JP7463493B2 - Information processing device and refrigeration system - Google Patents

Description

The present disclosure relates to an information processing device that determines the timing to start a defrosting operation in a refrigeration cycle device, and a refrigeration system that includes the refrigeration cycle device and the information processing device.

Conventionally, refrigeration cycle devices that perform defrosting operation are known. For example, JP 2011-127853 A (Patent Document 1) discloses a heat pump device that calculates the frost layer thickness of frost attached to an evaporator and performs a defrosting operation start determination based on the frost layer thickness. With this heat pump device, even if the operating conditions of the heat pump device change due to changes in the compressor frequency and load fluctuations, etc., it is possible to prevent erroneous determination in the defrosting operation start determination.

JP 2011-127853 A

In a refrigeration cycle device, the heat exchanger to be defrosted may be placed in various environments. The start timing of the defrosting operation that minimizes the power consumption of the refrigeration cycle device per unit time depends on the environment. In order to reduce the power consumption of the refrigeration cycle device, the start timing of the defrosting operation needs to be determined according to different criteria for each environment in which the heat exchanger to be defrosted is placed. However, the heat pump device disclosed in Patent Document 1 does not take into consideration the fact that different criteria are required for determining the start of the defrosting operation for each environment in which the evaporator is placed.

This disclosure has been made to solve the problems described above, and its purpose is to reduce the power consumption of a refrigeration cycle device performing defrosting operation.

An information processing device according to one aspect of the present disclosure outputs a start timing of a defrosting operation to a refrigeration cycle device. In the refrigeration cycle device, a refrigerant circulates through the compressor, the first heat exchanger, the expansion valve, and the second heat exchanger arranged in a specific space in that order. The information processing device includes a memory unit and an inference unit. The memory unit stores a defrost inference model that infers the start timing from specific information including information related to the amount of moisture contained in the air in the specific space. The inference unit is configured to determine the start timing from the specific information using the defrost inference model that has been trained by machine learning.

An information processing device according to another aspect of the present disclosure outputs a start timing of a defrosting operation to a refrigeration cycle device. In the refrigeration cycle device, a refrigerant circulates through the compressor, the first heat exchanger, the expansion valve, and the second heat exchanger arranged in a specific space, in that order. The information processing device includes a memory unit and a learning unit. The memory unit stores a defrosting inference model that infers the start timing from specific information including information related to the amount of moisture contained in the air in the specific space. The learning unit uses machine learning to make the defrosting inference model a learned model.

According to the information processing device of the present disclosure, the power consumption of a refrigeration cycle device performing defrosting operation can be reduced by using a defrosting inference model that infers the start timing from specific information including information regarding the amount of moisture contained in the air in a specific space.

1 is a block diagram showing a functional configuration of a refrigeration system including an information processing device according to an embodiment. 2 is a diagram showing an example of a refrigerant circulation path formed by the cooler and the heat source unit of FIG. 1 . 11 is a diagram showing an example of a relationship between an interval of a defrosting operation and an amount of power consumption of a refrigeration cycle device per unit time. FIG. FIG. 2 is a block diagram showing a functional configuration of the information processing apparatus shown in FIG. 1 . 5 is a flowchart showing a flow of a learning process performed by a learning unit in FIG. 4 . 5 is a flowchart showing the flow of an inference process performed by an inference unit in FIG. 4 . FIG. 5 is a block diagram showing a hardware configuration of the information processing device shown in FIG. 4. FIG. 11 is a block diagram showing a hardware configuration of an information processing device according to a first modified example of the embodiment. FIG. 11 is a block diagram showing a hardware configuration of an information processing device according to a second modified example of the embodiment.

Hereinafter, the embodiments of the present disclosure will be described in detail with reference to the drawings. Note that the same or corresponding parts in the drawings are designated by the same reference numerals, and their explanations will not be repeated in principle.

Figure 1 is a block diagram showing the functional configuration of a refrigeration system 1 including an information processing device 100 according to an embodiment. The refrigeration system 1 shown in Figure 1 includes an information processing device 100 and a refrigeration cycle device 200. The refrigeration cycle device 200 selectively performs normal operation and defrosting operation. The refrigeration cycle device 200 cools a specific space 300 in normal operation. The information processing device 100 outputs the start timing of the defrosting operation to the refrigeration cycle device 200. The information processing device 100 may be connected to the refrigeration cycle device 200 via, for example, a local area network or the like, or may be built into the refrigeration cycle device 200. The information processing device 100 may be placed in a cloud system on the Internet.

The refrigeration cycle device 200 starts a defrosting operation in response to a start timing from the information processing device 100. The refrigeration cycle device 200 includes a control device 210, a cooler 220, a heat source unit 230, and a heater 240. The cooler 220 is disposed in a specific space 300. The cooler 220 cools the specific space 300 by blowing cooled air Ca into the specific space 300.

2 is a diagram showing an example of a refrigerant circulation path formed by the cooler 220 and the heat source unit 230 of FIG. 1. As shown in FIG. 2, the cooler 220 includes an evaporator 221 (second heat exchanger) and an expansion valve 222. The heat source unit 230 includes a compressor 231 and a condenser 232 (first heat exchanger). The expansion valve 222 may be included in the heat source unit 230. The refrigerant circulates through the compressor 231, the condenser 232, the expansion valve 222, and the evaporator 221 in that order.

Referring again to FIG. 1, the control device 210 of the refrigeration cycle device 200 includes a control unit 211. The control unit 211 controls the heat source device 230 (for example, the driving frequency of the compressor 231). The control unit 211 controls the cooler 220 (for example, the airflow per unit time of a fan not shown). The control unit 211 operates the heater 240 in the defrosting operation to heat the evaporator 221, thereby melting the frost formed on the evaporator 221. Note that the method of the defrosting operation is not limited to the heater method, and may be, for example, a water spray method, a hot gas method, or an off-cycle method.

The specific space 300 has an opening/closing entrance Dw. An item 330 can move between the outside of the specific space 300 and the inside of the specific space 300 through the entrance Dw. The entrance Dw is configured to be openable and closable by sliding the door 320 in the direction D1. An item management terminal 310 is arranged in the specific space 300, and multiple items 330 are stored therein. The manager of the specific space 300 or the person in charge of transporting each of the multiple items 330 inputs information (for example, type or quantity) of each of the multiple items 330 into the item management terminal 310. Information about the items 330 may be automatically acquired from a barcode, QR code (registered trademark), or IC (Integrated Circuit) chip attached to the item 330 when the item 330 is carried in or out of the specific space 300, or may be acquired from a remote logistics system via a network.

If the cumulative time of normal operation performed between the previous defrost operation and the current defrost operation (the interval between defrost operations) is shortened, normal operation can be performed with a relatively small amount of frost on the evaporator, so the efficiency of normal operation increases, but the number of defrost operations per unit time increases. On the other hand, if the interval between defrost operations is lengthened, the number of defrost operations per unit time decreases, but the efficiency of normal operation decreases because normal operation continues even when the amount of frost on the evaporator is relatively large. Since there is a trade-off between the number of defrost operations and the efficiency of normal operation, the amount of power consumption of the refrigeration cycle device 200 is minimized by balancing the number of defrost operations and the efficiency of normal operation at a certain interval between defrost operations.

Figure 3 is a diagram showing an example of the relationship between the interval between defrosting operations and the amount of power consumed by the refrigeration cycle device 200 per unit time. The relationship shown by curve Wt1 and the relationship shown by curve Wt2 differ in the amount of moisture per unit volume of air in the specific space 300. The amount of moisture in curve Wt1 is greater than the amount of moisture in curve Wt2.

As shown in FIG. 3, in curve Wt1, power consumption is minimum at defrost operation interval Ot1. In curve Wt2, power consumption is minimum at defrost operation interval Ot2 (>Ot1). The relationship between the interval of defrost operation and power consumption differs depending on the amount of moisture per unit volume of air in specific space 300. Therefore, in order to reduce the power consumption of the refrigeration cycle device 200, a criterion is required to determine the start timing of defrost operation for each environment in which the heat exchanger to be defrosted is located.

Therefore, in the refrigeration system 1, a defrost inference model that has been trained by machine learning is used to determine the timing to start the defrost operation from moisture information (specific information) that includes information on the moisture amount per unit volume of air in the specific space 300. According to the refrigeration system 1, the defrost inference model is adapted to the characteristics of the specific space 300 in which the evaporator 221 is placed, so that the timing to start the defrost operation can be determined with high accuracy for each environment in which the evaporator 221 is placed.

In the following, a case where reinforcement learning is used as machine learning will be described. Reinforcement learning is a learning algorithm in which an agent (acting subject) in an environment where a reward is obtained according to a selected action learns a policy to maximize the expected value of the cumulative value of the reward by repeating action selection based on the state of the environment observed at each sampling time. Note that the learning algorithm is not limited to reinforcement learning, and it is also possible to use, for example, supervised learning, unsupervised learning, or semi-supervised learning. It is also possible to form a defrosting inference model as a neural network and apply deep learning to the defrosting inference model. Machine learning may be performed according to other known methods (for example, genetic programming, functional logic programming, or support vector machine).

1 again, the control device 210 of the refrigeration cycle device 200 further includes a state observation unit 212. The state observation unit 212 acquires moisture information Wi of the specific space 300 at each sampling time and outputs it to the information processing device 100. The moisture information Wi includes the power consumption Pw of the refrigeration cycle device 200 during the time period from the start timing of the previous defrosting operation to the current sampling time, the temperature Tout and humidity Hout of the space outside the specific space 300 at the current sampling time, information Gdi regarding a plurality of items 330 present in the specific space 300 at the current sampling time, the time Ot during which the entrance/exit Dw is open during the time period, information Evi regarding the structure of the evaporator 221, and the normal operation time Nt of the refrigeration cycle device during the time period.

When the entrance/exit Dw is open, air from the space outside the specific space 300 flows into the specific space 300 from the entrance/exit Dw, so the temperature Tout, humidity Hout, and time Ot affect the moisture content of the air in the specific space 300. In addition, the amount of moisture contained in the item 330 varies depending on the type of item 330 (for example, frozen item, room temperature item, seafood, or meat). Therefore, the information Gdi also affects the moisture content of the air in the specific space 300. Note that the moisture information Wi only needs to include at least one of the temperature Tout, humidity Hout, information Gdi, and time Ot.

The structure of the evaporator 221 affects, for example, the likelihood of frost formation in the evaporator 221, or the relationship between the amount of frost and the heat exchange efficiency in the evaporator 221. Therefore, the information Evi affects the need for a defrosting operation for the evaporator 221. By including the information Evi in the moisture information Wi, the start timing of the defrosting operation can be determined with higher accuracy.

The state observation unit 212 acquires the power consumption Pw from the power sensor Sp. The state observation unit 212 acquires the temperature Tout and humidity Hout from the temperature sensor St and humidity sensor Sh, respectively, arranged outside the specific space 300. The state observation unit 212 acquires information Gdi from the item management terminal 310. The state observation unit 212 acquires the time Ot from the opening/closing sensor Sd. The opening/closing sensor Sd includes, for example, a proximity sensor or a magnetic sensor. The state observation unit 212 stores the information Gdi in advance. The information Gdi includes, for example, the spacing between the fins, the type of coating material, the arrangement of the heat transfer tubes, and the type of heat exchanger. The state observation unit 212 acquires the time Nt from the control unit 211.

Fig. 4 is a block diagram showing a functional configuration of the information processing device 100 in Fig. 1. As shown in Fig. 4, the information processing device 100 includes a state acquisition unit 110, a learning unit 120, a storage unit 130, and an inference unit 140. The state acquisition unit 110 acquires moisture information Wi, and outputs a state s _t of the specific space 300 corresponding to the moisture information Wi to the learning unit 120 and the inference unit 140.

The learning unit 120 includes a reward calculation unit 121 and a function update unit 122. The function update unit 122 updates an action value function Q(s t , a t ) in which a state s _t , an action a _t that is either a defrosting operation or a normal operation, and a Q value (evaluation value) that is an evaluation value of the action a _t are associated with each other, using the following formula ( ₁ ) that is generally used in Q learning, which is an _example of reinforcement learning. The action value function Q is stored in the storage unit 130. In reinforcement learning, the action value function Q is included in the defrosting inference model 521.

In formula (1), state s _t represents the state of the specific space 300 at sampling time t, and is determined by the moisture information Wi observed at sampling time t. When action a _t is selected in state s _t , reward r _t+1 is obtained and the state of the specific space 300 transitions from state s _t to s _t+1 . Reward r _t+1 is calculated by the reward calculation unit 121. Reward r _t+1 is associated with action a _t in state s _t and stored in the storage unit 130. Action a is an action that can be selected in state s _t+1 . α is a learning rate, and γ is a discount rate. The learning rate α and the discount rate γ are hyperparameters. Note that the algorithm of reinforcement learning performed by the learning unit 120 is not limited to Q learning, and may be, for example, TD (Temporal Difference) learning.

Figure 5 is a flowchart showing the flow of the learning process performed by the learning unit 120 of Figure 4. The process shown in Figure 5 is called by a main routine (not shown) that controls the learning process in an integrated manner. Below, steps are simply abbreviated as S.

As shown in FIG. 5, in S11, the learning unit 120 determines whether the power consumption of the refrigeration cycle device 200 per unit time at the current sampling time (current power consumption) is equal to or less than the power consumption of the refrigeration cycle device 200 per unit time at the previous sampling time (previous power consumption). If the current power consumption is equal to or less than the previous power consumption (YES in S11), the learning unit 120 increases the reward corresponding to the behavior selected at the previous sampling time and proceeds to S14. If the current power consumption is greater than the previous power consumption (NO in S11), the learning unit 120 decreases the reward in S13 and proceeds to S14. In S14, the learning unit 120 updates the action value function Q and returns the process to the main routine. Note that S11 to S13 are performed by the reward calculation unit 121, and S14 is performed by the function update unit 122.

Figure 6 is a flowchart showing the flow of the inference process performed by the inference unit 140 of Figure 4. The process shown in Figure 6 is called by a main routine (not shown) that controls the inference process in an integrated manner.

6, in S21, the inference unit 140 determines whether the evaluation value of the defrosting operation based on the action value function Q is higher than the evaluation value of the normal operation. If the evaluation value of the defrosting operation is higher than the evaluation value of the normal operation (YES in S21), the inference unit 140 outputs a start command (start timing) of the defrosting operation to the refrigeration cycle device 200 in S22 and returns the process to the main routine. If the evaluation value of the defrosting operation is equal to or lower than the evaluation value of the normal operation (NO in S21), the inference unit 140 returns the process to the main routine.

Figure 7 is a block diagram showing the hardware configuration of the information processing device 100 of Figure 4. As shown in Figure 7, the information processing device 100 includes a processing circuit 51, a memory 52 (storage unit), and an input/output unit 53. The processing circuit 51 includes a CPU (Central Processing Unit) that executes a program stored in the memory 52. The processing circuit 51 may include a GPU (Graphics Processing Unit). The functions of the information processing device 100 are realized by software, firmware, or a combination of software and firmware. The software or firmware is written as a program and stored in the memory 52. The processing circuit 51 reads and executes the program stored in the memory 52. The CPU is also called a central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, processor, or DSP (Digital Signal Processor).

The memory 52 includes a non-volatile or volatile semiconductor memory (e.g., a random access memory (RAM), a read only memory (ROM), a flash memory, an erasable programmable read only memory (EPROM), or an electrically erasable programmable read only memory (EEPROM)), as well as a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, or a digital versatile disk (DVD). The memory 52 stores a defrost inference model 521, a defrost judgment program 522, and a machine learning program 523. The machine learning program 523 is a program for performing machine learning on the defrost inference model 521. The defrost inference model 521 is referenced in the defrost judgment program 522 and the machine learning program 523. The processing circuit 51 that executes the machine learning program 523 functions as the learning unit 120 in FIG. 4. The processing circuit 51 that executes the defrost judgment program 522 functions as the inference unit 140 in FIG. 4.

The input/output unit 53 receives operations from the user and outputs the processing results to the user. The input/output unit 53 includes, for example, a mouse, a keyboard, a touch panel, a display, and a speaker.

In the embodiment, an information processing device having both an inference function and a learning function has been described. Below, a case where an information processing device has either an inference function or a learning function will be described.

Figure 8 is a block diagram showing the hardware configuration of an information processing device 100A relating to a first variant of the embodiment. The configuration of the information processing device 100A is a configuration in which the machine learning program 523 has been removed from the information processing device 100 of Figure 7, and the defrosting inference model 521 has been replaced with 521A. The defrosting inference model 521A is a trained model that has undergone machine learning by a learning device separate from the information processing device 100A. As the rest is similar, the description will not be repeated. The information processing device 100A has an inference function among the inference function and the learning function.

Figure 9 is a block diagram showing the hardware configuration of an information processing device 100B relating to a second variant of the embodiment. The configuration of the information processing device 100B is the same as that of the information processing device 100 in Figure 7 except that the defrosting determination program 522 has been removed. As the rest of the configuration is the same, the description will not be repeated. The information processing device 100B has a learning function out of the inference function and learning function.

As described above, according to the information processing device of the embodiment and variants 1 and 2, it is possible to reduce the power consumption of a refrigeration cycle device performing defrosting operation.

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

1 Refrigeration system, 51 Processing circuit, 52 Memory, 53 Input/output unit, 100, 100A, 100B Information processing device, 110 Status acquisition unit, 120 Learning unit, 121 Reward calculation unit, 122 Function update unit, 130 Memory unit, 140 Inference unit, 200 Refrigeration cycle device, 210 Control device, 211 Control unit, 212 Status observation unit, 220 Cooler, 221 Evaporator, 222 Expansion valve, 230 Heat source unit, 231 Compressor, 232 Condenser, 240 Heater, 300 Specific space, 310 Item management terminal, 320 Door, 330 Item, 521, 521A Defrost inference model, 522 Defrost judgment program, 523 Machine learning program, Sd Open/close sensor, Sh Humidity sensor, Sp Power sensor, St Temperature sensor.

Claims (9)

An information processing device that outputs a defrosting operation start timing to a refrigeration cycle device that blows cooled air to a specific space,
In the refrigeration cycle device, a refrigerant circulates through a compressor, a first heat exchanger, an expansion valve, and a second heat exchanger disposed in the specific space in that order;
The information processing device includes:
a storage unit in which a defrost inference model is stored, the defrost inference model inferring the start timing from specific information including information regarding a moisture amount contained in the air of the specific space in which the second heat exchanger is disposed;
an inference unit configured to determine the start timing from the specific information by using the defrosting inference model trained by machine learning,
The inference unit is configured to update the specific information at each sampling time and determine whether to start the defrosting operation,
The specific space has an opening and closing entrance through which an article can move between the outside of the specific space and the inside of the specific space,
The specific information includes at least one of the temperature and humidity outside the specific space at the current sampling time, information about an item present in the specific space at the current sampling time, and a time during which the entrance/exit is open during a time period from the start timing of a previous defrosting operation to the current sampling time,
The specific information further includes the amount of power consumption of the refrigeration cycle device in a previous time period. An information processing device that outputs a defrosting operation start timing to a refrigeration cycle device that blows cooled air to a specific space,
In the refrigeration cycle device, a refrigerant circulates through a compressor, a first heat exchanger, an expansion valve, and a second heat exchanger disposed in the specific space in that order;
The information processing device includes:
a storage unit in which a defrost inference model is stored, the defrost inference model inferring the start timing from specific information including information regarding a moisture amount contained in the air of the specific space in which the second heat exchanger is disposed;
an inference unit configured to determine the start timing from the specific information by using the defrosting inference model trained by machine learning,
The inference unit is configured to update the specific information at each sampling time and determine whether to start the defrosting operation,
The specific space has an opening and closing entrance through which an article can move between the outside of the specific space and the inside of the specific space,
The specific information includes at least one of the temperature and humidity outside the specific space at the current sampling time, information about an item present in the specific space at the current sampling time, and a time during which the entrance/exit is open during a time period from the start timing of a previous defrosting operation to the current sampling time,
The information processing device, wherein the specific information further includes information regarding a structure of the second heat exchanger. The information processing device according to claim 1 , wherein the specific information further includes an operation time of the refrigeration cycle device in the time period. An information processing device that outputs a defrosting operation start timing to a refrigeration cycle device that blows cooled air to a specific space,
In the refrigeration cycle device, a refrigerant circulates through a compressor, a first heat exchanger, an expansion valve, and a second heat exchanger disposed in the specific space in that order;
The information processing device includes:
a storage unit in which a defrost inference model is stored, the defrost inference model inferring the start timing from specific information including information regarding a moisture amount contained in the air of the specific space in which the second heat exchanger is disposed;
an inference unit configured to determine the start timing from the specific information by using the defrosting inference model trained by machine learning,
An information processing device, wherein the defrosting inference model includes a function that associates the specific information with an evaluation value of the start timing. The information processing device according to claim 1 , further comprising a learning unit configured to make the defrosting inference model a trained model by the machine learning. An information processing device that outputs a defrosting operation start timing to a refrigeration cycle device that blows cooled air to a specific space,
In the refrigeration cycle device, a refrigerant circulates through a compressor, a first heat exchanger, an expansion valve, and a second heat exchanger disposed inside the specific space in which an article is stored, in that order;
The information processing device includes:
A storage unit in which a defrost inference model is stored, the defrost inference model inferring the start timing from specific information including information regarding a moisture amount contained in the air of the specific space;
and a learning unit configured to make the defrosting inference model a trained model by machine learning. An information processing device that outputs a defrosting operation start timing to a refrigeration cycle device that blows cooled air to a specific space,
In the refrigeration cycle device, a refrigerant circulates through a compressor, a first heat exchanger, an expansion valve, and a second heat exchanger disposed in the specific space in that order;
The information processing device includes:
a storage unit in which a defrost inference model is stored, the defrost inference model inferring the start timing from specific information including information regarding a moisture amount contained in the air of the specific space in which the second heat exchanger is disposed;
An inference unit configured to determine the start timing from the specific information by using the defrosting inference model trained by machine learning;
A learning unit configured to make the defrosting inference model a trained model by the machine learning,
The learning unit updates an evaluation value of the start timing in accordance with a change in power consumption per unit time of the refrigeration cycle device controlled in accordance with the start timing. An information processing device that outputs a defrosting operation start timing to a refrigeration cycle device that blows cooled air to a specific space,
In the refrigeration cycle device, a refrigerant circulates through a compressor, a first heat exchanger, an expansion valve, and a second heat exchanger disposed in the specific space in that order;
The information processing device includes:
a storage unit in which a defrost inference model is stored, the defrost inference model inferring the start timing from specific information including information regarding a moisture amount contained in the air of the specific space in which the second heat exchanger is disposed;
A learning unit configured to make the defrosting inference model a trained model by machine learning,
The learning unit updates an evaluation value of the start timing in accordance with a change in power consumption per unit time of the refrigeration cycle device controlled in accordance with the start timing. The refrigeration cycle device;
A refrigeration system comprising the information processing device according to any one of claims 1 to 8 .

Images