Machine learning device, air conditioning system, and machine learning method

Abstract

Provided is a machine learning device that optimizes the movement of a quantity of heat. The machine learning device learns the temperature and / or the flow rate when a heat conveyance device conveys a heat medium, in an air conditioning system including a heat supply-side device, a heat use-side device, and the heat conveyance device which conveys the heat medium from the heat supply-side device to the heat use-side device. The machine learning device includes: a state variable acquisition unit that acquires a state variable including an operation condition of the heat supply-side device, an operation condition of the heat use-side device, and a value correlated to the quantity of heat required by the heat use-side device; a learning unit that learns by associating the state variable and at least one of the temperature and the flow rate; and a remuneration calculation unit that calculates a remuneration on the basis of the total value of the consumed power of the heat supply-side device, the consumed power of the heat use-side device, and the consumed power of the heat conveyance device. The learning unit learns by using the remuneration.

Description

[Technical Field]

[0001] The present disclosure relates to a machine learning device, an air conditioning system, and a machine learning method. [Background technology]

[0002] In general, an air conditioning system is a system that moves heat by transporting a heat medium and adjusts the temperature or humidity in a target space, and configurations have been proposed to optimize the transfer of heat (optimizing the flow rate and temperature of the heat medium).

[0003] For example, Patent Document 1 listed below proposes a configuration in which the energy consumption of an air conditioning system is simulated and the transfer of heat is optimized so as to reduce the energy consumption. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2004-293844 [Patent Document 2] Japanese Patent Application Laid-Open No. 2004-53127 [Patent Document 3] Japanese Patent Application Publication No. 2018-173221 Summary of the Invention [Problem to be solved by the invention]

[0005] On the other hand, in the case of air conditioning systems, the optimal heat transfer (optimal heat medium flow rate and temperature) varies depending on the combination of the operating conditions of the equipment supplying the heat and the operating conditions or load of the equipment using the supplied heat.

[0006] Therefore, when trying to optimize the transfer of heat by simulating energy consumption as described above, it is necessary to obtain energy consumption data for a huge number of combinations in advance and build a model that corresponds to the characteristics of each device in the air conditioning system, which is a heavy workload.

[0007] The present disclosure provides a machine learning device, an air conditioning system, and a machine learning method for optimizing the transfer of heat quantity. [Means for solving the problem]

[0008] A machine learning device according to a first aspect of the present disclosure includes: A machine learning device for learning at least one of a temperature and a flow rate of a heat medium when the heat transfer device transfers the heat medium in an air conditioning system having a heat supply-side device, a heat utilization-side device, and a heat transfer device that transfers a heat medium from the heat supply-side device to the heat utilization-side device, comprising: a state variable acquisition unit that acquires state variables including an operating condition of the heat supplying device, an operating condition of the heat utilizing device, and a value that correlates with the amount of heat required for the heat utilizing device; a learning unit that learns the state variable in association with at least one of the temperature and the flow rate; a remuneration calculation unit that calculates a remuneration based on a total value of the power consumption of the heat supplying device, the power consumption of the heat using device, and the power consumption of the heat transfer device, The learning unit learns using the reward.

[0009] According to a first aspect of the present disclosure, a machine learning device that optimizes heat transfer can be provided.

[0010] A second aspect of the present disclosure is the machine learning device according to the first aspect, The operating conditions of the heat supplying equipment include any one of the outdoor air temperature, outdoor wet-bulb temperature, and underground temperature, which affect the processing capacity of the heat supplying equipment.

[0011] A third aspect of the present disclosure is the machine learning device according to the first aspect, The operating conditions of the heat-using equipment include either the intake air temperature or the chilled water return temperature, which affect the processing capacity of the heat-using equipment.

[0012] A fourth aspect of the present disclosure is the machine learning device according to the third aspect, The operating conditions of the heat-using equipment further include either the air volume or the chilled water flow rate.

[0013] A fifth aspect of the present disclosure is the machine learning device according to the first aspect, The value correlated with the amount of heat required by the heat-using device includes either the supply air temperature or the chilled water supply temperature.

[0014] A sixth aspect of the present disclosure is the machine learning device according to the fifth aspect, The temperature at which the heat transfer device transports the heat medium includes the cold water supply temperature and the cooling water supply temperature, and the flow rate at which the heat transfer device transports the heat medium includes either the cold water flow rate or the cooling water flow rate.

[0015] A seventh aspect of the present disclosure is the machine learning device according to the first aspect, The heat supplying equipment includes an air-cooled chiller, the heat utilizing equipment includes an air conditioning device, and the heat transporting device includes a chilled water pump.

[0016] An eighth aspect of the present disclosure is the machine learning device according to the first aspect, The heat supply side equipment includes a cooling tower, the heat utilization side equipment includes a water-cooled chiller, and the heat transfer device includes a cooling water pump.

[0017] A ninth aspect of the present disclosure is the machine learning device according to the first aspect, The heat supply side equipment includes a geothermal heat exchanger, the heat utilization side equipment includes a water-cooled chiller, and the heat transfer device includes a cooling water pump.

[0018] A tenth aspect of the present disclosure is the machine learning device according to the first aspect, The heat supply side equipment includes a cooling tower, a cooling water pump, and a water-cooled chiller, the heat utilization side equipment includes an air conditioning system, and the heat transfer device includes a chilled water pump.

[0019] An eleventh aspect of the present disclosure is the machine learning device according to the first aspect, The heat supply side equipment includes a cooling tower, the heat utilization side equipment includes a water-cooled chiller, a chilled water pump, and an air conditioning device, and the heat transfer device includes a chilled water pump.

[0020] A twelfth aspect of the present disclosure is the machine learning device according to the first aspect, If the risk to the air conditioning system increases due to the heat supply side equipment being operated based on at least one of the temperature and flow rate learned by the learning unit, the reward calculation unit reduces the reward.

[0021] A thirteenth aspect of the present disclosure is the machine learning device according to the first aspect, If at least one of the temperature and flow rate learned by the learning unit exceeds a predetermined upper limit or lower limit, the heat supply side equipment is operated based on the predetermined upper limit or lower limit.

[0022] In addition, an air conditioning system according to a fourteenth aspect of the present disclosure includes: An air conditioning system having a heat supply-side device, a heat utilization-side device, a heat transfer device that transfers a heat medium from the heat supply-side device to the heat utilization-side device, and a machine learning device that learns at least one of a temperature and a flow rate when the heat transfer device transfers the heat medium, The machine learning device includes: a state variable acquisition unit that acquires state variables including an operating condition of the heat supplying device, an operating condition of the heat utilizing device, and a value that correlates with the amount of heat required for the heat utilizing device; a learning unit that learns the state variable in association with at least one of the temperature and the flow rate; a remuneration calculation unit that calculates a remuneration based on a total value of the power consumption of the heat supplying device, the power consumption of the heat using device, and the power consumption of the heat transfer device, The learning unit learns using the reward.

[0023] According to the fourteenth aspect of the present disclosure, an air conditioning system that optimizes the transfer of heat can be provided.

[0024] Further, a machine learning method according to a fifteenth aspect of the present disclosure includes: A machine learning method for learning at least one of a temperature and a flow rate of a heat medium when the heat transfer device transfers the heat medium in an air conditioning system having a heat supply-side device, a heat utilization-side device, and a heat transfer device that transfers the heat medium from the heat supply-side device to the heat utilization-side device, comprising: a state variable acquisition step of acquiring state variables including an operating condition of the heat supplying device, an operating condition of the heat utilizing device, and a value correlating with the amount of heat required for the heat utilizing device; a learning step of associating the state variable with at least one of the temperature and the flow rate and learning the state variable; a remuneration calculation step of calculating a remuneration based on a total value of the power consumption of the heat supplying device, the power consumption of the heat using device, and the power consumption of the heat transfer device, The learning step involves learning using the reward.

[0025] According to a fifteenth aspect of the present disclosure, a machine learning method for optimizing heat transfer can be provided.

[0026] A machine learning device according to a sixteenth aspect of the present disclosure, Water-cooled chiller and a cooling water pump that supplies cooling water that cools the refrigerant by heat exchange in the water-cooled chiller; a cooling tower that cools the cooling water transported from the water-cooled chiller by bringing it into contact with outside air; An air conditioning device; a chilled water pump that supplies chilled water cooled by the refrigerant through heat exchange in the water-cooled chiller to the air conditioning device, a machine learning device that learns at least one pair of a temperature of the chilled water supplied by the chilled water pump and a temperature of the chilled water supplied by the chilled water pump, or a pair of a flow rate of the chilled water supplied by the chilled water pump and a flow rate of the chilled water supplied by the chilled water pump, a state variable acquisition unit that acquires state variables including an operating condition of the cooling tower, an operating condition of the air conditioning device, and a load of the air conditioning device; a learning unit that learns the state variable by associating it with at least one of the sets; a reward calculation unit that calculates a reward based on the total value of power consumption of the cooling tower, the water-cooled chiller, the cooling water pump, the chilled water pump, and the air conditioning device, The learning unit learns using the reward.

[0027] According to a sixteenth aspect of the present disclosure, a machine learning device that optimizes heat transfer can be provided.

[0028] Furthermore, a machine learning device according to a seventeenth aspect of the present disclosure includes: a chiller unit for heating or cooling water; a water pump for supplying water heated or cooled by the chiller unit; a heat exchanger that exchanges heat between air passing through the heat exchanger and water supplied by the pump, and an air conditioner that sends the air that has passed through the heat exchanger to a target space; and a machine learning device that learns at least one of the temperature and flow rate of water supplied by the water pump, a state variable acquisition unit that acquires state variables including an operating condition of the chiller unit, an operating condition of the air conditioning device, and a load of the air conditioning device; a learning unit that learns the state variable in association with at least one of the temperature and the flow rate; a reward calculation unit that calculates a reward based on a total value of the power consumption of the chiller unit, the air conditioning device, and the water pump, The learning unit learns using the reward.

[0029] According to a seventeenth aspect of the present disclosure, a machine learning device that optimizes heat transfer can be provided. [Brief explanation of the drawings]

[0030] [Figure 1] FIG. 1 is a diagram illustrating an example of a system configuration of an air conditioning system. [Figure 2] FIG. 2 is a diagram illustrating an example of a hardware configuration of a machine learning device. [Figure 3] FIG. 3 is a first diagram illustrating an example of the functional configuration of a machine learning device. [Figure 4] FIG. 4 is a first flowchart showing the flow of reinforcement learning processing by the machine learning device. [Figure 5] FIG. 5 is a first diagram showing specific examples of heat supplying side equipment and heat using side equipment. [Figure 6] FIG. 6 is a second diagram showing specific examples of the heat supplying side equipment and the heat using side equipment. [Figure 7] FIG. 7 is a third diagram showing specific examples of the heat supplying side equipment and the heat using side equipment. [Figure 8] FIG. 8 is a fourth diagram showing specific examples of the heat supplying side equipment and the heat using side equipment. [Figure 9] FIG. 9 is a fifth diagram showing specific examples of heat supplying side equipment and heat using side equipment. [Figure 10] FIG. 10 is a sixth diagram showing specific examples of heat supplying side equipment and heat using side equipment. [Figure 11] FIG. 11 is a second diagram illustrating an example of the functional configuration of the machine learning device. [Figure 12] FIG. 12 is a second flowchart showing the flow of reinforcement learning processing by the machine learning device. [Figure 13] FIG. 13 is a third diagram illustrating an example of the functional configuration of the machine learning device. [Figure 14]FIG. 14 is a third flowchart showing the flow of reinforcement learning processing by the machine learning device. [Figure 15] FIG. 15 is a first diagram showing an example of a system configuration of an air conditioning system including a coolant circuit and a chilled water circuit. [Figure 16] FIG. 16 is a diagram showing the details of the cooling water circuit. [Figure 17] FIG. 17 is a first diagram illustrating the function of the heat medium control device. [Figure 18] FIG. 18 is a fourth diagram illustrating an example of the functional configuration of the machine learning device. [Figure 19] FIG. 19 is a fourth flowchart showing the flow of reinforcement learning processing by the machine learning device. [Figure 20] FIG. 20 is a first diagram showing an example of the system configuration of an air conditioning system including a water circuit. [Figure 21] FIG. 21 is a first diagram showing a detailed configuration of an air conditioner. [Figure 22] FIG. 22 is a second diagram illustrating the function of the heat medium control device. [Figure 23] FIG. 23 is a fifth diagram illustrating an example of the functional configuration of the machine learning device. [Figure 24] FIG. 24 is a fifth flowchart showing the flow of reinforcement learning processing by the machine learning device. [Figure 25] FIG. 25 is a second diagram showing an example of the system configuration of an air conditioning system including a water circuit. [Figure 26] FIG. 26 is a second diagram showing the detailed configuration of the air conditioner. [Figure 27] FIG. 27 is a first diagram showing an installation mode of the fan coil unit in the target space. [Figure 28] FIG. 28 is a third diagram showing an example of the system configuration of an air conditioning system including a water circuit. [Figure 29] FIG. 29 is a third diagram showing the detailed configuration of the air conditioner. [Figure 30] FIG. 30 is a second diagram showing the installation mode of the fan coil unit in the target space. DETAILED DESCRIPTION OF THE INVENTION

[0031] Hereinafter, each embodiment will be described with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functional configuration are designated by the same reference numerals, and redundant description will be omitted.

[0032] [First embodiment] <Air conditioning system configuration> First, the system configuration of an air conditioning system according to the first embodiment will be described. Fig. 1 is a diagram showing an example of the system configuration of an air conditioning system. As shown in Fig. 1, an air conditioning system 100 includes air conditioning equipment 110 and a machine learning device 150.

[0033] The air conditioning equipment 110 includes multiple devices, which can be broadly divided into devices on the heat supply side 120, devices on the heat utilization side 140, and a heat transfer device 130. Here, the heat utilization side 140 refers to the side that receives the supply of heat and is closer to the target space to be air-conditioned in the direction of the heat medium transport. The heat supply side 120 refers to the side that supplies heat and is farther from the target space to be air-conditioned in the direction of the heat medium transport.

[0034] The heat transfer device 130 is a device that transfers heat quantity by transporting a heat medium from equipment on the heat supply side 120 to equipment on the heat utilization side 140. The heat quantity is determined based on the temperature and flow rate of the heat medium.

[0035] As shown in FIG. 1, the equipment on the heat supply side 120 operates under predetermined operating conditions to achieve the target value of at least one of the temperature and flow rate of the heat medium transmitted from the machine learning device 150.

[0036] In addition, the equipment on the heat utilization side 140 operates to achieve target values ​​previously set for the equipment on the heat utilization side 140 under predetermined operating conditions and the temperature and flow rate of the heat medium transported from the equipment on the heat supply side 120.

[0037] On the other hand, the machine learning device 150 -Operating conditions of the heat supply side 120 equipment, -Operating conditions of 140 heat utilization equipment, The load of the heat-using device 140 (a value correlating with the amount of heat required to achieve the target value in the heat-using device 140), is acquired from the air conditioning equipment 110 as a “state variable”.

[0038] In addition, the machine learning device 150 obtains from the air conditioning equipment 110 the power consumption of the equipment on the heat supply side 120, the power consumption of the equipment on the heat use side 140, and the power consumption of the heat transfer device 130 over a specified period, and calculates the "total power consumption."

[0039] Furthermore, the machine learning device 150 calculates a target value for at least one of the temperature and flow rate of the heat medium based on the state variables and total power consumption acquired from the air conditioning equipment 110, and transmits the calculated target value to the equipment on the heat supply side 120. Specifically, the machine learning device 150 uses the reward calculated based on the acquired total power consumption to associate the acquired state variables with the target values ​​for at least one of the temperature and flow rate of the heat medium, and performs learning. Furthermore, based on the results of learning, the machine learning device 150 calculates a target value for at least one of the temperature and flow rate of the heat medium associated with the current state variables, and transmits the calculated target value to the equipment on the heat supply side 120.

[0040] In this way, the machine learning device 150 performs reinforcement learning using a reward calculated based on the total power consumption of the air conditioning equipment 110, thereby changing the model parameters to reduce power consumption and calculating the target value for at least one of the temperature and flow rate of the heat medium. This makes it possible to optimize the transfer of heat quantity in the air conditioning system 100.

[0041] Furthermore, the machine learning device 150 automatically constructs a model that associates state variables, including operating conditions and loads, with target values ​​for at least one of the temperature and flow rate of the heat medium while acquiring actual data. This allows the machine learning device 150 to easily construct a highly accurate model.

[0042] <Hardware configuration of machine learning device> Next, the hardware configuration of the machine learning device 150 will be described. FIG. 2 is a diagram showing an example of the hardware configuration of a machine learning device. As shown in FIG. 2, the machine learning device 150 has a CPU (Central Processing Unit) 201, a ROM (Read Only Memory) 202, and a RAM (Random Access Memory) 203. The CPU 201, ROM 202, and RAM 203 form a so-called computer. The machine learning device 150 also has an auxiliary storage device 204, a display device 205, an operation device 206, and an I / F (Interface) device 207. The various hardware components of the machine learning device 150 are connected to one another via a bus 208.

[0043] The CPU 201 is a computing device that executes various programs (for example, a machine learning program, which will be described later) installed in the auxiliary storage device 204. The ROM 202 is a non-volatile memory. The ROM 202 functions as a main storage device and stores various programs and data necessary for the CPU 201 to execute the various programs installed in the auxiliary storage device 204. Specifically, the ROM 202 stores boot programs such as a BIOS (Basic Input / Output System) and an EFI (Extensible Firmware Interface).

[0044] The RAM 203 is a volatile memory such as a dynamic random access memory (DRAM) or a static random access memory (SRAM). The RAM 203 functions as a main storage device and provides a working area in which various programs installed in the auxiliary storage device 204 are expanded when the CPU 201 executes them.

[0045] The auxiliary storage device 204 stores various programs and information used when the various programs are executed.

[0046] Display device 205 is a display device that displays the internal state of machine learning device 150. Operation device 206 is an operation device that allows, for example, an administrator of machine learning device 150 to perform various operations on machine learning device 150. I / F device 207 is a connection device that connects to devices included in air conditioning equipment 110 and transmits and receives data to and from the devices included in air conditioning equipment 110.

[0047] <Functional configuration of machine learning device> Next, the functional configuration of the machine learning device 150 will be described in detail. Figure 3 is a first diagram showing an example of the functional configuration of a machine learning device. As described above, a machine learning program is installed in the machine learning device 150, and by executing this program, the machine learning device 150 functions as a power consumption acquisition unit 310, a reward calculation unit 320, a state variable acquisition unit 330, and a reinforcement learning unit 340.

[0048] The power consumption acquisition unit 310 acquires the power consumption for a specified period by the equipment on the heat supply side 120, the power consumption for a specified period by the equipment on the heat use side 140, and the power consumption for a specified period by the heat transfer device 130, and notifies the remuneration calculation unit 320 of the total value.

[0049] The reward calculation unit 320 calculates a reward based on the total value notified by the power consumption acquisition unit 310 and notifies the reinforcement learning unit 340 of the calculated reward.

[0050] The state variable acquisition unit 330 acquires, as state variables, the operating conditions of the equipment on the heat supply side 120 for a predetermined period, the operating conditions of the equipment on the heat use side 140 for a predetermined period, and the load of the equipment on the heat use side 140 for a predetermined period from each equipment included in the air conditioning equipment 110. In addition, the state variable acquisition unit 330 notifies the reinforcement learning unit 340 of the acquired state variables.

[0051] The reinforcement learning unit 340 has a heat quantity model 341, and changes the model parameters of the heat quantity model 341 so as to maximize the reward notified by the reward calculation unit 320. In this way, the reinforcement learning unit 340 performs reinforcement learning on the heat quantity model 341 that associates the state variables with the target values ​​of at least one of the temperature and flow rate of the heat medium.

[0052] Furthermore, the reinforcement learning unit 340 acquires a target value of at least one of the temperature and flow rate of the heat medium, which is calculated by inputting the current state variables notified by the state variable acquisition unit 330 into the heat quantity model 341 whose model parameters have been changed. Furthermore, the reinforcement learning unit 340 transmits the acquired target value of at least one of the temperature and flow rate of the heat medium to the equipment on the heat supply side 120. As a result, the equipment on the heat supply side 120 operates to achieve the transmitted target value of at least one of the temperature and flow rate of the heat medium. As a result, the machine learning device 150 can reduce the power consumption of the air conditioning equipment 110.

[0053] <Reinforcement learning process flow> Next, we will explain the flow of reinforcement learning processing by the machine learning device 150. Figure 4 is a first flowchart showing the flow of reinforcement learning processing by the machine learning device.

[0054] In step S401, the state variable acquisition unit 330 acquires state variables for a predetermined period from each device included in the air conditioning equipment 110.

[0055] In step S402, the power consumption acquisition unit 310 acquires the power consumption for a predetermined period by the equipment on the heat supply side 120, the power consumption for a predetermined period by the equipment on the heat use side 140, and the power consumption for a predetermined period by the heat transfer device 130, and calculates the total value.

[0056] In step S403, the remuneration calculation unit 320 calculates the remuneration based on the calculated total value.

[0057] In step S404, the remuneration calculation unit 320 determines whether the calculated remuneration is equal to or greater than a predetermined threshold. If it is determined in step S404 that the calculated remuneration is not equal to or greater than the predetermined threshold (NO in step S404), the process proceeds to step S405.

[0058] In step S405, the reinforcement learning unit 340 performs machine learning on the heat quantity model 341 so as to maximize the calculated reward.

[0059] In step S406, the reinforcement learning unit 340 inputs the current state variables to the heat quantity model 341, thereby executing the heat quantity model 341. As a result, the reinforcement learning unit 340 outputs a target value for at least one of the temperature and flow rate of the heat medium.

[0060] In step S407, the reinforcement learning unit 340 transmits the output target value of at least one of the temperature and flow rate of the heat medium to the device on the heat supply side 120. Then, the process returns to step S401.

[0061] On the other hand, if it is determined in step S404 that the difference is equal to or greater than the predetermined threshold (YES in step S404), the reinforcement learning process ends.

[0062] <Summary> As is clear from the above description, the air conditioning system according to the first embodiment Heat supply equipment and -Heat utilization equipment and A heat transfer device that transfers a heat medium from a heat supplying device to a heat using device; a machine learning device that learns at least one of the temperature and the flow rate when the heat transfer device transfers the heat medium; It has.

[0063] In addition, the machine learning device · Acquire state variables including the operating conditions of the heat supply side equipment, the operating conditions of the heat use side equipment, and the load of the heat use side equipment. · Learning is performed by associating state variables with at least one of temperature and flow rate. Compensation is calculated based on the total power consumption of the heat supply equipment, the heat use equipment, and the heat transport equipment. When learning by associating the state variables with at least one of the temperature and flow rate of the heat transfer medium, the calculated reward is used.

[0064] In this way, the machine learning device performs reinforcement learning using a reward calculated based on the total power consumption of each device, thereby changing model parameters to reduce power consumption and calculating target values ​​for at least one of the temperature and flow rate of the heat medium. The machine learning device also automatically builds a model that associates state variables, including operating conditions, with target values ​​for at least one of the temperature and flow rate of the heat medium while acquiring actual data.

[0065] As a result, according to the first embodiment, it is possible to easily construct a highly accurate model and to optimize the transfer of heat quantity.

[0066] [Second embodiment] In the first embodiment described above, no specific examples of the equipment on the heat supply side 120 and the equipment on the heat utilization side 140 were mentioned. However, the air conditioning equipment 110 includes various equipment, and there are various combinations of the equipment on the heat supply side 120 and the equipment on the heat utilization side 140.

[0067] Therefore, in the second embodiment, specific examples of the equipment on the heat supply side 120 and the equipment on the heat utilization side are 140A learning dataset used in reinforcement learning will be described in detail using specific examples of the above. In the second embodiment, a case will first be described in which the devices included in the heat supply side 120 or the heat utilization side 140 are formed in multiple stages. However, in the second embodiment, for the sake of simplicity, the description will be centered on the differences from the first embodiment.

[0068] <Example 1 of heat supply equipment and heat use equipment> Fig. 5 is a first diagram showing specific examples of heat supplying side equipment and heat using side equipment. In the specific example shown in Fig. 5, the equipment on the heat supplying side 120 includes a cooling tower 501, a water-cooled chiller 502, and a cooling water pump 503. The cooling tower 501 includes, for example, an open-type cooling tower and a closed-type cooling tower.

[0069] 5, the equipment on the heat utilization side 140 includes an air conditioner 511. The air conditioner 511 includes, for example, an outdoor air handling unit, an indoor fan coil unit, and a water source heat pump air conditioner.

[0070] In the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat supply side 120, which are included in the state variables and affect the processing capacity of the equipment on the heat supply side 120. -Outside air wet bulb temperature of open cooling tower, -Outside air wet bulb temperature of closed cooling tower When acquiring the outside air wet-bulb temperature from an open-type cooling tower or a closed-type cooling tower, the machine learning device 150 may acquire the outside air wet-bulb temperature by replacing it with the outside air temperature.

[0071] Furthermore, in the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat utilization side 140, which are included in the state variables and affect the processing capacity of the equipment on the heat utilization side 140, as follows: - Intake air temperature, intake air humidity (and air volume) of the outdoor air handling unit Indoor fan coil unit intake air temperature, intake air humidity (and air volume), - Water source heat pump air conditioner intake air temperature, intake air humidity (and air volume), The machine learning device 150 acquires data such as the above. When acquiring the intake air temperature and intake air humidity from the outdoor air-conditioning air handling unit, the machine learning device 150 may replace the intake air temperature and intake air humidity with the outdoor air temperature and outdoor air humidity. When acquiring the intake air temperature and intake air humidity from the indoor fan coil unit, the machine learning device 150 may replace the intake air temperature and intake air humidity with the indoor temperature and indoor humidity. When acquiring the intake air temperature and intake air humidity from the water-source heat pump air conditioner, the machine learning device 150 may replace the intake air temperature and intake air humidity with the outdoor air temperature and outdoor air humidity, or the indoor temperature and indoor humidity. Alternatively, the machine learning device 150 may replace the intake air temperature and intake air humidity with the refrigerant pressure, refrigerant temperature, and compressor operating status.

[0072] Furthermore, in the above specific example, the machine learning device 150 calculates the load of the equipment on the heat utilization side 140, which is included in the state variables and correlates with the amount of heat required to achieve the target value in the equipment on the heat utilization side 140. - Supply air temperature and humidity of the outdoor air handling unit, or intake air temperature and humidity of the intake air (and air volume), Indoor fan coil unit supply air temperature, supply air humidity, or intake air temperature, intake air humidity (and air volume), - Compressor load factor of water source heat pump air conditioner, The machine learning device 150 may calculate data correlated with the amount of heat required to achieve a target value in the equipment on the heat utilization side 140 from the chilled water side instead of acquiring the data from the air side or the compressor load factor.

[0073] Furthermore, in the above specific example, the machine learning device 150 uses the following as the power consumption of the heat supply side 120 for calculating the total power consumption: · Power consumption of cooling tower 501, ·Power consumption of water-cooled chiller 502, Power consumption of the cooling water pump 503, The data is acquired as the power consumption of the heat utilization side 140 for calculating the total power consumption. Power consumption of the air conditioner 511, The data such as the above is acquired, and the power consumption of the heat transfer device 130 for calculating the total power consumption is calculated as follows: · Power consumption of cold water pump 521, Obtain data such as:

[0074] Furthermore, in the above specific example, the machine learning device 150 determines the target value of at least one of the temperature and flow rate of the heat medium as follows: - Target value of chilled water supply temperature, · Target chilled water flow rate, Calculate.

[0075] <Example 2 of heat supply equipment and heat use equipment> Fig. 6 is a second diagram showing specific examples of heat supplying side equipment and heat using side equipment. In the specific example shown in Fig. 6, the equipment on the heat supplying side 120 includes a cooling tower 601. The cooling tower 601 includes, for example, an open-type cooling tower and a closed-type cooling tower.

[0076] 6, the equipment on the heat utilization side 140 includes a water-cooled chiller 611, an air conditioner 612, and a chilled water pump 613. The air conditioner 612 includes, for example, an outdoor air handling unit, an indoor fan coil unit, and a water-source heat pump air conditioner.

[0077] In the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat supply side 120, which are included in the state variables and affect the processing capacity of the equipment on the heat supply side 120. -Outside air wet bulb temperature of open cooling tower, -Outside air wet bulb temperature of closed cooling tower When acquiring the outside air wet-bulb temperature from an open-type cooling tower or a closed-type cooling tower, the machine learning device 150 may acquire the outside air wet-bulb temperature by replacing it with the outside air temperature.

[0078] Furthermore, in the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat utilization side 140, which are included in the state variables and affect the processing capacity of the equipment on the heat utilization side 140, as follows: - Intake air temperature, intake air humidity (and air volume) of the outdoor air handling unit Indoor fan coil unit intake air temperature, intake air humidity (and air volume), - Water source heat pump air conditioner intake air temperature, intake air humidity (and air volume), The machine learning device 150 acquires data such as the above. When acquiring the intake air temperature and intake air humidity from the outdoor air-conditioning air handling unit, the machine learning device 150 may replace the intake air temperature and intake air humidity with the outdoor air temperature and outdoor air humidity. When acquiring the intake air temperature and intake air humidity from the indoor fan coil unit, the machine learning device 150 may replace the intake air temperature and intake air humidity with the indoor temperature and indoor humidity. When acquiring the intake air temperature and intake air humidity from the water-source heat pump air conditioner, the machine learning device 150 may replace the intake air temperature and intake air humidity with the outdoor air temperature and outdoor air humidity, or the indoor temperature and indoor humidity. Alternatively, the machine learning device 150 may replace the intake air temperature and intake air humidity with the refrigerant pressure, refrigerant temperature, and compressor operating status.

[0079] Furthermore, in the above specific example, the machine learning device 150 calculates the load of the equipment on the heat utilization side 140, which is included in the state variables and correlates with the amount of heat required to achieve the target value in the equipment on the heat utilization side 140. - Supply air temperature and humidity of the outdoor air handling unit, or intake air temperature and humidity of the intake air (and air volume), Indoor fan coil unit supply air temperature, supply air humidity, or intake air temperature, intake air humidity (and air volume), - Compressor load factor of water source heat pump air conditioner, The machine learning device 150 may calculate data correlated with the amount of heat required to achieve a target value in the equipment on the heat utilization side 140 from the chilled water side instead of acquiring the data from the air side or the compressor load factor.

[0080] Furthermore, in the above specific example, the machine learning device 150 uses the following as the power consumption of the heat supply side 120 for calculating the total power consumption: ·Cooling tower 601 power consumption, The data is acquired as the power consumption of the heat utilization side 140 for calculating the total power consumption. ·Power consumption of water-cooled chiller 611, Power consumption of the air conditioner 612 · Power consumption of cold water pump 613, The data such as the above is acquired, and the power consumption of the heat transfer device 130 for calculating the total power consumption is calculated as follows: · Power consumption of cooling water pump 621, Obtain data such as:

[0081] Furthermore, in the above specific example, the machine learning device 150 determines the target value of at least one of the temperature and flow rate of the heat medium as follows: - Target value of cooling water supply temperature, - Target value of cooling water flow rate, Calculate.

[0082] <Summary> As is clear from the above description, the machine learning device 150 can perform reinforcement learning using the learning data sets shown in the specific examples 1 and 2 above.

[0083] [Third embodiment] In the second embodiment, the devices included in the heat supplying side 120 or the heat utilizing side 140 are configured in multiple stages. In contrast, in the third embodiment, the devices included in the heat supplying side 120 and the heat utilizing side 140 are both configured in a single stage.

[0084] <Example 1 of heat supply equipment and heat use equipment> 7 is a third diagram showing specific examples of heat supplying side equipment and heat using side equipment. In the specific example shown in FIG. 7, the equipment on the heat supplying side 120 includes an air-cooled chiller 701 (an example of a chiller unit). The air-cooled chiller 701 includes, for example, an air-cooled heat pump refrigerator.

[0085] 7, the equipment on the heat utilization side 140 includes an air conditioner 711. The air conditioner 711 includes, for example, an outdoor air handling unit, an indoor fan coil unit, and a water source heat pump air conditioner.

[0086] In the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat supply side 120, which are included in the state variables and affect the processing capacity of the equipment on the heat supply side 120. -Outside air temperature when cooling with an air-cooled heat pump refrigerator, -Wet bulb temperature of outside air when heating with an air-cooled heat pump refrigerator When acquiring the outdoor air temperature or the outdoor wet-bulb temperature from the air-cooled heat pump refrigerator, the machine learning device 150 may acquire the outdoor air temperature or the outdoor wet-bulb temperature by interpreting it as the operating status of the compressor and the fan.

[0087] Furthermore, in the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat utilization side 140, which are included in the state variables and affect the processing capacity of the equipment on the heat utilization side 140, as follows: - Intake air temperature, intake air humidity (and air volume) of the outdoor air handling unit Indoor fan coil unit intake air temperature, intake air humidity (and air volume), - Water source heat pump air conditioner intake air temperature, intake air humidity (and air volume), The machine learning device 150 acquires data such as the above. When acquiring the intake air temperature and intake air humidity from the outdoor air handling unit, the intake air temperature and intake air humidity may be read as outdoor air temperature and outdoor air humidity. When acquiring the intake air temperature and intake air humidity from the indoor fan coil unit, the machine learning device 150 may acquire the intake air temperature and intake air humidity as indoor temperature and indoor humidity. When acquiring the intake air temperature and intake air humidity from the water-source heat pump air conditioner, the machine learning device 150 may acquire the intake air temperature and intake air humidity as outdoor air temperature and outdoor air humidity, or indoor temperature and indoor humidity. Alternatively, the machine learning device 150 may acquire the intake air temperature and intake air humidity as refrigerant pressure, refrigerant temperature, and compressor operating status.

[0088] Furthermore, in the above specific example, the machine learning device 150 calculates the load of the equipment on the heat utilization side 140, which is included in the state variables and correlates with the amount of heat required to achieve the target value in the equipment on the heat utilization side 140. - Supply air temperature and humidity of the outdoor air handling unit, or intake air temperature and humidity of the intake air (and air volume), Indoor fan coil unit supply air temperature, supply air humidity, or intake air temperature, intake air humidity (and air volume), - Compressor load factor of water source heat pump air conditioner, The machine learning device 150 may calculate data correlated with the amount of heat required to achieve a target value in the equipment on the heat utilization side 140 from the chilled water side instead of acquiring the data from the air side or the compressor load factor.

[0089] Furthermore, in the above specific example, the machine learning device 150 uses the following as the power consumption of the heat supply side 120 for calculating the total power consumption: ·Power consumption of air-cooled chiller 701, The data is acquired as the power consumption of the heat utilization side 140 for calculating the total power consumption. Power consumption of the air conditioner 711, The data such as the above is acquired, and the power consumption of the heat transfer device 130 for calculating the total power consumption is calculated as follows: · Power consumption of chilled water pump 721, Obtain data such as:

[0090] Furthermore, in the above specific example, the machine learning device 150 determines the target value of at least one of the temperature and flow rate of the heat medium as follows: - Target value of chilled water supply temperature, · Target chilled water flow rate, Calculate.

[0091] <Example 2 of heat supply equipment and heat use equipment> Fig. 8 is a fourth diagram showing specific examples of heat supply-side equipment and heat use-side equipment. In the specific example shown in Fig. 8, the equipment on the heat supply side 120 includes a cooling tower or underground heat exchanger 801. The cooling tower or underground heat exchanger 801 includes, for example, an open-type cooling tower, a closed-type cooling tower, and an underground heat exchanger.

[0092] 8, the equipment on the heat utilization side 140 includes a water-cooled chiller 811. The water-cooled chiller 811 includes, for example, a water-cooled heat pump refrigerator and an absorption refrigerator.

[0093] In the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat supply side 120, which are included in the state variables and affect the processing capacity of the equipment on the heat supply side 120. -Outside air wet bulb temperature of open cooling tower, -Outside air wet bulb temperature of closed cooling tower - Ground temperature of the ground heat exchanger, The machine learning device 150 acquires data such as the above. When acquiring the outdoor wet-bulb temperature from an open-type cooling tower or a closed-type cooling tower, the machine learning device 150 may acquire the outdoor wet-bulb temperature by replacing it with the outdoor air temperature. When acquiring the underground temperature from a underground heat exchanger, the machine learning device 150 may acquire the underground temperature by replacing it with the underground heat exchanger outlet temperature.

[0094] Furthermore, in the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat utilization side 140, which are included in the state variables and affect the processing capacity of the equipment on the heat utilization side 140, as follows: - Chilled water return temperature (or chilled water return temperature and chilled water flow rate) of water-cooled heat pump refrigeration equipment, - Absorption chiller chiller return water temperature (or chilled water return water temperature and chilled water flow rate), Obtain data such as:

[0095] Furthermore, in the above specific example, the machine learning device 150 calculates the load of the equipment on the heat utilization side 140, which is included in the state variables and correlates with the amount of heat required to achieve the target value in the equipment on the heat utilization side 140. - Load factor of water-cooled heat pump refrigeration unit, chilled water supply temperature (or chilled water return temperature and chilled water flow rate), Absorption chiller chiller chilled water supply temperature (or chilled water return temperature, chilled water flow rate), Obtain data such as:

[0096] Furthermore, in the above specific example, the machine learning device 150 uses the following as the power consumption of the heat supply side 120 for calculating the total power consumption: Cooling tower power consumption, The data is acquired as the power consumption of the heat utilization side 140 for calculating the total power consumption. ·Power consumption of water-cooled chiller 811, The data such as the above is acquired, and the power consumption of the heat transfer device 130 for calculating the total power consumption is calculated as follows: · Power consumption of cooling water pump 821, Obtain data such as:

[0097] Furthermore, in the above specific example, the machine learning device 150 determines the target value of at least one of the temperature and flow rate of the heat medium as follows: - Target value of cooling water supply temperature, - Target value of cooling water flow rate, Calculate.

[0098] <Summary> As is clear from the above description, the machine learning device 150 can perform reinforcement learning using the learning data sets shown in the specific examples 1 and 2 above.

[0099] [Fourth embodiment] In the third embodiment, the devices included in the heat supplying side 120 or the heat utilizing side 140 are configured in a single stage. In contrast, in the fourth embodiment, the devices included in the heat supplying side 120 and the heat utilizing side 140 are both configured in parallel in a single stage.

[0100] <Example 1 of heat supply equipment and heat use equipment> Fig. 9 is a fifth diagram showing specific examples of heat supplying side equipment and heat using side equipment. In the specific example shown in Fig. 9, the equipment on the heat supplying side 120 includes an air-cooled chiller 901 and a cooling tower 902. Note that the types of equipment included in the air-cooled chiller 901 and the cooling tower 902 have already been explained, so explanation will be omitted here.

[0101] 9, the devices on the heat utilization side 140 include an air conditioner 911. Note that the types of devices included in the air conditioner 911 have already been explained, and therefore will not be explained here.

[0102] In the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat supply side 120, which are included in the state variables and affect the processing capacity of the equipment on the heat supply side 120. -Outside air temperature when cooling with an air-cooled heat pump refrigerator, -Wet bulb temperature of outside air when heating with an air-cooled heat pump refrigerator -Outside air wet bulb temperature of open cooling tower, -Outside air wet bulb temperature of closed cooling tower The machine learning device 150 acquires data such as the above. Note that the interpretation of these operating conditions when acquired by the machine learning device 150 has already been explained, and therefore the explanation will be omitted here.

[0103] Furthermore, in the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat utilization side 140, which are included in the state variables and affect the processing capacity of the equipment on the heat utilization side 140, as follows: - Intake air temperature, intake air humidity (and air volume) of the outdoor air handling unit Indoor fan coil unit intake air temperature, intake air humidity (and air volume), - Water source heat pump air conditioner intake air temperature, intake air humidity (and air volume), Obtain data such as:

[0104] Furthermore, in the above specific example, the machine learning device 150 calculates the load of the equipment on the heat utilization side 140, which is included in the state variables and correlates with the amount of heat required to achieve the target value in the equipment on the heat utilization side 140. - Supply air temperature and humidity of the outdoor air handling unit, or intake air temperature and humidity of the intake air (and air volume), Indoor fan coil unit supply air temperature, supply air humidity, or intake air temperature, intake air humidity (and air volume), - Compressor load factor of water source heat pump air conditioner, Obtain data such as:

[0105] Furthermore, in the above specific example, the machine learning device 150 uses the following as the power consumption of the heat supply side 120 for calculating the total power consumption: ·Power consumption of air-cooled chiller 901, · Power consumption of cooling tower 902, The data is acquired as the power consumption of the heat utilization side 140 for calculating the total power consumption. · Power consumption of the air conditioning unit 911, The data such as the above is acquired, and the power consumption of the heat transfer device 130 for calculating the total power consumption is calculated as follows: · Power consumption of cold water pump 921, Obtain data such as:

[0106] Furthermore, in the above specific example, the machine learning device 150 determines the target value of at least one of the temperature and flow rate of the heat medium as follows: - Target value of chilled water supply temperature, · Target chilled water flow rate, Calculate.

[0107] <Example 2 of heat supply equipment and heat use equipment> Fig. 10 is a sixth diagram showing specific examples of heat supplying side equipment and heat using side equipment. In the specific example shown in Fig. 10, the equipment on the heat supplying side 120 includes an air-cooled chiller 1001. Note that the types of equipment included in the air-cooled chiller 1001 have already been explained, so explanation will be omitted here.

[0108] 10, the equipment on the heat utilization side 140 includes an air conditioner 1011, a water-cooled building multi-air conditioner 1012, and a water-cooled building multi-air conditioner 1013. The types of equipment included in the air conditioner 1011 have already been explained, so explanations will be omitted here. The water-cooled building multi-air conditioners 1012 and 1013 include, for example, heat source units and indoor units of the water-cooled building multi-air conditioners.

[0109] In the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat supply side 120, which are included in the state variables and affect the processing capacity of the equipment on the heat supply side 120. -Outside air temperature when cooling with an air-cooled heat pump refrigerator, -Wet bulb temperature of outside air when heating with an air-cooled heat pump refrigerator The machine learning device 150 acquires data such as the above. Note that the interpretation of these operating conditions when acquired by the machine learning device 150 has already been explained, and therefore the explanation will be omitted here.

[0110] Furthermore, in the above specific example, the machine learning device 150 determines the operating conditions of the equipment on the heat utilization side 140, which are included in the state variables and affect the processing capacity of the equipment on the heat utilization side 140, as follows: - Intake air temperature, intake air humidity (and air volume) of the outdoor air handling unit Indoor fan coil unit intake air temperature, intake air humidity (and air volume), - Water source heat pump air conditioner intake air temperature, intake air humidity (and air volume), -Indoor temperature, indoor humidity, Obtain data such as:

[0111] Furthermore, in the above specific example, the machine learning device 150 calculates the load of the equipment on the heat utilization side 140, which is included in the state variables and correlates with the amount of heat required to achieve the target value in the equipment on the heat utilization side 140. - Supply air temperature and humidity of the outdoor air handling unit, or intake air temperature and humidity of the intake air (and air volume), Indoor fan coil unit supply air temperature, supply air humidity, or intake air temperature, intake air humidity (and air volume), - Compressor load factor of water source heat pump air conditioner, -Compressor load factor (or indoor unit operating capacity) of water-cooled multi-air conditioner for buildings, Obtain data such as:

[0112] Furthermore, in the above specific example, the machine learning device 150 uses the following as the power consumption of the heat supply side 120 for calculating the total power consumption: ·Power consumption of air-cooled chiller 1001, The data is acquired as the power consumption of the heat utilization side 140 for calculating the total power consumption. Power consumption of the air conditioning unit 1011, Power consumption of water-cooled multi-air conditioners for buildings 1012 and 1013, The data such as the above is acquired, and the power consumption of the heat transfer device 130 for calculating the total power consumption is calculated as follows: · Power consumption of chilled water pump 1021, Obtain data such as:

[0113] Furthermore, in the above specific example, the machine learning device 150 determines the target value of at least one of the temperature and flow rate of the heat medium as follows: - Target value of chilled water supply temperature, - Target value of total chilled water flow rate, Calculate.

[0114] <Summary> As is clear from the above description, the machine learning device 150 can perform reinforcement learning using the learning data sets shown in the specific examples 1 and 2 above.

[0115] [Fifth embodiment] In the above first to fourth embodiments, the reward calculation unit 320 is described as calculating a reward based on the total value notified by the power consumption acquisition unit 310 and notifying the reinforcement learning unit 340. However, the method of calculating the reward by the reward calculation unit 320 is not limited to this.

[0116] For example, if any abnormality occurs in the equipment on the heat supply side 120 or the equipment on the heat use side 140, a signal indicating the occurrence of the abnormality may be acquired as risk information, and the acquired risk information may be taken into account in calculating the reward. The fifth embodiment will be described below, focusing on the differences from the first embodiment.

[0117] <Functional configuration of machine learning device> First, the functional configuration of the machine learning device 150 according to the fifth embodiment will be described in detail. Fig. 11 is a second diagram showing an example of the functional configuration of the machine learning device. The difference from Fig. 3 is that the machine learning device 150 according to the fifth embodiment includes a risk information acquisition unit 1110.

[0118] When any abnormality occurs in the equipment on the heat supply side 120 or the equipment on the heat use side 140, the risk information acquisition unit 1110 acquires a signal indicating the occurrence of the abnormality as risk information and notifies the remuneration calculation unit 320.

[0119] The risk information acquisition unit 1110 may acquire information on, for example, the following: High pressure abnormality (high temperature, low flow rate), low pressure abnormality (low temperature, low flow rate), ·Heat exchanger freezing (low temperature, low flow rate), heat exchanger drift (low flow rate) Signals indicating the above are acquired as risk information.

[0120] Alternatively, the risk information acquisition unit 1110 may acquire information on, for example, the following: High pressure abnormality (high temperature, low flow rate), low pressure abnormality (low temperature, low flow rate), heat exchanger freezing (low temperature, low flow rate), heat exchanger drift (low flow rate), Signals indicating the above are acquired as risk information.

[0121] Alternatively, the risk information acquisition unit 1110 may acquire, for example, ·Heat exchanger freezing (low temperature, low flow rate), heat exchanger drift (low flow rate), Signals indicating the above are acquired as risk information.

[0122] <Reinforcement learning process flow> Next, we will explain the flow of reinforcement learning processing by the machine learning device 150. Figure 12 is a second flowchart showing the flow of reinforcement learning processing by the machine learning device. The differences from the flowchart shown in Figure 4 are steps S1201 and S1202.

[0123] In step S1201, the risk information acquisition unit 1110 determines whether or not a signal indicating that some abnormality has occurred has been acquired as risk information from the equipment on the heat supply side 120 or the equipment on the heat use side 140.

[0124] In step S1202, the remuneration calculation unit 320 subtracts the remuneration calculated in step S403 according to the acquired risk information.

[0125] <Summary> As is clear from the above explanation, the machine learning device 150 subtracts the reward calculated based on the total power consumption in accordance with the acquired risk information. This makes it possible for the machine learning device 150 to perform reinforcement learning using the reward calculated taking into account the risk information.

[0126] [Sixth embodiment] In the above first to fourth embodiments, the reinforcement learning unit 340 has been described as transmitting the target value of at least one of the temperature and flow rate of the heat medium calculated by the heat quantity model 341 to the device on the heat supply side 120. However, the method of outputting the target value of at least one of the temperature and flow rate of the heat medium is not limited to this.

[0127] For example, if at least one of the calculated target values ​​of the temperature and flow rate of the heat medium exceeds a predetermined upper or lower limit, the upper or lower limit may be output. The sixth embodiment will be described below, focusing on the differences from the first embodiment.

[0128] <Functional configuration of machine learning device> First, the functional configuration of the machine learning device 150 according to the sixth embodiment will be described in detail. Fig. 13 is a third diagram showing an example of the functional configuration of a machine learning device. The difference from Fig. 3 is that the machine learning device 150 according to the sixth embodiment includes an upper / lower limit value restriction unit 1310.

[0129] The upper and lower limit limiting unit 1310 acquires the target value of at least one of the temperature and flow rate of the heat medium calculated by the heat quantity model 341 and output by the reinforcement learning unit 340. The upper and lower limit limiting unit 1310 also determines whether the acquired target value of at least one of the temperature and flow rate of the heat medium exceeds the upper and lower limit values ​​of the flow rate and temperature that can be set for the devices on the heat supply side 120. If the upper and lower limit limiting unit 1310 determines that the upper and lower limit values ​​are exceeded, it transmits the upper or lower limit value to the devices on the heat supply side 120. On the other hand, if the upper and lower limit limiting unit 1310 determines that the upper and lower limit values ​​are not exceeded, it transmits the acquired target value of at least one of the temperature and flow rate of the heat medium to the devices on the heat supply side 120.

[0130] <Reinforcement learning process flow> Next, we will explain the flow of reinforcement learning processing by the machine learning device 150. Figure 14 is a third flowchart showing the flow of reinforcement learning processing by the machine learning device. The differences from the flowchart shown in Figure 4 are steps S1401 and S1402.

[0131] In step S1401, the upper and lower limit value restriction unit 1310 determines whether the target value of at least one of the temperature and flow rate of the heat medium output by the reinforcement learning unit 340 in step S406 exceeds the upper and lower limit values ​​that can be set for the equipment on the heat supply side 120.

[0132] In step S1401, if it is determined that the upper or lower limit is exceeded (YES in step S1401), the process proceeds to step S1402.

[0133] Step S140 2 In the above, the upper / lower limit limiting unit 1310 outputs the upper limit or lower limit to the device on the heat supply side 120 .

[0134] On the other hand, if it is determined in step S1401 that the upper or lower limit value is not exceeded (NO in step S1401), the process proceeds to step S407. In this case, the upper or lower limit limiting unit 1310 outputs the target value of at least one of the temperature and flow rate of the heat medium output by the reinforcement learning unit 340 in step S406 to the device on the heat supply side 120.

[0135] <Summary> As is clear from the above explanation, the machine learning device 150 limits, according to upper and lower limits, the target value of at least one of the temperature and flow rate of the heat medium sent to the heat supplying side 120. In this way, the machine learning device 150 can safely operate the equipment on the heat supplying side 120.

[0136] [Seventh embodiment] In the above-described first to sixth embodiments, when describing the system configuration of the air conditioning system, the specific circuit configuration for transporting the heat medium is omitted. Therefore, in the following embodiments, the system configuration of the air conditioning system will be described while clearly indicating the specific circuit configuration for transporting the heat medium.

[0137] In this embodiment, the specific circuit configuration (cooling water circuit, chilled water circuit) for transporting a heat medium in an air conditioning system (the air conditioning system shown in the second embodiment) equipped with a cooling tower, a water-cooled chiller, and an air conditioner will be clearly described.

[0138] <Air conditioning system configuration> Fig. 15 is a diagram showing an example of the system configuration of an air conditioning system including a coolant circuit and a chilled water circuit. Fig. 16 is a diagram showing details of the coolant circuit. Below, the configuration of the entire air conditioning system 100, and the configuration of each device and circuit of the air conditioning system 100 will be described using Figs. 15 and 16.

[0139] (1) Overall configuration of the air conditioning system The air conditioning system 100 shown in FIG. 15 is installed in a relatively large building such as a building, a factory, a hospital, or a hotel.

[0140] In the air conditioning system 100 shown in FIG. 15, the air conditioning equipment 110 includes a heat medium control device 1500, a cooling tower 1510, a group of water-cooled chillers 1520, a primary-side chilled water pump 1530, a secondary-side chilled water pump 1540, and an air conditioning device 1550.

[0141] As shown in FIG. 15, a cooling water circuit 1560 is formed between the cooling tower 1510 and the water-cooled chiller group 1520 , and a chilled water circuit 1570 is formed between the water-cooled chiller group 1520 and the air conditioner 1550 .

[0142] (2) Cooling tower and cooling water circuit The cooling tower 1510 cools the cooling water circulating in the cooling water circuit 1560 by bringing it into contact with the outside air, with a fan 1513 in a main body 1512 operating under the control of the heat medium control device 1500. The cooling tower 1510 is provided with a wet-bulb thermometer 1511, which measures the wet-bulb temperature of the outside air and outputs the measured wet-bulb temperature of the outside air to the heat medium control device 1500.

[0143] The cooling water circuit 1560 is provided with a cooling water pump 1562 (see FIG. 15). The cooling water pump 1562 is a pump that can adjust the discharge flow rate, and circulates the cooling water in the cooling water circuit 1560. The cooling water pump 1562 is inverter-driven by the heat medium control device 1500.

[0144] Furthermore, the coolant circuit 1560 is provided with temperature detection sensors 1575a, 1575b, 1576a, and 1576b (see FIG. 16). Temperature detection sensor 1575a is attached near the inlet of water-cooled chiller 1521a and measures coolant temperature T3a. Temperature detection sensor 1575b is attached near the inlet of water-cooled chiller 1521b and measures coolant temperature T3b.

[0145] Furthermore, a temperature detection sensor 1576a is attached near the outlet of the water-cooled chiller 1521a to measure a cooling water temperature T4a, and a temperature detection sensor 1576b is attached near the inlet of the water-cooled chiller 1521b to measure a cooling water temperature T4b.

[0146] The coolant temperatures measured by the temperature detection sensors 1575 a , 1575 b , 1576 a , and 1576 b are output to the heat medium control device 1500 .

[0147] A flow meter 1563 is attached to the cooling water circuit 1560 (see FIG. 16). The flow meter 1563 is attached to the outlet side of the water-cooled chillers 1521a and 1521b and to the inlet side of the cooling tower 1510, and measures the flow rate of the cooling water circulating through the cooling water circuit 1560.

[0148] (3) Water-cooled chillers The water-cooled chiller group 1520 includes water-cooled chillers 1521a and 1521b, which are water-cooled heat source devices. The water-cooled chillers 1521a and 1521b are connected in parallel to each other in a cooling water circuit 1560 (see FIG. 16).

[0149] The water-cooled chillers 1521a and 1521b have a refrigerant circuit (not shown) to which a compressor (not shown), radiators 1522a and 1522b (see FIG. 16), a chiller-side expansion valve (not shown), and evaporators 1523a and 1523b (see FIG. 15) are connected in sequence. The refrigerant circuit is filled with refrigerant. The cooling water circulating through the cooling water circuit 1560 exchanges heat with the refrigerant inside the refrigerant circuit by passing through the radiators 1522a and 1522b. Furthermore, the refrigerant inside the refrigerant circuit exchanges heat with the chilled water circulating through the chilled water circuit 1570 in the evaporators 1523a and 1523b.

[0150] In this way, the water-cooled chillers 1521a and 1521b cool or heat the cold water as a heat medium through the refrigerant circuit.

[0151] (4) Chilled water circuit The chilled water circuit 1570 is formed by interconnecting pipes L1 to L4 filled with chilled water as a heat medium (see FIG. 15).

[0152] Specifically, the number of communication pipes L1 provided corresponds to the number of water-cooled chillers 1521a, 1521b (two in the case of FIG. 15), and one end is connected to evaporators 1523a, 1523b that are the outlet sides of the water-cooled chillers 1521a, 1521b. The other end of the communication pipe L1 is connected to the inlet of the tank 1574.

[0153] One end of the connecting pipe L2 is connected to the outlet of the tank 1574, and the other end is connected to the user-side heat exchanger 1551, which is the inlet side of the air conditioner 1550. One end of the connecting pipe L3 is connected to the outlet side of the air conditioner 1550, and the other end is connected to the header 1571. The number of connecting pipes L4 provided corresponds to the number of water-cooled chillers 1521a, 1521b (two in the case of FIG. 15), and they are connected to the evaporators 1523a, 1523b.

[0154] In this way, the connecting pipes L1 to L4 connect the water-cooled chillers 1521a and 1521b to the air conditioning device 1550 in a circular manner. In this embodiment, a connecting pipe L5 is further provided that directly connects the header 1571 and the tank 1574 without passing through the water-cooled chillers 1521a and 1521b.

[0155] The connecting pipe L4 is provided with primary side chilled water pumps 1531 and 1532 as primary side chilled water pumps 1530, the number of which corresponds to the number of water-cooled chillers 1521a and 1521b (two in the case of FIG. 15).

[0156] The primary-side chilled water pumps 1531, 1532 are variable-displacement pumps whose capacity and discharge capacity are adjustable, and are inverter-driven by the heat medium control device 1500. The primary-side chilled water pumps 1531, 1532 transport chilled water flowing out from the use-side heat exchanger 1551 of the air conditioner 1550 to the water-cooled chillers 1521a, 1521b, thereby circulating the chilled water in the chilled water circuit 1570. In other words, the primary-side chilled water pumps 1531, 1532 circulate the chilled water in the connection pipes L1 to L4 between the water-cooled chillers 1521a, 1521b and the air conditioner 1550.

[0157] Furthermore, a tank 1574 is provided between the connecting pipes L1 and L2. The tank 1574 is connected to the water-cooled chillers 1521a and 1521b via the connecting pipe L1, and is connected to the air conditioner 1550 via the connecting pipe L2. The tank 1574 stores chilled water that has been heated or cooled by the water-cooled chillers 1521a and 1521b.

[0158] Further, a secondary-side chilled water pump 1540 is provided in the connecting pipe L2. Similar to the primary-side chilled water pumps 1531 and 1532, the secondary-side chilled water pump 1540 is a variable-displacement pump whose capacity and discharge capacity are adjustable, and is inverter-driven by the heat medium control device 1500. The secondary-side chilled water pump 1540 transports chilled water from the water-cooled chillers 1521a and 1521b to the air conditioning device 1550, thereby circulating the chilled water in the chilled water circuit 1570.

[0159] Furthermore, a flow meter 1575 is attached to the connecting pipe L3. The flow meter 1575 is attached to the connecting pipe L3 on the near side of the header 1571. The flow meter 1575 measures the flow rate of the chilled water circulating through the chilled water circuit 1570.

[0160] Furthermore, temperature detection sensors 1573a and 1573b are attached to connecting pipe L1, and temperature detection sensors 1572a and 1572b are attached to connecting pipe L4. Of these, temperature detection sensor 1573a measures a chilled water temperature T1a near the outlet of water-cooled chiller 1521a, and temperature detection sensor 1573b measures a chilled water temperature T1b near the outlet of water-cooled chiller 1521b.

[0161] Furthermore, the temperature detection sensor 1572a measures a chilled water temperature T2a near the inlet of the water-cooled chiller 1521a, and the temperature detection sensor 1572b measures a chilled water temperature T2b near the inlet of the water-cooled chiller 1521b.

[0162] The chilled water temperatures measured by the temperature detection sensors 1572a, 1572b, 1573a, and 1573b are output to the heat medium control device 1500.

[0163] (5) Air conditioning equipment 15, the air conditioner 1550 is connected to the air-conditioned space RM via a duct, etc. The air conditioner 1550 has a use-side heat exchanger 1551, a blower fan 1552, a temperature detection sensor 1553, and a humidity detection sensor 1554.

[0164] The utilization-side heat exchanger 1551 heats or cools the air taken in from the air-conditioned space RM by exchanging heat between the air and the chilled water in the chilled water circuit 1570. The utilization-side heat exchanger 1551 is, for example, a fin-and-tube heat exchanger having a plurality of heat transfer fins and heat transfer tubes that pass through the heat transfer fins.

[0165] The blower fan 1552 is an inverter-driven blower that can adjust the amount of heated or cooled air it blows by changing its rotation speed in stages. The blower fan 1552 forms a flow of air that passes through the utilization-side heat exchanger 1551 and is blown into the air-conditioned space RM.

[0166] The temperature detection sensor 1553 measures the temperature of the air sucked into the air conditioner 1550 and outputs the result to the heat medium control device 1500. The humidity detection sensor 1554 measures the humidity of the air sucked into the air conditioner 1550 and outputs the result to the heat medium control device 1500.

[0167] <Functions of the heat medium control device> Next, a brief description will be given of the functions of the heat medium control device 1500. Fig. 17 is a first diagram for explaining the functions of the heat medium control device. The heat medium control device 1500 is a device for comprehensively controlling the air conditioning system 100, and is electrically connected to various sensors and various driving devices (pumps, fans, valves) that constitute the air conditioning equipment 110.

[0168] In this embodiment, the heat medium control device 1500 identifies the operating conditions of the cooling tower 1510 and the operating conditions of the air conditioning device 1550 based on the outputs of various sensors, etc. The heat medium control device 1500 also identifies the load of the air conditioning device 1550 based on the outputs of various sensors, etc. The heat medium control device 1500 then transmits "state variables" including the operating conditions of the cooling tower 1510, the operating conditions of the air conditioning device 1550, and the load of the air conditioning device 1550 to the machine learning device 150.

[0169] The operating conditions of the cooling tower 1510 included in the state variables are as follows: · The outside air wet bulb temperature measured by the wet bulb thermometer 1511 of the cooling tower 1510; Includes:

[0170] In addition, the operating conditions of the air conditioner 1550 included in the state variables include: The intake air temperature measured by the temperature detection sensor 1553 of the air conditioning device 1550; The intake air humidity measured by the humidity detection sensor 1554 of the air conditioning device 1550, Includes:

[0171] The load of the air conditioning device 1550 included in the state variables is as follows: The amount of heat currently required by the air conditioning device 1550 (calculated using the intake air temperature, the target intake air temperature, and the air volume), Includes:

[0172] The state variables may include the load of the cooling tower 1510 and the load of the water-cooled chiller group 1520.

[0173] The load on the cooling tower 1510 is If the current operating capacity of the air conditioning system 100 satisfies the load (target supply air temperature value) of the air conditioner 1550, or When the current operating capacity of the air conditioning system 100 is being operated so as to approach the load of the air conditioner 1550, that is, when the operating state of the air conditioning system 100 transitions to a steady operation in which "load = operating capacity" after startup, For example, it is calculated by multiplying the temperature difference between the return and return cooling water by the cooling water flow rate.

[0174] The temperature difference between the return and return cooling water is calculated using the difference between cooling water temperatures T3a and T3b measured by temperature detection sensors 1575a and 1575b and cooling water temperatures T4a and T4b measured by temperature detection sensors 1576a and 1576b. The cooling water flow rate is calculated from the voltage and frequency of cooling water pump 1562 when the inverter is driving it, or is measured by flow meter 1563.

[0175] On the other hand, the load on the water-cooled chiller group 1520 is When the operating state of the air conditioning system 100 is a steady operation where "load = operating capacity", It is calculated from the product of the temperature difference between the return and return chilled water and the chilled water flow rate, and the load factor of the water-cooled chiller group 1520.

[0176] The temperature difference between the return and return chilled water is calculated using the difference between chilled water temperatures T1a, T1b measured by temperature detection sensors 1573a, 1573b and chilled water temperatures T2a, T2b measured by temperature detection sensors 1572a, 1572b, for example.

[0177] The chilled water flow rate is calculated from the voltage and frequency when primary-side chilled water pumps 1531 and 1532 or secondary-side chilled water pump 1540 are driven by an inverter, or is measured by flow meter 1575, for example.

[0178] In this embodiment, the heat medium control device 1500 identifies the power consumption of each type of driving device and transmits it to the machine learning device 150. The power consumption of each type of driving device includes the following: ·Cooling tower 1510 power consumption, Power consumption of water-cooled chiller group 1520, · Power consumption of cooling water pump 1562, Power consumption of the primary chilled water pump 1530 and secondary chilled water pump 1540 ·Power consumption of air conditioning unit 1550, Includes:

[0179] Furthermore, in this embodiment, the heat medium control device 1500 outputs the state variables and power consumption to the machine learning device 150, and the machine learning device 150 outputs the following: A pair of a target cooling water temperature and a target chilled water temperature, or A pair of a target value of the cooling water flow rate and a target value of the chilled water flow rate, At least one of the above is acquired.

[0180] The heat medium control device 1500 is Based on the acquired set of the target coolant temperature and the target chilled water temperature, or Based on a set of a target value of the cooling water flow rate and a target value of the chilled water flow rate, It controls the air conditioner 110 (fan 1513, cooling water pump 1562, primary side chilled water pump 1530, secondary side chilled water pump 1540, chiller side expansion valve 1710, blower fan 1552, etc.).

[0181] <Functional configuration of machine learning device> Next, a description will be given of the functional configuration of the machine learning device 150. Fig. 18 is a fourth diagram showing an example of the functional configuration of the machine learning device.

[0182] As described above, a machine learning program is installed in the machine learning device 150, and by executing this program, the machine learning device 150 functions as a power consumption acquisition unit 310, a reward calculation unit 320, a state variable acquisition unit 330, and a reinforcement learning unit 340.

[0183] The power consumption acquisition unit 310 receives the following from the heat medium control device 1500: ·Cooling tower 1510 power consumption, Power consumption of water-cooled chiller group 1520, · Power consumption of cooling water pump 1562, ·Power consumption of primary chilled water pump 1530, Power consumption of secondary chilled water pump 1540, ·Power consumption of air conditioning unit 1550, The power consumption obtaining unit 310 also notifies the remuneration calculation unit 320 of the calculated total power consumption.

[0184] The reward calculation unit 320 calculates a reward based on the total power consumption notified by the power consumption acquisition unit 310 and notifies the reinforcement learning unit 340 of the calculated reward.

[0185] The state variable acquisition unit 330 acquires state variables (operating conditions of the cooling tower 1510, operating conditions of the air conditioner 1550, and load of the air conditioner 1550) from the heat medium control device 1500 and notifies the reinforcement learning unit 340 of them.

[0186] The reinforcement learning unit 340 has a heat quantity model 341, and changes the model parameters of the heat quantity model 341 so as to maximize the reward notified by the reward calculation unit 320. As a result, the reinforcement learning unit 340 · State variables and At least one pair of target values ​​of a pair of a cooling water temperature target value and a chilled water temperature target value, or a pair of a cooling water flow rate target value and a chilled water flow rate target value, Reinforcement learning is performed on the heat quantity model 341 that associates the above.

[0187] Further, the reinforcement learning unit 340 calculates the heat quantity model 341 by inputting the current state variables notified by the state variable acquisition unit 330 into the heat quantity model 341 whose model parameters have been changed. A pair of a target coolant temperature and a target chilled water temperature, or A pair of a target value of the cooling water flow rate and a target value of the chilled water flow rate, The machine learning device 150 acquires a set of at least one of the target values. Furthermore, the reinforcement learning unit 340 transmits the acquired set of target values ​​to the air conditioning equipment 110. As a result, the air conditioning equipment 110 operates to realize the transmitted set of target values. As a result, the machine learning device 150 can reduce the power consumption of the air conditioning equipment 110.

[0188] <Reinforcement learning process flow> Next, we will explain the flow of reinforcement learning processing by the machine learning device 150. Figure 19 is a fourth flowchart showing the flow of reinforcement learning processing by the machine learning device.

[0189] In step S1901, the state variable acquisition unit 330 acquires state variables for a predetermined period from the heat medium control device 1500.

[0190] In step S1902, the power consumption acquisition unit 310 acquires the power consumption for each unit over a predetermined period from the heat medium control device 1500 and calculates the total value.

[0191] In step S1903, the remuneration calculation unit 320 calculates the remuneration based on the calculated total power consumption.

[0192] In step S1904, the remuneration calculation unit 320 determines whether the calculated remuneration is equal to or greater than a predetermined threshold. If it is determined in step S1904 that the calculated remuneration is not equal to or greater than the predetermined threshold (NO in step S1904), the process proceeds to step S1905.

[0193] In step S1905, the reinforcement learning unit 340 performs machine learning on the heat quantity model 341 so as to maximize the calculated reward.

[0194] In step S1906, reinforcement learning unit 340 inputs the current state variables to heat quantity model 341, thereby executing heat quantity model 341. As a result, reinforcement learning unit 340 outputs at least one set of target values, a set of target values ​​for the coolant temperature and the chilled water temperature, or a set of target values ​​for the coolant flow rate and the chilled water flow rate.

[0195] In step S1907, the reinforcement learning unit 340 transmits the set of output target values ​​to the heat medium control device 1500. Then, the process returns to step S1901.

[0196] On the other hand, if it is determined in step S1904 that the difference is equal to or greater than the predetermined threshold (YES in step S1904), the reinforcement learning process ends.

[0197] <Summary> As is clear from the above description, in the seventh embodiment, in an air conditioning system including a cooling tower, a water-cooled chiller, and an air conditioner, it is possible to optimize the transfer of heat in the coolant circuit and the chilled water circuit.

[0198] [Eighth embodiment] In the seventh embodiment, the specific circuit configuration of the heat medium in the air conditioning system (the second embodiment) that includes a cooling tower, a water-cooled chiller, and an air conditioning device and optimizes the transfer of heat in the cooling water circuit and the chilled water circuit was clarified.

[0199] In contrast, the eighth embodiment specifies the specific circuit configuration of the heat medium in an air conditioning system (the third embodiment described above) that includes an air-cooled chiller (chiller unit) and an air conditioning unit (air conditioning unit) and optimizes the transfer of heat in the chilled water circuit (water circuit).

[0200] <Air conditioning system configuration> 20 is a first diagram showing an example of the system configuration of an air conditioning system including a water circuit. The air conditioning system 100 is a central air conditioning system that provides air conditioning in a target space SP within a structure such as a house, building, factory, or public facility. In this embodiment, a case will be described in which the air conditioning system 100 is applied to a building BL that includes multiple (here, three) target spaces SP (SP1, SP2, and SP3).

[0201] 20, the air conditioning system 100 includes air conditioning equipment 110 and a machine learning device 150. The air conditioning equipment 110 takes in outside air OA, conditions it, and supplies it to the target space SP, thereby performing air conditioning such as cooling, heating, ventilation, dehumidification, and / or humidification in the target space SP. The outside air OA is air outside the target space SP, and in this embodiment, it is outdoor air.

[0202] 20, the air conditioning equipment 110 further includes a heat medium control device 2000, a chiller unit 2010, an air handling unit 2020, and a remote control 2030. The heat medium control device 2000 controls the operation of each device according to commands input to the remote control 2030 (commands related to start / stop, operation type, set temperature, set humidity, set air volume, etc.) and the load status (temperature and humidity of outdoor air OA, temperature and humidity of indoor air IA, etc.).

[0203] <Air conditioning equipment configuration> Next, the configuration of each device (here, chiller unit 2010, air handling unit 2020, and remote control 2030) that comprises air conditioning equipment 110 and the configuration of each circuit will be described.

[0204] (1) Water circuit and refrigerant circuit First, a specific circuit configuration for transporting the heat medium in the air conditioning equipment 110 will be described. Fig. 21 is a first diagram showing the detailed configuration of the air conditioning equipment. As shown in Fig. 21, the air handling unit 2020 and the chiller unit 2010 have a water circuit C1 and a refrigerant circuit C2.

[0205] The water circuit C1 is a circuit through which a heat medium (water: "W" in FIG. 21) that exchanges heat with the outside air OA circulates. The water circuit C1 is configured across the chiller unit 2010 and the air handling unit 2020. The water circuit C1 is mainly configured by connecting an air heat exchanger 2133 arranged in the air handling unit 2020, and a water heat exchanger 2122 and a water pump Pa arranged in the chiller unit 2010 with a first pipe P1.

[0206] By controlling the operation of the water pump Pa, water as a heat transfer medium is transported in a predetermined direction (the direction indicated by the two-dot chain arrow d1 in Figure 21) in the water circuit C1. The flow rate of water in the water circuit C1 is adjusted mainly by the rotation speed of the water pump Pa. Although not shown in Figure 21, the water circuit C1 (on the first piping P1) is assumed to be equipped with equipment such as a header pipe for merging and dividing the water, an on-off valve for blocking the water flow, and a pump other than the water pump Pa.

[0207] The refrigerant circuit C2 is a circuit through which a refrigerant that serves as a cooling source for the water in the water circuit C1 circulates. The refrigerant circuit C2 is configured within the chiller unit 2010. The refrigerant circuit C2 is mainly configured by connecting a compressor 2121, a water heat exchanger 2122, an expansion valve 2123, an outdoor heat exchanger 2124, and a four-way selector valve 2125, which are all arranged within the chiller unit 2010, via second piping P2. By controlling the operation of the compressor 2121 and the aperture of the expansion valve 2123, the refrigerant as a heat medium is transported in a predetermined direction in the refrigerant circuit C2, and a vapor compression refrigeration cycle is performed. Note that the predetermined direction refers to the direction indicated by the two-dot chain arrow d2 in FIG. 21 during forward cycle operation, and refers to the opposite direction to d2 during reverse cycle operation.

[0208] (2) Chiller unit The chiller unit 2010 is an example of a "heat source device." The chiller unit 2010 cools or heats water (W) in the water circuit C1 by causing the refrigerant circuit C2 to perform a refrigeration cycle, and discharges the cooled or heated water (W) to supply it to the air handling unit 2020 in operation. The chiller unit 2010 mainly includes a compressor 2121, a water heat exchanger 2122, an expansion valve 2123, an outdoor heat exchanger 2124, a four-way switching valve 2125, an outdoor fan 2126, and a water pump Pa. Note that the chiller unit 2010 may be replaced with another heat source device such as a refrigerator or a boiler.

[0209] The compressor 2121 is a device that compresses a low-pressure refrigerant to a high pressure in the refrigeration cycle. Here, a compressor of a sealed structure with a built-in compressor motor is used as the compressor 2121. A positive displacement compression element (not shown), for example, of a scroll type or the like, is housed inside the compressor 2121, and the compression element is rotationally driven by the compressor motor. The compressor motor is inverter-driven, thereby controlling the capacity of the compressor 2121. In other words, the compressor 2121 has a variable capacity.

[0210] The water heat exchanger 2122 is a device that exchanges heat between the water in the water circuit C1 and the refrigerant in the refrigerant circuit C2, thereby cooling or heating the water. The water heat exchanger 2122 is formed with a water flow path that communicates with the water circuit C1 and a refrigerant flow path that communicates with the refrigerant circuit C2, allowing the water heat exchanger 2122 to exchange heat between the water in the water flow path and the refrigerant in the refrigerant flow path. Specifically, during forward cycle operation (cooling operation or dehumidification operation), the water heat exchanger 2122 functions as an evaporator of low-pressure refrigerant to cool cold water. During reverse cycle operation (heating operation), the water heat exchanger 2122 functions as a condenser of high-pressure refrigerant to heat hot water.

[0211] The expansion valve 2123 functions as a refrigerant pressure reducing means or a flow rate adjusting means. In this embodiment, the expansion valve 2123 is an electric expansion valve whose opening degree can be controlled.

[0212] The outdoor heat exchanger 2124 is a device that exchanges heat between the refrigerant and air in the refrigerant circuit C2, and transfers heat to or absorbs heat from the air. The outdoor heat exchanger 2124 has heat transfer tubes and heat transfer fins that communicate with the refrigerant circuit C2. The outdoor heat exchanger 2124 exchanges heat between the air passing around the heat transfer tubes and heat transfer fins (outdoor air flow, described below) and the refrigerant passing through the heat transfer tubes. The outdoor heat exchanger 2124 functions as a condenser for high-pressure refrigerant during forward cycle operation, and as an evaporator for low-pressure refrigerant during heating operation.

[0213] The four-way switching valve 2125 is a valve that switches the flow of the refrigerant circuit C2. The four-way switching valve 2125 has four connection ports, which are connected to the suction pipe and discharge pipe of the compressor 2121, the gas side of the refrigerant flow path of the water heat exchanger 2122, and the gas side of the outdoor heat exchanger 2124, respectively. In this way, the four-way switching valve 2125 switches between a first state and a second state.

[0214] The first state is a state in which the gas side of the refrigerant flow path of the water heat exchanger 2122 is connected to the suction piping of the compressor 2121, and the discharge piping of the compressor 2121 is connected to the gas side of the outdoor heat exchanger 2124 (see the solid line of the four-way switching valve 2125 in Figure 21).

[0215] On the other hand, the second state is a state in which the discharge piping of the compressor 2121 communicates with the gas side of the refrigerant flow path of the water heat exchanger 2122, and the gas side of the outdoor heat exchanger 2124 communicates with the suction piping of the compressor 2121 (see the dashed line of the four-way switching valve 2125 in FIG. 21). The four-way switching valve 2125 is controlled to the first state during forward cycle operation, and to the second state during reverse cycle operation.

[0216] The outdoor fan 2126 is a blower that generates an outdoor airflow. The outdoor airflow is a flow of air that flows into the chiller unit 2010, passes through the outdoor heat exchanger 2124, and flows out of the chiller unit 2010. The outdoor airflow is a cooling source for the refrigerant in the outdoor heat exchanger 2124 during forward cycle operation, and is a heating source for the refrigerant in the outdoor heat exchanger 2124 during reverse cycle operation. The outdoor fan 2126 has a fan motor, and its rotation speed is adjusted by inverter driving the fan motor. In other words, the outdoor fan 2126 has a variable airflow rate.

[0217] The water pump Pa is disposed in the water circuit C1. During operation, the water pump Pa draws in and discharges water. The water pump Pa has a motor as its drive source, and the rotation speed of the motor is adjusted by being inverter-driven. In other words, the discharge flow rate of the water pump Pa is variable.

[0218] (3) Air handling unit The air handling unit 2020 is an example of an "air conditioner." The air handling unit 2020 cools, dehumidifies, heats, and / or humidifies the outside air OA. The air handling unit 2020 is placed outdoors (outside the target space SP).

[0219] The air handling unit 2020 mainly includes an air heat exchanger 2133, a humidifier 2135, and an air supply fan 2138.

[0220] The air heat exchanger 2133 (heat exchanger) is a device that functions as a cooler or heater for outdoor air OA. The air heat exchanger 2133 is arranged in the water circuit C1. The air heat exchanger 2133 has heat transfer tubes and heat transfer fins that communicate with the water circuit C1. The air heat exchanger 2133 exchanges heat between the outdoor air OA that passes around the heat transfer tubes and heat transfer fins and water that passes through the heat transfer tubes.

[0221] The humidifier 2135 is a device for humidifying the outside air OA that has passed through the air heat exchanger 2133. There are no particular limitations on the method or model of the humidifier 2135, but it is assumed here that a general natural evaporation type humidifier is employed.

[0222] The supply air fan 2138 (air conditioning fan) is a blower that takes in outside air OA into the air handling unit 2020 and sends it to the target space SP via duct D1. There are no particular limitations on the type of supply air fan 2138, but in this embodiment, a sirocco fan is used as the supply air fan 2138. Here, the air handling unit 2020 has an air flow path FP through which the outside air OA flows (see the dashed arrow "FP" in FIG. 21), and when the supply air fan 2138 is operating, the outside air OA flows along the air flow path FP. The supply air fan 2138 has a fan motor, and the rotation speed of the fan motor is adjusted by being inverter-driven. In other words, the supply air fan 2138 has variable airflow.

[0223] In the air handling unit 2020, an air heat exchanger 2133, a humidifier 2135, and an air supply fan 2138 are arranged in this order from the windward side to the leeward side of the air flow path FP. The leeward end of the air flow path FP is connected to a duct D1.

[0224] Various sensors are also provided in the air handling unit 2020. The various sensors provided in the air handling unit 2020 include, for example, an outside air temperature sensor S1 that measures the temperature of the outside air OA drawn into the air handling unit 2020, and an outside air humidity sensor S2 that measures the humidity. Further included is an intake air temperature sensor S3 that measures the temperature (intake air temperature) of the intake air SA sent to the duct D1 (i.e., the target space SP).

[0225] (4) Remote Control The remote control 2030 allows the user to control the air conditioning system 100 to start and stop the air handling unit 2020, the operation type, the set temperature, and the set humidity. 、 It is an input device for inputting various commands such as the set air volume, etc. The remote control 2030 also functions as a display device for displaying predetermined information (for example, the operating state of the air conditioning system 100, the temperature and humidity of the inside air IA, or the temperature and humidity of the outside air OA).

[0226] The remote controller 2030 also includes various sensors, such as: An indoor temperature sensor S4 (Fig. 22) that measures the temperature of the indoor air (indoor air IA) in the target space SP. - Indoor humidity sensor S5 (Fig. 22) for measuring humidity; A carbon dioxide concentration sensor S6 (Fig. 22) for measuring the carbon dioxide concentration; etc. will be placed.

[0227] (5) Operation of each device that makes up the air conditioning equipment Next, we will explain the operation of each device that constitutes the air conditioner 110 during operation. When the air conditioner 110 is operating, the water pump Pa is normally driven, and water circulates in the water circuit C1. Also, the compressor 2121 is driven, and refrigerant circulates in the refrigerant circuit C2.

[0228] Furthermore, during operation of the air conditioning equipment 110, the water in the water circuit C1 is cooled or heated to the target water temperature Tw,up by heat exchange with the refrigerant in the refrigerant circuit C2 in the water heat exchanger 2122. The water cooled or heated in the water heat exchanger 2122 is supplied to the air handling unit 2020, and is heated or cooled by heat exchange with the outside air OA in the air heat exchanger 2133. The water that has passed through the air heat exchanger 2133 is transported back to the water heat exchanger 2122.

[0229] During cooling operation, the refrigerant in the refrigerant circuit C2 is compressed in the compressor 2121 and discharged as high-pressure refrigerant. The high-pressure refrigerant discharged from the compressor 2121 is condensed or releases heat by exchanging heat with air (outdoor airflow generated by the outdoor fan 2126) in the outdoor heat exchanger 2124. The refrigerant that has passed through the outdoor heat exchanger 2124 is reduced in pressure in the expansion valve 2123 to become low-pressure refrigerant, and then transported to the water heat exchanger 2122. The low-pressure refrigerant transported to the water heat exchanger 2122 is evaporated or heated by exchanging heat with water in the water circuit C1. The low-pressure refrigerant that has passed through the water heat exchanger 2122 is transported back to the compressor 2121.

[0230] On the other hand, during heating operation, the refrigerant in the refrigerant circuit C2 is compressed in the compressor 2121 and discharged as high-pressure refrigerant. The high-pressure refrigerant discharged from the compressor 2121 is condensed or releases heat by exchanging heat with water in the water circuit C1 in the water heat exchanger 2122. The refrigerant that has passed through the water heat exchanger 2122 is reduced in pressure in the expansion valve 2123 to become low-pressure refrigerant, and then transported to the outdoor heat exchanger 2124. The low-pressure refrigerant transported to the outdoor heat exchanger 2124 is evaporated or heated by exchanging heat with air (outdoor airflow generated by the outdoor fan 2126). The low-pressure refrigerant that has passed through the outdoor heat exchanger 2124 is transported back to the compressor 2121.

[0231] In the air heat exchanger 2133 of the operating air handling unit 2020, outside air OA is cooled, dehumidified, or heated by heat exchange with water. The air that has passed through the air heat exchanger 2133 is sent to the target space SP as supply air SA. At this time, if the humidifier 2135 is operating, the air that has passed through the air heat exchanger 2133 is humidified by the humidifier 2135 and then sent to the target space SP as supply air SA.

[0232] <Functions of the heat medium control device> Next, a brief description will be given of the functions of the heat medium control device 2000. Fig. 22 is a second diagram for explaining the functions of the heat medium control device. The heat medium control device 2000 is a device for comprehensively controlling the air conditioning system 100, and is electrically connected to various sensors and various driving devices (humidifier, compressor, pump, fan, valve) that constitute the air conditioning equipment 110.

[0233] Specifically, the heat medium control device 2000 calculates the target value of the temperature of the water circulating through the water circuit C1 (target water temperature Tw , Based on the temperature and humidity, the operation of various driving devices (humidifiers, compressors, pumps, fans, valves, etc.) is controlled.

[0234] For example, the heat medium control device 2000 controls the capacity of the compressor 2121, the opening of the expansion valve 2123, the state of the four-way switching valve 2125, the rotation speed of the outdoor fan 2126, the rotation speed of the water pump Pa, starting and stopping of the humidifier 2135, or the rotation speed of the air supply fan 2138. In controlling the operation of the various driving devices, the heat medium control device 2000 acquires outputs from various sensors and transmits and receives signals to and from the remote control 2030.

[0235] The heat medium control device 2000 also identifies the operating conditions of the chiller unit 2010 and the operating conditions of the air handling unit 2020 based on the outputs of various sensors, etc. The heat medium control device 2000 also identifies the load of the air handling unit 2020 based on the outputs of various sensors, etc. Furthermore, the heat medium control device 2000 transmits "state variables" including the operating conditions of the chiller unit 2010, the operating conditions of the air handling unit 2020, and the load of the air handling unit 2020 to the machine learning device 150.

[0236] The operating conditions of the chiller unit 2010 included in the state variables are as follows: Outdoor air temperature during cooling, outdoor air wet-bulb temperature during heating (however, in reality, these values ​​are interpreted based on the frequency of the compressor 2121, the state of the four-way switching valve 2125, the rotation speed of the outdoor fan 2126, and the rotation speed of the water pump Pa), Includes:

[0237] In addition, the operating conditions of the Air Handling Unit 2020 included in the state variables are as follows: - The coil inlet air temperature of the air handling unit 2020 (or the outdoor air temperature outside the target space SP, or the indoor air temperature inside the target space SP), The coil inlet air temperature is measured by an outside air temperature sensor S1. The temperature of the inside air IA in the target space SP is measured by an indoor temperature sensor S4.

[0238] The operating conditions of the air handling unit 2020 included in the state variables are further as follows: - Air humidity at the coil inlet of the air handling unit 2020 (or the humidity of the outside air OA outside the target space SP, or the humidity of the inside air IA inside the target space SP), The humidity of the air at the coil inlet is measured by an outside air humidity sensor S2. The humidity of the inside air IA in the target space SP is measured by an indoor humidity sensor S5.

[0239] Furthermore, the operating conditions of the air handling unit 2020 included in the state variables are as follows: ·Air intake fan 2138 air volume, may be included.

[0240] In addition, the load of the Air Han Unit 2020 is Air handling unit 2020 supply air temperature setting, The supply air set temperature is determined based on a command input to the remote control 2030 (a command related to the set temperature).

[0241] In addition, the load of the Air Handling Unit 2020 is further Air handling unit 2020 supply air humidity setting, The supply air set humidity is determined based on a command input to the remote control 2030 (a command related to the set humidity).

[0242] In this embodiment, the heat medium control device 2000 identifies the power consumption of each type of driving device and transmits it to the machine learning device 150. The power consumption of each type of driving device includes the following: Power consumption of Chiller Unit 2010 Power consumption of Air Han Unit 2020 ·Power consumption of water pump Pa, Includes:

[0243] Furthermore, in this embodiment, the heat medium control device 2000 outputs the state variables and the power consumption to the machine learning device 150, and the machine learning device 150 outputs the following: - Target water temperature (target water temperature), or · Target water flow rate (target water volume), At least one of the above is acquired.

[0244] The heat medium control device 2000 is Based on the target water temperature, or Based on the target water flow rate, It controls the operation of the air conditioner 110 (compressor 2121, expansion valve 2123, four-way switching valve 2125, outdoor fan 2126, water pump Pa, air supply fan 2138, etc.).

[0245] <Functional configuration of machine learning device> Next, a description will be given of the functional configuration of the machine learning device 150. Fig. 23 is a fifth diagram showing an example of the functional configuration of the machine learning device.

[0246] As described above, a machine learning program is installed in the machine learning device 150, and by executing this program, the machine learning device 150 functions as a power consumption acquisition unit 310, a reward calculation unit 320, a state variable acquisition unit 330, and a reinforcement learning unit 340.

[0247] The power consumption acquisition unit 310 acquires each power consumption (power consumption of the chiller unit 2010, power consumption of the air handling unit 2020, and power consumption of the water pump Pa) from the heat medium control device 1500. The power consumption acquisition unit 310 also calculates the total value of the acquired power consumption and notifies the remuneration calculation unit 320 of the calculated total power consumption.

[0248] The reward calculation unit 320 calculates a reward based on the total power consumption notified by the power consumption acquisition unit 310 and notifies the reinforcement learning unit 340 of the calculated reward.

[0249] The state variable acquisition unit 330 acquires state variables (operating conditions of the chiller unit 2010, operating conditions of the air handling unit 2020, and load of the air handling unit 2020) from the heat medium control device 1500 and notifies the reinforcement learning unit 340 of the state variables.

[0250] The reinforcement learning unit 340 has a heat quantity model 341, and changes the model parameters of the heat quantity model 341 so as to maximize the reward notified by the reward calculation unit 320. As a result, the reinforcement learning unit 340 · State variables and At least one of a target water temperature and a target water flow rate; Reinforcement learning is performed on the heat quantity model 341 that associates the above.

[0251] Further, the reinforcement learning unit 340 calculates the heat quantity model 341 by inputting the current state variables notified by the state variable acquisition unit 330 into the heat quantity model 341 whose model parameters have been changed. - Target water temperature, or Target water flow rate, Furthermore, the reinforcement learning unit 340 transmits the acquired target value to the air conditioning equipment 110. As a result, the air conditioning equipment 110 operates to achieve the transmitted target value. As a result, the machine learning device 150 can reduce the power consumption of the air conditioning equipment 110.

[0252] <Reinforcement learning process flow> Next, we will explain the flow of reinforcement learning processing by the machine learning device 150. Figure 24 is a fifth flowchart showing the flow of reinforcement learning processing by the machine learning device.

[0253] In step S2401, the state variable acquisition unit 330 acquires state variables for a predetermined period from the heat medium control device 2000.

[0254] In step S2402, the power consumption obtaining unit 310 obtains each power consumption from the heat medium control device 2000 and calculates the total value.

[0255] In step S2403, the remuneration calculation unit 320 calculates the remuneration based on the calculated total power consumption.

[0256] In step S2404, the remuneration calculation unit 320 determines whether the calculated remuneration is equal to or greater than a predetermined threshold. If it is determined in step S2404 that the calculated remuneration is not equal to or greater than the predetermined threshold (NO in step S2404), the process proceeds to step S2405.

[0257] In step S2405, the reinforcement learning unit 340 performs machine learning on the heat quantity model 341 so as to maximize the calculated reward.

[0258] In step S2406, reinforcement learning unit 340 inputs the current state variables to heat quantity model 341, thereby executing heat quantity model 341. As a result, reinforcement learning unit 340 outputs at least one of the target value of the water temperature and the target value of the water flow rate.

[0259] In step S2407, the reinforcement learning unit 340 transmits the output target value to the heat medium control device 2000, and then the process returns to step S2401.

[0260] On the other hand, if it is determined in step S2404 that the difference is equal to or greater than the predetermined threshold (YES in step S2404), the reinforcement learning process ends.

[0261] <Summary> As is clear from the above description, in the eighth embodiment, in an air conditioning system including a chiller unit and an air handling unit, it is possible to optimize the transfer of heat in the water circuit.

[0262] [Ninth embodiment] In the eighth embodiment, an air conditioning system including a chiller unit and an air handling unit has been described. In contrast, in the ninth embodiment, as a modification of the eighth embodiment, an air conditioning system including a fan coil unit instead of an air handling unit will be described.

[0263] <System configuration of air conditioning system (variation 1)> Fig. 25 is a second diagram showing an example of the system configuration of an air conditioning system including a water circuit, and Fig. 26 is a second diagram showing the detailed configuration of the air conditioning equipment.

[0264] 25, the air conditioning system 100a has multiple fan coil units 2020a (here, the same number as the target space SP) instead of an air handling unit 2020. During operation, the fan coil units 2020a of the air conditioning system 100a take in air from the target space SP (inside air IA), cool, heat, or dehumidify it, and supply it to the target space SP as supply air SA. Note that the humidification function does not have to be provided in this embodiment.

[0265] 26, the chiller unit 2010 and the fan coil unit 2020a have a water circuit C1' instead of the water circuit C1. The water circuit C1' is mainly configured by connecting the air heat exchanger 2133 arranged in each fan coil unit 2020a to the water heat exchanger 2122 and water pump Pa arranged in the chiller unit 2010 via a first pipe P1.

[0266] The fan coil unit 2020a is an example of an "air conditioner" and cools, dehumidifies, and heats the inside air IA. The fan coil unit 2020a is placed in the target space SP.

[0267] Fig. 27 is a first diagram showing an installation mode of fan coil units in a target space. In this embodiment, each fan coil unit 2020a is a ceiling-embedded type that is installed on the ceiling CL of the target space SP. As shown in Fig. 27, each fan coil unit 2020a is installed in the target space SP so that its air outlet is exposed from the ceiling CL.

[0268] Each fan coil unit 2020a has an air heat exchanger 2133a and an air supply fan 2138a, similar to the air handling unit 2020. The air heat exchanger 2133a and the air supply fan 2138a are arranged in this order from the upwind side to the downwind side of the air flow path FP' through which the room air IA flows.2 In fan coil unit 2020a, the downwind end of air flow path FP' communicates with target space SP. Unlike air handling unit 2020, fan coil unit 2020a is not connected to duct D1. Fan coil unit 2020a does not take in outside air OA to send supply air SA to target space SP, but rather takes in indoor air IA, cools, dehumidifies, and heats it, and then sends the supply air SA to target space SP.

[0269] In addition, in the air conditioning system 100a having the fan coil unit 2020a shown in Figure 25, the machine learning device 150 can output optimal target values ​​that reduce power consumption as the target value of water temperature or the target value of water flow rate using a method similar to that of the eighth embodiment.

[0270] Furthermore, even if the fan coil unit 2020a and the air handling unit 2020 are mixed and arranged in the air conditioning system 100a, the machine learning device 150 can output the optimal target value using a method similar to that of the eighth embodiment described above.

[0271] The number of fan coil units 2020a does not need to be the same as the number of target spaces SP, and may be more or less than the number of target spaces SP. For example, multiple fan coil units 2020a may be arranged in one target space SP.

[0272] <System configuration of air conditioning system (variation 2)> Next, other modified examples will be described. Fig. 28 is a third diagram showing an example of the system configuration of an air conditioning system including a water circuit. Fig. 29 is a third diagram showing the detailed configuration of the air conditioning equipment. Below, the differences from the air conditioning system 100a shown in Fig. 25 and the detailed configuration of the air conditioning equipment shown in Fig. 26 will be mainly described.

[0273] As shown in FIG. 28, the air conditioning system 100b includes a fan coil unit 20 2In place of the fan coil unit 200a, the air conditioning system 100b has a fan coil unit 2020b. During operation, the fan coil unit 2020b takes in outside air OA through a duct D2, cools or heats, or dehumidifies or humidifies the air, and sends the supply air SA to the target space SP.

[0274] The fan coil unit 2020b is an example of an "air conditioner", and like the fan coil unit 2020a, has an air heat exchanger 2133, a humidifier 2135, and an air supply fan 2138 (see FIG. 29).

[0275] 29, the air heat exchanger 2133, the humidifier 2135, and the supply air fan 2138 are arranged in this order from the upwind side to the downwind side of the air flow path FP through which the outside air OA flows. Also, unlike the fan coil unit 2020a, the upwind end of the air flow path FP of the fan coil unit 2020b is connected to a duct D2. Also, the fan coil unit 2020b takes in the outside air OA via the duct D2 and cools, dehumidifies, heats, or humidifies it before sending the supply air SA to the target space SP.

[0276] In this embodiment, the fan coil unit 2020b is associated with one of the target spaces SP and is installed in the corresponding target space SP. FIG. 30 is a second diagram showing an installation mode of the fan coil unit in the target space. In this embodiment, each fan coil unit 2020b is a ceiling-embedded type that is installed in the ceiling CL of the target space SP. As shown in FIG. 30, each fan coil unit 2020b is installed in the target space SP so that its air outlet is exposed from the ceiling CL.

[0277] 30, the duct D2 is a member that forms a flow path for the outside air OA. One end of the duct D2 is connected to the corresponding fan coil unit 2020b so that the outside air OA is taken in by the air supply fan 2138 being driven. 2 0b. The other end is connected to an intake port H2 (see FIG. 28) formed in the target space SP.

[0278] In addition, in the air conditioning system 100b having the fan coil unit 2020b shown in Figure 28, the machine learning device 150 can output optimal target values ​​that reduce power consumption as the target value of water temperature or the target value of water flow rate using a method similar to that of the eighth embodiment.

[0279] In addition, in the air conditioning system 100b, Fan coil unit 2020b and Air handling unit 2020 and / or fan coil unit 2020a, Even when these are mixed and arranged, the machine learning device 150 can output an optimal target value using a method similar to that of the eighth embodiment.

[0280] <Summary> As is clear from the above description, in the ninth embodiment, in an air conditioning system equipped with a chiller unit and a fan coil unit, it is possible to optimize the transfer of heat in the water circuit.

[0281] [Other embodiments] In the above embodiments, the timing of acquiring the power consumption used to calculate the reward is not mentioned. However, for example, the power consumption acquiring unit 310 may acquire the power consumption after a predetermined period has elapsed since the reinforcement learning unit 340 transmitted the target value of at least one of the temperature and flow rate of the heat medium.

[0282] In addition, in the above embodiments, the remuneration is calculated using the total amount of power consumption, but the total amount of consumed energy used to calculate the remuneration is not limited to the total amount of power consumption. For example, the remuneration may be calculated using the total amount of consumed energy, such as the energy consumption efficiency (COP: Coefficient of Performance), carbon dioxide emissions, and energy costs (electricity bills, gas bills).

[0283] Furthermore, although the above embodiments do not specifically mention the details of the model (calorie model) used when performing machine learning, any type of model may be applied when performing machine learning. Specifically, any type of model may be applied, such as a neural network (NN) model, a random forest model, or a support vector machine (SVM) model.

[0284] Furthermore, the air conditioning system 100 in the eighth embodiment may be configured not to perform heating operation, that is, the chiller unit 2010 does not have to be a heat pump type.

[0285] Furthermore, in the air conditioning system 100b in the ninth embodiment, the humidifier 2135 may be omitted as appropriate. That is, the air conditioning system 100b may be configured not to perform a humidifying operation.

[0286] Furthermore, in the above ninth embodiment, a case has been described in which the air conditioning system 100b is applied to a building BL including three target spaces SP. However, the installation environment of the air conditioning system 100 is not limited to this. For example, the air conditioning system 100b may be applied to a building including four or more target spaces SP. Furthermore, for example, the air conditioning system 100b may be applied to a building including two or less target spaces SP (including one). In such cases, the number of fan coil units 2020b may be changed appropriately depending on the number of target spaces SP. Furthermore, multiple fan coil units may be arranged in one target space SP.

[0287] Furthermore, in the above eighth embodiment, the air conditioning system 100 has been described as having one chiller unit 2010 and one air handling unit 2020. However, the number of chiller units 2010 and air handling units 2020 included in the air conditioning system 100 is not necessarily limited to one, and can be changed as appropriate depending on the installation environment and design specifications. In other words, the air conditioning system 100 may have a plurality of chiller units 2010 and / or air handling units 2020. Note that the number of chiller units 2010 and the number of air handling units 2020 do not necessarily have to be the same.

[0288] In the above eighth embodiment, the case where all the air flowing into the air handling unit 2020 is outside air has been described. However, the air flowing into the air handling unit 2020 is not limited to this. For example, the air flowing into the air handling unit 2020 may be a mixture of outside air and return air, or may be all inside air.

[0289] The configuration of the refrigerant circuit (C2) configured in the eighth and ninth embodiments can be modified as appropriate depending on the installation environment and design specifications. For example, if heating operation is omitted, the four-way switching valve 2125 may be omitted. Alternatively, a water heat exchanger may be provided in place of the outdoor heat exchanger 2124, and the refrigerant may be cooled or heated by exchanging heat between the refrigerant and water in the water heat exchanger. The configuration of the water circuit C1 configured in the air handling unit 2020 can also be modified as appropriate depending on the installation environment and design specifications.

[0290] In the eighth and ninth embodiments, the refrigerant circulating through the refrigerant circuit (C2) is assumed to be an HFC refrigerant such as R32 or R410A, but it does not necessarily have to be an HFC refrigerant. For example, other refrigerants (such as HFO1234yf, HFO1234ze(E), CO2, or ammonia) may also be used. Furthermore, the heat medium circulating through the water circuit C1 does not necessarily have to be water, and other fluids may also be used.

[0291] Furthermore, the locations of the various sensors included in the air conditioning system 100 are not necessarily limited to those in the eighth and ninth embodiments, and can be changed as appropriate. For example, the outdoor air temperature sensor S1, the outdoor air humidity sensor S2, and the supply air temperature sensor S3 do not necessarily have to be disposed in the air handling unit 2020, but may be disposed in another unit or disposed independently. Furthermore, the indoor temperature sensor S4, the indoor humidity sensor S5, and / or the carbon dioxide concentration sensor S6 do not necessarily have to be disposed in the remote control 2030, but may be disposed in another unit or disposed independently.

[0292] Furthermore, in the above eighth and ninth embodiments, no particular explanation was given about the installation mode of the machine learning device 150, but the installation mode of the machine learning device 150 can be selected as appropriate. For example, the heat medium control device 2000 may be placed in a management office of the building BL, or may be placed in a remote location communicably connected via a WAN or LAN.

[0293] The configurations of the heat medium control device 2000 and the machine learning device 150 may also be changed as appropriate. For example, the heat medium control device 2000 and the machine learning device 150 may be arranged integrally or separately and connected via a communication network. When arranged integrally, the heat medium control device 2000 and the machine learning device 150 may be configured by a single computer or by connecting multiple devices (for example, PCs, smartphones, etc.).

[0294] Although the embodiments have been described above, it will be understood that various changes in form and details can be made without departing from the spirit and scope of the claims.

[0295] This application claims priority based on Japanese Patent Application No. 2019-086781 filed on April 26, 2019, Japanese Patent Application No. 2019-086782 filed on April 26, 2019, and Japanese Patent Application No. 2019-086785 filed on April 26, 2019, the entire contents of which are incorporated herein by reference. [Explanation of symbols]

[0296] 100: Air conditioning system 110: Air conditioning equipment 120: Heat supply side 130: Heat transfer device 140: Heat utilization side 150: Machine learning device 310: Power consumption acquisition section 320: Reward calculation unit 330: State variable acquisition unit 340: Reinforcement Learning Department 341: Heat model 1110: Risk Information Acquisition Department 1310: Upper and lower limit value limit section 1500: Heat medium control device 1510 :Cooling tower 1520: Water-cooled chiller group 1530: Primary side chilled water pump 1540: Secondary chilled water pump 1550: Air conditioning equipment 2000: Heat transfer medium control device 2010: Chiller unit 2020: Aircraft unit 2020a, 2020b: Fan coil unit

Claims

1. A machine learning device that learns at least one of a temperature and a flow rate of a heat medium when the heat transfer device transfers the heat medium in an air conditioning system having a heat supply-side device, a heat utilization-side device, and a heat transfer device that transfers a heat medium from the heat supply-side device to the heat utilization-side device, a state variable acquisition unit that acquires state variables including an operating condition of the heat supplying device, an operating condition of the heat utilizing device, and a value that correlates with the amount of heat required for the heat utilizing device; a learning unit that learns the state variable in association with at least one of the temperature and the flow rate; A heat medium control device controls the heat supplying side device, the heat utilizing side device, and the heat transfer device based on at least one of the temperature and the flow rate output by the learning unit when the state variable is input, and in accordance with the control, Power consumption, which is the power consumed in the heat supplying device; Power consumption, which is the power consumed in the heat utilization device; power consumption, which is the power consumed in the heat transfer device; a reward calculation unit that calculates a reward based on the total value of the power consumption when the power consumption acquisition unit acquires the total value of the power consumption; The learning unit is a machine learning device that repeats learning based on the reward calculated by the reward calculation unit each time the heat medium control device controls the heat supply side equipment, the heat utilization side equipment, and the heat transport device, thereby maximizing the reward.

2. The machine learning device according to claim 1 , wherein the operating conditions of the heat supplying equipment include any one of an outdoor air temperature, an outdoor wet-bulb temperature, and an underground temperature, which affect the processing capacity of the heat supplying equipment.

3. The machine learning device according to claim 1 , wherein the operating conditions of the heat-using equipment include either an intake air temperature or a chilled water return temperature, which affect the processing capacity of the heat-using equipment.

4. The machine learning device according to claim 3 , wherein the operating conditions of the heat-using equipment further include either an air volume or a chilled water flow rate.

5. The machine learning device according to claim 1 , wherein the value correlated with the amount of heat required by the heat-using device includes either an intake air temperature or a chilled water supply temperature.

6. 6. The machine learning device according to claim 5, wherein the temperature when the heat transfer device transports the heat medium includes a cold water supply temperature and a cooling water supply temperature, and the flow rate when the heat transfer device transports the heat medium includes either a cold water flow rate or a cooling water flow rate.

7. The machine learning device according to claim 1 , wherein the heat supply side equipment includes an air-cooled chiller, the heat utilization side equipment includes an air conditioning device, and the heat transfer device includes a chilled water pump.

8. The machine learning device according to claim 1 , wherein the heat supply side equipment includes a cooling tower, the heat utilization side equipment includes a water-cooled chiller, and the heat transfer device includes a cooling water pump.

9. The machine learning device according to claim 1 , wherein the heat supply side equipment includes a geothermal heat exchanger, the heat utilization side equipment includes a water-cooled chiller, and the heat transport device includes a cooling water pump.

10. The machine learning device according to claim 1 , wherein the heat supply side equipment includes a cooling tower, a cooling water pump, and a water-cooled chiller, the heat utilization side equipment includes an air conditioning device, and the heat transport device includes a chilled water pump.

11. The machine learning device according to claim 1 , wherein the heat supply side equipment includes a cooling tower, the heat utilization side equipment includes a water-cooled chiller, a chilled water pump, and an air conditioning device, and the heat transport device includes a chilled water pump.

12. 2. The machine learning device according to claim 1, wherein the reward calculation unit reduces the reward when a risk to the air conditioning system increases as a result of the heat medium control device controlling the heat supply side equipment based on at least one of the temperature and the flow rate output by the learning unit when the state variable is input.

13. 2. The machine learning device according to claim 1, wherein when at least one of the temperature and the flow rate output by the learning unit due to the input of the state variable exceeds a predetermined upper limit value or a predetermined lower limit value, the heat medium control device controls the heat supply side equipment based on the predetermined upper limit value or the predetermined lower limit value.

14. An air conditioning system having a heat supply-side device, a heat utilization-side device, a heat transfer device that transfers a heat medium from the heat supply-side device to the heat utilization-side device, and a machine learning device that learns at least one of a temperature and a flow rate when the heat transfer device transfers the heat medium, The machine learning device includes: a state variable acquisition unit that acquires state variables including an operating condition of the heat supplying device, an operating condition of the heat utilizing device, and a value that correlates with the amount of heat required for the heat utilizing device; a learning unit that learns the state variable in association with at least one of the temperature and the flow rate; A heat medium control device controls the heat supplying side device, the heat utilizing side device, and the heat transfer device based on at least one of the temperature and the flow rate output by the learning unit when the state variable is input, and in accordance with the control, Power consumption, which is the power consumed in the heat supplying device; Power consumption, which is the power consumed in the heat utilization device; power consumption, which is the power consumed in the heat transfer device; a reward calculation unit that calculates a reward based on the total value of the power consumption when the power consumption acquisition unit acquires the total value of the power consumption; The learning unit repeats learning based on the reward calculated by the reward calculation unit each time the heat medium control device controls the heat supply side equipment, the heat use side equipment, and the heat transfer device, thereby maximizing the reward.

15. A machine learning method for learning at least one of a temperature and a flow rate of a heat medium when the heat transfer device transfers the heat medium in an air conditioning system having a heat supply-side device, a heat utilization-side device, and a heat transfer device that transfers the heat medium from the heat supply-side device to the heat utilization-side device, comprising: a state variable acquisition step of acquiring state variables including an operating condition of the heat supplying device, an operating condition of the heat utilizing device, and a value correlating with the amount of heat required for the heat utilizing device; a learning step of associating the state variable with at least one of the temperature and the flow rate and learning the state variable; A heat medium control device controls the heat supplying device, the heat utilizing device, and the heat transfer device based on at least one of the temperature and the flow rate output in the learning step by inputting the state variables, and in accordance with the control, in a power consumption acquisition step: Power consumption, which is the power consumed in the heat supplying device; Power consumption, which is the power consumed in the heat utilization device; power consumption, which is the power consumed in the heat transfer device; and a reward calculation step of calculating a reward based on the total value of the power consumption when the power consumption is acquired. The learning process is a machine learning method in which the heat medium control device repeats learning based on the reward calculated in the reward calculation process each time it controls the heat supply side equipment, the heat use side equipment, and the heat transport device, thereby maximizing the reward.

16. Water-cooled chiller and a cooling water pump that supplies cooling water that cools the refrigerant by heat exchange in the water-cooled chiller; a cooling tower that cools the cooling water transported from the water-cooled chiller by bringing it into contact with outside air; An air conditioning device; a chilled water pump that supplies chilled water cooled by the refrigerant through heat exchange in the water-cooled chiller to the air conditioning device, a machine learning device that learns at least one of a first set that is a set of a temperature of the chilled water supplied by the chilled water pump and a temperature of the chilled water supplied by the chilled water pump, or a second set that is a set of a flow rate of the chilled water supplied by the chilled water pump and a flow rate of the chilled water supplied by the chilled water pump, a state variable acquisition unit that acquires state variables including an operating condition of the cooling tower, an operating condition of the air conditioning device, and a load of the air conditioning device; a learning unit that learns the state variables by associating them with at least one of the first set and the second set; A heat medium control device controls the cooling tower, the water-cooled chiller, the cooling water pump, the chilled water pump, and the air conditioning device based on at least one of the first set and the second set that are output by the learning unit when the state variables are input, and in conjunction with this control, Consumed energy is energy consumed in the cooling tower; Consumed energy is energy consumed in the water-cooled chiller; Consumed energy is energy consumed in the cooling water pump; Consumed energy is energy consumed in the chilled water pump; Consumed energy is energy consumed in the air conditioning device; a reward calculation unit that calculates a reward based on the total value of the consumed energy when the consumed energy acquisition unit acquires the total value of the consumed energy; The learning unit is a machine learning device that repeats learning based on the reward calculated by the reward calculation unit each time the heat medium control device controls the cooling tower, the water-cooled chiller, the cooling water pump, the chilled water pump, and the air conditioning device, thereby maximizing the reward.

17. a heat source device that heats or cools a heat medium; a pump that discharges the heat medium heated or cooled by the heat source device; a heat exchanger that exchanges heat between air passing through the heat exchanger and a heat medium discharged by the pump, and an air conditioner that sends the air that has passed through the heat exchanger to a target space; a state variable acquisition unit that acquires state variables including the operating conditions of the heat source device, the operating conditions of the air conditioning device, and the load of the air conditioning device; a learning unit that learns the state variable in association with at least one of the temperature and the flow rate of the heat medium; a heat medium control device that controls the heat source device, the air conditioning device, and the pump based on at least one of the temperature and the flow rate of the heat medium that is output by the learning unit in response to the input of the state variables, and that performs the following control: Consumed energy is energy consumed in the heat source device; Consumed energy is energy consumed in the air conditioning device; Consumed energy is energy consumed in the pump; a reward calculation unit that calculates a reward based on the total value of the consumed energy when the consumed energy acquisition unit acquires the total value of the consumed energy; The learning unit is a machine learning device that repeats learning based on the reward calculated by the reward calculation unit each time the heat medium control device controls the heat source device, the air conditioning device, and the pump, thereby maximizing the reward.

Images