Patent Application
20230304691
The present invention relates to an air conditioning load learning apparatus and an air conditioning load prediction apparatus.
To implement optimum facility design, optimum control design, and the like, at the time of updating an air conditioner, there is a need to provide a highly accurate air conditioning load prediction technique in consideration of the operation of a proposed building. In PTL 1 (Japanese Patent No. 5943255), the difference between the actual air conditioning load measurement value and the external thermal load (excluding the internal heat generation) obtained by the physical model is determined so that the internal heat generation related to the operation of the building is estimated and used for air conditioning load prediction,
In order to predict the air conditioning load by using the method disclosed in PTL 1, there is a need to input the information about the operation of the building so as to reflect the effect of the operation of the building on the internal heat generation, or the like. However, the information about the operation of the building is typically hard to obtain and is often incomplete even if it is obtained. Therefore, the method disclosed in PTL 1 has an issue that it is difficult to grasp the effect of the operation of the building on the air conditioning load.
An air conditioning load learning apparatus according to a first aspect includes an actual load acquisition unit, a first information acquisition unit, and a learning unit. The actual load acquisition unit acquires an actual air conditioning load that is an actual air conditioning load in a target space inside a target building. The first information acquisition unit acquires first information. The first information is information about an operation of the target building. The learning unit generates a learning model using at least the first information as an explanatory variable and using a value regarding the actual air conditioning load as an objective variable.
In the air conditioning load learning apparatus according to the first aspect, the learning unit generates the learning model using at least the first information as an explanatory variable and using the value regarding the actual air conditioning load as an objective variable. Therefore, the air conditioning load learning apparatus can associate the first information with the value regarding the actual air conditioning load to learn the effect of the operation of the building on the internal heat generation, etc. As a result, the air conditioning load learning apparatus can finally grasp the effect of the operation of the building on the air conditioning load.
An air conditioning load learning apparatus according to a second aspect is the air conditioning load learning apparatus according to the first aspect and further includes a prediction load acquisition unit. The prediction load acquisition unit acquires a prediction air conditioning load that is an air conditioning load in the target space predicted by a physical model from at least a thermal property of the target building. The learning unit generates the learning model using the first information as an explanatory variable and using a difference load, which is a difference between the actual air conditioning load and the prediction air conditioning load, as an objective variable.
With such a configuration, the air conditioning load learning apparatus according to the second aspect can associate the first information with the difference load (the value regarding the actual air conditioning load) to learn the effect of the operation of the building on the internal heat generation, etc. As a result, the air conditioning load learning apparatus can finally grasp the effect of the operation of the building on the air conditioning load.
An air conditioning load learning apparatus according to a third aspect is the air conditioning load learning apparatus according to the second aspect, wherein the actual air conditioning load is an air conditioning load per unit time.
With such a configuration, the air conditioning load learning apparatus according to the third aspect can learn the effect of the operation of the building on the air conditioning load in more detail.
An air conditioning load learning apparatus according to a fourth aspect is the air conditioning load learning apparatus according to the second aspect or the third aspect and further includes a second information acquisition unit. The second information acquisition unit acquires second information. The second information is at least one of an indoor humidity, an indoor temperature, an air conditioning operating time, a post air conditioning operation start elapsed time, an outside air temperature, an outside air humidity, and solar radiation. The learning unit generates the learning model further using the second information as an explanatory variable.
With such a configuration, the air conditioning load learning apparatus according to the fourth aspect can learn the effect of the operation of the building on the outside air introduction load, the heat storage load, the solar radiation load, etc. As a result, the air conditioning load
learning apparatus can finally grasp the effect of the operation of the building on the air conditioning load.
An air conditioning load learning apparatus according to a fifth aspect is the air conditioning load learning apparatus according to any one of the second aspect to the fourth aspect, wherein the learning unit generates the learning model further using the prediction air conditioning load as an explanatory variable.
With such a configuration, the air conditioning load learning apparatus according to the fifth aspect can generate the learning model having a higher prediction accuracy.
An air conditioning load prediction apparatus according to a sixth aspect includes a difference load prediction unit and a load prediction unit. The difference load prediction unit uses the learning model of the air conditioning load learning apparatus according to the second aspect or the third aspect to predict the difference load from the first information. The load prediction unit predicts an air conditioning load in the target space based on the predicted difference load and the prediction air conditioning load.
With such a configuration, the air conditioning load prediction apparatus according to the sixth aspect can use the learning model having learned the effect of the operation of the building on internal heat generation, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
An air conditioning load prediction apparatus according to a seventh aspect is the air conditioning load prediction apparatus according to the sixth aspect, wherein the learning model is generated based on the first information, the actual air conditioning load, and the prediction air conditioning load in a short period. The difference load prediction unit predicts, from the first information in a long period that is a period longer than the short period, the difference load in the long period, The load prediction unit predicts an air conditioning load in the long period in the target space based on the predicted difference load in the long period and the prediction air conditioning load in the long period.
With such a configuration, the air conditioning load prediction apparatus according to the seventh aspect can use the first learning model having learned with the data in the short period to predict the air conditioning load in the long period.
An air conditioning load prediction apparatus according to an eighth aspect includes a difference load prediction unit and a load prediction unit. The difference load prediction unit uses the learning model of the air conditioning load learning apparatus according to the fifth aspect to predict the difference load from the first information and the second information or the prediction air conditioning load. The load prediction unit predicts an air conditioning load in the target space based on the predicted difference load and the prediction air conditioning load.
With such a configuration, the air conditioning load prediction apparatus according to the eighth aspect can use the learning model having learned the effect of the operation of the building on the internal heat generation, the outside air introduction load, the heat storage load, the solar radiation load, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
An air conditioning load prediction apparatus according to a ninth aspect is the air conditioning load prediction apparatus according to the eighth aspect, wherein the learning model is generated based on the first information, the second information, the actual air conditioning load, and the prediction air conditioning load in a short period. The difference load prediction unit predicts, from the first information and the second information or the prediction air conditioning load in a long period that is a period longer than the short period, the difference load in the long period. The load prediction unit predicts an air conditioning load in the long period in the target space based on the predicted difference load in the long period and the prediction air conditioning load in the long period.
With such a configuration, the air conditioning load prediction apparatus according to the ninth aspect can use the learning model having learned with the data in the short period to predict the air conditioning load in the long period.
An air conditioning load learning apparatus according to a tenth aspect is the air conditioning load learning apparatus according to the first aspect and further includes a prediction load acquisition unit. The prediction load acquisition unit acquires a prediction air conditioning load that is an air conditioning load in the target space predicted by a physical model from at least a thermal property of the target building. The learning unit generates the learning model using the prediction air conditioning load and the first information as explanatory variables and using the actual air conditioning load as an objective variable.
With such a configuration, the air conditioning load learning apparatus according to the tenth aspect can associate the first information with the actual air conditioning load (the value regarding the actual air conditioning load) to learn the effect of the operation of the building on the internal heat generation, etc. As a result, the air conditioning load learning apparatus can finally grasp the effect of the operation of the building on the air conditioning load.
An air conditioning load learning apparatus according to an eleventh aspect is the air conditioning load learning apparatus according to the tenth aspect, wherein the actual air conditioning load is an air conditioning load per unit time.
With such a configuration, the air conditioning load learning apparatus according to the eleventh aspect can learn the effect of the operation of the building on the air conditioning load in more detail.
An air conditioning load learning apparatus according to a twelfth aspect is the air conditioning load learning apparatus according to the tenth aspect or the eleventh aspect and further includes a second information acquisition unit. The second information acquisition unit acquires second information. The second information is at least one of an indoor humidity, an indoor temperature, an air conditioning operating time, a post air conditioning operation start elapsed time, an outside air temperature, an outside air humidity, and solar radiation. The learning unit generates the learning model further using the second information as an explanatory variable.
With such a configuration, the air conditioning load teaming apparatus according to the twelfth aspect can learn the effect of the operation of the building on the outside air introduction load, the heat storage load, the solar radiation load, etc. As a result, the air conditioning load learning apparatus can finally grasp the effect of the operation of the building on the air conditioning load.
An air conditioning load prediction apparatus according to a thirteenth aspect includes a load prediction unit, The load prediction unit uses the learning model generated by the learning unit in the air conditioning load learning apparatus according to the tenth aspect or the eleventh aspect to predict an air conditioning load in the target space from the prediction air conditioning load and the first information.
With such a configuration, the air conditioning load prediction apparatus according to the thirteenth aspect can use the learning model having learned the effect of the operation of the building on internal heat generation, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
An air conditioning load prediction apparatus according to a fourteenth aspect is the air conditioning load prediction apparatus according to the thirteenth aspect, wherein the learning model is generated based on the prediction air conditioning load, the first information, and the actual air conditioning load in a short period. Based on the prediction air conditioning load and the first information in a long period that is a period longer than the short period, the load prediction unit predicts an air conditioning load in the long period in the target space.
With such a configuration, the air conditioning load prediction apparatus according to the fourteenth aspect can use the learning model having learned with the data in the short period to predict the air conditioning load in the long period.
An air conditioning load prediction apparatus according to a fifteenth aspect includes a load prediction unit. The load prediction unit uses the learning model generated by the learning unit in the air conditioning load learning apparatus according to the twelfth aspect to predict an air conditioning load in the target space from the prediction air conditioning load, the first information, and the second information.
With such a configuration, the air conditioning load prediction apparatus according to the fifteenth aspect can use the learning model having learned the effect of the operation of the building on the internal heat generation, the outside air introduction load, the heat storage load, the solar radiation load, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
An air conditioning load prediction apparatus according to a sixteenth aspect is the air conditioning load prediction apparatus according to the fifteenth aspect, wherein the learning model is generated based on the prediction air conditioning load, the first information, the second information, and the actual air conditioning load in a short period. Based on the prediction air conditioning load, the first information, and the second information in a long period that is a period longer than the short period, the load prediction unit predicts an air conditioning load in the long period in the target space.
With such a configuration, the air conditioning load prediction apparatus according to the sixteenth aspect can use the learning model having learned with the data in the short period to predict the air conditioning load in the long period.
An air conditioning load learning apparatus according to a seventeenth aspect is the air conditioning load learning apparatus according to the first aspect and further includes an input value acquisition unit. The input value acquisition unit acquires a first input value and a second input value. The first input value is an input value including at least a thermal property of the target building to a physical model, which outputs a prediction air conditioning load that is a predicted air conditioning load in the target space. The second input value is an input value calculated by inverse calculation of the physical model using the actual air conditioning load, The learning unit generates the learning model using the first information as an explanatory variable and using a difference input value, which is a difference between the first input value and the second input value, as an objective variable.
With such a configuration, the air conditioning load learning apparatus according to the seventeenth aspect can associate the first information with the difference input value (the value regarding the actual air conditioning load) to learn the effect of the operation of the building on the internal heat generation, etc. As a result, the air conditioning load learning apparatus can finally grasp the effect of the operation of the building on the air conditioning load.
An air conditioning load learning apparatus according to an eighteenth aspect is the air conditioning load learning apparatus according to the seventeenth aspect, wherein the learning unit generates the learning model further using the first input value as an explanatory variable.
With such a configuration, the air conditioning load learning apparatus according to the eighteenth aspect can generate the learning model having a higher prediction accuracy.
An air conditioning load prediction apparatus according to a nineteenth aspect includes a difference input value prediction unit and a prediction load acquisition unit. The difference input value prediction unit uses the learning model of the air conditioning load learning apparatus according to the seventeenth aspect or the eighteenth aspect to predict the difference input value from at least the first information. The prediction load acquisition unit acquires a prediction air conditioning load that is an air conditioning load in the target space predicted by the physical model. The prediction air conditioning load is predicted by the physical model using, as an input, a third input value. The third input value is obtained by correcting the first input value using the predicted difference input value.
With such a configuration, the air conditioning load prediction apparatus according to the nineteenth aspect can use the learning model having learned the effect of the operation of the building on internal heat generation, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
An air conditioning load learning apparatus according to a twentieth aspect is the air conditioning load learning apparatus according to the first aspect and further includes a prediction load acquisition unit and a difference input value acquisition unit. The prediction load acquisition unit acquires a prediction air conditioning load that is an air conditioning load in the target space predicted by a physical model from at least a thermal property of the target building. The difference input value acquisition unit acquires a difference input value calculated by inverse calculation of the physical model using a difference load that is a difference between the actual air conditioning load and the prediction air conditioning load, The learning unit generates the learning model using the first information as an explanatory variable and using the difference input value as an objective variable.
With such a configuration, the air conditioning load learning apparatus according to the twentieth aspect can associate the first information with the difference input value (the value regarding the actual air conditioning load) to learn the effect of the operation of the building on the internal heat generation, etc. As a result, the air conditioning load learning apparatus can finally grasp the effect of the operation of the building on the air conditioning load.
An air conditioning load learning apparatus according to a twenty-first aspect is the air conditioning load learning apparatus according to the twentieth aspect, wherein the learning unit generates the learning model further using an input value to the physical model as an explanatory variable.
With such a configuration, the air conditioning load learning apparatus according to the twenty-first aspect can generate the learning model having a higher prediction accuracy.
An air conditioning load prediction apparatus according to a twenty-second aspect includes a difference input value prediction unit, a prediction difference load acquisition unit, and a load prediction unit. The difference input value prediction unit uses the learning model of the air conditioning load learning apparatus according to the twentieth aspect or the twenty-first aspect to predict the difference input value from at least the first information. The prediction difference load acquisition unit acquires the difference load predicted by the physical model using the predicted difference input value as an input. The load prediction unit predicts an air conditioning load in the target space based on the acquired difference load and the prediction air conditioning load.
With such a configuration, the air conditioning load prediction apparatus according to the twenty-second aspect can use the learning model having learned the effect of the operation of the building on internal heat generation, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
An air conditioning load learning apparatus according to a twenty-third aspect is the air conditioning load learning apparatus according to the first aspect and further includes an input value acquisition unit. The input value acquisition unit acquires a first input value and a second input value. The first input value is an input value including at least a thermal property of the target building to a physical model, which outputs a prediction air conditioning load that is a. predicted air conditioning load in the target space. The second input value is an input value calculated by inverse calculation of the physical model using the actual air conditioning load. The learning unit generates the learning model using the first information and the first input value as explanatory variables and using the second input value as an objective variable.
With such a configuration, the air conditioning load learning apparatus according to the twenty-third aspect can associate the first information with the second input value (the value regarding the actual air conditioning load) to learn the effect of the operation of the building on the internal heat generation, etc. As a result, the air conditioning load learning apparatus can finally grasp the effect of the operation of the building on the air conditioning load.
An air conditioning load prediction apparatus according to a twenty-fourth aspect includes an input value prediction unit and a prediction load acquisition unit. The input value prediction unit uses the learning model of the air conditioning load learning apparatus according to the twenty-Third aspect to predict the second input value from the first information and the first input value. The prediction load acquisition unit acquires a prediction air conditioning load that is an air conditioning load in the target space predicted by the physical model using the predicted second input value as an input.
With such a configuration, the air conditioning load prediction apparatus according to the twenty-fourth aspect can use the learning model having learned the effect of the operation of the building on internal heat generation, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
<first Embodiment>
(1) Overall Configuration
An air conditioning load learning apparatus 110 learns an air conditioning load of an air conditioning system 30 installed in a target building 10 and generates a learning model LM1. Hereinafter, a learning process by the air conditioning load learning apparatus 110 is referred to as an air conditioning load learning process. An air conditioning load prediction apparatus 120 uses the learning model LM1 generated by the air conditioning load learning apparatus 110 to predict the air conditioning load of the air conditioning system 30 installed in the target building 10. Hereinafter, a prediction process by the air conditioning load prediction apparatus 120 is referred to as an air conditioning load prediction process. The target building 10 is, for example, an office building. The target space 11 of the air conditioning system 30 in the target building 10 is, for example, an office space.
During the air conditioning load learning process and the air conditioning load prediction process, the air conditioning load learning apparatus 110 and the air conditioning load prediction apparatus 120 use thermal load calculation results by a physical model calculation device 70. The physical model calculation device 70 calculates the thermal load of the target building 10 based on a physical model PM. Hereinafter, a calculation process by the physical model calculation device 70 is referred to as a physical model calculation process.
(2) Detailed Configuration
(2-1) Air Conditioning System
The air conditioning system 30 forms a vapor compression refrigeration cycle to air-condition the target space 11. According to the present embodiment, the air conditioning system 30 is a multi-air conditioning system for buildings. However, the air conditioning system 30 is not limited thereto, and the air conditioning system 30 may be of any type.
The indoor unit 40 and the outdoor unit 50 are coupled via a liquid refrigerant connection pipe 31 and a gas refrigerant connection pipe 32 to form a refrigerant circuit 33. The refrigerant circuit 33 includes an indoor expansion valve 41 and an indoor heat exchanger 42 of the indoor unit 40. Furthermore, the refrigerant circuit 33 includes a compressor 51, a flow direction switching mechanism 52, an outdoor heat exchanger 53, and an outdoor expansion valve 54 of the outdoor unit 50.
The air conditioning system 30 has, as primary operating modes for the air conditioning operation, a cooling operating mode for performing a cooling operation and a heating operating mode for performing a heating operation. The cooling operation is an operation for causing the outdoor heat exchanger 53 to function as a condenser of the refrigerant and causing the indoor heat exchanger 42 to function as an evaporator of the refrigerant to cool the air in the target space 11 where the indoor unit 40 is installed. The heating operation is an operation for causing the outdoor heat exchanger 53 to function as an evaporator of the refrigerant and causing the indoor heat exchanger 42 to function as a condenser of the refrigerant to heat the air in the target space 11 where the indoor unit 40 is installed.
(2-1-1) Indoor Unit
The indoor unit 40 is a unit installed in the target space 11. For example, the indoor unit 40 is a ceiling-embedded unit. As illustrated in
The indoor unit 40 primarily includes the indoor expansion valve 41, the indoor heat exchanger 42, an indoor fan 43, various sensors, and an indoor control unit 44. Various sensors included in the indoor unit 40 will be described below.
(2-1-1-1) Indoor Expansion Valve The indoor expansion valve 41 is a mechanism that adjusts the pressure and the flow rate of the refrigerant flowing through the indoor-side refrigerant circuit 33a. The indoor expansion valve 41 is provided in a refrigerant pipe that couples the liquid side of the indoor heat exchanger 42 and the liquid refrigerant connection pipe 31. The indoor expansion valve 41 is, for example, an electronic expansion valve whose opening degree is adjustable.
(2-1-1-2) Indoor Heat Exchanger
In the indoor heat exchanger 42, heat is exchanged between the refrigerant flowing through the indoor heat exchanger 42 and the air in the target space 11. The indoor heat exchanger 42 is, for example, a fin-and-tube heat exchanger including a plurality of heat transfer tubes and fins.
One end of the indoor heat exchanger 42 is coupled to the liquid refrigerant connection pipe 31 via a refrigerant pipe. The other end of the indoor heat exchanger 42 is coupled to the gas refrigerant connection pipe 32 via a refrigerant pipe. During the cooling operation, the refrigerant flows into the indoor heat exchanger 42 from the liquid refrigerant connection pipe 31 side, and the indoor heat exchanger 42 functions as an evaporator of the refrigerant. During the heating operation, the refrigerant flows into the indoor heat exchanger 42 from the gas refrigerant connection pipe 32 side, and the indoor heat exchanger 42 functions as a condenser of the refrigerant.
(2-1-1-3) Indoor Fan
The indoor fan 43 is a fan that supplies the air to the indoor heat exchanger 42. The indoor fan 43 is, for example, a cross-flow fan. The indoor fan 43 is driven by an indoor fan motor 43a. The number of revolutions of the indoor fan motor 43a can be controlled by an inverter.
(2-1-1-4) Sensor
As illustrated in
The indoor temperature sensor 45a is provided on an air intake side of a casing (not illustrated) of the indoor unit 40. The indoor temperature sensor 45a detects the temperature of the air in the target space Il that flows into the casing of the indoor unit 40 (the suction temperature of the indoor unit 40).
The indoor humidity sensor 45b is provided on the air intake side of the casing (not illustrated) of the indoor unit 40. The indoor humidity sensor 45b detects the humidity of the air in the target space 11 that flows into the casing of the indoor unit 40. According to the present embodiment, the indoor humidity sensor 45h detects a relative humidity. However, this is not a limitation, and the indoor humidity sensor 45b may detect an absolute humidity.
(2-1-1-5) Indoor Control Unit
The indoor control unit 44 controls the operation of each unit included in the indoor unit 40. The indoor control unit 44 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. Furthermore, the indoor control unit 44 includes a timer. According to the present embodiment, the timer is installed in the indoor control unit 44, but may be installed in an outdoor control unit 57 described below.
As illustrated in
The indoor control unit 44 is configured to enable reception of various signals transmitted from a remote controller 46 to operate the indoor unit 40. The various signals transmitted from the remote controller 46 include signals for giving instructions to operate or stop the indoor unit 40 and signals regarding various settings. The signals regarding various settings include, for example, switching signals for the operating modes and signals regarding a setting temperature or a setting humidity for the cooling operation and the heating operation,
The indoor control unit 44 is coupled to the outdoor control unit 57 of the outdoor unit 50 and a total heat exchange control unit 21 of the total heat exchanger 20 via a transmission line 61 in a state where control signals, and the like, can be exchanged. Further, instead of being coupled via the physical transmission line 61, the indoor control unit 44, the outdoor control unit 57, and the total heat exchange control unit 21 may be coupled wirelessly to enable communications. The indoor control unit 44, the outdoor control unit 57, and the total heat exchange control unit 21 cooperate with each other to function as the control unit 60 that controls the overall operation of the air conditioning system 30. The control unit 60 will be described below.
(2-1-2) Outdoor Unit
The outdoor unit 50 is installed outside the target space 11. The outdoor unit 50 is installed, for example, on a roof floor of the target building 10, in which the air conditioning system 30 is provided, or is installed adjacent to the target building 10. As illustrated in
The outdoor unit 50 primarily includes the compressor 51, the flow direction switching mechanism 52, the outdoor heat exchanger 53, the outdoor expansion valve 54, an accumulator 55, an outdoor fan 56, various sensors, and the outdoor control unit 57. Various sensors included in the outdoor unit 50 will be described below.
Further, the outdoor unit 50 includes an intake pipe 34a, a discharge pipe 34b, a first gas refrigerant pipe 34c, a liquid refrigerant pipe 34d, a second gas refrigerant pipe 34e, a liquid-side shutoff valve 35, and a gas-side shutoff valve 36. The intake pipe 34a couples the flow direction switching mechanism 52 and an intake side of the compressor 51. The intake pipe 34a is provided with the accumulator 55. The discharge pipe 34b couples a discharge side of the compressor 51 and the flow direction switching mechanism 52. The first gas refrigerant pipe 34c couples the flow direction switching mechanism 52 and a gas side of the outdoor heat exchanger 53. The liquid refrigerant pipe 34d couples a liquid side of the outdoor heat exchanger 53 and the liquid refrigerant connection pipe 31. The liquid refrigerant pipe 34d is provided with the outdoor expansion valve 54. The liquid-side shutoff valve 35 is provided in a coupling portion between the liquid refrigerant pipe 34d and the liquid refrigerant connection pipe 31, The second gas refrigerant pipe 34e couples the flow direction switching mechanism 52 and the gas refrigerant connection pipe 32. The gas-side shutoff valve 36 is provided in a coupling portion between the second gas refrigerant pipe 34e and the gas refrigerant connection pipe 32.
(2-1-2-1) Compressor
As illustrated in
The compressor 51 is, for example, a rotary type or scroll type volume compressor. The compression mechanism of the compressor 51 is driven by a compressor motor 51a. When the compression mechanism is driven by the compressor motor 51a, the refrigerant is compressed by the compression mechanism. The compressor motor 51a is a motor whose number of revolutions can be controlled by an inverter. By controlling the number of revolutions of the compressor motor 51a, the volume of the compressor 51 is controlled.
(2-1-2-2) Flow Direction Switching Mechanism
The flow direction switching mechanism 52 is a mechanism that switches the flow direction of the refrigerant to change the state of the refrigerant circuit 33 between a first state and a second state.
When the refrigerant circuit 33 is in the first state, the outdoor heat exchanger 53 functions as a condenser of the refrigerant, and the indoor heat exchanger 42 functions as an evaporator of the refrigerant. The flow direction switching mechanism 52 sets the state of the refrigerant circuit 33 to the first state during the cooling operation. In other words, during the cooling operation, the flow direction switching mechanism 52 causes the intake pipe 34a to communicate with the second gas refrigerant pipe 34e and causes the discharge pipe 34h to communicate with the first gas refrigerant pipe 34c as indicated in the solid lines in the flow direction switching mechanism 52 in
When the refrigerant circuit 33 is in the second state, the outdoor heat exchanger 53 functions as an evaporator of the refrigerant, and the indoor heat exchanger 42 functions as a condenser of the refrigerant. The flow direction switching mechanism 52 sets the state of the refrigerant circuit 33 to the second state during the heating operation. In other words, during the heating operation, the flow direction switching mechanism 52 causes the intake pipe 34a to communicate with the first gas refrigerant pipe 34c and causes the discharge pipe 34b to communicate with the second gas refrigerant pipe 34e as indicated in the broken lines in the flow direction switching mechanism 52 in
According to the present embodiment, the flow direction switching mechanism 52 is a four-way switching valve.
(2-1-2-3) Outdoor Heat Exchanger
In the outdoor heat exchanger 53, heat is exchanged between the refrigerant flowing inside the outdoor heat exchanger 53 and the air in the outdoor area where the outdoor unit 50 is installed. The outdoor heat exchanger 53 is, for example, a fin-and-tube heat exchanger including a plurality of heat transfer tubes and fins.
One end of the outdoor heat exchanger 53 is coupled to the liquid refrigerant pipe 34d. The other end of the outdoor heat exchanger 53 is coupled to the first gas refrigerant pipe 34c.
The outdoor heat exchanger 53 functions as a condenser of the refrigerant during the cooling operation and functions as an evaporator of the refrigerant during the heating operation.
(2-1-2-4) Outdoor Expansion Valve
The outdoor expansion valve 54 is a mechanism that adjusts the pressure and the flow rate of the refrigerant flowing through the liquid refrigerant pipe 34d. As illustrated in
(2-1-2-5) Accumulator
The accumulator 55 has a gas-liquid separation function to separate the flowing refrigerant into a gas refrigerant and a liquid refrigerant. Furthermore, the accumulator 55 is a container that has the function to store an excess refrigerant generated in accordance with fluctuations in the operation load of the indoor unit 40, etc. As illustrated in
(2-1-2-6) Outdoor Fan
The outdoor fan 56 is a fan that supplies air to the outdoor heat exchanger 53. Specifically, the outdoor fan 56 is a fan that takes in the heat source air outside the outdoor unit 50 into the casing (not illustrated) of the outdoor unit 50, supplies the heat source air to the outdoor heat exchanger 53, and discharges the air, which has exchanged heat with the refrigerant in the outdoor heat exchanger 53, to the outside of the casing of the outdoor unit 50. The outdoor fan 56 is, for example, a propeller fan. The outdoor fan 56 is driven by an outdoor fan motor 56a. The number of revolutions of the outdoor fan motor 56a can be controlled by an inverter.
(2-1-2-7) Sensor
As illustrated in
The outdoor temperature sensor 58a measures the temperature of the air in the outdoor area where the outdoor unit 50 is installed.
The outdoor humidity sensor 58b measures the humidity of the air in the outdoor area where the outdoor unit 50 is installed. According to the present embodiment, the outdoor humidity sensor 58b detects the relative humidity. However, this is not a limitation, and the outdoor humidity sensor 58b may detect the absolute humidity.
The outdoor solar radiation sensor 58c measures the solar radiation in the outdoor area where the outdoor unit 50 is installed.
Furthermore, these sensors are some of the sensors, and the outdoor unit 50 also includes sensors that measure the refrigerant temperature, the refrigerant pressure, etc.
(2-1-2-8) Outdoor Control Unit
The outdoor control unit 57 controls the operation of each unit included in the outdoor unit 50. The outdoor control unit 57 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program.
As illustrated in
The outdoor control unit 57 is coupled to the indoor control unit 44 of the indoor unit 40 and the total heat exchange control unit 21 of the total heat exchanger 20 via the transmission line 61 in a state where control signals, and the like, can be exchanged. The outdoor control unit 57, the indoor control unit 44, and the total heat exchange control unit 21 cooperate with each other to function as the control unit 60 that controls the overall operation of the air conditioning system 30. The control unit 60 will be described below.
(2-1-3) Total Heat Exchanger
The total heat exchanger 20 is a unit installed in the target space 11. For example, the total heat exchanger 20 is a ceiling-embedded unit, The total heat exchanger 20 performs total heat exchange operation. The total heat exchange refers to the exchange of sensible heat and the exchange of latent heat. The total heat exchange operation is an operation to perform total heat exchange between the exhaust air discharged from the target space 11 to outside the room and the intake air taken into the target space 11 as fresh air from outside the room.
The total heat exchanger 20 includes the total heat exchange control unit 21. The total heat exchange control unit 21 controls the operation of each unit included in the total heat exchanger 20. The total heat exchange control unit 21 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program.
The total heat exchange control unit 21 is configured to enable reception of various signals transmitted from the remote controller 46 to operate the total heat exchanger 20. The various signals transmitted from the remote controller 46 include signals for giving instructions to operate or stop the total heat exchanger 20 and signals regarding various settings.
(2-1-4) Control Unit
As illustrated in
As illustrated in
The control unit 60 controls the operation and stoppage of the air conditioning system 30 and the operations of various devices of the air conditioning system 30 based on measurement signals of the various sensors 45a, 45b, 58a, 58b, 58c, and the like, commands received by the indoor control unit 44 and the total heat exchange control unit 21 from the remote controller 46, etc.
Furthermore, the control unit 60 acquires information (air conditioning information 37) about the air conditioning system 30 from the air conditioning system 30 at predetermined time intervals and transmits the information to the air conditioning database 80. According to the present embodiment, the predetermined time interval is per unit time. For the verification described below, the unit time is set to one hour. The air conditioning database 80 stores the transmitted air conditioning information 37.
The time information I11 and the date information I12 are the time and date when the control unit 60 acquires the air conditioning information 37. The time information I11 is, for example, a 6-digit numerical value indicating a time, such as 18 (hour) 00 (minute) 00 (second).
The date information I12 is, for example, an 8-digit numerical value indicating a date, such as 2018 (year) 01 (month) 01 (day).
The indoor humidity I21 is a measurement value of the indoor humidity sensor 45b when the control unit 60 acquires the air conditioning information 37.
The indoor temperature I22 is a measurement value of the indoor temperature sensor 45a when the control unit 60 acquires the air conditioning information 37.
The air conditioning operating time I23 is a time during which the air conditioning operation is performed. In other words, the air conditioning operating time I23 is a time in a state where the indoor unit 40 is on in the remote controller 46. According to the present embodiment, the air conditioning operating time I23 is a time during which the air conditioning operation is performed in a unit time from the previous acquisition of the air conditioning information 37 by the control unit 60. The time up to a unit time is stored as the air conditioning operating time I23.
The post air conditioning operation start elapsed time 124 is a cumulative operating time (e.g., up to three hours) from the start of the most recent air conditioning operation after the stoppage of the air conditioning operation for a long time (e.g., five hours or more) at the time when the control unit 60 acquires the air conditioning information 37. When the air conditioning operation has stopped when the control unit 60 acquires the air conditioning information 37, 0 is stored as the post air conditioning operation start elapsed time I24. Even when the air conditioning operation has started, 0 is stored as the post air conditioning operation start elapsed time I24 in a case where the operation does not correspond to the operation after the stoppage of the air conditioning operation for a long time or in a case where the cumulative operating time from the start of the air conditioning operation is more than a predetermined time (e.g., three hours described above). When the air conditioning operation has started, the operation corresponds to the operation after the stoppage of the air conditioning operation for a long time, and the cumulative operating time from the start of the air conditioning operation falls within a predetermined time (e.g., three hours described above), the time from the start of the most recent air conditioning operation to the time when the control unit 60 acquires the air conditioning information 37 is stored as the post air conditioning operation start elapsed time I24.
The outside air temperature I25 is a measurement value of the outdoor temperature sensor 58a when the control unit 60 acquires the air conditioning information 37.
The outside air humidity I26 is a measurement value of the outdoor humidity sensor 58b when the control unit 60 acquires the air conditioning information 37.
The solar radiation I27 is a measurement value of the outdoor solar radiation sensor 58c when the control unit 60 acquires the air conditioning information 37.
The actual air conditioning load 140 is an actual air conditioning load processed by the air conditioning system 30 in the target space 11 inside the target building 10. According to the present embodiment, the actual air conditioning load 140 is an air conditioning capacity actual measurement value of the outdoor unit 50. The air conditioning capacity actual measurement value of the outdoor unit 50 is calculated using a compressor curve method (CC method) based on the actual measurement value of an internal sensor of the outdoor unit 50. The CC method is a method for calculating a supply capacity by using a refrigerant enthalpy difference obtained from sensor information (refrigerant temperature, refrigerant pressure, etc.) inside the multi-air conditioning system for buildings and a refrigerant circulation amount obtained from the number of revolutions of the compressor motor 51a using a compressor property curve. As the cooling load actual measurement value included in the actual air conditioning load 140, a cooling capacity actual measurement value of the outdoor unit 50 is used as it is. As the heating load actual measurement value included in the actual air conditioning load 140, the value is used, which is obtained by removing heat loss due to the liquid refrigerant connection pipe 31 and the gas refrigerant connection pipe 32 between the outdoor unit 50 and the indoor unit 40 from the heating capacity actual measurement value of the outdoor unit 50. The cooling load actual measurement value and the heating load actual measurement value are calculated for each system of the outdoor unit 50.
The total heat exchange operating time 131 is a time during which the total heat exchange operation is performed. In other words, the total heat exchange operating time 131 is a time in a state where the total heat exchanger 20 is on in the remote controller 46. According to the present embodiment, the total heat exchange operating time 131 is a time during which the total heat exchange operation is performed in a unit time from the previous acquisition of the air conditioning information 37 by the control unit 60. The time up to a unit time is stored as the total heat exchange operating time 131.
(2-2) Physical Model Calculation Device
The physical model calculation device 70 includes a control arithmetic device and a. storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program.
The physical model calculation device 70 calculates the thermal load of the target building 10 based on the physical model PM. FIG, 4 is a configuration diagram of a physical model calculation process. As illustrated in
The input value 130 includes at least a thermal property of the target building 10. The thermal property of the target building 10 is, for example, thermal insulation property, heat capacity, building shape, solar radiation shielding performance, amount of ventilation of the target building 10, the thermal property of building materials used for obtaining them, Q value, etc. The content of the input value 130 depends on the physical model PM and dynamic thermal load calculation software. In addition to the thermal property of the target building 10, the input value 130 may include, for example, outside air condition, solar radiation condition, indoor condition (e.g., indoor temperature and indoor humidity), condition of air conditioning or outside air introduction (e.g., air conditioning, on/off or ventilation introduction on/off), and information about internal heat generation, etc.
The physical model PM may be a model including a simple heat balance equation or may be a model included in existing dynamic thermal load calculation software. Examples of the model including a simple heat balance equation include the following Equation 1.
Q
solar
+Q
skin
+Q
buil
+Q
air
+Q
outair
+I
gΦHVAC (1)
Here, Qsolar is a solar radiation load from a window, Qoutair is a through-flow thermal load from a wall, Qbuil is heat storage of indoor thermal mass. Qair is heat storage of air, Qsolar is a ventilation load, Ig is internal heat generation, and ΦHVAC is an air conditioning load in the target space 11. The linear physical model PM such as Math. 1 makes it possible to inversely calculate the corresponding input value 130 from the air conditioning load in the target space 11. Furthermore, the linear physical model PM such as Math. 1 makes it possible to calculate the difference in the air conditioning load in the target space 11 from the difference in the input value 130 and to inversely calculate the difference in the corresponding input value 130 from the difference in the air conditioning load in the target space 11.
(2-3) Air Conditioning Load Learning Apparatus
The air conditioning load learning apparatus 110 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The actual load acquisition unit 119, the prediction load acquisition unit 113, the first information acquisition unit 111, the second information acquisition unit 112, and the learning unit 114 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-3-1) Actual Load Acquisition Unit
As illustrated in
(2-3-2) Prediction Load Acquisition Unit
As illustrated in
(2-3-3) First Information Acquisition Unit
As illustrated in
According to the present embodiment, the time information I11 and the date information I12 are acquired from the air conditioning information 37 in the air conditioning database 80. The day-of-week information I13 and the holiday information I14 are calculated based on the date information I12. The day-of-week information I13 is a numerical value corresponding to the day of week, such as “1” for Monday and “2” for Tuesday. The holiday information I14 is a numerical value, for example, “0” for a day that is not a holiday and “1” for a holiday.
(2-3-4) Second Information Acquisition Unit
As illustrated in
According to the present embodiment, the second information I2 is acquired from the air conditioning information 37 in the air conditioning database 80.
(2-3-5) Learning Unit
As illustrated in
The period (learning data period) of data used for the air conditioning load learning process of the air conditioning load learning apparatus 110 may be shorter than the period (prediction data period) of data used for the air conditioning load prediction process of the air conditioning load prediction apparatus 120. Specifically, the learning model LM1 may be generated based on the first information I1, the second information I2, the actual air conditioning load 140, and the first prediction air conditioning load 161 in a period shorter than the prediction data period.
According to the present embodiment, the learning model LM1 is a regression-type multilayer perceptron. However, the learning model LM1 is not limited thereto, and the learning model LM1 may be any regression model. The regression-type multilayer perceptron will be described in the verification below,
(2-4) Air Conditioning Load Prediction Apparatus
The air conditioning load prediction apparatus 120 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The prediction load acquisition unit 123, the first information acquisition unit 121, the second information acquisition unit 122, the difference load prediction unit 124, and the load prediction unit 125 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-4-1) Prediction Load Acquisition Unit
As illustrated in
(24-2) First Information Acquisition Unit
As illustrated in
(2-4-3) Second Information Acquisition Unit
As illustrated in
(2-4-4) Difference Load Prediction Unit
As illustrated in
From the first information I1 and the second information I2 in a period longer than the learning data period, the difference load prediction unit 124 may predict the difference load 170 in the same long period.
(2-4-5) Load Prediction Unit
As illustrated in
Based on the predicted difference load 170 in the long period and the first prediction air conditioning load 16-1 in the same long period, the load prediction unit 125 may predict the air conditioning load in the same long period in the target space 11.
(3) Process
(3-1) Air Conditioning Load Learning Process
The air conditioning load learning process will be described using the flowchart of
As described in Step S111 the air conditioning load learning apparatus 110 acquires the actual air conditioning load 140 from the air conditioning information 37 in the air conditioning database 80.
After acquiring the actual air conditioning load 140, the air conditioning load learning apparatus 110 acquires the first prediction air conditioning load 161 from the physical model calculation device 70 as described in Step S112.
After acquiring the first prediction air conditioning load 161, the air conditioning load learning apparatus 110 acquires the first information I1 from the air conditioning information 37 in the air conditioning database 80 as described in Step S113.
After acquiring the first information I1, the air conditioning load teaming apparatus 110 acquires the second information I2 from the air conditioning information 37 in the air conditioning database 80 as described in Step S114.
After acquiring the second information I2, the air conditioning load learning apparatus 110 calculates the difference load 170, which is the difference between the actual air conditioning load 140 and the first prediction air conditioning load 161, as described in Step S115.
After calculating the difference load 170, the air conditioning load learning apparatus 110 generates the learning model LM1 using the first information I1 and the second information I2 as explanatory variables and using the difference load 170 as an objective variable as described in Step S116.
(3-2) Air Conditioning Load Prediction Process
The air conditioning load prediction process will be described using the flowchart of
As described in Step S121, the air conditioning load prediction apparatus 120 acquires the first prediction air conditioning load 161 from the physical model calculation device 70.
After acquiring the first prediction air conditioning load 161, the air conditioning load prediction apparatus 120 acquires the first information I1 from the air conditioning information 37 in the air conditioning database 80 as described in Step S122.
After acquiring the first information I1, the air conditioning load prediction apparatus 120 acquires the second information I2 from the air conditioning information 37 in the air conditioning database 80 as described in Step S123.
After acquiring the second information I2, the air conditioning load prediction apparatus 120 predicts the difference load 170 from the first information I1 and the second information I2 using the learning model LM1 as described in Step S124.
After predicting the difference load 170, the air conditioning load prediction apparatus 120 adds or subtracts the difference load 170 to or from the first prediction air conditioning load 161 to calculate the second prediction air conditioning load 162 as described in Step S125.
(4) Verification of Prediction Accuracy
(4-1) Verification Flow
The air conditioning load of the target building 10 is predicted using the air conditioning load learning apparatus 110 and the air conditioning load prediction apparatus 120, and the prediction accuracy is verified.
The learning data period of this verification is the entire period of 2018 and a partial period of 2018, The prediction data period for this verification is the entire period of 2019,
In this verification, first, the air conditioning load learning apparatus 110 generates the learning model LM1 from the first information I1, the second information I2, the actual air conditioning load 140, and the first prediction air conditioning load 161 of 2018. Subsequently, the air conditioning load prediction apparatus 120 inputs the first information I1, and the second information I2 of 2019 to the learning model LM1, from which the difference load 170 of 2019 is output. Further, the first prediction air conditioning load 161 of 2019 is added to the difference load 170 of 2019 to calculate the second prediction air conditioning load 162 of 2019. Finally, the actual air conditioning load 140 of 2019 is compared with the second prediction air conditioning load 162 of 2019 to verify the prediction accuracy.
(4-2) Target Building, Target Space
Table 1 below is an overview of the target building 10 used for this verification.
As described in Table 1, the target space 11 of the target building 10 is offices on the second to ninth floors.
(4-3) Air Conditioning System
The air conditioning system 30 for this verification is a multi-air conditioning system for buildings. As illustrated in
(4-4) Various Types of Data
For this verification, hourly data is used.
(4-4-1) Actual Air conditioning Load
The actual air conditioning load 140 is acquired from the air conditioning information 37 in the air conditioning database 80. In this verification, as there are two systems of the outdoor units 50 per floor, the sums of the cooling load actual measurement value and the heating load actual measurement value of the two systems are used as the cooling load actual measurement value and the heating load actual measurement value of each floor.
(4-4-2) First Prediction Air Conditioning Load
For this verification, Energyplus is used as dynamic thermal load calculation software for calculating the first prediction air conditioning load 161. Trimble SketchUp is used to model the chamber shape. Table 2 below describes the wall and window structure of the target building 10 for the dynamic thermal load calculation.
Table 3 below describes various input conditions to the dynamic thermal load calculation software (the input value 130 to the physical model PM).
(4-4-3) Explanatory Variable of Learning Model
Table 4 below describes the explanatory variables of the learning model LM1 for this verification.
(4-4-4) Objective Variable of Learning Model
The objective variable of the learning model LM1 in this verification is the difference load 170 obtained by subtracting the first prediction air conditioning load 161 from the actual air conditioning load 140.
(4-5) Learning model
In this verification, the learning model LM1 uses a multilayer perceptron. The multilayer perceptron is a model that enables nonlinear regression by providing a hidden layer between an input layer and an output layer for linear regression and applying a nonlinear function to a result of a weighted sum.
To create the model, the data set of 2018 is divided into a training set and a validation set, and hyperparameters (the number of hidden layers, the number of nodes in the hidden layer, and a regularization parameter) are adjusted such that the accuracy in the validation set becomes the highest. Learning is then performed again with the adjusted parameters using all the data set of 2018.
(4-6) Evaluation Index
As an error index, the CVRMSE (Coefficient of Variation of the Root Mean Square Error) described below is used.
Here, mi is an actual measurement value, si is a prediction value,
(4-7) Verification Result
A result of annual cooling load prediction in the future (one year of 2019) from the annual (one year of 2018) actual measurement and the short-period (one month of July 2018) actual measurement is described here.
(5) Feature
(5-1)
Conventionally, the difference between the actual air conditioning load measurement value and the external thermal load (excluding the internal heat generation) obtained by the physical model is determined so that the internal heat generation related to the operation of the building is estimated and used for air conditioning load prediction. In order to predict the air conditioning load according to this method, there is a need to input the information about the operation of the building so as to reflect the effect of the operation of the building on the internal heat generation, etc.
However, the information about the operation of the building is typically hard to obtain and is often incomplete even if it is obtained. Therefore, the conventional method has an issue that it is difficult to grasp the effect of the operation of the building on the air conditioning load.
In the air conditioning load learning apparatus 110 according to the present embodiment, the actual load acquisition unit 119 acquires the actual air conditioning load 140 that is an actual air conditioning load in the target space 11 inside the target building 10. The first information acquisition unit 111 acquires the first information I1. The first information i1 is information about the operation of the target building 10. The prediction load acquisition unit 113 acquires the first prediction air conditioning load 161 that is an air conditioning load in the target space 11 predicted by the physical model PM from at least the thermal property of the target building 10. The second information acquisition unit 112 acquires the second information I2. The second information I2 is at least one of the indoor humidity I21, the indoor temperature I22, the air conditioning operating time I23, the post air conditioning operation start elapsed time I24, the outside air temperature I25, the outside air humidity 126, and the solar radiation 127, The learning unit 114 generates the learning model LM1 using the first information I1 and the second information I2 as explanatory variables and using the difference load 170. which is the difference between the actual air conditioning load 140 and the first prediction air conditioning load 161, as an objective variable.
Therefore, the air conditioning load learning apparatus 110 associates the first information I1 and the second information I2 with the difference load 170 (the value regarding the actual air conditioning load 140) so as to learn the effect of the operation of the building on the internal heat generation, the outside air introduction load, the heat storage load, the solar radiation load, etc. As a result, the air conditioning load learning apparatus 110 can finally grasp the effect of the operation of the building on the air conditioning load.
(5-2)
In the air conditioning load learning apparatus 110 according to the present embodiment, the actual air conditioning load 140 is an air conditioning load per unit time.
As a result, the air conditioning load learning apparatus 110 can learn the effect of the operation of the building on the air conditioning load in more detail.
(5-3)
In the air conditioning load prediction apparatus 120 according to the present embodiment, the difference load prediction unit 124 uses the learning model LM1 to predict the difference load 170 from the first information I1 and the second information I2. The load prediction unit 125 predicts the air conditioning load (the second prediction air conditioning load 162) in the target space 11 based on the predicted difference load 170 and the first prediction air conditioning load 161.
As a result, the air conditioning load prediction apparatus 120 can use the learning model LM1 having learned the effect of the operation of the building on the internal heat generation, the outside air introduction load, the heat storage load, the solar radiation load, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
(5-4)
In the air conditioning load prediction apparatus 120 according to the present embodiment, the learning model LM1 is generated based on the first information I1, the second information I2, the actual air conditioning load 140, and the first prediction air conditioning load 161 in the short period. From the first information I1 and the second information I2 in the long period that is a period longer than the short period, the difference load prediction unit 124 predicts the difference load 170 in the same long period. Based on the predicted difference load 170 in the long period and the first prediction air conditioning load 161 in the same long period, the load prediction unit 125 predicts the air conditioning load (the second prediction air conditioning load 162) in the same long period in the target space 11.
As a result, the air conditioning load prediction apparatus 120 can use the learning model LM1 having learned with the data in the short period to predict the air conditioning load in the long period.
(6) Modification
(6-1) Modification 1A
According to the present embodiment, the first information I1, the second information I2, and the actual air conditioning load 140 are acquired from the air conditioning information 37 in the air conditioning database 80. However, these pieces of information may be directly acquired from the air conditioning system 30.
(6-2) Modification 1B
According to the present embodiment, the outside air temperature I25, the outside air humidity I26, and the solar radiation I27, which are the second information I2, are acquired from the air conditioning information 37 in the air conditioning database 80. However, these pieces of information may be acquired from weather data (AMEDAS) on the neighborhood of the target building 10, etc.
(6-3) Modification 1C
According to the present embodiment, the learning unit 114 generates the learning model LM1 using the first information I1 and the second information I2 as explanatory variables and using the difference load 170, which is the difference between the actual air conditioning load 140 and the first prediction air conditioning load 161, as an objective variable. However, the learning unit 114 may generate the learning model LM1 using the first prediction air conditioning load 161 as an explanatory variable instead of the second information I2. Further, the learning unit 114 may generate the learning model LM1 using the first information I1, the second information I2, and the first prediction air conditioning load 161 as explanatory variables. As a result, the air conditioning load learning apparatus 110 can generate the learning model LM1 with a higher prediction accuracy.
In this case, the difference load prediction unit 124 uses the learning model LM1 to predict the difference load 170 from the first information I1 and the second information I2 or the first prediction air conditioning load 161. Furthermore, from the first information I1 and the second information I2 or the first prediction air conditioning load 161 in the long period that is a period longer than the short period, the difference load prediction unit 124 predicts the difference load 170 in the same long period.
(6-4)
Although the embodiment of the present disclosure has been described above, it is understood that various modifications may be made to forms and details without departing from the spirit and scope of the present disclosure described in claims.
<Second Embodiment>
A different part from the first embodiment is primarily described below. Therefore, the present embodiment is basically the same as the first embodiment except for the contents described according to the present embodiment.
(1) Overall Configuration
(2) Detailed Configuration
(2-1) Air Conditioning System
The air conditioning system 30 and the air conditioning database 80 are the same as those according to the first embodiment.
(2-2) Physical Model Calculation Device
The physical model calculation device 70 calculates the thermal load of the target building 10 based on the physical model PM.
(2-3) Air Conditioning Load Learning Apparatus
The air conditioning load learning apparatus 300 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program, Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The actual load acquisition unit 219, the prediction load acquisition unit 213, the first information acquisition unit 211, the second information acquisition unit 212, and the learning unit 214 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-3-1) Actual Load Acquisition Unit
As illustrated in
(2-3-2) Prediction Load Acquisition Unit
As illustrated in
(2-3-3) First Information Acquisition Unit
As illustrated in
(2-3-4) Second Information Acquisition Unit
As illustrated in
(2-3-5) Learning Unit
As illustrated in
The period (learning data period) of data used for the air conditioning load learning process of the air conditioning load learning apparatus 210 may be shorter than the period (prediction data period) of data used for the air conditioning load prediction process of the air conditioning load prediction apparatus 220. Specifically, the learning model LM2 may be generated based on the first prediction air conditioning load 261, the first information I1, the second information I2, and the actual air conditioning load 240 in a period shorter than the prediction data period.
(2-4) Air Conditioning Load Prediction Apparatus
The air conditioning load prediction apparatus 220 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The prediction load acquisition unit 223, the first information acquisition unit 221, the second information acquisition unit 222, and the load prediction unit 225 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-4-1) Prediction Load Acquisition Unit
As illustrated in
(2-4-2) First Information Acquisition Unit
As illustrated in
(2-4-3) Second Information Acquisition Unit
As illustrated in
(2-4-4) Load Prediction Unit
As illustrated in
Based on the first prediction air conditioning load 261, the first information I1, and the second information I2 in a period longer than the learning data period, the load prediction unit 225 may predict the air conditioning load in the same long period in the target space 11.
(3) Process
(3-1) Air Conditioning Load Learning Process
The air conditioning load learning process will be described using the flowchart of
As described in Step S211, the air conditioning load learning apparatus 210 acquires the actual air conditioning load 240 from the air conditioning information 37 in the air conditioning database 80.
After acquiring the actual air conditioning load 240, the air conditioning load learning apparatus 210 acquires the first prediction air conditioning load 261 from the physical model calculation device 70 as described in Step S212.
After acquiring the first prediction air conditioning load 261, the air conditioning load learning apparatus 210 acquires the first information I1 from the air conditioning information 37 in the air conditioning database 80 as described in Step S213.
After acquiring the first information I1, the air conditioning load learning apparatus 210 acquires the second information I2 from the air conditioning information 37 in the air conditioning database 80 as described in Step S214.
After acquiring the second information I2, the air conditioning load learning apparatus 210 generates the learning model LM2 using the first prediction air conditioning load 261, the first information I1, and the second information I2 as explanatory variables and using the actual air conditioning load 240 as an objective variable as described in Step S215.
(3-2) Air Conditioning Load Prediction Process
The air conditioning load prediction process will be described using the flowchart of
As described in Step S221, the air conditioning load prediction apparatus 220 acquires the first prediction air conditioning load 261 from the physical model calculation device 70.
After acquiring the first prediction air conditioning load 261, the air conditioning load prediction apparatus 220 acquires the first information I1 from the air conditioning information 37 in the air conditioning database 80 as described in Step S222.
After acquiring the first information I1, the air conditioning load prediction apparatus 220 acquires the second information I2 from the air conditioning information 37 in the air conditioning database 80 as described in Step S223.
After acquiring the second information I2, the air conditioning load prediction apparatus 220 uses the learning model LM2 to predict the second prediction air conditioning load 262 in the target space 11 from the first prediction air conditioning load 261, the first information I1, and the second information I2 as described in Step S224.
(4) Feature
(4-1)
In the air conditioning load learning apparatus 210 according to the present embodiment, the actual load acquisition unit 219 acquires the actual air conditioning load 240, which is an actual air conditioning load in the target space 11 inside the target building 10. The first information acquisition unit 211 acquires the first information I1. The first information I1 is information about the operation of the target building 10. The prediction load acquisition unit 213 acquires the first prediction air conditioning load 261 that is an air conditioning load in the target space 11 predicted by the physical model PM from at least the thermal property of the target building 10. The second information acquisition unit 212 acquires the second information I2. The second information I2 is at least one of the indoor humidity I21, the indoor temperature I22, the air conditioning operating time I23, the post air conditioning operation start elapsed time I24. the outside air temperature I25, the outside air humidity I26, and the solar radiation I27. The learning unit 214 generates the learning model LM2 using the first prediction air conditioning load 261, the first information I1, and the second information I2 as explanatory variables and using the actual air conditioning load 240 as an objective variable.
Thus, the air conditioning load learning apparatus 210 associates the first information I1 and the second information I2 with the actual air conditioning load 240 (the value regarding the actual air conditioning load 240) so as to learn the effect of the operation of the building on the internal heat generation, the outside air introduction load, the heat storage load, the solar radiation load, etc. As a result, the air conditioning load learning apparatus 210 can finally grasp the effect of the operation of the building on the air conditioning load.
(4-2)
In the air conditioning load learning apparatus 210 according to the present embodiment, the actual air conditioning load 240 is an air conditioning load per unit time.
As a result, the air conditioning load learning apparatus 210 can learn the effect of the operation of the building on the air conditioning load in more detail.
(4-3)
In the air conditioning load prediction apparatus 220 according to the present embodiment, the load prediction unit 225 uses the learning model LM2 to predict the air conditioning load (the second prediction air conditioning load 262) in the target space 11 from the first prediction air conditioning load 261, the first information I1, and the second information I2.
As a result, the air conditioning load prediction apparatus 220 can use the learning model LM2 having learned the effect of the operation of the building on the internal heat generation, the outside air introduction load, the heat storage load, the solar radiation load, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
(4-4)
In the air conditioning load prediction apparatus 220 according to the present embodiment, the learning model LM2 is generated based on the first prediction air conditioning load 261, the first information I1, the second information I2, and the actual air conditioning load 240 in a short period. Based on the first prediction air conditioning load 261, the first information I1, and the second information I2 in the long period that is a period longer than the short period, the load prediction unit 225 predicts the air conditioning load (the second prediction air conditioning load 262) in the same long period in the target space 11.
As a result, the air conditioning load prediction apparatus 220 can use the learning model LM2 having learned with the data in the short period to predict the air conditioning load in the long period.
(5) Modification
Although the embodiment of the present disclosure has been described above, it is understood that various modifications may be made to forms and details without departing from the spirit and scope of the present disclosure described in claims.
<Third Embodiment>
A different part from the first embodiment is primarily described below. Therefore, the present embodiment is basically the same as the first embodiment except for the contents described according to the present embodiment.
(1) Overall Configuration
(2) Detailed Configuration
(2-1) Air Conditioning System
The air conditioning system 30 and the air conditioning database 80 are the same as those according to the first embodiment.
(2-2) Physical Model Calculation Device
The physical model calculation device 70 calculates the thermal load of the target building 10 based on the physical model PM. The present embodiment uses the physical model PM that can inversely calculate the corresponding input value from the air conditioning load in the target space 11.
(2-3) Air Conditioning Load Learning Apparatus
The air conditioning load learning apparatus 310 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The actual load acquisition unit 319, the input value acquisition unit 316, the first information acquisition unit 311, and the learning unit 314 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-3-1) Actual Load Acquisition Unit
As illustrated in
(2-3-2) Input Value Acquisition Unit
As illustrated in
(2-3-3) First Information Acquisition Unit
As illustrated in
(2-3-4) Learning Unit
As illustrated in
(2-4) Air Conditioning Load Prediction Apparatus
The air conditioning load prediction apparatus 320 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program,
Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The first information acquisition unit 321, the difference input value prediction unit 326, the prediction load acquisition unit 323, and the input value acquisition unit 329 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-4-1) First Information Acquisition Unit
As illustrated in
(2-4-2) Difference Input Value Prediction Unit
As illustrated in
(2-4-3) Input Value Acquisition Unit
As illustrated in
(2-4-4) Prediction Load Acquisition Unit
As illustrated in
(3) Process
(3-1) Air Conditioning Load Learning Process
The air conditioning load learning process will be described using the flowchart of
As described in Step S311, the air conditioning load learning apparatus 310 acquires the first input value 331 from the physical model calculation device 70.
After acquiring the first input value 331, the air conditioning load learning apparatus 310 acquires the actual air conditioning load 340 corresponding to the first input value 331 from the air conditioning information 37 in the air conditioning database 80 as described in Step S312.
After acquiring the actual air conditioning load 340, the air conditioning load learning apparatus 310 transmits the actual air conditioning load 340 to the physical model calculation device 70 and acquires the second input value 332 as described in Step S313.
After acquiring the second input value 332, the air conditioning load learning apparatus 310 calculates the difference input value 350 that is the difference between the first input value 331 and the second input value 332 as described in Step S314.
After calculating the difference input value 350, the air conditioning load learning apparatus 310 acquires the first information I1 from the air conditioning information 37 in the air conditioning database 80 as described in Step S315.
After calculating the difference input value 350 and acquiring the first information I1, the air conditioning load learning apparatus 310 generates the learning model LM3 using the first information I1 as an explanatory variable and using the difference input value 350 as an objective variable as described in Step S316.
(3-2) Air Conditioning Load Prediction Process
The air conditioning load prediction process will be described using the flowchart of
The air conditioning load prediction apparatus 320 acquires the first information I1 from the air conditioning information 37 in the air conditioning database 80 as described in Step S321.
After acquiring the first information I1, the air conditioning load prediction apparatus 320 uses the learning model LM3 to predicts the difference input value 350 from the first information I1 as described in Step S322.
After predicting the difference input value 350, the air conditioning load prediction apparatus 320 acquires the first input value 331 from the physical model calculation device 70 as described in Step S323.
After acquiring the first input value 331, the air conditioning load prediction apparatus 320 adds or subtracts the difference input value 350 to or from the first input value 331 to calculate the third input value 333 as described in Step S324.
After calculating the third input value 333, the air conditioning load prediction apparatus 320 acquires, from the physical model calculation device 70, the prediction air conditioning load 360 predicted by the physical model PM using the third input value 333 as an input as described in Step S325.
(4) Feature
(4-1)
In the air conditioning load learning apparatus 310 according to the present embodiment, the actual load acquisition unit 319 acquires the actual air conditioning load 340 that is an actual air conditioning load in the target space 11 inside the target building 10. The first information acquisition unit 311 acquires the first information I1. The first information I1 is information about the operation of the target building 10. The input value acquisition unit 316 acquires the first input value 331 and the second input value 332. The first input value 331 is an input value including at least the thermal property of the target building 10 to the physical model PM that outputs the prediction air conditioning load 360 that is a predicted air conditioning load in the target space 11. The second input value 332 is an input value calculated by inverse calculation of the physical model PM using the actual air conditioning load 340. The learning unit 314 generates the learning model LM3 using the first information I1 as an explanatory variable and using the difference input value 350, which is the difference between the first input value 331 and the second input value 332, as an objective variable.
Therefore, the air conditioning load learning apparatus 310 associates the first information I1 with the difference input value 350 (the value regarding the actual air conditioning load 340) so as to learn the effect of the operation of the building on internal heat generation, etc. As a result, the air conditioning load learning apparatus 310 can finally grasp the effect of the operation of the building on the air conditioning load.
(4-2)
In the air conditioning load prediction apparatus 320 according to the present embodiment, the difference input value prediction unit 326 uses the learning model LM3 to predict the difference input value 350 from the first information I1. The prediction load acquisition unit 323 acquires the prediction air conditioning load 360 that is an air conditioning load in the target space 11 predicted by the physical model PM. The prediction air conditioning load 360 is predicted by the physical model PM using the third input value 333 as an input. The third input value 333 is obtained by correcting the first input value 331 using the predicted difference input value 350.
As a result, the air conditioning load prediction apparatus 320 can use the learning model LM2 having learned the effect of the operation of the building on internal heat generation, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
(5) Modification
(5-1) Modification 3A
According to the present embodiment, the learning unit 314 generates the learning model LM3 using the first information I1 as an explanatory variable and using the difference input value 350, which is the difference between the first input value 331 and the second input value 332, as an objective variable. However, the learning unit 314 may further generate the learning model LM3 using the first input value 331 as an explanatory variable.
As a result, the air conditioning load learning apparatus 310 can generate the learning model LM3 haying a higher prediction accuracy.
In this case, the difference input value prediction unit 326 uses the learning model LM3 to predict the difference input value 350 from the first information I1 and the first input value 331.
(5-2)
Although the embodiment of the present disclosure has been described above, it is understood that various modifications may be made to forms and details without departing from the spirit and scope of the present disclosure described in claims.
<Fourth Embodiment>
A different part from the first embodiment is primarily described below. Therefore, the present embodiment is basically the same as the first embodiment except for the contents described according to the present embodiment.
(1) Overall Configuration
(2) Detailed Configuration
(2-1) Air Conditioning System
The air conditioning system 30 and the air conditioning database 80 are the same as those according to the first embodiment.
(2-2) Physical Model Calculation Device
The physical model calculation device 70 calculates the thermal load of the target building 10 based on the physical model PM. The present embodiment uses the physical model PM that can inversely calculate the corresponding input value from the air conditioning load in the target space I1. Furthermore, the present embodiment uses the physical model PM that can calculate the difference in the air conditioning load in the target space 11 from the difference in the input value and can also inversely calculate the difference in the corresponding input value from the difference in the air conditioning load in the target space 11.
(2-3) Air Conditioning Load Learning Apparatus
The air conditioning load learning apparatus 410 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The actual load acquisition unit 419, the prediction load acquisition unit 413, the difference input value acquisition unit 417, the first information acquisition unit 411, and the learning unit 414 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-3-1) Actual Load Acquisition Unit
As illustrated in
(2-3-2) Prediction Load Acquisition Unit
As illustrated in
(2-3-3) Difference Input Value Acquisition Unit
As illustrated in
(2-3-4) First Information Acquisition Unit
As illustrated in
(2-3-5) Learning Unit
As illustrated in
(2-4) Air Conditioning Load Prediction Apparatus
The air conditioning load prediction apparatus 420 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The first information acquisition unit 421, the difference input value prediction unit 426, the prediction difference load acquisition unit 427, the prediction load acquisition unit 423, and the load prediction unit 425 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-4-1) First Information Acquisition Unit
As illustrated in
(2-4-2) Difference Input Value Prediction Unit
As illustrated in
(2-4-3) Prediction Difference Load Acquisition Unit
As illustrated in
(2-4-4) Prediction Load Acquisition Unit
As illustrated in
(2-4-5) Load Prediction Unit
As illustrated in
(3) Process
(3-1) Air Conditioning Load Learning Process
The air conditioning load learning process will be described using the flowchart of
As described in Step S411, the air conditioning load learning apparatus 410 acquires the actual air conditioning load 440 from the air conditioning information 37 in the air conditioning database 80.
After acquiring the actual air conditioning load 440, the air conditioning load learning apparatus 410 acquires the first prediction air conditioning load 461 from the physical model calculation device 70 as described in Step S412.
After acquiring the first prediction air conditioning load 461, the air conditioning load learning apparatus 410 calculates the difference load 470, which is the difference between the actual air conditioning load 440 and the first prediction air conditioning load 460, as described in Step S413.
After calculating the difference load 470, the air conditioning load learning apparatus 410 acquires the difference input value 450 calculated by inverse calculation of the physical model PM using the difference load 470 as described in Step S414.
After calculating the difference input value 450, the air conditioning load learning apparatus 410 acquires the first information 11 from the air conditioning information 37 in the air conditioning database 80 as described in Step S415.
After acquiring the first information I1, the air conditioning load learning apparatus 410 generates the learning model LM4 using the first information I1 as an explanatory variable and using the difference input value 450 as an objective variable as described in Step S416.
(3-2) Air Conditioning Load Prediction Process
The air conditioning load prediction process will be described using the flowchart of
The air conditioning load prediction apparatus 420 acquires the first information I1 from the air conditioning information 37 in the air conditioning database 80 as described in Step S421.
After acquiring the first information I1, the air conditioning load prediction apparatus 420 uses the learning model LM4 to predict the difference input value 450 from the first information I1 as described in Step S422.
After predicting the difference input value 450, the air conditioning load prediction apparatus 420 acquires the difference load 470 predicted by the physical model PM using the difference input value 450 as an input as described in Step S423.
After acquiring the difference load 470, the air conditioning load prediction apparatus 420 acquires the first prediction air conditioning load 461 from the physical model calculation device 70 as described in Step S424.
After acquiring the first prediction air conditioning load 461, the air conditioning load prediction apparatus 420 adds or subtracts the difference load 470 to or from the first prediction air conditioning load 461 to calculate the second prediction air conditioning load 462 as described in Step S425.
(4) Feature
(4-1)
In the air conditioning load learning apparatus 410 according to the present embodiment, the actual load acquisition unit 419 acquires the actual air conditioning load 440 that is an actual air conditioning load in the target space 11 inside the target building 10. The first information acquisition unit 411 acquires the first information I1. The first information I1 is information about the operation of the target building 10. The prediction load acquisition unit 413 acquires the first prediction air conditioning load 460 that is an air conditioning load in the target space 11 predicted by the physical model PM from at least the thermal property of the target building 10. The difference input value acquisition unit 417 acquires the difference input value 450 calculated by, inverse calculation of the physical model PM using the difference load 470. The learning unit 414 generates the learning model LM4 using the first information I1 as an explanatory variable and using the difference input value 450 as an objective variable.
Therefore, the air conditioning load learning apparatus 410 associates the first information I1 with the difference input value 450 (the value regarding the actual air conditioning load 440) so as to learn the effect of the operation of the building on internal heat generation, etc. As a result, the air conditioning load learning apparatus 410 can finally grasp the effect of the operation of the building on the air conditioning load.
(4-2)
In the air conditioning load prediction apparatus 420 according to the present embodiment, the difference input value prediction unit 426 uses the learning model LM4 to predict the difference input value 450 from the first information I1. The prediction difference load acquisition unit 427 acquires the difference load 470 predicted by the physical model PM using the predicted difference input value 450 as an input. The load prediction unit 425 predicts an air conditioning load (the second prediction air conditioning load 462) in the target space I1 based on the acquired difference load 470 and the first prediction air conditioning load 461.
As a result, the air conditioning load prediction apparatus 420 can use the learning model LM4 having learned the effect of the operation of the building on internal heat generation, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
(5) Modification
(5-1) Modification 4A
According to the present embodiment, the learning unit 414 generates the learning model LM4 using the first information I1 as an explanatory variable and using the difference input value 450 as an objective variable. However, the learning unit 414 may further generate the learning model LM4 using the input value 430 to the physical model PM as an explanatory variable.
As a result, the air conditioning load learning apparatus 410 can generate the learning model LM4 having a higher prediction accuracy.
In this case, the difference input value prediction unit 426 uses the learning model LM4 to predict the difference input value 450 from the first information I1 and the input value 430.
(5-2)
Although the embodiment of the present disclosure has been described above, it is understood that various modifications may be made to forms and details without departing from the spirit and scope of the present disclosure described in claims.
<Fifth Embodiment>
A different part from the first embodiment is primarily described below. Therefore, the present embodiment is basically the same as the first embodiment except for the contents described according to the present embodiment.
(1) Overall Configuration
(2) Detailed Configuration
(2-1) Air Conditioning System
The air conditioning system 30 and the air conditioning database 80 are the same as those according to the first embodiment.
(2-2) Physical Model Calculation Device
The physical model calculation device 70 calculates the thermal load of the target building 10 based on the physical model PM. The present embodiment uses the physical model PM that may inversely calculate the corresponding input value from the air conditioning load in the target space 11.
(2-3) Air Conditioning Load Learning Apparatus
The air conditioning load learning apparatus 310 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The actual load acquisition unit 519, the input value acquisition unit 516, the first information acquisition unit 511, and the learning unit 514 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-3-1) Actual Load Acquisition Unit
As illustrated in
(2-3-2) Input Value Acquisition Unit
As illustrated in
(2-3-3) First Information Acquisition Unit
As illustrated in
(2-34) Learning Unit
As illustrated in
(2-4) Air Conditioning Load Prediction Apparatus
The air conditioning load prediction apparatus 520 includes a control arithmetic device and a storage device. A processor such as a CPU or a GPU may be used as the control arithmetic device. The control arithmetic device reads a program stored in the storage device and performs predetermined image processing and arithmetic processing in accordance with the program. Further, the control arithmetic device can write an arithmetic result in the storage device and read information stored in the storage device in accordance with the program. The first information acquisition unit 521, the input value prediction unit 528, the prediction load acquisition unit 523, and the input value acquisition unit 529 are various functional blocks implemented by the control arithmetic device and the storage device.
(2-4-1) First Information Acquisition Unit
As illustrated in
(2-4-2) Input Value Acquisition Unit
As illustrated in
(2-4-3) Input Value Prediction Unit
As illustrated in
(2-4-4) Prediction Load Acquisition Unit
As illustrated in
(3) Process
(3-1) Air Conditioning Load Learning Process
The air conditioning load learning process will be described using the flowchart of
As described in Step S511, the air conditioning load learning apparatus 510 acquires the first input value 531 from the physical model calculation device 70.
After acquiring the first input value 531, the air conditioning load learning apparatus 510 acquires the actual air conditioning load 540 corresponding to the first input value 531 from the air conditioning information 37 in the air conditioning database 80 as described in Step S512.
After acquiring the actual air conditioning load 540, the air conditioning load learning apparatus 510 transmits the actual air conditioning load 540 to the physical model calculation device 70 and acquires the second input value 532 as described in Step S513.
After acquiring the second input value 532, the air conditioning load learning apparatus 510 acquires the first information I1 from the air conditioning information 37 in the air conditioning database 80 as described in Step S514. After calculating the difference input value 350 and acquiring the first information I1, the air conditioning load learning apparatus 510 generates the learning model LM5 using the first information I1 and the first input value 531 as explanatory variables and using the second input value 532 as an objective variable as described in Step S515.
(3-2) Air Conditioning Load Prediction Process
The air conditioning load prediction process will be described using the flowchart of
The air conditioning load prediction apparatus 520 acquires the first information I1 from the air conditioning information 37 in the air conditioning database 80 as described in Step S521.
After acquiring the first information I1, the air conditioning load prediction apparatus 520 acquires the first input value 531 from the physical model calculation device 70 as described in Step S522.
After acquiring the first input value 531, the air conditioning load prediction apparatus 520 uses the learning model LM5 to predict the second input value 532 from the first information I1 and the first input value 531 as described in Step S523.
After predicting the second input value 532, the air conditioning load prediction apparatus 520 acquires the prediction air conditioning load 560 that is an air conditioning load in the target space 11 predicted by the physical model PM using the second input value 532 as an input as described in Step S524.
(4) Feature
(4-1)
In the air conditioning load learning apparatus 510 according to the present embodiment, the actual load acquisition unit 519 acquires the actual air conditioning load 540 that is an actual air conditioning load in the target space 11 inside the target building 10. The first information acquisition unit 511 acquires the first information I1. The first information I1 is information about the operation of the target building 10. The input value acquisition unit 516 acquires the first input value 531 and the second input value 532 from the physical model calculation device 70. The first input value 531 is an input value including at least the thermal property of the target building 10 to the physical model PM that outputs the prediction air conditioning load 560 that is a predicted air conditioning load in the target space 11. The second input value 532 is an input value calculated by inverse calculation of the physical model PM using the actual air conditioning load 540. The learning unit 514 generates the learning model LM5 using the first information I1 and the first input value 531 as explanatory variables and using the second input value 532 as an objective variable.
Therefore, the air conditioning load learning apparatus 510 associates the first information I1 with the second input value 532 (the value regarding the actual air conditioning load 540) so as to learn the effect of the operation of the building on internal heat generation, etc. As a result, the air conditioning load learning apparatus 510 can finally grasp the effect of the operation of the building on the air conditioning load.
(4-2)
In the air conditioning load prediction apparatus 520 according to the present embodiment, the input value prediction unit 528 uses the learning model LM5 to predict the second input value 532 from the first information I1 and the first input value 531, The prediction load acquisition unit 523 acquires the prediction air conditioning load 560 that is an air conditioning load in the target space 11 predicted by the physical model PM using the second input value 532 as an input.
As a result, the air conditioning load prediction apparatus 520 can use the learning model LM5 having learned the effect of the operation of the building on internal heat generation, and the like, to predict the air conditioning load in consideration of the effect of the operation of the building.
(5) Modification
Although the embodiment of the present disclosure has been described above, it is understood that various modifications may be made to forms and details without departing from the spirit and scope of the present disclosure described in claims.
10 Target building
11 Target space
110, 210, 310, 410, 510 Air conditioning load learning apparatus
111, 211, 311, 411, 511 First information acquisition unit
112, 212 Second information acquisition unit
113, 213, 323, 413, 523 Prediction load acquisition unit
114, 214, 314, 414, 514 Learning unit
119, 219, 319, 419, 519 Actual load acquisition unit
120, 220, 320, 420, 520 Air conditioning load prediction apparatus
124 Difference load prediction unit
125, 225, 425 Load prediction unit
140, 240, 340, 440, 540 Actual air conditioning load
161, 261, 360, 461, 560 Prediction air conditioning load
170, 470 Difference load
316, 516 Input value acquisition unit
326, 426 Difference input value prediction unit
331, 531 First input value
332, 532 Second input value
333 Third input value
350, 450 Difference input value
417 Difference input value acquisition unit
427 Prediction difference load acquisition unit
430 Input value
528 Input value prediction unit
11 First information
12 Second information
LM1 to LM5 Learning model
PM Physical model
PTL 1: U.S. Pat. No. 5,943,255
| Number | Date | Country | Kind |
|---|---|---|---|
| 2020-149959 | Sep 2020 | JP | national |
| 2021-123224 | Jul 2021 | JP | national |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/JP2021/032779 | 9/7/2021 | WO |
