Prediction device, refrigeration system, prediction method, and prediction program product
By generating learning to complete the model, and using cluster analysis and labeled data, the frost condition of the refrigeration heat exchanger is predicted, which solves the problem of resource consumption and time consumption in the existing technology and realizes efficient defrosting action control.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- DAIKIN INDUSTRIES LTD
- Filing Date
- 2022-12-27
- Publication Date
- 2026-07-07
AI Technical Summary
In existing technologies, using reinforcement learning to determine the defrosting start time requires a large amount of data and time, resulting in a resource-intensive and inefficient learning process.
The model is completed through generative learning. It uses cluster analysis and labeled data to predict the frost condition of the refrigeration unit's heat exchanger and outputs control information to determine the start timing of the defrosting action. The model includes data inputs of refrigerant circuit status and environmental status, and is learned using a random forest model.
It enables the defrosting action to start accurately at the designated time when frost forms, improving the precision and efficiency of defrosting and reducing the need for learning data and time.
Smart Images
Figure CN119213269B_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to a prediction device, a refrigeration system, a prediction method, and a prediction program product. Background Technology
[0002] As an example of defrosting in a refrigeration cycle device, for example, Patent Document 1 below discloses the use of reinforcement learning to determine the start timing of the defrosting action.
[0003] <Prior art documents>
[0004] <Patent Documents>
[0005] Patent Document 1: International Publication No. 2021 / 176689 Summary of the Invention
[0006] <Problem to be solved by this invention>
[0007] However, in the case of defrosting using reinforcement learning as described above, the learning process to improve the accuracy of the start timing is time-consuming and requires a large amount of data.
[0008] The purpose of this disclosure is to enable defrosting to begin at a time corresponding to the frost formation.
[0009] <Methods for solving problems>
[0010] The first aspect of this disclosure provides a prediction device that outputs information controlling the start timing of defrosting operations of a heat exchanger acting as a heat absorber via a refrigeration unit. The refrigeration unit has a refrigerant circuit that connects a compressor, a radiator, an expansion valve, and the heat absorber in a loop and circulates refrigerant. The control unit of the prediction device, having generated a learning completion model based on learning data, inputs current operating data of the refrigeration unit into the learning completion model. The learning data is obtained by clustering historical records of past operating data of the refrigeration unit into multiple groups and assigning each group data indicating whether or not the heat exchanger is frosted. The device outputs information predicting whether or not the heat exchanger is frosted, as determined by the learning completion model.
[0011] According to the first aspect of this disclosure, defrosting can be initiated at a time corresponding to frost formation.
[0012] Furthermore, a second aspect of this disclosure is a prediction device according to the first aspect, wherein the operating data includes data representing the state of the refrigerant circuit.
[0013] Furthermore, a third aspect of this disclosure is a prediction device according to the second aspect, wherein the operating data includes data representing the state of the environment.
[0014] Furthermore, a fourth aspect of this disclosure is a prediction device according to the third aspect, wherein the data indicating the state of the refrigerant circuit includes any one of the compressor frequency, the pressure data of the low-pressure side of the compressor, the opening degree data of the expansion valve, the heating capacity data, and the evaporation temperature data, and the data indicating the state of the environment includes ambient humidity data.
[0015] Furthermore, the fifth aspect of this disclosure is a prediction device according to any one of the first to fourth aspects, wherein the control unit outputs information indicating the start timing of the defrosting operation when the learning completion model predicts that frost is present, as information indicating whether the heat exchanger is frosted or not.
[0016] Furthermore, the sixth aspect of this disclosure is a prediction device according to any one of the first to fifth aspects, wherein the control unit generates the learning data by clustering historical records of past operating data in the refrigeration unit into multiple groups and assigning data indicating whether or not the heat exchanger is frosted to each group.
[0017] Furthermore, the seventh aspect of this disclosure is a prediction device according to any one of the first to sixth aspects, wherein the control unit learns the learning model by inputting historical records of past operating data classified into groups of the learning data into the learning model, so that the output data output from the learning model is close to the data representing whether the heat exchanger is frosted or not assigned to the classified groups, thereby generating the learning completion model.
[0018] The eighth aspect of this disclosure provides a refrigeration system having the refrigerator, which controls the timing of the start of a defrosting operation using information output from a prediction device according to any one of the first to seventh aspects, indicating whether or not the heat exchanger is frosted.
[0019] Furthermore, the ninth aspect of this disclosure is a refrigeration system according to the eighth aspect, wherein the predictive device is implemented in any of the plurality of units comprising the refrigeration unit.
[0020] Furthermore, the tenth aspect of this disclosure is a refrigeration system according to the eighth aspect, wherein a server device is connected to the refrigeration unit via a network, and the prediction device is implemented in the server device.
[0021] The eleventh aspect of this disclosure provides a prediction method for a prediction device, the prediction device outputting information for controlling the start timing of defrosting operations of a heat exchanger acting as a heat absorber via a refrigeration unit, the refrigeration unit having a refrigerant circuit that connects a compressor, a radiator, an expansion valve, and the heat absorber in a loop and circulates refrigerant, the prediction method comprising the steps of: inputting current operating data of the refrigeration unit into a learning completion model generated based on learning data, the learning data being obtained by clustering historical records of past operating data of the refrigeration unit into multiple groups and assigning each group data indicating whether or not the heat exchanger is frosted; and outputting information indicating whether or not the heat exchanger is frosted, predicted by the learning completion model.
[0022] The 12th aspect of this disclosure provides a prediction program for a prediction device that outputs information controlling the start timing of defrosting operations on a heat exchanger acting as a heat absorber via a refrigeration unit. The refrigeration unit has a refrigerant circuit that connects a compressor, a radiator, an expansion valve, and the heat absorber in a loop and circulates refrigerant. The prediction program causes the control unit of the prediction device to perform the following steps: inputting current operating data of the refrigeration unit into a learning completion model generated based on learning data obtained by clustering historical records of past operating data of the refrigeration unit into multiple groups and assigning data indicating whether or not the heat exchanger is frosted; and outputting information indicating whether or not the heat exchanger is frosted, predicted by the learning completion model. Attached Figure Description
[0023] [ Figure 1A ] Figure 1A This is a diagram illustrating an example of the system configuration of a refrigeration system during the learning phase.
[0024] [ Figure 1B ] Figure 1B This is a diagram illustrating an example of the system configuration of the system related to the collection and processing of operational data in the refrigeration system during the learning phase.
[0025] [ Figure 2 ] Figure 2 This is a diagram illustrating an example of the hardware configuration of a learning device.
[0026] [ Figure 3 ] Figure 3 This is a diagram illustrating an example of the functional configuration of a system related to learning processing within a freezing system during the learning phase.
[0027] [ Figure 4 ] Figure 4 This is a graph showing an example of operational and observational data.
[0028] [ Figure 5 ] Figure 5 This is a diagram illustrating a specific example of the processing in the cluster analysis unit.
[0029] [ Figure 6 ] Figure 6 This is a diagram illustrating a specific example of the processing of the annotation section.
[0030] [ Figure 7 ] Figure 7 This is a diagram illustrating a specific example of the learning department's processing.
[0031] [ Figure 8 ] Figure 8 This is a flowchart illustrating the learning process.
[0032] [ Figure 9 ] Figure 9 This is a diagram illustrating an example of the system configuration of a refrigeration system during the prediction phase.
[0033] [ Figure 10 ] Figure 10 This is a diagram illustrating an example of the hardware configuration of a server device.
[0034] [ Figure 11 ] Figure 11 This is a diagram illustrating an example of the functional configuration of the system related to predictive processing in the refrigeration system during the predictive phase.
[0035] [ Figure 12 ] Figure 12 This is a flowchart illustrating the predictive processing flow.
[0036] [ Figure 13 ] Figure 13 This is a diagram illustrating a specific example of the defrosting process. Detailed Implementation
[0037] Hereinafter, various embodiments will be described with reference to the accompanying drawings. It should be noted that in this specification and the accompanying drawings, constituent elements having substantially the same functional configuration are marked with the same symbols to omit repeated descriptions.
[0038] [First Implementation]
[0039] <System Composition of the Refrigeration System in the Learning Phase>
[0040] First, the system configuration of the refrigeration system according to the first embodiment during the learning phase will be described. Figure 1A This is a diagram illustrating an example of the system configuration of a refrigeration system during the learning phase.
[0041] like Figure 1A As shown, the refrigeration system 100 according to the first embodiment includes a refrigerator (a refrigerator composed of multiple units including indoor unit 110, indoor unit 120, and outdoor unit 130) and a learning device 150 during the learning phase. It should be noted that in this embodiment, the refrigerator is described as having two indoor units, but the number of indoor units can be one or more.
[0042] The refrigeration unit has a refrigerant circuit 140 for refrigerant circulation, and achieves indoor air conditioning by connecting the compressor, radiator, expansion valve and heat absorber in a loop along the refrigerant circuit 140.
[0043] It should be noted that during the learning phase, the refrigeration system 100 collects operational data related to frost formation in the outdoor unit 130 (specifically, the heat exchanger 144) during the operation of the refrigeration unit. Therefore, in Figure 1A In the system configuration shown, the device required for collecting operating data related to frost formation in the heat exchanger 144 of the outdoor unit 130 is described, and the device required for the operation of the refrigeration unit itself is described in a simplified manner.
[0044] The outdoor unit 130 includes an electronic expansion valve 141, a plate heat exchanger 142, a liquid receiver 143, a heat exchanger 144, a flow path switching mechanism 145, a compression mechanism 146, a pressure sensor 147, an ambient humidity sensor 148, and an outdoor controller 149.
[0045] The electronic expansion valve 141 adjusts the amount of refrigerant supplied to the indoor units 110 and 120 between heat exchangers 111 and 121 when acting as heat absorbers and heat exchanger 144 when acting as heat absorbers (i.e., during cooling or refrigeration operation). Alternatively, the electronic expansion valve 141 adjusts the refrigerant flow rate between heat exchangers 111 and 121 when acting as heat absorbers and heat exchanger 144 when acting as heat absorbers (i.e., during heating or warming operation). The plate heat exchanger 142 cools the refrigerant. The receiver 143 stores the refrigerant.
[0046] Heat exchanger 144 is an air heat exchanger that exchanges heat between outdoor air supplied by an outdoor fan (not shown) and refrigerant.
[0047] Temperature sensor 144t measures the evaporation temperature of the refrigerant in heat exchanger 144 when it functions as a heat absorber.
[0048] The flow path switching mechanism 145 switches the flow path of the refrigerant. For example,
[0049] During cooling or refrigeration operation, the flow path switching mechanism 145 draws the refrigerant that has evaporated (absorbed heat) in the heat exchanger 111 of the indoor unit 110 and the heat exchanger 121 of the indoor unit 120 into the compression mechanism 146. Additionally, the flow path switching mechanism 145 transfers the refrigerant compressed in the compression mechanism 146 to the heat exchanger 144. Thus, the refrigerant dissipates heat in the heat exchanger 144, and after the supply is adjusted using the electronic expansion valve 141, it evaporates again in the heat exchangers 111 of the indoor unit 110 and the heat exchanger 121 of the indoor unit 120.
[0050] During heating or warming operation, the flow path switching mechanism 145 transfers the refrigerant compressed in the compression mechanism 146 to the indoor units 110 and 120. As a result, the refrigerant dissipates heat in heat exchangers 111 and 121, and after the flow rate is adjusted using the electronic expansion valve 141, it evaporates (absorbs heat) in heat exchanger 144. Additionally, the flow path switching mechanism 145 draws the refrigerant evaporated (absorbs heat) in heat exchanger 144 into the compression mechanism 146. Thus, the refrigerant is compressed again in the compression mechanism 146.
[0051] • During defrosting operation, the flow path switching mechanism 145 operates in the same manner as during refrigeration or cooling operation. That is, the refrigerant compressed in the compression mechanism 146 dissipates heat in the heat exchanger 144. As a result, the frost on the surface of the heat exchanger 144 is heated from the inside, and the heat exchanger 144 is defrosted.
[0052] Compression mechanism 146 is equipped with multiple compressors (in Figure 1A In the example, the three compressors (146_1 to 146_3) compress the refrigerant. These compressors are variable-capacity compressors whose operating frequency or speed is adjustable.
[0053] Pressure sensor 147 measures the pressure of the refrigerant at the inlet of the compression unit 146 (low-pressure side pressure). Ambient humidity sensor 148 measures the humidity of the external environment.
[0054] The outdoor controller 149 controls the operation of each device in the outdoor unit 130. Additionally, the outdoor controller 149 collects measurement data from each sensor in the outdoor unit 130.
[0055] The indoor unit 110 is a unit for indoor air conditioning and includes a heat exchanger 111, a temperature sensor 112, and an indoor controller 113.
[0056] Heat exchanger 111 is an air heat exchanger that exchanges heat between indoor air supplied by an indoor fan (not shown) and refrigerant.
[0057] Temperature sensor 122 measures the evaporation temperature of the refrigerant in heat exchanger 111 when it acts as a heat absorber.
[0058] The indoor controller 113 controls the operation of each device in the indoor unit 110. In addition, the indoor controller 113 acquires the measurement data of each sensor in the indoor unit 110 and notifies the outdoor controller 149.
[0059] The indoor unit 120 is a unit for cooling indoor air and includes a heat exchanger 121, a temperature sensor 122, and an indoor controller 123.
[0060] Heat exchanger 121 is an air heat exchanger that exchanges heat between indoor air supplied by an indoor fan (not shown) and refrigerant.
[0061] Temperature sensor 112 measures the evaporation temperature of the refrigerant in heat exchanger 121 when it acts as a heat absorber.
[0062] The indoor controller 123 controls the operation of each device in the indoor unit 120. In addition, the indoor controller 123 acquires the measurement data of each sensor in the indoor unit 120 and notifies it to the outdoor controller 149.
[0063] The learning device 150 acquires operational data related to frost formation in the heat exchanger 144 of the outdoor unit 130, collected by the outdoor controller 149 during the operation of the refrigerator. Additionally, at each time point the acquired operational data is measured, the learning device 150 acquires observational data indicating whether the heat exchanger 144 of the outdoor unit 130 has actually frostted (or not). Furthermore, the learning device 150 uses the operational data and observational data to generate a learning completion model.
[0064] <System Structure of Systems Related to Operational Data Collection and Processing>
[0065] Next, the system configuration of the system related to the operation data collection and processing system in the refrigeration system 100 during the learning phase, which is used to collect operation data related to frost in the heat exchanger 144 of the outdoor unit 130, will be described. Figure 1B This is a diagram illustrating an example of the system configuration of the system related to the collection and processing of operational data in the refrigeration system during the learning phase.
[0066] like Figure 1BAs shown, the system related to the collection and processing of operational data includes an indoor controller 113, an outdoor controller 149, and an indoor controller 123.
[0067] The indoor controller 113 has a microcomputer 161 mounted on a control board and a memory device 162 storing programs for making the microcomputer 161 operate.
[0068] Similarly, the indoor controller 123 has a microcomputer 181 mounted on a control board and a memory device 182 storing programs for making the microcomputer 181 operate.
[0069] The outdoor controller 149 includes a microcomputer 171 mounted on a control board, a memory device 172 storing programs for operating the microcomputer 171, and a communication device 173. In the outdoor controller 149, during the operation of the chiller, data is collected...
[0070] • Valve opening degree of electronic expansion valve 141 (expansion valve opening degree data)
[0071] • The total frequency calculated based on the combined rotational speeds of compressors 146_1 to 146_3 (compressor frequency data).
[0072] • The pressure of the refrigerant at the inlet of the compressor unit 146, measured by pressure sensor 147 (low-pressure side pressure data); • The humidity of the external environment, measured by ambient humidity sensor 148 (ambient humidity data).
[0073] • Heating capacity of the refrigeration unit (heating capacity data)
[0074] • The evaporation temperature of the refrigerant in the heat exchanger 144, as measured by temperature sensor 144t (evaporation temperature data).
[0075] Therefore, in the outdoor controller 149 according to this embodiment, as operating data related to frost formation, it is possible to obtain low-pressure side pressure data, which represents the state of the refrigerant circuit 140.
[0076] • Expansion valve opening data
[0077] • Compressor frequency data,
[0078] Heating capacity data,
[0079] Evaporation temperature data
[0080] And data representing the state of the environment
[0081] • Ambient humidity data
[0082] The data is collected. Furthermore, the collected operational data can be transmitted to the learning device 150 via the communication device 173. It should be noted that, in this embodiment, the microcomputer and memory devices of each controller in the refrigeration unit are sometimes collectively referred to as the "control unit".
[0083] <Hardware Components of the Learning Device>
[0084] Next, the hardware configuration of the learning device 150 will be explained. Figure 2 This is a diagram illustrating an example of the hardware configuration of a learning device.
[0085] like Figure 2 As shown, the learning device 150 includes a processor 201, a memory 202, an auxiliary storage device 203, a display device 204, an operation device 205, and a communication device 206. It should be noted that the various hardware components of the learning device 150 are interconnected via a bus 207.
[0086] The processor 201 has various computing devices such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor 201 reads various programs (e.g., learning programs described later) into the memory 202 and executes them.
[0087] The memory 202 includes main storage devices such as ROM (Read Only Memory) and RAM (Random Access Memory). The processor 201 and the memory 202 form what is called a computer. The computer performs various functions by executing various programs read from the memory 202 by the processor 201.
[0088] The auxiliary storage device 203 stores various programs or data used when the processor 201 executes various programs.
[0089] Display device 204 is a display device that displays the internal processing results of learning device 150. Operation device 205 is an input device used when the user inputs various instructions to learning device 150.
[0090] The communication device 206 is connected, for example, to the outdoor controller 149 of the chiller and receives operating data.
[0091] <System Structure of Systems Related to Learning Processing>
[0092] Next, the functional configuration of the systems related to learning processing in the freezing system 100 during the learning phase will be explained. Figure 3This is a diagram illustrating an example of the functional configuration of a system related to learning processing within a freezing system during the learning phase.
[0093] As described above, the outdoor controller 149 functions as an operation data collection unit 301 during the learning phase. The operation data collection unit 301 collects operation data during the operation of the chiller and sends it to the learning device 150.
[0094] In addition, as described above, a learning program is installed in the learning device 150. By executing the program, the learning device 150 functions as a data acquisition unit 311, a cluster analysis unit 312, a labeling unit 313, and a learning unit 314.
[0095] The data acquisition unit 311 acquires operational data sent from the outdoor controller 149. Specifically, the data acquisition unit 311 acquires low-pressure side pressure data, which represents the state of the refrigerant circuit 140 at various times.
[0096] • Expansion valve opening data
[0097] • Compressor frequency data,
[0098] Heating capacity data,
[0099] Evaporation temperature data
[0100] And data representing the state of the environment at various times.
[0101] • Ambient humidity data
[0102] As operational data.
[0103] In addition, the data acquisition unit 311 acquires observation data. As described above, the observation data indicates whether or not frost has formed on the heat exchanger 144 at each time the operating data is measured, and the results obtained by an observer observing whether or not frost has formed are input as observation data.
[0104] It should be noted that the data acquisition unit 311 stores the acquired operational data and observation data in the data storage unit 321.
[0105] The clustering analysis unit 312 reads the operating data from the data storage unit 321 and performs clustering analysis, for example, based on the k-means method, to classify the operating data at each time point into multiple groups. As described above, the operating data includes six types of data: low-pressure side pressure data, expansion valve opening data, compressor frequency data, heating capacity data, evaporation temperature data, and ambient humidity data. Therefore, the clustering analysis unit 312 groups the operating data at each time point, which is distributed in a 6-dimensional feature space, so that operating data that are close in location belong to the same group.
[0106] In addition, the cluster analysis unit 312 will notify the annotation unit 313 of the operational data for each time period that has been classified into groups.
[0107] The labeling unit 313 attaches labels to the operation data for each time period that is classified into groups. Specifically, the labeling unit 313 reads the observation data from the data storage unit 321 and determines whether the operation data for each time period that is classified into groups is operation data for times when frost was observed or operation data for times when frost was not observed.
[0108] In addition, when the operating data for each time period of the group classified as the processing object is the operating data of the time period in which frost was observed, the labeling unit 313 adds a frost label to all operating data for each time period of the group classified as the processing object.
[0109] In addition, if the operating data for each time period of the group classified as the processing object is the operating data of the time period when no frost was observed, the labeling unit 313 adds a no-frost label to all operating data for each time period of the group classified as the processing object.
[0110] In addition, the labeling unit 313 stores the operation data for each time period, with or without frosting labels attached, as learning data in the learning data storage unit 322.
[0111] The learning unit 314 learns from the learning model by using the learning data stored in the learning data storage unit 322, and generates a learning completion model.
[0112] <Specific examples of operational and observational data>
[0113] Next, specific examples of the operational data and observation data acquired by the data acquisition unit 311 will be explained. Figure 4 This is a graph showing an example of operational and observational data.
[0114] like Figure 4 As shown, the operating data 410 includes data items (low-pressure side pressure, expansion valve opening, compressor frequency, heating capacity, evaporation temperature, ambient humidity) as information items. Additionally, the operating data 410 includes time intervals (t1, t2, t3, t4, t5, ..., t...). x (This refers to items that are considered information.)
[0115] The operation data 410 stores operation data for each time period in its various columns. For example, in the column where data item = "Low-pressure side pressure" and time = "t1" are interleaved, the low-pressure side pressure data at time t1 is stored. Similarly, in the column where data item = "Expansion valve opening" and time = "t1" are interleaved, the expansion valve opening data at time t1 is stored. Similarly, in the column where data item = "Compressor frequency" and time = "t1" are interleaved, the compressor frequency data at time t1 is stored. Similarly, in the column where data item = "Heating capacity" and time = "t1" are interleaved, the heating capacity data at time t1 is stored. Similarly, in the column where data item = "Evaporation temperature" and time = "t1" are interleaved, the evaporation temperature data at time t1 is stored. Similarly, in the column where data item = "Ambient humidity" and time = "t1" are interleaved, the ambient humidity data at time t1 is stored.
[0116] It should be noted that, as Figure 4 As shown, the operational data at time = "t1" is plotted as point P1 in a 6-dimensional feature space.
[0117] Similarly, at each time point (t2, t3, t4, t5, ..., t... x The operational data of ) are plotted in a 6-dimensional feature space as points P2, P3, P4, P5, ..., P6. x .
[0118] In addition, such as Figure 4 As shown, observation data 420 includes data items (whether or not frost has formed) as information items. Additionally, observation data 420 includes time intervals (t1, t2, t3, t4, t5, ..., t...). x (This refers to items that are considered information.)
[0119] The columns in observation data 420 store observation data for each time period. For example, in the column where data item = "Frost present or absent" and time = "t1" are interleaved, data indicating whether frost occurred at time t1 is stored. It should be noted that in Figure 4 In the example, the columns shown by the shading line represent data with frost, and the columns not shown by the shading line represent data without frost.
[0120] <Specific examples of cluster analysis processing>
[0121] Next, a specific example of the processing of the cluster analysis unit 312 will be explained. Figure 5 This is a diagram illustrating a specific example of the processing in the cluster analysis unit.
[0122] As described above, the cluster analysis unit 312 reads the operation data for each time period from the data storage unit 321 and classifies the operation data into multiple groups by performing cluster analysis.
[0123] exist Figure 5 In the feature space 510, the operational data (points P1, P2, P3, P4, P5, ..., P6) read from the data storage unit 321 at various times are plotted. x In this case, it should be noted that, for ease of explanation, the feature space 510 is represented as a 3-dimensional feature space (low-pressure side pressure data, compressor frequency data, heating capacity data).
[0124] In the cluster analysis unit 312, the points distributed in the feature quantity space 510 are grouped in such a way that the operational data that are close in distribution belong to the same group.
[0125] exist Figure 5 In the diagram, feature space 520 illustrates the case where group 1 and group 2 are generated by grouping. It should be noted that in cluster analysis unit 312, all points distributed within feature space 510 are classified into any one group.
[0126] <Specific examples of annotation processing>
[0127] Next, a specific example of the processing of annotation section 313 will be explained. Figure 6 This is a diagram illustrating a specific example of the processing of the annotation section.
[0128] As described above, the labeling unit 313 attaches labels to the operational data for each time period, which are categorized into different groups. Figure 6 In the clustering result 610, a list of points representing operational data at each time point that are classified into groups is provided. Figure 6 The example shows the case where the operational data at each time point is classified into α groups from group 1 to group α.
[0129] In addition, Figure 6 In the example, label 620 shows the cases where frosting or non-frosting labels were attached to the operational data for each time period that were classified into groups. Figure 6 The example shows the case where the operation data for each time period classified into group 1, group 3, and group 4 is labeled with "1" for frosting, and the operation data for each time period classified into group 2 and group α is labeled with "0" for no frosting.
[0130] In this way, by labeling the operational data using groups generated through clustering analysis, the same effect can be achieved as when training data is generated even for operational data in locations not plotted in the 6-dimensional feature space. Therefore, in the learning unit 314, a learning-completed model with high prediction accuracy can be generated with less training data and less learning time.
[0131] <Specific examples of how the Learning Department handles this>
[0132] Next, a specific example of the processing in Study Section 314 will be explained. Figure 7 This is a diagram illustrating a specific example of the learning department's processing.
[0133] As described above, the learning unit 314 uses the learning data stored in the learning data storage unit 322 to learn the learning model and generate a completed learning model. Figure 7 In the image, learning data 700 shows an example of learning data read from the learning data storage unit 322. For example... Figure 7 As shown, in the learning data 700, the operational data for each time period was labeled.
[0134] The learning unit 314 inputs the operational data at each time point from the learning data 700 as input data into a random forest, which serves as an example of a learning model. Additionally, the learning unit 314 inputs the labels at each time point from the learning data 700 as ground truth data into the random forest, which also serves as an example of a learning model. Thus, the learning unit 314 can learn from the random forest by making the output data produced by the input data approximate the ground truth data.
[0135] Therefore, the learning unit 314 can generate a learning-complete model that has learned the relationship between operating data and whether or not the heat exchanger 144 is frosted.
[0136] <Learning Process>
[0137] Next, the learning process performed by the learning device 150 will be explained. Figure 8 This is a flowchart illustrating the learning process.
[0138] In step S801, the learning device 150 acquires operational data and observation data.
[0139] In step S802, the learning device 150 performs cluster analysis on the acquired operation data and classifies the operation data at each time into multiple groups.
[0140] In step S803, the learning device 150, based on the acquired observation data, attaches a frost label or a non-frost label to the operation data for each time period classified into groups.
[0141] In step S804, the learning device 150 uses the tagged operational data at various times to learn the learning model and generate a completed learning model.
[0142] <System Composition of the Refrigeration System in the Prediction Phase>
[0143] Next, the system configuration in the prediction stage of the refrigeration system according to the first embodiment will be described. Figure 9 This is a diagram illustrating an example of the system configuration of a refrigeration system during the prediction phase.
[0144] and utilization Figure 1A The difference between the refrigeration system 100 in the learning phase and the refrigeration system 900 in the prediction phase is that the refrigeration unit is connected to the server device 910 via network 920.
[0145] Server device 910 is an example of a predictive device that receives operating data from the outdoor unit 130 of a running refrigerator via network 920. Furthermore, based on the received operating data, server device 910 predicts whether frost will form on the heat exchanger 144 of the outdoor unit 130, and if frost is predicted, sends information indicating frost formation to the outdoor unit 130.
[0146] In this way, by configuring the system to predict whether there is frost in the heat exchanger 144 based on operating data, defrosting can be initiated in the refrigeration unit based on whether frost has formed.
[0147] It should be noted that the defrosting action refers to the operation of the various devices in the refrigeration unit used to perform the aforementioned defrosting operation.
[0148] <Hardware Components of Server Devices>
[0149] Next, the hardware configuration of the server device 910 will be explained. Figure 10 This is a diagram illustrating an example of the hardware configuration of a server device.
[0150] like Figure 10 As shown, the server device 910 includes a processor 1001, a memory 1002, an auxiliary storage device 1003, a display device 1004, an operation device 1005, and a communication device 1006. It should be noted that the various hardware components of the server device 910 are interconnected via a bus 1007.
[0151] The processor 1001 has various computing devices such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor 1001 reads various programs (e.g., prediction programs described later) into the memory 1002 and executes them.
[0152] The memory 1002 includes main storage devices such as ROM (Read Only Memory) and RAM (Random Access Memory). The processor 1001 and the memory 1002 form a so-called computer (sometimes referred to as a "control unit" in this embodiment). By executing various programs read from the memory 1002 by the processor 1001, the computer performs various functions.
[0153] The auxiliary storage device 1003 stores various programs and various data used when the processor 1001 executes the various programs.
[0154] Display device 1004 is a display device that displays the results of internal processing of server device 910. Operation device 1005 is an input device used by the user to input various instructions to server device 910.
[0155] The communication device 1006 communicates with the outdoor controller 149 of the chiller, for example, via network 920.
[0156] <Functional Composition of Systems Related to Predictive Processing>
[0157] Next, the functional configuration of the systems related to the prediction process in the refrigeration system 900 during the prediction phase will be explained. Figure 11 This is a diagram illustrating an example of the functional configuration of the system related to predictive processing in the refrigeration system during the predictive phase.
[0158] like Figure 11 As shown in (a), the outdoor controller 149 functions as an operation data collection unit 1101 and a defrost control unit 1102 during the prediction phase and the operation of the chiller.
[0159] In addition, as described above, a prediction program is installed in the server device 910. By executing the program, the server device 910 functions as a data acquisition unit 1111 and a learning and completion model 1112 during the operation of the refrigerator.
[0160] The operation data collection unit 1101 collects operation data and sends it to the server device 910 via the network 920.
[0161] The data acquisition unit 1111 acquires operating data from the outdoor controller 149 of the refrigeration system 900. Specifically, the data acquisition unit 1111 acquires low-pressure side pressure data, which represents the state of the refrigerant circuit 140 at various times.
[0162] • Expansion valve opening data
[0163] • Compressor frequency data,
[0164] Heating capacity data,
[0165] Evaporation temperature data
[0166] And data representing the state of the environment at various times.
[0167] • Ambient humidity data
[0168] As operational data.
[0169] By inputting operational data acquired at various times by the data acquisition unit 1111, the learning and completion model 1112 predicts whether or not frost will form on the heat exchanger 144 at each time. Furthermore, if frost is predicted, the learning and completion model 1112 sends information indicating frost formation to the outdoor controller 149.
[0170] When the server device 910 receives information indicating that frost has formed, the defrosting control unit 1102 controls the heat exchanger 144, which functions as a heat absorber, to begin a defrosting operation. That is, in this embodiment, the information indicating that frost has formed sent from the server device 910 is information indicating the start timing of the defrosting operation of the heat exchanger 144, which functions as a heat absorber.
[0171] It should be noted that, Figure 11 (a) shows a configuration in which a refrigeration system 900 has a server device 910 and a function that predicts whether or not there is frost in the heat exchanger 144 in the control unit of the server device 910 by installing a prediction program in the server device 910.
[0172] However, in the refrigeration system 900, the installation destination of the prediction program is not limited to the server device 910. For example, an information processing device can be set up near the refrigeration unit, and the prediction program can be installed in that information processing device. In this case, the information processing device functions as a prediction device, possessing the same characteristics as... Figure 11 (a) has the same functional configuration as the server device 910.
[0173] Alternatively, the installation destination of the prediction program can be any one of the outdoor controller 149, indoor controller 113, or indoor controller 123 of the chiller. In this case, the control unit of the controller that is the installation destination functions as the prediction device.
[0174] Figure 11 (b) Shows the functional configuration of the outdoor controller 149 when the prediction program is installed on the outdoor controller 149. For example... Figure 11 As shown in (b), by executing a prediction program in the outdoor controller 149, the control unit of the outdoor controller 149 functions as the operation data collection unit 1101, the learning and completion model 1112, and the defrosting control unit 1102. It should be noted that details regarding the functions of each unit have already been explained, so they are omitted here.
[0175] <Predictive Processing Flow>
[0176] Next, the process of prediction processing performed by server device 910 will be explained. Figure 12 This is a flowchart illustrating the predictive processing flow.
[0177] In step S1201, the server device 910 acquires operational data. In step S1202, the server device 910 uses a learned model to predict whether the heat exchanger 144 will frost based on the acquired operational data.
[0178] In step S1203, the server device 910 determines whether the model has predicted frost formation through learning. If frost formation is predicted (if "yes" in step S1203), the process proceeds to step S1204. In step S1204, the server device 910 sends information indicating frost formation to the refrigerator.
[0179] On the other hand, if it is determined that there is no frost (if "No" is true in step S1203), proceed to step S1205. In step S1205, the server device 910 determines whether to end the prediction process. In step S1205, if it is determined that the prediction process should continue (if "No" is true in step S1205), return to step S1201. On the other hand, if it is determined that the prediction process should end in step S1205 (if "Yes" is true in step S1205), end the prediction process.
[0180] <Specific examples of defrosting procedures>
[0181] Next, a specific example of the defrosting process performed by the refrigeration system 900 will be described. Figure 13 This is a diagram illustrating a specific example of the defrosting process.
[0182] exist Figure 13 In the graph 1300, the horizontal axis represents time, the left vertical axis represents the start timing of the defrosting operation of the heat exchanger 144, which functions as a heat absorber, and the right vertical axis represents operating data. Additionally, the lines in the graph represent low-pressure side pressure data, compressor frequency data, and heating capacity data. Furthermore, the thick lines in the graph indicate the start timing of the defrosting operation of the heat exchanger 144, which functions as a heat absorber.
[0183] like Figure 13 As shown, in the case of the refrigeration system 900 according to this embodiment, a learning-completed model that has learned the relationship between operating data and whether or not frost has formed on the heat exchanger 144 is present predicts whether or not frost will form on the heat exchanger 144 based on the operating data. Furthermore, in the case of the refrigeration system 900 according to this embodiment, a defrosting operation is initiated when frost is predicted. Therefore, unlike the case where defrosting is initiated based on operating data regardless of whether frost has actually formed, the refrigeration system 900 according to this embodiment can initiate a defrosting operation only when frost has actually formed. Thus, the number of defrosting operations can be reduced by the refrigeration system 900 according to this embodiment.
[0184] Summary
[0185] As can be seen from the above description, in the refrigeration system according to the first embodiment,
[0186] • It has a refrigeration unit with a refrigerant circuit that connects the compressor, radiator, expansion valve and the absorber in a ring and provides refrigerant circulation.
[0187] • During the learning phase, a learning completion model is generated based on the learning data. This learning data is obtained by clustering historical data of past operation data in the refrigeration unit into multiple groups and assigning each group data indicating whether the heat exchanger is frosted or not.
[0188] • During the prediction phase, the current operating data of the chiller is input into the learned model, and the model outputs information indicating whether the heat exchanger is frosted or not.
[0189] Thus, in the first embodiment, a learned model is configured to predict and output whether or not the heat exchanger will frost based on the operating data, by learning the relationship between operating data and the presence or absence of frost on the heat exchanger. Therefore, according to the first embodiment, defrosting can be initiated for the heat exchanger, which functions as a heat absorber, based on information indicating the presence of frost. Consequently, according to the first embodiment, the refrigerator can initiate defrosting at a time corresponding to the frost formation.
[0190] [Second Implementation]
[0191] In the first embodiment described above, the case of generating a learning completion model 1112 using the learning data 700 was explained. However, the learning completion model generated using the learning data 700 is not limited to one. For example, the learning data 700 can be divided into segments according to each season, and learning completion models can be generated separately using the learning data corresponding to each season.
[0192] Alternatively, the learning data 700 can be divided into segments according to each region (climate) where the refrigeration unit is located, and the learning data corresponding to each region (climate) can be used to generate learning models separately.
[0193] Furthermore, in the first embodiment described above, the server device 910 is configured to send information indicating the presence of frost to the refrigerator when frost is detected. In the refrigerator, upon receiving the information indicating the presence of frost, a defrosting operation is initiated on the heat exchanger, which functions as a heat absorber. However, it is not necessary to use the information indicating the presence of frost as information indicating the start timing of defrosting; the refrigerator may, for example, be configured to initiate the defrosting operation based on the fulfillment of other conditions. That is, the information indicating the presence of frost output by the server device 910 can be used not only as information indicating the start timing of the defrosting operation, but also broadly as "information for controlling the start timing when the heat exchanger, which functions as a heat absorber, begins defrosting."
[0194] Furthermore, in the first embodiment described above, six types of data are exemplified as operational data, but operational data is not limited to these six types; it is sufficient to include at least one of the six types of data. Additionally, operational data may also include data other than the six types. For example, in the operational data representing the state of the environment, in addition to ambient humidity data, ambient temperature data may also be included.
[0195] Furthermore, in the first embodiment described above, the prediction of whether or not frost will form on the heat exchanger 144 (more precisely, the heat transfer fins) of the outdoor unit 130 was explained. However, the object of predicting whether or not frost will form is not limited to the heat exchanger 144 of the outdoor unit 130; for example, it can also predict whether or not frost will form on the indoor unit 120. In this case, the refrigeration system 900 can be configured to use the same operating data as in the case of predicting whether or not frost will form on the outdoor unit 130, or it can be configured to use different operating data.
[0196] Furthermore, in the first embodiment described above, the case where the learning device and the prediction device are configured as independent devices was explained; however, the learning device and the prediction device can also be configured as a single integrated device. For example, the server device 910, which is an example of a prediction device, can be configured to perform both learning processing and prediction processing. In this case, the control unit of the server device 910 performs the learning processing (see [reference]). Figure 3 ) and predictive processing (refer to Figure 11 The functions of both. Alternatively, for example, in a refrigerator, which is an example of a predictive device, both learning processing and predictive processing can be performed. In this case, the refrigerator's control unit performs the learning processing (see...). Figure 3 ) and predictive processing (refer to Figure 11 The functions of both.
[0197] Furthermore, while the first embodiment described above shows a learning phase and a prediction phase, the refrigeration system can also be configured to further perform a relearning phase. For example, the refrigeration system can be configured to relearn the learned model by using the operational data collected in the prediction phase as learning data. It should be noted that, in the case of relearning, the learning device and the prediction device can be installed as an integrated device or as separate devices within the refrigeration system.
[0198] Furthermore, in the first embodiment described above, a case of using random forest as an example of a learning model was shown. However, the learning model is not limited to random forest, and any other arbitrary learning model for supervised learning can also be used.
[0199] Although the embodiments have been described above, it should be understood that various changes in form or detail may be made without departing from the spirit and scope of the claims.
[0200] This application claims priority to Japanese Patent Application No. 2022-082891, filed on May 20, 2022, the entire contents of which are incorporated herein by reference.
[0201] Symbol Explanation
[0202] 100: Refrigeration system
[0203] 110: Indoor Unit
[0204] 111: Heat exchanger
[0205] 112: Temperature sensor
[0206] 113: Indoor controller
[0207] 120: Indoor Unit
[0208] 121: Heat exchanger
[0209] 122: Temperature sensor
[0210] 123: Indoor controller
[0211] 130: Outdoor Unit
[0212] 140: Refrigerant circuit
[0213] 141: Electronic expansion valve
[0214] 144: Heat exchanger
[0215] 144t: Temperature sensor
[0216] 146: Compression mechanism
[0217] 147: Pressure sensor
[0218] 148: Ambient humidity sensor
[0219] 149: Outdoor controller
[0220] 311: Data Acquisition Department
[0221] 312: Cluster Analysis Department
[0222] 313: Annotation Section
[0223] 314: Study Department
[0224] 410: Operational Data
[0225] 420: Observational Data
[0226] 510, 520: Feature space
[0227] 610: Clustering Results
[0228] 620: Annotation Results
[0229] 700: Learning with Data
[0230] 900: Refrigeration System
[0231] 910: Server device
[0232] 1101: Operational Data Collection Department
[0233] 1102: Defrosting Control Department
[0234] 1111: Data Acquisition Department
[0235] 1112: Learning completes the model.
Claims
1. A predictive device that outputs information to control the timing of the start of defrosting operation of a heat exchanger acting as a heat absorber via a refrigeration unit, the refrigeration unit having a refrigerant circuit that connects a compressor, a radiator, an expansion valve, and the heat absorber in a loop and circulates refrigerant. The control unit of the prediction device Having generated a completed learning model based on the learning data, the current operating data of the chiller is input into the completed learning model. The learning data is obtained by clustering the historical operating data of the chiller, classifying the operating data containing multiple types of data at different times into multiple groups, such that operating data that are close in distribution in the multidimensional feature space corresponding to the types of data belong to the same group. Then, using the classified group as the unit, all operating data at each time within each group are assigned the same label indicating whether the heat exchanger is frosted or not. The output is information about whether or not the heat exchanger is frosted, as predicted by the learned model.
2. The prediction device according to claim 1, wherein, The operational data includes data indicating the state of the refrigerant circuit.
3. The prediction device according to claim 2, wherein, The operational data includes data representing the state of the environment.
4. The prediction device according to claim 3, wherein, The data indicating the state of the refrigerant circuit includes any one of the following: the compressor frequency, the compressor low-pressure side pressure data, the expansion valve opening data, the heating capacity data, and the evaporation temperature data. The data representing the state of the environment includes ambient humidity data.
5. The prediction device according to claim 1, wherein, When the learning-completed model predicts that there is frost, the control unit outputs information indicating the start timing of the defrosting operation, as information indicating whether there is frost on the heat exchanger.
6. The prediction device according to claim 1, wherein, The control unit generates the learning data by clustering historical records of past operation data in the refrigeration unit into multiple groups, and assigning the same data indicating whether or not the heat exchanger is frosted to each group.
7. The prediction device according to claim 1, wherein, The control unit learns from the learning model by inputting historical records of past operating data, which are categorized into groups, into the learning model, so that the output data from the learning model approximates the data representing whether or not the heat exchanger is frosted, assigned to each group in the categorized groups, thereby generating the completed learning model.
8. A refrigeration system having the refrigerator, the refrigerator controlling the timing of the start of defrosting operation using information output by the predictive device according to any one of claims 1 to 7 indicating whether or not the heat exchanger is frosted.
9. The refrigeration system according to claim 8, wherein, The prediction device is implemented in any of the multiple units of the refrigeration unit.
10. The refrigeration system according to claim 8, wherein, It has a server device that is connected to the freezer via a network. The prediction device is implemented in the server device.
11. A prediction method for a prediction device, the prediction device outputting information controlling the start timing of defrosting operations of a heat exchanger acting as a heat absorber via a refrigeration unit, the refrigeration unit having a refrigerant circuit that connects a compressor, a radiator, an expansion valve, and the heat absorber in a loop and circulates refrigerant, the prediction method comprising the following steps: In the case where a learning completion model has been generated based on the learning data, the process of inputting the current operating data of the chiller into the learning completion model is as follows: the learning data is obtained by clustering the historical records of the chiller's past operating data, classifying the operating data containing multiple types of data at each time into multiple groups such that operating data with similar distribution positions in the multidimensional feature space corresponding to the types of data belong to the same group, and assigning the same label indicating whether the heat exchanger is frosted or not to all operating data at each time within each group as a unit of classification; and The process of outputting information about whether or not the heat exchanger is frosted, as predicted by the learned model.
12. A prediction program product for a prediction device, the prediction device outputting information controlling the start timing of defrosting operations on a heat exchanger acting as a heat absorber via a refrigeration unit, the refrigeration unit having a refrigerant circuit that connects a compressor, a radiator, an expansion valve, and the heat absorber in a loop and circulates refrigerant, the prediction program product being used to cause the control unit of the prediction device to perform the following steps: In the case where a learning completion model has been generated based on the learning data, the process of inputting the current operating data of the chiller into the learning completion model is as follows: the learning data is obtained by clustering the historical records of the chiller's past operating data, classifying the operating data containing multiple types of data at each time into multiple groups such that operating data with similar distribution positions in the multidimensional feature space corresponding to the types of data belong to the same group, and assigning the same label indicating whether the heat exchanger is frosted or not to all operating data at each time within each group as a unit of classification; and The process of outputting information about whether or not the heat exchanger is frosted, as predicted by the learned model.