Determination device, determination method, and determination program

Abstract

This determination device is provided with a model information storage unit (212), an information acquisition unit (231), a determination unit (233), and a correction unit (232). The model information storage unit stores a neural network model constructed with respect to the operation of a second device that is a different model than a first device to be controlled and that operates on the same operating principle as the first device. The information acquisition unit acquires input data of a first item relating to a device operating on the operating principle, a first coefficient for calculating, in the first device, output data of a second item from the input data of the first item, and a second coefficient for calculating, in the second device, output data of the second item from the input data of the first item. The determination unit determines output data of the second item for the first device from the input data of the first item using the stored neural network model. The correction unit multiplies the input data of the first item or the output data of the second item used by the determination unit by the ratio of the first coefficient to the second coefficient to correct the respective data.

Description

[Technical Field]

[0001] The present invention relates to a determination device, a determination method, and a determination program. [Background technology]

[0002] In recent years, advances in IoT (Internet of Things) technology have made it possible to collect vast amounts of data in a variety of fields. Accordingly, in the industrial field, technology has been developed to build neural network models (NN models) that use data collected using IoT technology to determine information about the operation of equipment.

[0003] By using the NN model, each device can take into account a large amount of information related to operation and accurately and efficiently determine information to be used for control. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 6791429 Summary of the Invention

[0005] As mentioned above, to build an NN model, a large amount of data acquired when the relevant equipment is operated is required as training data.As a technology to deal with cases where it is difficult to acquire a large amount of data when the equipment is operated, a method has been proposed in which an existing NN model is reused to limit the amount of training data required to build a new NN model.

[0006] However, there is a problem in that such a method cannot be used for new product devices that have no operational track record before being released to the market, since there is no data that can be used as learning data.

[0007] The present invention has been made in consideration of the above circumstances, and aims to provide a determination device, a determination method, and a determination program that are capable of determining information regarding the operation of equipment that has no operational history using a neural network model.

[0008] The determination device of the present invention has been made in consideration of the above circumstances, and comprises a model information storage unit that stores a neural network model constructed for the operation of a second device that operates on the same operating principle as a first device to be controlled but is a different model; an information acquisition unit that acquires input data for a first item related to the device that operates on the same operating principle, a first coefficient for calculating output data for a second item from the input data for the first item in the first device, and a second coefficient for calculating output data for the second item from the input data for the first item in the second device; a determination unit that determines the output data for the second item related to the first device from the input data for the first item using the neural network model stored in the model information storage unit; and a correction unit that corrects the corresponding data by multiplying the input data for the first item or the output data for the second item used in the determination unit by the ratio of the first coefficient to the second coefficient.

[0009] Furthermore, the determination method of the present invention stores a neural network model constructed for the operation of a second device that operates according to the same operating principle as a first device to be controlled but is a different model, obtains input data for a first item related to the device that operates according to the operating principle, a first coefficient for calculating output data for a second item from the input data for the first item in the first device, and a second coefficient for calculating output data for the second item from the input data for the first item in the second device, corrects the input data for the first item by multiplying it by the ratio of the first coefficient to the second coefficient, and determines the output data for the second item related to the first device from the corrected input data for the first item using the neural network model, or determines the output data for the second item from the input data for the first item using the neural network model, and then multiplies the determined output data for the second item by the ratio of the first coefficient to the second coefficient to correct it so that it corresponds to the first device.

[0010] Furthermore, the determination program of the present invention causes a computer to perform the following functions: storing a neural network model constructed for the operation of a second device that operates according to the same operating principle as a first device to be controlled but is a different model; acquiring input data for a first item related to the device that operates according to the operating principle, a first coefficient for calculating output data for a second item from the input data for the first item in the first device, and a second coefficient for calculating output data for the second item from the input data for the first item in the second device; and correcting the input data for the first item by multiplying the ratio of the first coefficient to the second coefficient, and determining the output data for the second item related to the first device from the corrected input data for the first item using the neural network model, or determining the output data for the second item from the input data for the first item using the neural network model, and then multiplying the determined output data for the second item by the ratio of the first coefficient to the second coefficient to correct it so that it corresponds to the first device. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 is a block diagram showing the configuration of an air conditioner using a first control device as a determination device according to a first embodiment of the present invention. [Figure 2] FIG. 2 is a block diagram showing the configuration of an air conditioner corresponding to the neural network model used in the determination device according to the first and second embodiments of the present invention. [Figure 3] FIG. 3 is a flowchart showing the flow of processing executed by the first control device as the determination device according to the first embodiment of the present invention. [Figure 4] FIG. 4 is a block diagram showing the configuration of an air conditioner using a third control device as a determination device according to a second embodiment of the present invention. [Figure 5] FIG. 5 is a flowchart showing the flow of processing executed by a third control device as a determination device according to the second embodiment of the present invention. [Figure 6] Figure 6(a) is a graph showing, using a set of points, the positions indicating the state in which the judgment electronic expansion valve opening prediction value determined without data correction using a neural network model constructed for the operation of an air conditioner model with a proven track record under various operating conditions corresponds to the electronic expansion valve opening prediction value actually measured on the air conditioner model being verified. Figure 6(b) is a graph showing, using a set of points, the positions indicating the state in which the judgment electronic expansion valve opening prediction value determined after data correction using a neural network model constructed for the operation of an air conditioner model with a proven track record under various operating conditions corresponds to the electronic expansion valve opening prediction value actually measured on the air conditioner model being verified. DETAILED DESCRIPTION OF THE INVENTION

[0012] First Embodiment An air conditioner using a determination device according to a first embodiment of the present invention will be described. In the first embodiment, a first model air conditioner (first device) that has been newly developed and has no operational history is used to operate the devices inside the air conditioner using a neural network model (NN model) that has already been constructed for the operation of a second model air conditioner (second device) that has operational history.

[0013] <Configuration of an air conditioner using the determination device according to the first embodiment> The configuration of an air conditioner using a determination device according to a first embodiment of the present invention will be described with reference to Fig. 1. The air conditioner 1 according to this embodiment is a newly developed first model air conditioner that has no operational track record.

[0014] The air conditioner 1 includes a first compressor 11, a first evaporator 12, a first electronic expansion valve 13, a first condenser 14, and a first control device 20 as a determination device, which constitute a first refrigeration cycle 10.

[0015] The first compressor 11 draws in refrigerant, compresses the drawn refrigerant, and discharges it in a high-temperature, high-pressure state. The high-temperature, high-pressure refrigerant discharged from the first compressor 11 is condensed by releasing heat through heat exchange in the first condenser 14. The condensed refrigerant is reduced in pressure by the first electronic expansion valve 13 to a low-temperature, low-pressure state. The low-temperature, low-pressure refrigerant absorbs heat through heat exchange in the first evaporator 12 and is evaporated. The evaporated, gasified refrigerant is returned to the first compressor 11 via an accumulator (not shown).

[0016] The first control device 20 has a first storage unit 21, a first input unit 22, and a first CPU 23, and controls the devices 11, 12, 13, and 14 in the first refrigeration cycle 10. The first storage unit 21 has a first setting information storage unit 211 and a first NN model information storage unit 212.

[0017] The first setting information storage unit 211 stores setting information relating to the devices 11, 12, 13, and 14 in the first refrigeration cycle 10.

[0018] The first NN model information storage unit 212 stores an NN model M that has already been constructed by learning operational data regarding the refrigeration cycle used in a second model air conditioner that has a proven operational track record.

[0019] NN model M is information constructed by performing machine learning with a neural network based on various data collected from each device in the second model air conditioner using IoT technology when the second model air conditioner was in operation. The neural network used to construct NN model M includes three layers: an input layer, an intermediate layer, and an output layer. The configuration of the second model air conditioner will be described later.

[0020] The first input unit 22 inputs commands related to the operation of the devices 11, 12, 13, and 14 in the first refrigeration cycle 10, and information used in processing corresponding to the commands. Specifically, the first input unit 22 inputs input data for a first item related to the operation of the devices in the first refrigeration cycle 10, and a command to output output data for a second item related to the operation of the devices in the first refrigeration cycle 10 from the input data for the first item. The first input unit 22 also inputs a second coefficient for calculating output data for the second item from the input data for the first item, related to the refrigeration cycle used in a second model air conditioner, as information used in processing corresponding to the commands.

[0021] The first CPU 23 includes a first information acquisition unit 231, a first correction unit 232, a first determination unit 233, and a first operation control unit 234.

[0022] The first information acquisition unit 231 acquires a command input from the first input unit 22 and information used in processing corresponding to the command. The first information acquisition unit 231 also acquires information used in processing corresponding to the acquired command from the information stored in the first setting information storage unit 211.

[0023] The first correction unit 232 corrects the input data of the corresponding first item regarding the operation of the first refrigeration cycle 10 acquired by the first information acquisition unit 231 by multiplying the input data of the corresponding first item by the ratio of the first coefficient to the second coefficient.

[0024] The first determination unit 233 determines output data of the second item related to the operation of the first refrigeration cycle 10 using the NN model M from the input data of the first item corrected by the first correction unit 232.

[0025] The first operation control unit 234 controls the operation of each device 11, 12, 13, or 14 in the first refrigeration cycle 10 based on the output data of the second item regarding the first refrigeration cycle 10 determined by the first judgment unit 233 and the information stored in the first setting information memory unit 211.

[0026] Next, the configuration of the second model air conditioner will be described with reference to Figure 2. The second model air conditioner 2 operates on the same operating principle as the first model air conditioner 1, but uses different parts, etc.

[0027] The air conditioner 2 includes a second compressor 31, a second evaporator 32, a second electronic expansion valve 33, a second condenser 34, and a second control device 40, which constitute a second refrigeration cycle 30. The functions of the devices 31, 32, 33, and 34 that constitute the second refrigeration cycle 30 are similar to the functions of the devices 11, 12, 13, and 14 that constitute the first refrigeration cycle 10, and therefore detailed description thereof will be omitted.

[0028] The second control device 40 has a second storage unit 41, a second input unit 42, and a second CPU 43, and controls the devices 31, 32, 33, and 34 in the second refrigeration cycle 30. The second storage unit 41 has a second setting information storage unit 411 and a second NN model information storage unit 412.

[0029] The second setting information storage unit 411 stores setting information about the devices 31, 32, 33, and 34 in the second refrigeration cycle 30. The second NN model information storage unit 412 stores the NN model M described above.

[0030] The second input unit 42 inputs commands related to the operation of the devices 31, 32, 33, and 34 in the second refrigeration cycle 30, and information used in processing corresponding to the commands. Specifically, the second input unit 42 inputs a first item of input data related to the operation of the devices in the second refrigeration cycle 30, and a command to output a second item of output data related to the operation of the devices in the second refrigeration cycle 30 from the first item of input data.

[0031] The second CPU 43 includes a second information acquisition unit 431, a second determination unit 432, and a second operation control unit 433.

[0032] The second information acquisition unit 431 acquires a command input from the second input unit 42 and information used in the process corresponding to the command. The second information acquisition unit 431 also acquires information used in the process corresponding to the acquired command from the information stored in the second setting information storage unit 411.

[0033] The second determination unit 432 determines output data of a second item related to the operation of the second refrigeration cycle 30 using the NN model M from the input data of the first item acquired by the second information acquisition unit 431.

[0034] The second operation control unit 433 controls the operation of each device 31, 32, 33, or 34 in the second refrigeration cycle 30 based on the output data of the second item regarding the second refrigeration cycle 30 determined by the second judgment unit 432 and the information stored in the second setting information memory unit 411.

[0035] <Operation of air conditioner using determination device according to first embodiment> The operation of the air conditioner 1 according to this embodiment will be described below. Here, the operation of the air conditioner 1 will be described using as an example a regression model in which the input data of the first item, namely, the operating frequency (compressor frequency Hz) of the motor used in the first compressor 11 in the first refrigeration cycle 10 and the specific volume of the intake gas of the first compressor 11 (compressor intake gas specific volume v), are used as explanatory variables, and the output data of the second item, namely, the refrigerant circulation amount G in the first refrigeration cycle 10, is used as a response variable.

[0036] The first model air conditioner 1, which has no operational track record, and the second model air conditioner 2, which has an operational track record, operate on the same operating principle, use the same type of refrigerant, and have different displacement volumes set for compressors 11, 31 in refrigeration cycles 10, 30. Because air conditioners 1 and 2 operate on the same principle, it is presumed that the logic for determining the refrigerant circulation amount G from the compressor frequency Hz and the compressor suction gas specific volume v is similar.

[0037] Specifically, the refrigerant circulation rate G1 [kg / s] of the first refrigeration cycle 10 is determined by the compressor frequency Hz [1 / s] and the compressor intake gas specific volume v [m 3 / kg], and the excluded volume V1 [m 3 ] as a coefficient. The refrigerant circulation amount G2 [kg / s] of the second refrigeration cycle 30 is calculated by the following equation (2) based on the compressor frequency Hz [1 / s] and the compressor intake gas specific volume v [m 3 / kg], and the excluded volume V2 [m 3 ] is considered to be expressed as a formula with coefficients.

[0038]

number

[0039] In this embodiment, information about the rejected volume V1 set for the first compressor 11 is stored in a first setting information storage unit 211. Information about the rejected volume V2 set for the second compressor 31 is stored in a second setting information storage unit 411.

[0040] Here, since there is no operational experience for the first refrigeration cycle 10, it is not possible to construct an NN model for it. Therefore, we consider using the NN model M constructed for the second refrigeration cycle 30, which operates on the same operating principle as the first refrigeration cycle 10, for operating the equipment in the first refrigeration cycle 10.

[0041] In this case, because the NN model M is constructed based on the operating characteristics of the air conditioner 2, it cannot be used as is to determine the operation of the first refrigeration cycle 10 of the air conditioner 1. Specifically, because the first compressor 11 of the first refrigeration cycle 10 and the second compressor 31 of the second refrigeration cycle 30 have different displacement volumes, even if the NN model M is used as is, it is not possible to appropriately determine the refrigerant circulation amount G1 of the first refrigeration cycle 10 from the compressor frequency Hz and the compressor suction gas specific volume v.

[0042] Therefore, the first control device 20 performs a process of appropriately determining the refrigerant circulation amount G1 of the first refrigeration cycle 10 using the NN model M by correcting one of the data used to calculate the refrigerant circulation amount G1.

[0043] The following describes a process in which the first control device 20 determines the refrigerant circulation amount G1 for the first refrigeration cycle 10 using the NN model M by correcting the compressor frequency Hz, which is one of the first items of input data for the operation of the first refrigeration cycle 10. Fig. 3 is a flowchart showing the flow of the process executed by the first control device 20.

[0044] First, the first input unit 22 inputs a command to determine the refrigerant circulation amount G1. At that time, the first input unit 22 inputs information on the compressor frequency Hz and the compressor suction gas specific volume v, which are input data of the first item, as explanatory variables to be given to the NN model M to determine the refrigerant circulation amount G1. The first input unit 22 also inputs correction target information indicating that the data to be corrected when using the NN model M is the compressor frequency Hz, and information on the displacement volume V2 of the second compressor 31 to be used in the correction process of the compressor frequency Hz.

[0045] The information input from the first input unit 22 is acquired by the first information acquisition unit 231. When the first information acquisition unit 231 acquires a command to determine the refrigerant circulation amount G1 of the first refrigeration cycle 10 ("YES" in S1), it acquires, as information to be used in the determination process, information on the compressor frequency Hz and the compressor suction gas specific volume v, which are input data of the first item acquired together with the command, information on the excluded volume V2 of the second compressor 31, and information on the excluded volume V1 of the first compressor 11 stored in the first setting information storage unit 211 (S2).

[0046] As described above, it is predicted that the displacement volume is a coefficient when determining the refrigerant circulation volume from the compressor frequency and the compressor suction gas specific volume. In other words, in the NN model M, it is considered that the refrigerant circulation volume is determined by multiplying the input data, which are the compressor frequency and the compressor suction gas specific volume, by a value equivalent to the displacement volume. Although the displacement volumes differ between the first compressor 11 and the second compressor 31, in the NN model M, the value equivalent to the displacement volume cannot be corrected because it is processed in the black box in the intermediate layer.

[0047] Therefore, based on the correction target information acquired by the first information acquisition unit 231, the first correction unit 232 corrects the compressor frequency Hz to the compressor frequency Hz1 by multiplying the compressor frequency Hz by the ratio of the displacement volume V1 of the first compressor 11 to the displacement volume V2 of the second compressor 31, as shown in the following equation (3) (S3).

[0048]

number

[0049] The first determination unit 233 provides the compressor frequency Hz1 corrected by the first correction unit 232 and the compressor suction gas specific volume v acquired by the first information acquisition unit 231 as explanatory variables to the input layer of the neural network. Then, the first determination unit 233 determines the objective variable output by the NN model M based on the provided explanatory variables as the refrigerant circulation amount G1 of the first refrigeration cycle 10 (S4). Since the NN model M used here corresponds to the second refrigeration cycle 30, the determined refrigerant circulation amount G1 is expressed as the upper part of the following equation (4) using information on the excluded volume V2 of the second compressor 31.

[0050]

number

[0051] This refrigerant circulation amount G1 can be converted as shown in the lower part of the above formula (4). That is, by using the compressor frequency correction value Hz1, the first determination unit 233 can determine the refrigerant circulation amount G1 of the first refrigeration cycle 10 corresponding to the displacement volume V1 of the first compressor 11 using the NN model M.

[0052] The first operation control unit 234 controls the operation of each device 11, 12, 13, or 14 in the first refrigeration cycle 10 based on the refrigerant circulation amount G1 determined by the first judgment unit 233 and various setting information stored in the first setting information storage unit 211 (S5).

[0053] According to the first embodiment described above, the information used to operate a first model air conditioner that has no operational history can be accurately determined by correcting the input data using an NN model that has already been constructed for a second model air conditioner that has an operational history.

[0054] In the above-described determination process, the first CPU 23 corrects the compressor frequency Hz, which is one of the first items of input data, and determines the refrigerant circulation amount G1 based on the corrected value Hz1 of the compressor frequency Hz and the input compressor suction gas specific volume v. However, this is not limited to this, and the compressor suction gas specific volume v, which is one of the first items of input data, may be similarly corrected, and the refrigerant circulation amount G1 may be determined based on the corrected value of the compressor suction gas specific volume v and the input compressor frequency Hz.

[0055] Second Embodiment An air conditioner using a determination device according to a second embodiment of the present invention will be described. In the second embodiment, a newly developed third model air conditioner with no operational track record is controlled using an NN model that has already been constructed for controlling a second model air conditioner with operational track record.

[0056] <Configuration of an air conditioner using a determination device according to the second embodiment> The configuration of an air conditioner using a determination device according to a second embodiment of the present invention will be described with reference to Fig. 4. The air conditioner 3 according to this embodiment is a newly developed third model air conditioner that has no operational track record.

[0057] The air conditioner 3 includes a third compressor 51, a third evaporator 52, a third electronic expansion valve 53, a third condenser 54, and a third control device 60 as a determination device, which constitute a third refrigeration cycle 50. The functions of the devices 51, 52, 53, and 54 that constitute the third refrigeration cycle 50 are similar to the functions of the devices 11, 12, 13, and 14 described in the first embodiment, and therefore detailed description thereof will be omitted.

[0058] The third control device 60 has a third storage unit 61, a third input unit 62, and a third CPU 63, and controls the devices 51, 52, 53, and 54 in the third refrigeration cycle 50. The third storage unit 61 has a third setting information storage unit 611 and a third NN model information storage unit 612.

[0059] The third setting information storage unit 611 stores setting information about the devices 51, 52, 53, and 54 in the third refrigeration cycle 50. The third NN model information storage unit 612 stores the above-mentioned NN model M.

[0060] The third input unit 62 inputs commands related to the operation of the devices 51, 52, 53, and 54 in the third refrigeration cycle 50, and information used in processing corresponding to the commands. Specifically, the third input unit 62 inputs a first item of input data related to the operation of the devices in the third refrigeration cycle 50, and a command to output a second item of output data related to the operation of the devices in the third refrigeration cycle 50 from the first item of input data.

[0061] The third CPU 63 includes a third information acquisition unit 631, a third determination unit 632, a third correction unit 633, and a third operation control unit 634.

[0062] The third information acquisition unit 631 acquires a command input from the third input unit 62 and information used in the processing corresponding to the command. The third information acquisition unit 631 also acquires information used in the processing corresponding to the acquired command from the information stored in the third setting information storage unit 611.

[0063] The third determination unit 632 determines the output data of the second item from the input data of the first item acquired by the third information acquisition unit 631 using the NN model M.

[0064] The third correction unit 633 corrects the output data of the corresponding second item determined by the third determination unit 632 by multiplying the output data of the corresponding second item by the ratio of the first coefficient to the second coefficient, so that the output data corresponds to the third refrigeration cycle 50.

[0065] The third operation control unit 634 controls the operation of each device 51, 52, 53, or 54 in the third refrigeration cycle 50 based on the output data of the second item regarding the third refrigeration cycle 50 corrected by the third judgment unit 632 and the information stored in the second setting information memory unit 411.

[0066] <Operation of air conditioner using determination device according to second embodiment> The operation of the air conditioner 3 according to this embodiment will be explained using a regression model as an example, in which the compressor frequency Hz used in the third compressor 51 in the third refrigeration cycle 50 and the compressor suction gas specific volume v of the third compressor 51 are used as explanatory variables, and the refrigerant circulation amount G3 in the third refrigeration cycle 50 is used as the objective variable.

[0067] The following describes a process in which the third control device 60 determines the refrigerant circulation amount G3 related to the third refrigeration cycle 50 by correcting the refrigerant circulation amount G2 output by the NN model M from the compressor frequency Hz and the compressor suction gas specific volume v related to the operation of the third refrigeration cycle 50. Fig. 5 is a flowchart showing the flow of the process executed by the third control device 60.

[0068] First, the third input unit 62 inputs a command to determine the refrigerant circulation amount G3. At that time, the third input unit 62 inputs information on the compressor frequency Hz and the compressor suction gas specific volume v, which are input data of the first item, as explanatory variables to be given to the NN model M to determine the refrigerant circulation amount G3. The third input unit 62 also inputs correction target information indicating that the data to be corrected when using the NN model M is the refrigerant circulation amount G2, and information on the excluded volume V2 of the second compressor 31 to be used in the correction process of the refrigerant circulation amount G2.

[0069] The information input from the third input unit 62 is acquired by the third information acquisition unit 631. When the third information acquisition unit 631 acquires a command to determine the refrigerant circulation amount G3 of the third refrigeration cycle 50 ("YES" in S11), it acquires, as information to be used in the determination process, information on the compressor frequency Hz and the compressor suction gas specific volume v, which are input data of the first item acquired together with the command, information on the excluded volume V2 of the second compressor 31, and information on the excluded volume V3 of the third compressor 51 stored in the third setting information storage unit 611 (S12).

[0070] Next, the third determination unit 632 provides the compressor frequency Hz and the compressor suction gas specific volume v acquired by the third information acquisition unit 631 as explanatory variables to the NN model M. Then, the third determination unit 632 determines the objective variable output by the NN model M based on the provided explanatory variables as the refrigerant circulation amount G2 (S13).

[0071] Next, the third correction unit corrects the refrigerant circulation amount G2 determined by the third judgment unit 632 to the ratio of the refrigerant circulation volume V3 of the third compressor 51 to the refrigerant circulation volume V2 of the second compressor 31, as shown in the following equation (5) (S14).

[0072]

number

[0073] The third operation control unit 634 controls the operation of each device 51, 52, 53, or 54 in the third refrigeration cycle 50 based on the refrigerant circulation amount G3 corrected by the third correction unit 633 and various setting information stored in the third setting information storage unit 611 (S15).

[0074] According to the second embodiment described above, a first model air conditioner that has no operational history can be properly controlled by correcting the output data using an NN model that has already been constructed for a second model air conditioner that has operational history.

[0075] In the first and second embodiments described above, when determining the refrigerant circulation amount using an NN model from the compressor frequency and the compressor suction gas specific volume in the refrigeration cycle of an air conditioner, a process of correcting any of the relevant information has been described. However, this is not limited to this, and similar techniques can be used to determine information about the operation of other actuators whose operation amount relative to predetermined input parameters varies depending on the model of the equipment.

[0076] For example, in the refrigeration cycle of an air conditioner, at least one of the high pressure, low pressure, discharge gas temperature, intake gas temperature, inlet water temperature, outside air temperature, water flow rate, compressor frequency, and fan speed may be used as input data for the first item, and an NN model may be used to output the electronic expansion valve opening, fan speed, refrigerant leakage determination information, or malfunction determination information, etc., as output data for the second item. In these cases, when the refrigeration cycle operates based on specified input parameters, by correcting either the compressor frequency, fan speed, or electronic expansion valve opening, which differ depending on the model, it is possible to determine information about the air conditioner to be controlled using an existing NN model for other models of air conditioners.

[0077] [Verification results] This section explains the results of verifying the opening degree of the electronic expansion valve under various conditions for the target air conditioner, model B, using an NN model constructed for the operation of model A air conditioner, which has a proven operational track record.

[0078] Figure 6(a) is a graph showing, with a set of points, the positions indicating the corresponding states between the opening of the electronic expansion valve determined using the NN model for air conditioner A without correcting any of the data (predicted electronic expansion valve opening value) and the opening of the electronic expansion valve actually measured for air conditioner B (actual measured electronic expansion valve opening value) for various operating conditions.

[0079] Figure 6(b) is a graph showing, with a set of points, the positions indicating the state in which the predicted value of the electronic expansion valve opening determined for various operating conditions by using an NN model for air conditioner A and correcting any data related to the opening of the electronic expansion valve to correspond to air conditioner B model as described in the above embodiment, corresponds to the actual measured value of the electronic expansion valve opening actually measured in air conditioner B model under the respective corresponding conditions.

[0080] In Figures 6(a) and (b), when the predicted value of the electronic expansion valve opening and the actual measured value of the electronic expansion valve opening match under specified operating conditions, a point corresponding to one of the positions on the dotted line L in the graph is shown.

[0081] In Figure 6(a), the set of points deviates significantly from dotted line L, whereas in Figure 6(b), the set of points is concentrated in a position that is close to dotted line L. This verified that by correcting any of the data related to the electronic expansion valve opening, it was possible to accurately determine the electronic expansion valve opening for model B air conditioner using the NN model for model A air conditioner.

[0082] Although an embodiment of the present invention has been described, this embodiment is presented as an example and is not intended to limit the scope of the invention. This novel embodiment can be embodied in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. This embodiment and its modifications are included within the scope and spirit of the invention, and are also included in the invention described in the claims and their equivalents.

[0083] It is also possible to program the functional configuration of the first control device 20 or the third control device 60 and incorporate it into a computer, thereby constructing a judgment program that causes the computer to function as the first control device 20 or the third control device 60.

Claims

1. a model information storage unit that stores a neural network model constructed for the operation of a second device that operates on the same operating principle as the first device to be controlled and is a different model; an information acquisition unit that acquires input data of a first item related to a device that operates according to the operating principle, a first coefficient for calculating output data of a second item from the input data of the first item in the first device, and a second coefficient for calculating output data of the second item from the input data of the first item in the second device; a determination unit that determines output data of the second item related to the first device from input data of the first item using a neural network model stored in the model information storage unit; a correction unit that corrects the corresponding data by multiplying the input data of the first item or the output data of the second item used in the judgment unit by the ratio of the first coefficient to the second coefficient.

2. 2. The determination device according to claim 1, wherein the correction unit corrects data relating to an actuator whose operating amount with respect to a predetermined input parameter differs depending on the model of the device, among the input data of the first item or the output data of the second item used by the determination unit.

3. the information acquisition unit acquires information as input data for the first item, which is at least one of a high pressure, a low pressure, a discharge gas temperature, an intake gas temperature, an inlet water temperature, an outside air temperature, a water flow rate, a compressor frequency, and a fan rotation speed, and acquires information as output data for the second item, which is an electronic expansion valve opening, a fan rotation speed, information on whether a refrigerant leaks, or information on whether a malfunction occurs; The determination device according to claim 1 or 2, wherein the correction unit corrects any one of a compressor frequency, a fan rotation speed, and an electronic expansion valve opening.

4. A determination device comprising: storing a neural network model constructed for the operation of a second device that operates according to the same operating principle as the first device to be controlled but is a different model; acquiring input data for a first item related to a device operating according to the operating principle, a first coefficient for calculating output data for a second item from the input data for the first item in the first device, and a second coefficient for calculating output data for the second item from the input data for the first item in the second device; a determination method comprising: correcting the input data of the first item by multiplying it by the ratio of the first coefficient to the second coefficient; and determining output data of the second item related to the first device from the corrected input data of the first item using the neural network model; or determining output data of the second item from the input data of the first item using the neural network model, and then multiplying the determined output data of the second item by the ratio of the first coefficient to the second coefficient to correct it so that it corresponds to the first device.

5. On the computer, a function of storing a neural network model constructed for the operation of a second device that operates on the same operating principle as the first device to be controlled and is a different model; a function of acquiring input data of a first item related to a device operating on the operating principle, a first coefficient for calculating output data of a second item from the input data of the first item in the first device, and a second coefficient for calculating output data of the second item from the input data of the first item in the second device; a function of correcting the input data of the first item by multiplying the ratio of the first coefficient to the second coefficient, and determining the output data of the second item related to the first device from the corrected input data of the first item using the neural network model, or determining the output data of the second item from the input data of the first item using the neural network model, and then multiplying the determined output data of the second item by the ratio of the first coefficient to the second coefficient to correct it so that it corresponds to the first device; A judgment program for executing the above.

Images