Performance determination device, performance determination method, and cooling water circulation device

Abstract

The present invention reduces erroneous detection of abnormalities in a refrigerant circuit by a cooling water circulation device having a plurality of circuits.　The present disclosure is a performance determination device that determines performance deterioration of a cooling water circulation device. Piping through which a liquid cooled by a refrigerant of the cooling water circulation device circulates passes through one or more evaporators provided to a plurality of circuits in which the refrigerant circulates. The performance determination device has a control unit that inputs the operation mode and operation data of the plurality of circuits to a model obtained by learning a correspondence between an objective variable and an explanatory variable, the objective variable being a performance index value of one circuit among the plurality of circuits, the explanatory variable being the operation mode and operation data of the plurality of circuits, and the control unit outputting a prediction value of the performance index value. The control unit calculates an actual measurement value of the performance index value from the operation data of one circuit among the plurality of circuits, and determines the performance deterioration from a deviation between the prediction value and the actual measurement value.

Description

Performance judgment device, performance judgment method, cooling water circulation device

[0001] The present disclosure relates to a performance determination device, a cooling method, and a cooling water circulation device.

[0002] The performance of a cooling water circulation system (chiller) may deteriorate during operation. For example, if the amount of refrigerant charged in the refrigerant circuit of a heat source unit falls below a specified level, the cooling capacity decreases. For this reason, a method for detecting loss of refrigerant charge (refrigerant leakage) based on a value that serves as an index of the refrigerant amount (hereinafter also referred to as a refrigerant amount index value) is known.

[0003] Patent Document 1 discloses a technology for estimating the proportion of leaked refrigerant using at least the compressor rotation speed, the compressor refrigerant discharge temperature, the heat exchanger temperature, the expansion valve opening, and the outside air temperature, which are operating state quantities that indicate the operating state during operation, and an estimation model that estimates the amount of refrigerant remaining in the refrigerant circuit.

[0004] Patent No. 7124851

[0005] However, in conventional techniques, when a cooling water circulation device has multiple circuits through which refrigerant circulates, there are cases where an abnormality in the refrigerant circuit is falsely detected. For example, in a cooling water circulation device, when a first refrigerant circuit stops, the load on the second refrigerant circuit increases, and there is a case where a refrigerant leak from the second refrigerant circuit is falsely detected even though there is no leak.

[0006] In view of the above-described problems, the present disclosure provides a technique for reducing false detection of an abnormality in a refrigerant circuit in a coolant circulation device having multiple circuits.

[0007] A first aspect of the present disclosure is a performance determination device for determining performance degradation of a cooling water circulation device, wherein piping for circulating a liquid cooled by the refrigerant passes through one or more evaporators provided in multiple circuits through which the refrigerant of the cooling water circulation device circulates, and the device has a control unit that sets a performance index value of one of the multiple circuits as a dependent variable and an operating mode and operating data of the multiple circuits as explanatory variables, inputs the operating mode and operating data of the multiple circuits into a model that has learned the correspondence between the dependent variable and the explanatory variables, and outputs a predicted value of the performance index value, and the control unit calculates an actual measurement value of the performance index value from the operating data of one of the multiple circuits, and determines the performance degradation from the discrepancy between the predicted value and the actual measurement value.

[0008] According to the first aspect of the present disclosure, it is possible to reduce false detection of an abnormality in a refrigerant circuit in a coolant circulation device having multiple circuits.

[0009] A second aspect of the present disclosure is the performance determination device according to the first aspect, wherein the operation mode is status information of a circuit that is output for each circuit through which a refrigerant circulates.

[0010] A third aspect of the present disclosure is the performance determination device according to the first or second aspect, wherein the performance index value is a refrigerant amount index value related to leakage of refrigerant circulating through one of the circuits.

[0011] A fourth aspect of the present disclosure is a performance determination device for determining performance degradation of a cooling water circulation device, wherein one or more evaporators are provided in multiple circuits through which refrigerant of the cooling water circulation device circulates, and pipes through which liquid cooled by the refrigerant circulates pass; a model is generated for each number of circuits in a predetermined operating state among the multiple circuits, the model selecting a model according to the number of predetermined operating states from multiple models that have learned the correspondence between the dependent variable and the explanatory variables, using a performance index value of one of the circuits in the predetermined operating state as a dependent variable and operating data of the multiple circuits as an explanatory variable; and a control unit that inputs the operating data of the multiple circuits into the selected model and outputs a predicted value of the performance index value; the control unit calculates an actual measurement value of the performance index value from the operating data of one of the circuits in the predetermined operating state, and determines the performance degradation from the deviation between the predicted value and the actual measurement value.

[0012] According to the fourth aspect of the present disclosure, it is possible to reduce false detection of an abnormality in a refrigerant circuit in a coolant circulation device having multiple circuits.

[0013] A fifth aspect of the present disclosure is a performance determination method performed by a performance determination device that determines performance degradation of a cooling water circulation device, wherein one or more evaporators are provided in multiple circuits through which refrigerant of the cooling water circulation device is circulated, and pipes through which liquid cooled by the refrigerant circulates pass; a performance index value of one of the multiple circuits is used as a dependent variable, and operating modes and operating data of the multiple circuits are used as explanatory variables, and a control unit inputs the operating mode and operating data of the multiple circuits into a model that has learned the correspondence between the dependent variable and the explanatory variables, and outputs a predicted value of the performance index value; the control unit calculates an actual measurement value of the performance index value from the operating data of the one of the multiple circuits; and determines the performance degradation from the deviation between the predicted value and the actual measurement value.

[0014] According to the fifth aspect of the present disclosure, it is possible to reduce false detection of an abnormality in a refrigerant circuit in a coolant circulation device having multiple circuits.

[0015] A sixth aspect of the present disclosure is a cooling water circulation system equipped with a performance determination device that determines performance degradation of the cooling water circulation system, wherein piping through which liquid cooled by the refrigerant circulates passes through one or more evaporators provided in multiple circuits through which the refrigerant of the cooling water circulation system circulates, and the performance determination device has a control unit that sets a performance index value of one of the multiple circuits as a dependent variable and an operating mode and operating data of the multiple circuits as explanatory variables, inputs the operating mode and operating data of the multiple circuits into a model that has learned the correspondence between the dependent variable and the explanatory variables, and outputs a predicted value of the performance index value, and the control unit calculates an actual measurement value of the performance index value from the operating data of the one of the multiple circuits, and determines the performance degradation from the discrepancy between the predicted value and the actual measurement value.

[0016] According to the sixth aspect of the present disclosure, it is possible to reduce false detection of an abnormality in a refrigerant circuit in a coolant circulation device having multiple circuits.

[0017] 1 is a diagram illustrating the overall configuration of a cooling system including a cooling water circulation device and a performance assessment device. FIG. 2 is a diagram illustrating an example of a cooling water circulation device in which two refrigerant circuits each have one evaporator. FIG. 3 is a diagram illustrating the learning phase of a normal amount prediction model in a comparative technology. FIG. 4 is a diagram illustrating a method for determining a refrigerant leak using a normal amount prediction model in a comparative technology. FIG. 5 is a diagram illustrating an example of a refrigerant amount index value and a normal amount predicted value before removing a disturbance. FIG. 6 is a diagram illustrating an example of a difference ΔRLI between the refrigerant amount index value and the normal amount predicted value. FIG. 7 is a diagram illustrating a method for determining a leak in a comparative technology when there are two refrigerant circuits. FIG. 8 is a diagram illustrating a state in which the other of two refrigerant circuits has stopped. FIG. 9 is an example of a diagram illustrating changes in the expansion valve opening of a refrigerant circuit over time. FIG. 10 is an example of a diagram illustrating changes in the expansion valve opening of a refrigerant circuit over time. FIG. 11 is an example of a diagram illustrating the relationship between the difference ΔRLI and time in a refrigerant circuit. FIG. 12 is a diagram illustrating an example of an arrangement of a cooling water circulation device, a performance assessment device, and a learning device. 1 is a diagram illustrating an example of the arrangement of a cooling water circulation device, a performance determination device, and a learning device. FIG. 1 is a diagram illustrating an example of the arrangement of a cooling water circulation device, a performance determination device, and a learning device. FIG. 2 is a hardware configuration diagram of an example of a performance determination device. FIG. 3 is a functional block diagram of a learning device according to an embodiment of the present disclosure. FIG. 4 is a functional block diagram of a performance determination device according to an embodiment of the present disclosure. FIG. 5 is a diagram illustrating an example of sensors arranged in a refrigerant circuit and their positions. FIG. 6 is a diagram schematically illustrating a learning phase of a normal amount prediction model. FIG. 7 is a flowchart of an example of a process in which a learning device generates a normal amount prediction model. FIG. 8 is a flowchart of an example of a process in which a performance determination device determines a refrigerant leak from a refrigerant circuit using the normal amount prediction model. FIG. 9 is a diagram illustrating a ΔRLI of an example of a refrigerant circuit 1 in a comparative technology. FIG. 10 is a diagram illustrating a ΔRLI of an example of a refrigerant circuit 1 in the present disclosure. FIG. 11 is a diagram illustrating an overview of a learning phase in a second embodiment. FIG. 12 is a flowchart of an example of a process in which a refrigerant leak is determined using a normal amount prediction model.

[0018] Hereinafter, a performance determination method performed by a cooling water circulation device and a control device will be described as an example of an embodiment of the present disclosure.

[0019] <System Configuration Example> FIG. 1 illustrates the overall configuration of a cooling system 1000 including a cooling water circulation device 100 and a performance evaluation device 400. The cooling water circulation device 100, also known as a chiller, is a system that cools a space or object by cooling a liquid and circulating the liquid to a user side. However, it may also be applied to a refrigeration air-conditioner that cools gas. The cooling water circulation device 100 of the present disclosure has two refrigerant circuits 1 and 2 that share an evaporator 14. Having multiple refrigerant circuits (i.e., multiple circuits) increases the cooling capacity. Note that the refrigerants in refrigerant circuit 1 (an example of a first heat source unit) and refrigerant circuit 2 (an example of a second heat source unit) are independent, and the refrigerant in refrigerant circuit 1 does not flow through refrigerant circuit 2, and vice versa.

[0020] The refrigerant circulates through the refrigerant circuits 1 and 2 in the direction of arrow 18, repeating the cycle of "evaporation" → "compression" → "condensation" → "expansion." Figure 1 shows the main components, including an "evaporator 14," a "compressor 13," a "condenser 11," and an "expansion valve 12." Note that Figure 1 only shows the main components, and other components may be present.

[0021] The flow of refrigerant will now be described. The flow of refrigerant is the same in the two refrigerant circuits 1 and 2. The compressor 13 generates pressure to circulate the refrigerant, and compresses low-temperature, low-pressure gas refrigerant with a piston or the like to turn it into high-temperature, high-pressure gas refrigerant. The gas refrigerant drawn into the compressor 13 flows to the condenser 11.

[0022] The condenser 11 releases heat from the refrigerant, cooling the high-temperature, high-pressure gas refrigerant to become a room-temperature, high-pressure liquid refrigerant. That is, the gas refrigerant that enters the condenser 11 is cooled to become a room-temperature, high-pressure liquid, which then flows to the expansion valve 12. The condenser 11 can be an air-cooled condenser (plate type), a water-cooled condenser (shell-and-tube type), or the like. The outdoor unit of an air conditioner in cooling operation corresponds to the condenser 11.

[0023] The expansion valve 12 reduces the pressure of the liquid refrigerant. The expansion valve 12 passes the room-temperature, high-pressure liquid refrigerant through a small hole and adjusts the pressure and flow rate of the refrigerant to produce a low-temperature, low-pressure liquid refrigerant.

[0024] The evaporator 14 lowers the temperature of the object to be cooled (liquid in the case of a chiller) using the heat of vaporization when the liquid refrigerant evaporates. The circuit that lowers or raises the temperature in this way is called a heat source machine. The evaporator 14 vaporizes the low-temperature, low-pressure liquid refrigerant into a low-temperature, low-pressure gas refrigerant. The refrigerant absorbs external heat and becomes vapor, which is then drawn into the compressor 13. The indoor unit of an air conditioner in cooling operation corresponds to the evaporator 14.

[0025] In FIG. 1, the evaporator 14 is common to the refrigerant circuits 1 and 2. The evaporator 14 functions as a heat exchanger that exchanges heat with the object to be cooled. The object to be cooled is a liquid such as water or brine. This liquid circulates in a pipe 17 under pressure generated by a pump 16. The pipe 17 is installed so as to pass through an indoor heat exchanger 15, thereby cooling the indoor side. The object to be cooled on the indoor side can vary depending on the application, such as room air, a refrigerator, or a freezer.

[0026] FIG. 2 is a diagram showing an example of a cooling water circulation system 100 in which two refrigerant circuits 1 and 2 each have one evaporator 14a and 14b. The explanation of FIG. 2 will mainly focus on the differences from FIG. 1. As shown in FIG. 2, refrigerant circuit 1 has evaporator 14a, and refrigerant circuit 2 has evaporator 14b. Liquid cooled by evaporators 14a and 14b circulates through piping 17 under pressure generated by pump 16. Therefore, even with the structure of FIG. 2, indoor heat exchanger 15 can cool the indoor side using the cooling capacity of the two refrigerant circuits 1 and 2. The method for predicting a refrigerant amount index value described in this disclosure can be applied without any problems to either the circuit configuration of FIG. 1 or FIG. 2.

[0027] 1 and 2, the number of refrigerant circuits 1 and 2 is two, but the number of refrigerant circuits may be three or more.

[0028] <Terminology> A performance index value is a value that represents the performance of the cooling water circulation system 100. One type of performance is cooling capacity, and cooling capacity performance is known as a refrigerant quantity index value or COP (cooling capacity value per 1 kW of power consumption). In this disclosure, the refrigerant quantity index value is mainly used as an example, but the performance determination device 400 can predict COP in the same way as the refrigerant quantity index value by using COP as the objective variable described below. If the predicted COP deviates from the actual measured value, an abnormality can be detected.

[0029] <Comparative Techniques for Refrigerant Leak Detection> Next, comparative techniques for refrigerant leak detection will be described. Refrigerant leakage from refrigerant circuit 1 or 2 reduces the cooling capacity of the cooling water circulation system 100. Therefore, the performance assessment device 400 actually measures the refrigerant amount (the actually measured refrigerant amount may be referred to as a refrigerant amount index value, a refrigerant amount index value RLI, or simply RLI. Also, actual measurement may be referred to as calculation). In addition to actually measuring the refrigerant amount index value, the performance assessment device 400 creates a normal amount prediction model that predicts the refrigerant amount index value through machine learning, inputs operating data into the normal amount prediction model, and calculates a predicted value of the refrigerant amount index value (hereinafter referred to as a normal amount prediction value). In this way, the performance assessment device 400 detects refrigerant leakage in the cooling water circulation system 100 during actual operation.

[0030] The flow of refrigerant leak detection is as follows (see Figures 3A and 3B). (1) The developer first creates learning data. (2) The learning device extracts explanatory variables from the learning data and calculates a refrigerant quantity index value, which is a dependent variable, and creates a normal quantity prediction model that predicts the refrigerant quantity index value under normal conditions from the correspondence between the explanatory variables and dependent variables. (3) The performance assessment device 400 calculates a refrigerant quantity index value from the cooling water circulation device 100 for which a refrigerant leak is to be detected, acquires operating data, and calculates a normal quantity prediction value using the normal quantity prediction model. (4) The performance assessment device 400 determines whether a leak has occurred based on the degree of deviation between the refrigerant quantity index value and the normal quantity prediction value.

[0031] 3A is a diagram illustrating the learning phase of the normal amount prediction model in the comparative technology. Learning data 21 is prepared from the operating data of the cooling water circulation system 100. The performance assessment device 400 acquires, from the learning data 21, a refrigerant amount index value 23 in a state where there is no refrigerant leakage and explanatory variables (operating data) 24 in a state where there is no leakage. The explanatory variables 24 will be described later. The performance assessment device 400 learns the correspondence between the explanatory variables 24 and the refrigerant amount index value 23 using various algorithms, and generates a normal amount prediction model 22.

[0032] FIG. 3B is a diagram illustrating a method (inference phase) for determining a refrigerant leak using the normal amount prediction model 22 in the comparative technique. Operating data 30 is acquired from an actually operating cooling water circulation system 100. Preprocessing 25 (e.g., removal of outliers and normalization) is performed on the learning data. A refrigerant quantity index value 26 is calculated from the operating data after preprocessing 25. The operating data after preprocessing 25 is input to the normal amount prediction model 22, which outputs a normal amount prediction value 28. A performance determination device 400 calculates a difference ΔRLI 29 between the refrigerant quantity index value 26 and the normal amount prediction value 28. The performance determination device 400 outputs a determination result 31 based on the difference ΔRLI 29. Specifically, the performance determination device 400 determines a refrigerant leak (an example of performance degradation) when the difference ΔRLI 29 is greater than a threshold value. The ratio of the refrigerant quantity index value 26 to the normal amount prediction value 28 may be compared with a threshold value.

[0033] That is, whether or not there is a refrigerant leak is determined based on the degree of deviation between the refrigerant amount index value 26 and the predicted normal amount value 28. If the deviation becomes large, it is determined that there is a leak. By making a determination based on the degree of deviation, even if a disturbance occurs in the refrigerant amount index value due to the outside air temperature or the load on the compressor 13, the disturbance can be removed by taking the difference.

[0034] The effect of removing the disturbance from the refrigerant amount index value will be described with reference to Figures 4A and 4B. Figure 4A shows the refrigerant amount index value 33 and the predicted normal amount value 32 before removing the disturbance. The horizontal axis represents time, and the vertical axis represents the refrigerant amount index value 33 and the predicted normal amount value 32. Before removing the disturbance, there is a large fluctuation between the refrigerant amount index value 33 and the predicted normal amount value 32. Figure 4B shows the difference ΔRLI between the refrigerant amount index value 33 and the predicted normal amount value 32. It can be seen that removing the disturbance reduces the fluctuation of the difference ΔRLI.

[0035] Next, we will explain how to determine whether there is a refrigerant leak when there are two refrigerant circuits 1 and 2. When there are two refrigerant circuits 1 and 2 as shown in Figure 1 or 2, the performance determination device 400 inputs operating data for each of the refrigerant circuits 1 and 2 into a normal amount prediction model and determines whether there is a leak for each of the refrigerant circuits 1 and 2.

[0036] 5 is a diagram illustrating a leakage determination method in the comparative technology when the cooling water circulation system 100 has two refrigerant circuits 1 and 2. As shown in FIG. 5, a refrigerant amount index value 201 is acquired from refrigerant circuit 1, and the operating data acquired from refrigerant circuit 1 is input to a normal amount prediction model to output a normal amount prediction value 202. The performance determination device 400 compares the difference ΔRLI between the refrigerant amount index value 201 and the normal amount prediction value 202 with a threshold value to determine whether or not a refrigerant leak exists in refrigerant circuit 1 (determination result 205).

[0037] Similarly, a refrigerant amount index value 203 is acquired from the refrigerant circuit 2, and the operating data acquired from the refrigerant circuit 2 is input into a normal amount prediction model to output a normal amount prediction value 204. The performance determination device 400 compares the difference ΔRLI between the refrigerant amount index value 203 and the normal amount prediction value 204 with a threshold value to determine whether or not a refrigerant leak exists in the refrigerant circuit 2 (determination result 206).

[0038] However, the normal amount prediction model was generated without taking into account the operating status of the other refrigerant circuit. In other words, the normal amount prediction model was generated without taking into account that the evaporator 14 is shared by the two refrigerant circuits 1 and 2. As a result, there is a possibility that the judgment results 1 and 2 are affected by the operating status of the other refrigerant circuit. This will be explained with reference to Figures 6A to 6C.

[0039] FIG. 6A illustrates a state in which one of the two refrigerant circuits 1 and 2 in a cooling water circulation system 100 is stopped. While FIG. 6A shows refrigerant circuit 2 stopped, refrigerant circuit 1 may also be stopped. When two refrigerant circuits 1 and 2 are used to cool water or brine using a shared evaporator 14, even if the other refrigerant circuit 2 experiences an abnormality and stops, cooling of the water or brine may still be performed using only one refrigerant circuit 1. In this case, the water temperature must be brought close to the set value in one refrigerant circuit 1, which increases the load on the normal refrigerant circuit 1. As a result, the operating data of the normal refrigerant circuit 1 (e.g., the load factor of the compressor 13 and the opening of the expansion valve 12) changes significantly. The normal amount prediction value calculated by inputting such operating data into the normal amount prediction model may be a value that indicates an abnormality in the refrigerant circuit 1.

[0040] Figure 6B shows the change over time in the expansion valve opening degree of refrigerant circuit 1, and Figure 6C shows the change over time in the expansion valve opening degree of refrigerant circuit 2. During time T shown in Figure 6C, an alarm was generated in refrigerant circuit 2 (some kind of abnormality occurred), so the expansion valve opening degree of refrigerant circuit 2 is zero. Due to the stop of refrigerant circuit 2, a load is applied to refrigerant circuit 1, and during time T, the expansion valve opening degree of refrigerant circuit 1 increases, increasing the flow rate of refrigerant.

[0041] 7 shows the relationship between the difference ΔRLI and time for the refrigerant circuit 1. It can be seen that the difference ΔRLI fluctuates greatly and decreases at time T when the refrigerant circuit 2 shown in FIG. 6A is stopped. In other words, because the normal amount prediction model does not take into account the operating status of the other refrigerant circuit 2, the difference ΔRLI for the refrigerant circuit 1 increases when the refrigerant circuit 2 stops, leading to an erroneous determination that refrigerant is leaking from the refrigerant circuit 1 when there is no leak.

[0042] As described above, in a cooling water circulation system 100 with multiple circuits, there is a possibility of false detection when detecting an abnormality in a refrigerant circuit. This is because the comparative technology generates a normal amount prediction model during the learning phase without considering the operating status of the other refrigerant circuits. Because the normal amount prediction model does not detect that the load on one refrigerant circuit is high due to an abnormality in the other refrigerant circuit, when the other refrigerant circuit stops, the difference ΔRLI between the refrigerant amount index value and the predicted normal amount value becomes large, and there are cases where a refrigerant leak is erroneously determined even when there is no leak.

[0043] Therefore, in the present disclosure, in a coolant circulation system 100 having multiple circuits, a normal amount prediction model is created that takes into account the operating status of each refrigerant circuit, and the operating data of each refrigerant circuit is input into the normal amount prediction model to determine whether a refrigerant leak has occurred. This will be described in detail below.

[0044] 8A to 8D are diagrams illustrating an example of the arrangement of a cooling water circulation apparatus 100, a performance determination device 400, and a learning device 500. In FIG. 8A, the performance determination device 400 is arranged as a computer installed in the same building as the cooling water circulation apparatus 100. The learning device 500 is arranged on a cloud server connected to the cooling water circulation apparatus 100 and the performance determination device 400 via a network N. The learning device 500 is an information processing device that generates a normal amount prediction model in the learning phase.

[0045] 8B, the performance determination device 400 is disposed integrally with the cooling water circulation system 100 (for example, integrally with or near the control device of the cooling water circulation system 100). The learning device 500 is disposed on a cloud server connected to the cooling water circulation system 100 and the performance determination device 400 via a network N.

[0046] In FIG. 8C , the performance determination device 400 and the learning device 500 are arranged on a cloud server connected to the cooling water circulating device 100 via a network N.

[0047] In FIG. 8D, the performance determination device 400 and the learning device 500 are disposed integrally with the cooling water circulation device 100 (for example, integrally with or near the control device of the cooling water circulation device 100).

[0048] In either arrangement example, in the learning phase, the learning device 500 acquires operating data (learning data) from the cooling water circulation system 100 and generates a normal amount prediction model. In the inference phase, the performance determination device 400 inputs the operating data acquired from the cooling water circulation system 100 into the generated normal amount prediction model and determines an abnormality in the refrigerant circuit.

[0049] <Example of Hardware Configuration of Performance Determination Device and Learning Device> The performance determination device 400 is a device that determines the amount of refrigerant leakage from the cooling water circulation device 100. The performance determination device 400 is one or more information processing devices that are communicatively connected to the cooling water circulation device 100 via a network or a dedicated line.

[0050] 9 is a hardware configuration diagram of a performance determination device 400 according to an embodiment of the present disclosure. The learning device may have the same hardware configuration, but may be provided with a high-speed GPU (Graphics Processing Unit) for learning.

[0051] The performance determining device 400 includes a central processing unit (CPU) 1, a read only memory (ROM) 2, and a random access memory (RAM) 3. The CPU 1, the ROM 2, and the RAM 3 form a so-called computer.

[0052] The performance determining device 400 may also have an auxiliary storage device 4, a display device 5, an operation device 6, and an I / F (Interface) device 7. The hardware components of the performance determining device 400 are connected to each other via a bus 8.

[0053] The CPU 1 is a computing device that executes various programs installed in the auxiliary storage device 4 .

[0054] The ROM 2 is a non-volatile memory and functions as a main storage device that stores various programs, data, etc. required for the CPU 1 to execute various programs installed in the auxiliary storage device 4. Specifically, the ROM 2 functions as a main storage device that stores boot programs such as a Basic Input / Output System (BIOS) and an Extensible Firmware Interface (EFI).

[0055] The RAM 3 is a volatile memory such as a dynamic random access memory (DRAM) or a static random access memory (SRAM). The RAM 3 functions as a main storage device that provides a working area in which various programs installed in the auxiliary storage device 4 are expanded when the CPU 1 executes them.

[0056] The auxiliary storage device 4 is an auxiliary storage device that stores various programs and information used when the various programs are executed.

[0057] The display device 5 is a display device that displays the internal state of the performance determining device 400 and the like.

[0058] The operation device 6 is an input device through which the administrator of the performance determining device 400 inputs various instructions to the performance determining device 400 .

[0059] The I / F device 7 is a communication device that connects to various sensors 9 and a network and communicates with other terminals.

[0060] <Functions of the Learning Device> Next, functions of the learning device 500 will be described with reference to FIG. 10A . FIG. 10A is a functional block diagram of the learning device 500 according to an embodiment of the present disclosure. The learning device 500 has a learning data acquisition unit 501, a learning data storage unit 502, and a learning unit 503. These functional units of the learning device 500 are functions or means realized by the CPU 1 of the learning device 500 executing instructions of a program deployed in RAM 3. The learning data acquisition unit 501, the learning data storage unit 502, and the learning unit 503 are collectively referred to as a control unit 510.

[0061] The learning data acquisition unit 501 acquires learning data 505. The learning data 505 is normal operating data, operating mode, and refrigerant amount index value. Normal means a state in which there is no refrigerant leakage from refrigerant circuits 1 and 2. If one of the refrigerant circuits is stopped simply due to an alarm, the other refrigerant circuit is determined to be normal. The operating data and operating mode are explanatory variables, and the refrigerant amount index value is a target variable (sometimes called teacher data). Details of the operating data will be described later, but include the compressor load factors of multiple refrigerant circuits, the expansion valve opening degrees of multiple refrigerant circuits, etc.

[0062] The operating mode is information indicating whether refrigerant circuit 1 is operating or not, and whether refrigerant circuit 2 is operating or not. The operating mode can be acquired in the following two ways: (i) The operating mode is explicitly transmitted from refrigerant circuits 1 and 2 to the performance determination device 400 (a signal indicating whether refrigerant circuit 1 is operating or not, and whether refrigerant circuit 2 is operating or not). (ii) The learning data acquisition unit 501 determines the operating mode from the operating data. For example, the operating mode (whether or not the refrigerant circuits 1 and 2 are operating or not) can be determined from whether the number of operating compressors is 0 or not, or whether the expansion valve opening is 0 or not, etc.

[0063] If refrigerant circuit 1 or 2 stops, some kind of abnormality has occurred, and therefore the operation mode also contains information about the abnormality. The refrigerant amount index value, which is the objective variable, can be calculated from each of refrigerant circuits 1 and 2. However, since the refrigerant amount index value in a normal state is approximately the same for refrigerant circuits 1 and 2, the refrigerant amount index value of either one of the refrigerant circuits may be used.

[0064] The learning data storage unit 502 stores the learning data acquired by the learning data acquisition unit 501 .

[0065] The learning unit 503 learns the learning data using various machine learning algorithms to generate a normal amount prediction model 504. The normal amount prediction model 504 is correspondence information that associates the operating data and operating mode of refrigerant circuit 1 and the operating data and operating mode of refrigerant circuit 2 with the refrigerant amount index value of at least one of the refrigerant circuits 1 and 2. In other words, the normal amount prediction model 504 outputs a normal amount prediction value of the refrigerant circuit 1 or 2 in response to input of the operating data and operating mode of the refrigerant circuits 1 and 2. Such correspondence information can be realized by a regression model. Examples of regression models include multiple regression, neural network, ridge regression, lasso regression, and elastic net regression.

[0066] <Functions of the Performance Determination Device> Next, functions of the performance determination device 400 will be described with reference to FIG. 10B . FIG. 10B is a functional block diagram of the performance determination device 400 according to an embodiment of the present disclosure. The performance determination device 400 includes an operating data acquisition unit 401, a calculation unit 402, an inference unit 403, a determination unit 404, and an output unit 405. These functional units of the performance determination device 400 are functions or means realized when the CPU 1 of the performance determination device 400 executes instructions of a program deployed in the RAM 3. The operating data acquisition unit 401, the calculation unit 402, the inference unit 403, the determination unit 404, and the output unit 405 are collectively referred to as a control unit 520.

[0067] The operating data acquisition unit 401 acquires real-time operating data 406 (current operating data) of the cooling water circulation system 100 detected by various sensors (temperature sensor, pressure sensor, etc.) of the cooling water circulation system 100 .

[0068] The calculation unit 402 calculates a refrigerant amount index value 407 from the operating data acquired by the operating data acquisition unit 401. The refrigerant amount index value 407 is a value that serves as an index of the amount of refrigerant flowing through the refrigerant circuit 1 or 2, and is a value that is correlated with the refrigerant amount (details will be described later).

[0069] The inference unit 403 determines the operation mode of each refrigerant circuit by either method (i) or (ii) above. The inference unit 403 uses the normal amount prediction model 504 to infer a normal amount prediction value 408, which is a refrigerant amount index value under normal conditions, from the operation data and operation modes of the refrigerant circuits 1 and 2 acquired by the operation data acquisition unit 401.

[0070] The judgment unit 404 compares the difference or ratio between the refrigerant quantity index value 407 calculated by the calculation unit 402 and the normal quantity predicted value 408 inferred by the inference unit 403 with a threshold value to determine whether or not there is a refrigerant leak in the refrigerant circuits 1 and 2.

[0071] The output unit 405 outputs the result of the determination made by the determination unit 404. For example, the output unit 405 displays a warning on the display of the cooling water circulating device 100, or notifies the administrator of the refrigerant leak by email or the like.

[0072] <Regarding the refrigerant quantity index value> The refrigerant quantity index value will be described. The refrigerant quantity index value in the case of cooling operation is, for example, one or more of the following: Condensing temperature - outdoor heat exchanger outlet temperature (hereinafter also referred to as outdoor heat exchanger outlet subcooling degree. Note that the subcooling degree is also called SC, or subcool). The outdoor heat exchanger corresponds to the condenser 11. Intake superheat degree of the compressor 13 (note that the superheat degree is also called SH, or superheat) Discharge superheat degree of the compressor 13 A value based on the outdoor heat exchanger outlet subcooling degree, the intake superheat degree of the compressor 13, or the discharge superheat degree of the compressor 13 The refrigerant quantity index value can include at least one of the following.

[0073] For example, the value based on the outdoor heat exchanger outlet subcooling degree is a calculated value using the outdoor heat exchanger outlet subcooling degree, such as: Calculated value using the outdoor heat exchanger outlet subcooling degree = outdoor heat exchanger outlet subcooling degree / (condensing temperature - outdoor air temperature).

[0074] Furthermore, for example, the refrigerant quantity index value may include at least one of the following in addition to the above-mentioned refrigerant quantity index value, or instead of the outdoor heat exchanger outlet subcooling degree of the above-mentioned refrigerant quantity index value: - Subcooling degree at the subcooling heat exchanger outlet - A value based on the subcooling degree at the subcooling heat exchanger outlet.

[0075] <Regarding Explanatory Variables> Next, the explanatory variables will be explained with reference to FIG. 11 . FIG. 11 shows an example of sensors arranged in a refrigerant circuit and their locations. Because the sensor locations are the same for refrigerant circuits 1 and 2, FIG. 11 shows only one refrigerant circuit. As shown, refrigerant circuits 1 and 2 are equipped with various temperature sensors T1 to T5, a compressor sensor C1, an opening sensor EV, a rotation speed sensor R1, and pressure sensors P1 and P2. Note that the number and locations of the sensors are merely examples. The opening sensor EV detects the opening of the expansion valve. The compressor sensor C1 detects the compressor load factor (or rotation speed). The compressor sensor C1 can also detect the number of operating compressors. The rotation speed sensor R1 detects the fan rotation speed. The temperature sensor T1 detects the compressor suction pipe temperature. The temperature sensor T2 detects the compressor discharge pipe temperature. The temperature sensor T3 detects the heat exchanger outlet liquid pipe temperature. - Temperature sensor T4 is a sensor that detects the supply water temperature. - Temperature sensor T5 is a sensor that detects the return water temperature. - Pressure sensor P1 is a low-pressure sensor. The evaporation temperature, which will be described later, is a temperature calculated from the sensor value of the low-pressure sensor. - Pressure sensor P2 is a high-pressure sensor. The condensation temperature, which will be described later, is a temperature calculated from the sensor value of the high-pressure sensor.

[0076] Among the operating data acquired from the cooling water circulation system 100, operating data that can be used as explanatory variables will be described. The following operating data is an example of the parameters acquired from each of the refrigerant circuits 1 and 2. The following parameters are acquired from each of the refrigerant circuits 1 and 2: Heat source unit number Subcooling degree Superheat degree Number of compressors Number of operating compressors Compressor load (rate) Discharge superheat degree Condensation temperature Evaporation temperature High pressure Low pressure Expansion valve opening Fan rotation speed Control mode Operation mode Calculated values ​​of the above parameters Next, parameters common to the refrigerant circuits 1 and 2 (only one of which is acquired for each) will be described. Outdoor air temperature Current / set value of supply water temperature Return water temperature (temperature of the liquid returning to the evaporator 14) Overall compressor 13 load rate Calculated values ​​of common parameters (calculated by combining common parameters) <Method for generating a normal amount prediction model according to the present disclosure> Next, generation of a normal amount prediction model according to the present disclosure will be described. In the present disclosure, the learning device 500 takes into consideration the operation modes of the refrigerant circuits 1 and 2 when generating a normal amount prediction model.

[0077] Fig. 12 is a diagram schematically illustrating the learning phase of the normal amount prediction model. Fig. 13 is a flowchart illustrating the process of generating a normal amount prediction model by the learning device 500. The step numbers in Fig. 12 and Fig. 13 correspond to each other.

[0078] (S11) The learning data acquisition unit 501 of the learning device 500 acquires operating data 52 of the refrigerant circuit 1 and operating data 53 of the refrigerant circuit 2 from the operating data 51. The learning data acquisition unit 501 acquires, from the operating data 52 and the operating data 53 of the refrigerant circuit 2, a) operating data 54 in which both are operating b) operating data 55 in which only one is operating Whether both or only one is operating is determined based on the operating mode and the operating data.

[0079] The learning data acquisition unit 501 associates the explanatory variables a) and b) with the objective variables.

[0080] a) Operating data in which both are operating The explanatory variables are the parameters and operating mode of refrigerant circuit 1, the parameters and operating mode of refrigerant circuit 2, and common parameters. Furthermore, the learning data acquisition unit 501 calculates a refrigerant amount index value (objective variable) from the operating data of refrigerant circuit 1. The objective variable may also be calculated from the operating data of refrigerant circuit 2.

[0081] b) Operating data when only one side is operating When only refrigerant circuit 1 is operating and when only refrigerant circuit 2 is operating, the explanatory variables in both cases are the parameters and operating mode of refrigerant circuit 1, the parameters and operating mode of refrigerant circuit 2, and common parameters (i.e., the same as a). When only refrigerant circuit 1 is operating, the objective variable is the refrigerant amount index value calculated from refrigerant circuit 1. When only refrigerant circuit 2 is operating, the objective variable is the refrigerant amount index value calculated from refrigerant circuit 2. In this way, the explanatory variables can be the same regardless of a) or b).

[0082] (S12) Next, the learning data acquisition unit 501 adds the operation mode to the explanatory variables and associates the explanatory variables with the objective variables to create learning data 56. The operation mode exists for each of the refrigerant circuits 1 and 2. The operation mode indicates, for example, whether or not the refrigerant circuits are operating, and is 1 if they are operating and 0 if they are not operating.

[0083] The operation mode may be the operation status (operating or not) transmitted from the refrigerant circuits 1 and 2, or may be determined from the operation data by the learning data acquisition unit 501. In addition, since the refrigerant circuits 1 and 2 stop when an abnormality occurs, the operation mode may be determined from the presence or absence of an alarm.

[0084] In the case of a), the learning data acquisition unit 501 associates the explanatory variable with the refrigerant amount index value 57 of the refrigerant circuit 1. In the case of b), the learning data acquisition unit 501 associates the explanatory variable 58 with the refrigerant amount index value 57 of the operating refrigerant circuit 1 or 2.

[0085] Similarly, when there are three or more refrigerant circuits, the number of parameters simply increases by the number of refrigerant circuits, and the correspondence between the explanatory variables and the response variables remains the same.

[0086] (S13) Next, the learning unit 503 of the learning device 500 generates a normal quantity prediction model 59 using a learning method such as multiple regression or neural network. For example, in the case of multiple regression, the response variable and the explanatory variable are associated as follows: y = β ０ +β １ x １ +β ２ x ２ +...+β ｎ x ｎ y is the objective variable (refrigerant amount index value 57), x １ ~x ｎ are explanatory variables 58 (operation data and operation modes in this disclosure). In the case of two refrigerant circuits, two of the explanatory variables are the operation modes of the refrigerant circuits 1 and 2. β ０ ~β ｎ is the coefficient of the explanatory variable. ０ ~β ｎ is determined.

[0087] Furthermore, the refrigerant circuit number (heat source unit number) may be used as an explanatory variable corresponding to the operation mode. The explanatory variable corresponding to the refrigerant circuit number takes a value of 1 when refrigerant circuit 1 is operating, 0 when refrigerant circuit 2 is not operating, and 1 when refrigerant circuit 2 is operating, and 0 when refrigerant circuit 2 is not operating.

[0088] As described above, a normal amount prediction model 59 that learns the correspondence between the refrigerant amount index value (objective variable) and the operating data (explanatory variables) is generated. Therefore, the normal amount prediction model 59 can output a normal amount prediction value according to the operating mode and operating data of each refrigerant circuit under normal conditions.

[0089] In the present disclosure, one normal amount prediction model 59 is generated for each of the refrigerant circuits 1 and 2, but a normal amount prediction model 59 may be generated for each of the refrigerant circuits. For example, if the refrigerant circuits 1 and 2 are operating but are controlled so that their loads are different, it is conceivable to generate two normal amount prediction models that use the refrigerant amount index values ​​of the refrigerant circuits 1 and 2 as target variables.

[0090] Furthermore, when the learning unit 503 generates a normal amount prediction model using a neural network, explanatory variables are input to the input layer, and a normal amount prediction value is output from the output layer via one or more hidden layers. In the learning phase, the difference between the refrigerant amount index value and the normal amount prediction value is fed back to the weights between layers using the backpropagation method, and optimal weights are learned.

[0091] <Determining Refrigerant Leakage Using Normal Amount Prediction Model> Next, the determination of refrigerant leakage using a normal amount prediction model (inference phase) in the present disclosure will be described with reference to Fig. 14. Fig. 14 is a diagram schematically illustrating the inference phase in which the performance determination device 400 determines refrigerant leakage using the normal amount prediction model. Fig. 15 is a flowchart illustrating the process in which the performance determination device 400 determines refrigerant leakage from the refrigerant circuit using the normal amount prediction model. Step numbers in Fig. 14 and Fig. 15 correspond to each other.

[0092] (S21) The operating data acquisition unit 401 of the performance evaluation device 400 acquires operating data 41 of the refrigerant circuits 1 and 2 from the cooling water circulation device 100. The operating data acquisition unit 401 acquires operating data 42 of the refrigerant circuit 1 from the operating data 41, and acquires operating data 43 of the refrigerant circuit 2. The operating data acquisition unit 401 acquires the operating mode directly from the refrigerant circuits 1 and 2, or determines which refrigerant circuit is operating based on the operating data and generates the operating mode. The operating mode is also added to the operating data 42 and the operating data 43.

[0093] (S22) Next, the calculation unit 402 calculates the refrigerant amount index value 44 using the operating data 42. Therefore, the refrigerant amount index value 44 is calculated from the operating data of the refrigerant circuit 1, but it may also be calculated from the operating data of the refrigerant circuit 2. When only the refrigerant circuit 1 or 2 is operating, the refrigerant amount index value 44 may be calculated from the operating data of the operating refrigerant circuit.

[0094] (S23) The inference unit 403 inputs the operating data 42, 43 into the normal amount prediction model and outputs a normal amount prediction value 45. The data input to the inference unit 403 is the same regardless of the operating modes of the refrigerant circuits 1, 2. That is, the parameters and operating mode of refrigerant circuit 1, the parameters and operating mode of refrigerant circuit 2, and common parameters are input. The operating modes are as follows: When refrigerant circuits 1 and 2 are operating, the operating mode of refrigerant circuit 1 is "1", and the operating mode of refrigerant circuit 2 is "1". When only refrigerant circuit 1 is operating, the operating mode of refrigerant circuit 1 is "1", and the operating mode of refrigerant circuit 2 is "0". When only refrigerant circuit 2 is operating, the operating mode of refrigerant circuit 1 is "0", and the operating mode of refrigerant circuit 2 is "1".

[0095] (S24) The determination unit 404 compares the refrigerant amount index value 44 with the normal amount prediction value 45 to calculate ΔRLI 46. The determination unit 404 compares ΔRLI 46 with a threshold and outputs a determination result 47 indicating whether refrigerant is leaking from refrigerant circuits 1 and 2. That is, the determination unit 404 determines whether there is a refrigerant leak based on the degree of deviation between the refrigerant amount index value 44 and the normal amount prediction value 45. Because the normal amount prediction model learns each of the following situations: when refrigerant circuits 1 and 2 are operating, when only refrigerant circuit 1 is operating, and when only refrigerant circuit 2 is operating, the determination unit 404 simply compares the refrigerant amount index value 44 with the normal amount prediction value 45. When refrigerant circuits 1 and 2 are operating, the determination unit 404 can determine whether there is a leak in refrigerant circuit 1 or 2. To determine whether there is a leak in refrigerant circuit 1 or 2, the user can manually stop only one of refrigerant circuits 1 or 2. When only the refrigerant circuit 1 is operating, the determination unit 404 can determine whether there is a leak in the refrigerant circuit 1. When only the refrigerant circuit 2 is operating, the determination unit 404 can determine whether there is a leak in the refrigerant circuit 2.

[0096] (S25) If it is determined that there is a leak, the output unit 405 displays a warning of the refrigerant leak on the display together with the distinction between the refrigerant circuits 1 and 2, or notifies the relevant parties by e-mail.

[0097] By doing this, even if there is a sudden change in operating conditions when one of the refrigerant circuits stops, the error in the predicted normal amount prediction value can be reduced, and false detection of an abnormality in the refrigerant circuit that is not stopped can be suppressed.

[0098] 16A and 16B are diagrams showing the verification results of the normal amount prediction model. FIG. 16A shows the ΔRLI of the refrigerant circuit 1 in the comparative technology. FIG. 16B shows the ΔRLI of the refrigerant circuit 1 in the present disclosure. Refrigerant circuit 2 stops at time T. As can be seen by comparing arrows 211 and 212, the ΔRLI in FIG. 16B is smaller than the ΔRLI in FIG. 16A. This indicates that the error between the refrigerant amount index value and the normal amount prediction value is smaller, and that the performance determination device 400 of the present disclosure is less likely to erroneously detect an abnormality in the refrigerant circuit 1 when the refrigerant circuit 2 stops.

[0099] <Major Effects> The performance assessment device 400 of the present disclosure inputs the operating data of each refrigerant circuit into a normal quantity prediction model that takes into account the operating status of each heat source machine, thereby reducing errors in the predicted refrigerant quantity index value and preventing false detection of an abnormality in a refrigerant circuit that is not stopped, even in the case of a sudden change in operating conditions when one of the refrigerant circuits stops.

[0100] Second Embodiment In the first embodiment, one normal amount prediction model was generated using the operation modes of multiple refrigerant circuits as separate explanatory variables. This disclosure describes a learning device 500 that generates a normal amount prediction model for each operating state of multiple refrigerant circuits, and a performance determination device 400 that determines an abnormality using the normal amount prediction model generated for each operating state of the multiple refrigerant circuits.

[0101] In the present disclosure, the circuit configurations of FIGS. 1 and 2 and the functional block diagrams of FIGS. 10A and 10B described in the first embodiment will be described as being applicable.

[0102] <Overview of Learning Phase> Fig. 17 is a diagram illustrating an overview of the learning phase in the present disclosure. First, the explanatory variables and the objective variables will be described. The explanatory variables are "parameters of all refrigerant circuits and common parameters" regardless of the number of operating refrigerant circuits. Furthermore, the objective variable is "a refrigerant amount index value of any operating refrigerant circuit" regardless of the number of operating refrigerant circuits.

[0103] In the description of FIG. 17 , it is assumed that there are n refrigerant circuits 1 to n. The learning data acquisition unit 501 extracts operating data in which the same number of operating refrigerant circuits are in operation from the operating data 61 of all refrigerant circuits 1 to n. That is, the learning data acquisition unit 501 extracts operating data 62 in which all refrigerant circuits are in operation, operating data 63 in which only one circuit is in operation, operating data 64 in which only two circuits are in operation, ..., operating data 65 in which only n-1 circuits are in operation. For example, when only one circuit is in operation, any one of the refrigerant circuits 1 to n may be in operation. Similarly, when only two circuits are in operation, any two of the refrigerant circuits 1 to n may be in operation as long as only two of the refrigerant circuits 1 to n are in operation.

[0104] The explanatory variables are parameters for each of the n refrigerant circuits and common parameters, regardless of the number of operating refrigerant circuits. The objective variable is a refrigerant amount index value calculated from any of the operating refrigerant circuits. Note that the explanatory variables may differ depending on the number of operating refrigerant circuits. In this case, for example, the explanatory variables when only two circuits are operating will be greater than the explanatory variables when only one circuit is operating by the number of parameters.

[0105] In addition, in the present disclosure, the normal amount prediction model to be used changes depending on the operating status of the refrigerant circuits 1 to n, so the operation mode is not required as an explanatory variable. As described above, the learning unit 503 performs learning 81 to 84 using learning data prepared for each number of operating refrigerant circuits, and creates normal amount prediction models 1 to n (reference numerals 66 to 69) corresponding to the number of operating refrigerant circuits.

[0106] <Inference Phase> FIG. 18 is a flowchart illustrating the inference phase for determining whether a refrigerant leaks using normal amount prediction models 1 to n.

[0107] First, the operating data acquisition unit 401 of the performance determining device 400 acquires operating data of all the refrigerant circuits 1 to n from the cooling water circulating device 100 (S31).

[0108] Next, the operation data acquisition unit 401 determines which refrigerant circuits are operating based on the operation mode or the operation data, and determines the number of refrigerant circuits that are operating (S32).

[0109] If all the refrigerant circuits are operating, the calculation unit 402 calculates the refrigerant amount index value of an arbitrary refrigerant circuit (S33).

[0110] Then, the judgment unit 404 inputs the operating data of all refrigerant circuits into the normal amount prediction model 66 generated from the operating data when all refrigerant circuits are operating, and outputs a normal amount prediction value (S34).

[0111] If only one refrigerant circuit is operating, the calculation unit 402 calculates the refrigerant amount index value of the one operating refrigerant circuit (S35).

[0112] Then, the judgment unit 404 inputs the operating data of all refrigerant circuits into the normal amount prediction model 67 generated from the operating data when only one refrigerant circuit is operating, and outputs a normal amount prediction value (S36).

[0113] When only two refrigerant circuits are operating, the calculation unit 402 calculates the refrigerant amount index value of any one of the two operating refrigerant circuits (S37).

[0114] Then, the judgment unit 404 inputs the operating data of all refrigerant circuits into the normal amount prediction model 68 generated from the operating data when only two refrigerant circuits are operating, and outputs a normal amount prediction value (S38).

[0115] When only n-1 refrigerant circuits are operating, the calculation unit 402 calculates the refrigerant amount index value of an arbitrary refrigerant circuit among the n-1 refrigerant circuits that are operating (S39).

[0116] Then, the judgment unit 404 inputs the operating data of all refrigerant circuits into a normal quantity prediction model 69 generated from the operating data when only n-1 refrigerant circuits are operating, and outputs a normal quantity prediction value (S40).

[0117] The determination unit 404 determines whether there is a leak based on the degree of deviation between the refrigerant amount index value and the predicted normal amount value (S41). Below, we will explain the cases where a deviation is determined to exist. ・When all refrigerant circuits are operating, the determination unit 404 determines that there is a leak in one or more of the refrigerant circuits 1 to n. To identify which of the refrigerant circuits 1 to n is leaking, the user can manually stop each refrigerant circuit one by one. ・When only one refrigerant circuit is operating, the determination unit 404 determines that there is a leak in the operating refrigerant circuit. ・When only two refrigerant circuits are operating, the determination unit 404 determines that there is a leak in one or both of the two operating refrigerant circuits. ・When only n-1 refrigerant circuits are operating, the determination unit 404 determines that there is a leak in one or more of the n-1 operating refrigerant circuits.

[0118] If it is determined that there is a leak, the output unit 405 displays the refrigerant leak along with the refrigerant circuit on a display, or notifies relevant parties by email (S42).

[0119] <Major Effects> The performance assessment device of the present disclosure generates a normal quantity prediction model for each number of operating refrigerant circuits, thereby preventing false detection of an abnormality when one or more refrigerant circuits stop, even though no abnormality has occurred in the refrigerant circuits that are not stopped.

[0120] [Other Application Examples] Although the best mode for carrying out the present disclosure has been described above using examples, the present disclosure is not limited to these examples in any way, and various modifications and substitutions can be made within the scope that does not deviate from the gist of the present disclosure.

[0121] For example, although the present disclosure expresses the operation mode of each refrigerant circuit as 1 (operating) or 0 (stopped), the operation mode may reflect the operating capacity of the refrigerant circuit. For example, the operation mode of a refrigerant circuit operating at 50% capacity is 0.5.

[0122] The configuration examples shown in Figures 10A and 10B are divided according to main functions to facilitate understanding of the processing by the learning device 500 and the performance assessment device 400. The technology of the present disclosure is not limited by the way in which the processing units are divided or the names of the processing units. The processing by the learning device 500 and the performance assessment device 400 can also be divided into more processing units depending on the processing content. Furthermore, the processing can also be divided so that one processing unit includes more processes.

[0123] The functions of the present disclosure described above can be realized not only by software processing through the execution of a program, but also by one or more processing circuits. Here, the term "processing circuit" in this specification includes a processor programmed to perform each function by software, such as a processor implemented by an electronic circuit, as well as devices such as an ASIC (Application Specific Integrated Circuit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), and a conventional circuit module designed to perform each of the functions described above.

[0124] <Reasons for the Effects> - A first aspect of the present disclosure relates to a cooling water circulation system having multiple connected circuits, including a control unit that inputs the operating mode and operating data of the multiple circuits into a trained model and outputs a predicted value of the performance index value. The control unit calculates an actual measurement value of the performance index value from the operating data of one of the multiple circuits and determines the performance degradation based on the discrepancy between the predicted value and the actual measurement value. Therefore, even if one refrigerant circuit stops and the load on another refrigerant circuit increases, the predicted value takes into account the refrigerant circuit stoppage, thereby preventing false detection of a refrigerant leak from another refrigerant circuit. - A second aspect of the present disclosure relates to an operating mode that is status information of the circuit output for each circuit through which refrigerant circulates. Therefore, the learning model can learn the correspondence between the status of each circuit as an explanatory variable and the performance index value, which is a target variable. - A third aspect of the present disclosure relates to a performance index value that is a refrigerant quantity index value related to the leakage of refrigerant circulating through one of the circuits. Therefore, the learning model can output a refrigerant quantity index value, preventing false detection of a refrigerant leak. - A fourth aspect of the present disclosure is a cooling water circulation device in which multiple circuits are connected, which "has a control unit that selects a model from multiple models according to the number of predetermined operating conditions, inputs the operating data of the multiple circuits into the selected model, and outputs a predicted value of the performance index value, and the control unit calculates an actual measurement value of the performance index value from the operating data of one of the circuits in the predetermined operating condition, and determines the performance degradation from the discrepancy between the predicted value and the actual measurement value," so even if one refrigerant circuit stops and the load on another refrigerant circuit increases, the predicted value takes into account the fact that the refrigerant circuit has stopped, and therefore it is possible to prevent erroneous detection of refrigerant leakage from another refrigerant circuit.

[0125] This application claims priority based on Japanese Patent Application No. 2024-021969, filed with the Japan Patent Office on February 16, 2024, the entire contents of which are incorporated herein by reference.

[0126] 400 Performance determination device 500 Learning device

Claims

1. A performance assessment device for assessing performance degradation of a cooling water circulation device, wherein pipes for circulating a liquid cooled by the refrigerant pass through one or more evaporators provided in multiple circuits through which the refrigerant of the cooling water circulation device circulates, and the device has a control unit that uses a performance index value of one of the multiple circuits as a dependent variable and the operating mode and operating data of the multiple circuits as explanatory variables, inputs the operating mode and operating data of the multiple circuits into a model that has learned the correspondence between the dependent variable and the explanatory variables, and outputs a predicted value of the performance index value, and the control unit calculates an actual measurement value of the performance index value from the operating data of one of the multiple circuits, and assesses the performance degradation from the discrepancy between the predicted value and the actual measurement value.

2. The performance determination device according to claim 1, wherein the operation mode is status information of the circuit output for each circuit through which the refrigerant circulates.

3. A performance determination device according to claim 1 or 2, wherein the performance index value is a refrigerant amount index value relating to leakage of refrigerant circulating through one of the circuits.

4. A performance assessment device for assessing performance degradation of a cooling water circulation device, wherein pipes for circulating a liquid cooled by the refrigerant pass through one or more evaporators provided in multiple circuits through which the refrigerant of the cooling water circulation device circulates, and a model is generated for each number of circuits in a predetermined operating state among the multiple circuits, the model being generated from multiple models that have learned the correspondence between the objective variable and the explanatory variables, with the performance index value of one of the circuits in the predetermined operating state as an objective variable and the operating data of the multiple circuits as explanatory variables, and the device has a control unit that selects a model according to the number of the predetermined operating states from multiple models that have learned the correspondence between the objective variable and the explanatory variables, inputs the operating data of the multiple circuits into the selected model, and outputs a predicted value of the performance index value, and the control unit calculates an actual measurement value of the performance index value from the operating data of one of the circuits in the predetermined operating state, and assesses the performance degradation from the discrepancy between the predicted value and the actual measurement value.

5. A performance assessment method performed by a performance assessment device that assesses performance degradation of a cooling water circulation device, wherein pipes through which liquid cooled by the refrigerant circulates pass through one or more evaporators provided in multiple circuits through which the refrigerant of the cooling water circulation device circulates, the method comprising the steps of: a control unit inputting the performance index value of one of the multiple circuits as an objective variable and the operating mode and operating data of the multiple circuits as explanatory variables into a model that has learned the correspondence between the objective variable and the explanatory variables, and outputting a predicted value of the performance index value; a control unit calculating an actual measurement value of the performance index value from the operating data of one of the multiple circuits; and assessing the performance degradation from the discrepancy between the predicted value and the actual measurement value.

6. A cooling water circulation system equipped with a performance determination device that determines performance degradation of the cooling water circulation system, wherein piping through which liquid cooled by the refrigerant circulates passes through one or more evaporators provided in multiple circuits through which the refrigerant of the cooling water circulation system circulates, the performance determination device has a control unit that uses the performance index value of one of the multiple circuits as an objective variable and the operating mode and operating data of the multiple circuits as explanatory variables, inputs the operating mode and operating data of the multiple circuits into a model that has learned the correspondence between the objective variable and the explanatory variables, and outputs a predicted value of the performance index value, the control unit calculates an actual measurement value of the performance index value from the operating data of one of the multiple circuits, and determines the performance degradation from the discrepancy between the predicted value and the actual measurement value.

Images