example, the state variables include the room temperature at the time of step S2, the outside air temperature, the power consumption of the air conditioner 10, the time until the user returns to the room R, and the like.

In step S3, the reward setting unit 453 determines whether or not the magnitude of the difference between the room temperature and the set temperature at the time when the user returns to the room R is less than a predetermined value when the certain pre-cooling operation or the pre-warming operation is performed. When it is determined that the magnitude of the difference is smaller than the predetermined value, the reward setting unit 453 increments the reward in step S4. On the other hand, when it is determined that the magnitude of the difference is equal to or larger than the predetermined value, the reward setting unit 453 decreases the reward in step S5.

When the magnitude of the electric power amount required for the pre-cooling operation or the warm-up operation is determined to be smaller than the predetermined value at step S6, the reward setting unit 453 increases the reward at step S7. On the other hand, when it is determined that the magnitude of the amount of electric power is equal to or greater than the predetermined value, the reward setting unit 453 decreases the reward in step S8.

In step S9, the behavior cost function update unit 454 updates a function (behavior cost function) for calculating the control parameter, based on the state variable and the report.

In step S10, it is determined whether or not the determination condition is satisfied. The decision condition may be an input based on a command of the user. Alternatively, the determination condition may be a condition that satisfies a predetermined relational expression. If the determination condition is satisfied, the process proceeds to step S11. If the determination condition is not satisfied, the process returns to step S2.

In step S11, the behavior merit function update unit 454 determines the control parameter for the pre-cooling operation or the warm-up operation that is the most rewarded based on the function.

(5) Feature(s)

The operating conditions for the pre-cooling operation or the pre-heating operation are derived by machine learning. Therefore, user comfort or energy saving effect is improved.

(6) Modification example

- (a) difference between room temperature and set temperature at the time when the user returns to the room R.

Instead of this, the user can,

- (C) a difference time between a time when the user returns to the room R and a time when the room temperature actually reaches the set temperature.

May also be used for processing.

(7) Summary of the invention

While the embodiments of the present invention have been described above, it should be understood that various changes in the form and details may be made therein without departing from the spirit and scope of the present disclosure as set forth in the appended claims.

Description of the reference symbols

10: an air conditioner; 11: a compressor; 13: a heat source heat exchanger; 14: a heat source fan; 15: an expansion valve; 16: utilizing a heat exchanger; 17: utilizing a fan; 19 a: an outside air temperature sensor; 19 b: a room temperature sensor; 40: a control unit; 100: a mechanical learning device; 451: an acquisition unit; 452: a learning unit; 453: a return setting unit; 454: a behavior cost function update unit; 455: a behavior cost function holding unit; r: a room; ts: and setting temperature data.

Documents of the prior art

Patent document

Patent document 1: japanese patent No. 6270996.