Heat source machine system, trained model generation method and trained model

Abstract

To provide a heat source machine system, a generation method of a learned model, and the learned model for performing an operation with excellent responsiveness in a proper operating state according to change of an operating condition.SOLUTION: A heat source machine system 1 has: heat source devices 11-13 for cooling or heating a heat medium CH to be supplied to heat demand equipment 99; heat source auxiliary devices 21-23, 31-33, 41-43; and a control device 70 for adjusting operating states of the heat source devices and the heat source auxiliary devices. The control device 70 has a learned control model 80. On the control model 80, machine learning processing using teacher data is performed to output the operating state to keep a prescribed index at a value suited to a condition when the operating condition is input. The control device 70 controls the heat source devices and the heat source auxiliary devices to have the operating state output by the control model 80. A generation method of the learned model includes: generation of the teacher data by simulation; and generation of the learned model by performing the machine learning processing with the teacher data.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a heat source machine system, a method for generating a trained model, and a trained model, and in particular to a heat source machine system that controls equipment using a machine-learned model, a method for generating a trained model, and a trained model. [Background technology]

[0002] In order to supply air conditioners with chilled or hot water for cooling or heating the air, heat source systems are generally constructed by appropriately combining heat source devices and their auxiliary equipment. The heat source system adjusts the output of the equipment that makes up the heat source system and the operating state of the number of operating units according to the load on the air conditioners. Devices that can be useful for controlling the equipment that makes up the heat source system include a reproduction system that calls up and reproduces actual operating data, and a system that can perform simulations using past heat load data and facility environment data (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2011-163727 Summary of the Invention [Problem to be solved by the invention]

[0004] In order to control the devices that make up the heat source system with good response, it is preferable to shorten the calculation time required to determine an appropriate operating state.

[0005] In view of the above-mentioned problems, the present disclosure relates to providing a heat source machine system that operates responsively in an appropriate operating state in response to changes in operating conditions, a method for generating a trained model, and a trained model. [Means for solving the problem]

[0006] A heat source equipment system according to a first aspect of the present disclosure includes a heat source equipment that cools or heats a heat medium to be supplied to heat demand equipment, a heat source auxiliary equipment that operates in conjunction with the operation of the heat source equipment, and a control device that adjusts the operating states of the heat source equipment and the heat source auxiliary equipment, the control device having a trained control model, wherein the heat source auxiliary equipment includes a heat medium pump that causes the heat medium to flow through the heat source equipment, and a heat source fluid supply device that supplies the heat source equipment with a heat source fluid that exchanges heat directly or indirectly with the heat medium in the heat source equipment, and the operating state is determined by at least one of an operating status of the heat source equipment, a flow rate of the heat medium discharged by the heat medium pump, and a flow rate of the heat source fluid supplied by the heat source fluid supply device. and the control model has undergone machine learning processing using teacher data so that, when an operating condition that affects the operating state is input, the control model outputs the operating state in which a predetermined index becomes a value that satisfies the condition, the operating condition includes at least one of the heat demand of the heat demand facility or a physical quantity correlated thereto, and the outdoor air temperature or a physical quantity correlated thereto, and the predetermined index includes at least one of the power consumption of the heat source equipment and the heat source auxiliary equipment, the operating cost of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emission amount of the heat source equipment and the heat source auxiliary equipment, and the control device controls the heat source equipment and the heat source auxiliary equipment so as to achieve the operating state output by the control model.

[0007] With this configuration, the control model can output an appropriate operating state in response to changes in operating conditions, so the heat source system can be operated with good response.

[0008] Furthermore, a heat source machine system according to a second aspect of the present disclosure is a heat source machine system according to the first aspect of the present disclosure, wherein the specified indicators include multiple indicators among the power consumption of the heat source equipment and the heat source auxiliary equipment, the operating costs of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emissions of the heat source equipment and the heat source auxiliary equipment, and the control model is configured to select multiple operating states in which a first specified indicator among the multiple specified indicators takes a value that meets a condition, and output the operating state from the selected multiple operating states in which a second specified indicator different from the first specified indicator takes a value that meets the condition.

[0009] With this configuration, the heat source machine system can be operated in an appropriate operating state determined by comprehensively determining a plurality of predetermined indexes.

[0010] Furthermore, a heat source machine system according to a third aspect of the present disclosure is the heat source machine system according to the first or second aspect of the present disclosure, wherein at least one of the heat source equipment, the heat medium pump, and the heat source fluid supply device is composed of a plurality of units, the operating state includes the number of operating units of the plurality of units of the heat source equipment, the heat medium pump, and the heat source fluid supply device, and the control model includes a first control model that outputs the number of operating units as an integer value among the operating states to be output, and a second control model that inputs the integer value of the number of operating units output by the first control model as one of the operating conditions.

[0011] By configuring in this manner, it is possible to prevent the output of the control model for the number of operating units from becoming an unrealizable decimal number, and it is possible to operate the heat source machine system in an appropriate operating state that conforms to actual operation.

[0012] Furthermore, a heat source machine system according to a fourth aspect of the present disclosure is a heat source machine system according to any one of the first to third aspects of the present disclosure, wherein the control device uses the output of the control model for some of the operating states, including the heat processing amount of the heat source equipment, the flow rate of the heat medium discharged by the heat medium pump, and the flow rate of the heat source fluid supplied by the heat source fluid supply device, and determines the remaining operating states by simulation or rule-based determination.

[0013] With this configuration, the time required to output part of the operating state can be reduced by using the control model, while the remaining operating state can be output with higher accuracy by using the simulation or rule base.

[0014] Furthermore, a heat source machine system according to a fifth aspect of the present disclosure is a heat source machine system according to any one of the first to fourth aspects of the present disclosure, wherein the operating conditions include a pressure loss coefficient of the heat demand equipment.

[0015] With this configuration, the heat source system can be operated in an appropriate operating state even if the flow rate of the heat medium changes.

[0016] Furthermore, a heat source equipment system according to a sixth aspect of the present disclosure is a heat source equipment system according to any one of the first to fifth aspects of the present disclosure, wherein the operating conditions include at least one of the cost per unit power consumption of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emissions per unit power consumption of the heat source equipment and the heat source auxiliary equipment.

[0017] With this configuration, it is not necessary to recreate the control model even if the cost per unit of power consumption or the amount of carbon dioxide emissions changes.

[0018] Furthermore, a heat source machine system according to a seventh aspect of the present disclosure is a heat source machine system according to any one of the first to sixth aspects of the present disclosure, wherein the control model has a plurality of control models in which the specified indicator is one or more of the power consumption of the heat source equipment and the heat source auxiliary equipment, the operating cost of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emission amount of the heat source equipment and the heat source auxiliary equipment, and the control device uses an appropriate control model from the plurality of control models depending on the specified indicator desired by the user.

[0019] With this configuration, even if the predetermined index is changed by the user, the heat source machine system can be maintained in an appropriate operating state.

[0020] Furthermore, a method for generating a trained model according to an eighth aspect of the present disclosure is a method for generating a trained model used to control a heat source machine system including a heat source machine that cools or heats a heat medium to be supplied to heat demand equipment and a heat source auxiliary machine that operates in conjunction with the operation of the heat source machine, the method including: obtaining, by simulation, a plurality of predetermined indicators for an assumed operating state of the heat source machine system under assumed operating conditions while changing the operating state; defining the operating state when the predetermined indicator satisfies a condition as a relationship with the operating condition; and performing this for a plurality of operating conditions to obtain a plurality of pairs of relationships between the operating state when the predetermined indicator satisfies the condition and the operating condition, thereby generating training data; and performing machine learning processing using the training data to input the operating conditions and output the operating state. and generating a trained model for generating a trained model based on the trained model, wherein the heat source auxiliary equipment includes a heat medium pump that circulates the heat medium passing through the heat source equipment, and a heat source fluid supply device that supplies the heat source equipment with a heat source fluid that exchanges heat directly or indirectly with the heat medium in the heat source equipment, wherein the operating conditions include at least one of the heat demand of the heat demand facility or a physical quantity correlated thereto, and an outdoor air temperature or a physical quantity correlated thereto, the operating state includes at least one of the operating status of the heat source equipment, the flow rate of the heat medium discharged by the heat medium pump, and the flow rate of the heat source fluid supplied by the heat source fluid supply device, and the predetermined indicators include at least one of the power consumption of the heat source equipment and the heat source auxiliary equipment, the operating costs of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emissions of the heat source equipment and the heat source auxiliary equipment.

[0021] By configuring in this manner, a trained model can be obtained that can operate the heat source machine system in a responsive manner in an appropriate operating state in response to changes in operating conditions.

[0022] Furthermore, a method for generating a trained model according to a ninth aspect of the present disclosure is the method for generating a trained model according to the eighth aspect of the present disclosure, wherein at least one of the heat source equipment, the heat medium pump, and the heat source fluid supply device is composed of a plurality of units, the operating state includes the number of operating units of the plurality of units among the heat source equipment, the heat medium pump, and the heat source fluid supply device, and the step of generating the trained model includes the step of generating a first trained model including an integer value of the number of operating units in the operating state of output, and the step of generating a second trained model that inputs the integer value of the number of operating units as one of the operating conditions.

[0023] With this configuration, it is possible to prevent the output of the control model for the number of operating vehicles from becoming an unrealizable decimal number.

[0024] Furthermore, a method for generating a trained model according to a tenth aspect of the present disclosure is the method for generating a trained model according to the ninth aspect of the present disclosure, wherein the first trained model and the second trained model use neural networks, and the first trained model and the second trained model have a common intermediate layer, the driving state items that are the output of the first trained model include the driving state items that are the output of the second trained model, and the step of generating the trained model first performs machine learning processing on the first trained model to generate the first trained model, and then sets the initial values ​​of the weight coefficients of the intermediate layer of the second trained model to the weight coefficients of the intermediate layer of the first trained model, and performs machine learning processing on the resultant model to generate the second trained model.

[0025] This configuration makes it possible to shorten the learning time required to generate the second trained model.

[0026] Furthermore, a method for generating a trained model according to an eleventh aspect of the present disclosure is a method for generating a trained model according to any one of the eighth to tenth aspects of the present disclosure, wherein the step of generating the training data randomly determines at least one of the operating conditions and the operating state when performing a simulation.

[0027] This configuration makes it possible to prevent the amount of data to be handled from increasing more than necessary, and also makes it relatively easy to create additional training data.

[0028] Furthermore, a method for generating a trained model according to a twelfth aspect of the present disclosure is a method for generating a trained model according to any one of the eighth to eleventh aspects of the present disclosure, wherein the step of generating the teacher data includes, when data of the specified index for a driving state in a wider range than the driving state expected under a wider range of driving conditions than the expected driving conditions already exists, extracting the specified index for the driving state expected under the expected driving conditions from the already existing data and using it as the teacher data.

[0029] With this configuration, it is possible to create training data using existing data, thereby reducing the time required to create training data.

[0030] In addition, a method for generating a trained model according to a thirteenth aspect of the present disclosure is the method for generating a trained model according to the twelfth aspect of the present disclosure, wherein the step of generating the training data involves changing the values ​​of items in the unextracted data that do not match the expected driving conditions and driving states to matching values, and then performing a simulation to generate the training data.

[0031] This configuration allows efficient replenishment of insufficient training data.

[0032] Furthermore, a method for generating a trained model according to a fourteenth aspect of the present disclosure is a method for generating a trained model according to any one of the eighth to thirteenth aspects of the present disclosure, wherein the operating conditions include a pressure loss coefficient of the heat demand equipment, and the step of generating the training data involves calculating the pressure loss in the heat demand equipment based on the pressure loss coefficient and the flow rate of the heat medium in the simulation, and then obtaining the specified index.

[0033] By configuring in this manner, it is possible to generate a trained model that can output an appropriate operating state even when the flow rate of the heat transfer medium changes.

[0034] Furthermore, a trained model according to a fifteenth aspect of the present disclosure is a trained model to be installed in a computer used for controlling a heat source equipment system including a heat source equipment that cools or heats a heat medium to be supplied to heat demand equipment, and a heat source auxiliary equipment that operates in conjunction with the operation of the heat source equipment, the trained model comprising: an input layer to which operating conditions of the heat source equipment system are input; an output layer to which the operating state of the heat source equipment system is output; and an intermediate layer in which parameters are trained using teacher data that inputs the operating conditions and outputs the operating state in which a predetermined index has a value that meets the condition, and the heat source auxiliary equipment comprises a heat medium pump that circulates the heat medium passing through the heat source equipment, and a heat source fluid that supplies the heat source equipment with a heat source fluid that directly or indirectly exchanges heat with the heat medium in the heat source equipment. and a heat source fluid supply device that supplies heat to the heat demanding equipment, the operating conditions including at least one of the heat demand of the heat demanding equipment or a physical quantity correlated thereto, and the outdoor air temperature or a physical quantity correlated thereto, the operating state including at least one of the operating status of the heat source equipment, the flow rate of the heat medium discharged by the heat medium pump, and the flow rate of the heat source fluid supplied by the heat source fluid supply device, and the predetermined index including at least one of the power consumption of the heat source equipment and the heat source auxiliary equipment, the operating cost of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emission amount of the heat source equipment and the heat source auxiliary equipment, and the computer is caused to function so that the operating conditions are input to the input layer, calculated in the intermediate layer, and the operating state is output from the output layer.

[0035] When configured in this manner, the heat source machine system becomes a trained model that can responsively operate in an appropriate operating state in response to changes in operating conditions.

[0036] In addition, a heat source equipment system according to a sixteenth aspect of the present disclosure includes a control device having a trained model according to the fifteenth aspect of the present disclosure, the heat source equipment, and the heat source auxiliary equipment, and the control device controls the heat source equipment and the heat source auxiliary equipment so that they reach the operating state output by the trained model.

[0037] With this configuration, the heat source machine system can be operated in an appropriate operating state with good response in response to changes in operating conditions. [Effects of the Invention]

[0038] According to the present disclosure, a heat source machine system can be operated in an appropriate operating state with good responsiveness in response to changes in operating conditions. [Brief explanation of the drawings]

[0039] [Figure 1] 1 is a schematic system diagram of a heat source machine system according to one embodiment. [Figure 2] 10 is a flowchart illustrating a procedure for creating training data used in machine learning processing of a control model provided in a heat source machine system according to one embodiment. [Figure 3] FIG. 10 is a diagram illustrating an example of operating conditions assumed when creating training data. [Figure 4] FIG. 10 is a diagram illustrating an example of an operating state assumed when creating training data. [Figure 5] 10 is a flowchart illustrating a procedure for creating training data by reusing some existing data. [Figure 6] FIG. 2 is a schematic configuration diagram of a first control model. [Figure 7] FIG. 10 is a schematic configuration diagram of a second control model. [Figure 8]FIG. 2 is a block diagram showing a calculation flow of a control device of a heat source machine system according to one embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0040] Hereinafter, an embodiment will be described with reference to the drawings. In the drawings, identical or similar reference numerals are used to designate identical or corresponding components, and redundant explanations will be omitted.

[0041] <Example of heat source system configuration> First, a heat source machine system 1 according to one embodiment will be described with reference to FIG. 1. FIG. 1 is a schematic system diagram of the heat source machine system 1. The heat source machine system 1 mainly comprises three heat source machines 11, 12, and 13, three chilled / hot water pumps 21, 22, and 23, three cooling towers 31, 32, and 33, three cooling water pumps 41, 42, and 43, and a control device 70. The heat source machine system 1 is a system that supplies chilled / hot water CH cooled or heated by the heat source machines 11, 12, and 13 to heat demand equipment 99 in response to demand. Examples of the heat demand equipment 99 include air conditioning equipment such as an air handling unit and a fan coil unit. The chilled / hot water CH is a medium that transports cold or hot heat to the heat demand equipment 99 and corresponds to a heat medium. Chilled / hot water CH is a general term for cold water, which is a medium for cold heat, and hot water, which is a medium for heat, and typically becomes cold water when cooling is performed in the heat demand equipment 99, and becomes hot water when heating is performed in the heat demand equipment 99.

[0042] The three heat source units 11, 12, and 13 are devices that cool or heat the chilled or hot water CH and correspond to heat source devices. Hereinafter, to distinguish between the three heat source units, they may be referred to as the first heat source unit 11, the second heat source unit 12, and the third heat source unit 13, respectively. Although various types of devices can be used for the heat source units 11, 12, and 13, this embodiment will be described assuming that the first heat source unit 11 and the second heat source unit 12 are both variable-speed turbo chillers with the same characteristics, and the third heat source unit 13 is a fixed-speed turbo chiller. Each of the heat source units 11, 12, and 13 is a device that can cool or heat the chilled or hot water CH using input power. The amount of heat that each of the heat source units 11, 12, and 13 removes from the chilled or hot water CH or the amount of heat that each of them provides to the chilled or hot water CH will be referred to as the "processed heat amount." The first heat source unit 11 and the second heat source unit 12 are configured to adjust the processed heat amount by capacity control. The third heat source unit 13 operates by switching between an operating state and a stopped state, and the amount of heat processed is either rated or zero. The number of operating heat source units 11, 12, and 13 and the operating capacity of the first heat source unit 11 and the second heat source unit 12 are referred to as the "operating status." The amount of heat processed by each heat source unit 11, 12, and 13 is determined depending on the operating status. Each heat source unit 11, 12, and 13 is capable of detecting the temperature difference and pressure difference between the inflowing and outflowing chilled / hot water CH. Each heat source unit 11, 12, and 13 is also configured to provide the cooling water CD with heat removed from the chilled / hot water CH to cool the chilled / hot water CH, and to remove heat from the cooling water CD to heat the chilled / hot water CH. The cooling water CD is a fluid that indirectly exchanges heat with the chilled / hot water CH in each heat source unit 11, 12, and 13 via a refrigerant (not shown), and corresponds to a heat source fluid.

[0043] The three chilled / hot water pumps 21, 22, and 23 are devices that operate in conjunction with the operation of the heat source units 11, 12, and 13, and are a form of heat source auxiliary equipment. They are also devices that circulate chilled / hot water CH and correspond to heat medium pumps. Hereinafter, to distinguish between the three chilled / hot water pumps, they may be referred to as the first chilled / hot water pump 21, the second chilled / hot water pump 22, and the third chilled / hot water pump 23, respectively. Each of the chilled / hot water pumps 21, 22, and 23 is a device that can circulate chilled / hot water CH using input power. In this embodiment, each of the chilled / hot water pumps 21, 22, and 23 has an inverter and is configured to be able to continuously change the flow rate of the chilled / hot water CH that it discharges.

[0044] The three cooling towers 31, 32, and 33 are devices that operate in conjunction with the operation of the heat source units 11, 12, and 13, and are a form of heat source auxiliary equipment. They also supply cooling water CD to the heat source units 11, 12, and 13, and correspond to heat source fluid supply devices. Hereinafter, to distinguish between the three cooling towers, they may be referred to as the first cooling tower 31, the second cooling tower 32, and the third cooling tower 33, respectively. Each of the cooling towers 31, 32, and 33 is a device that uses input power to perform heat exchange between the cooling water CD and the atmosphere (outside air). In this embodiment, each of the cooling towers 31, 32, and 33 is operated / stopped in accordance with the operation / stop of the heat source units 11, 12, and 13, but may also be configured to allow the rotation speed of the fan to be adjusted continuously or in stages.

[0045] The three cooling water pumps 41, 42, and 43 are devices that operate in conjunction with the operation of the heat source units 11, 12, and 13 and are a form of heat source auxiliary equipment, and are devices that supply cooling water CD to the heat source units 11, 12, and 13 and are a form of heat source fluid supply device. Hereinafter, to distinguish between the three cooling water pumps, they may be referred to as the first cooling water pump 41, the second cooling water pump 42, and the third cooling water pump 43, respectively. Each cooling water pump 41, 42, and 43 is a device that can circulate the cooling water CD using input power. In this embodiment, each cooling water pump 41, 42, and 43 has an inverter and is configured to be able to continuously change the flow rate of the discharged cooling water CD.

[0046] The first heat source unit 11 is connected to the supply header 19 via a first chilled / hot water supply pipe 15, and is connected to the return header 29 via a first chilled / hot water return pipe 25. A first chilled / hot water pump 21 is provided in the first chilled / hot water return pipe 25. The first heat source unit 11 is also connected to the first cooling tower 31 via a first cooling water supply pipe 35 and a first cooling water return pipe 45. A first cooling water pump 41 is provided in the first cooling water supply pipe 35. The second heat source unit 12 is connected to the supply header 19 via a second chilled / hot water supply pipe 16, and is also connected to the return header 29 via a second chilled / hot water return pipe 26. A second chilled / hot water pump 22 is provided in the second chilled / hot water return pipe 26. The second heat source unit 12 is also connected to the second cooling tower 32 via a second cooling water supply pipe 36 and a second cooling water return pipe 46. A second cooling water pump 42 is provided in the second cooling water supply pipe 36. The third heat source unit 13 is connected to the supply header 19 via a third chilled / hot water supply pipe 17, and is connected to the return header 29 via a third chilled / hot water return pipe 27. A third chilled / hot water pump 23 is provided in the third chilled / hot water return pipe 27. The third heat source unit 13 is also connected to the third cooling tower 33 via a third cooling water supply pipe 37 and a third cooling water return pipe 47. A third cooling water pump 43 is provided in the third cooling water supply pipe 37.

[0047] The supply header 19 is a component that collects chilled or hot water CH that has been cooled or heated in each of the heat source units 11, 12, and 13. The supply header 19 is connected to a heat demand facility 99 via a supply pipe 91, and is capable of directing the chilled or hot water CH collected in the supply header 19 to the heat demand facility 99. The heat demand facility 99 and the return header 29 are connected via a recovery pipe 92. The return header 29 is capable of receiving the chilled or hot water CH, whose heat has been utilized in the heat demand facility 99, via the recovery pipe 92. The chilled or hot water CH that has flowed into the return header 29 is distributed to each of the heat source units 11, 12, and 13 depending on the operating status of each of the heat source units 11, 12, and 13. In this way, the return header 29 is a component that distributes the chilled or hot water CH, after its cold or hot energy has been utilized, to each of the heat source units 11, 12, and 13.

[0048] The control device 70 is a device that controls the operation of the heat source machine system 1. The control device 70 is connected to each of the heat source machines 11, 12, and 13 via communication lines. The control device 70 is configured to receive operation information data from each of the heat source machines 11, 12, and 13. Examples of receivable operation data include all or part of the outlet and inlet temperatures and pressures of the chilled and hot water CH, the outlet and inlet temperatures and pressures of the cooling water CD, and power consumption. In this embodiment, the flow rate can be calculated from the difference in inlet and outlet pressures of the fluid, but the flow rate may also be detected directly. The control device 70 is also configured to send control signals to each of the heat source machines 11, 12, and 13 to adjust the operating status of each of the heat source machines 11, 12, and 13. The control device 70 is also configured to control the operation of the heat source auxiliary machines via each of the heat source machines 11, 12, and 13 and the auxiliary machine power panel 75. The auxiliary machine power panel 75 is a device that controls the operation of the heat source auxiliary machines. The auxiliary power panel 75 is connected to each heat source auxiliary unit by a signal line. The auxiliary power panel 75 is configured to adjust the power supplied to each chilled / hot water pump 21, 22, 23 based on control signals received from each heat source unit 11, 12, 13, thereby controlling start / stop and the flow rate of chilled / hot water CH discharged. The auxiliary power panel 75 is also configured to adjust the power supplied to each cooling tower 31, 32, 33, thereby controlling start / stop, based on control signals received from each heat source unit 11, 12, 13. The auxiliary power panel 75 is also configured to adjust the power supplied to each cooling water pump 41, 42, 43, based on control signals received from each heat source unit 11, 12, 13, thereby controlling start / stop and the flow rate of chilled water CD discharged.

[0049] The control device 70 is further configured to receive a temperature information signal from an outdoor air thermometer 61 that detects the outdoor air temperature, thereby determining the outdoor air temperature. The outdoor air thermometer 61 is preferably arranged to detect the temperature of the outdoor air surrounding each of the cooling towers 31, 32, and 33. The control device 70 is also configured to receive information on the amount of heat demanded by the heat demand facility 99 as a signal, thereby determining the amount of heat demanded by the heat demand facility 99. The control device 70 also has a control model 80. The control model 80 is a model that has undergone machine learning processing in a computer and into which operating conditions are input in order to output an operating state. Here, the operating state refers to the operating state of the heat source equipment and heat source auxiliary equipment that can be controlled by the heat source equipment system 1. Specific examples include the on / off state (in other words, the number of operating units) and the amount of heat processed (in other words, output) of each of the heat source units 11, 12, and 13. Other examples of operating states include the discharge flow rate of chilled / hot water CH from each chilled / hot water pump 21, 22, 23 (a flow rate of 0 indicates a stopped state), the start / stop of each cooling tower 31, 32, 33, and the discharge flow rate of cooling water CD from each cooling water pump 41, 42, 43 (a flow rate of 0 indicates a stopped state). On the other hand, operating conditions are conditions that affect the operating state and are basically conditions that cannot be controlled by the heat source equipment system 1. Examples of operating conditions include the heat demand of the heat demand equipment 99, the outside air temperature, the target temperature of the chilled / hot water CH flowing from the forward header 19 to the heat demand equipment 99, and the pressure loss coefficient on the heat demand equipment 99 side. Operating conditions may also include the unit prices of resources (e.g., electricity, water, fuel, etc.) consumed during operation of the heat source equipment system 1, the amount of carbon dioxide emissions per unit consumption of these resources, etc. The sum of the unit prices of the various consumable resources used to power the equipment that constitutes the heat source equipment system 1 corresponds to the cost per unit of consumed power. Furthermore, the total amount of carbon dioxide emissions per unit consumption of the various consumable resources used to power the devices that make up the heat source machine system 1 corresponds to the amount of carbon dioxide emissions per unit of consumed power.

[0050] Incidentally, including the pressure loss coefficient on the heat demand facility 99 side in the operating conditions has the following advantages. First, as a premise, the pressure loss on the heat demand facility 99 side varies depending on the total flow rate of chilled / hot water CH, which varies depending on the number of operating air conditioners that make up the heat demand facility 99. This pressure loss is proportional to the square of the total flow rate of chilled / hot water CH. In this embodiment, this proportionality constant (coefficient) is referred to as the "pressure loss coefficient." As a result, the pressure loss on the heat demand facility 99 side can be calculated by multiplying the pressure loss coefficient by the square of the total flow rate of chilled / hot water CH. Then, the power of each chilled / hot water pump 21, 22, 23 can be calculated based on the calculated pressure loss on the heat demand facility 99 side. Therefore, even if the operating state changes and the flow rate of chilled / hot water CH on the heat demand facility 99 side changes, the power of each chilled / hot water pump 21, 22, 23 can be accurately estimated. The reason why the control device 70 has the control model 80 is as follows.

[0051] Conventional system control techniques involve performing simulations under various assumed operating conditions for the actual operating conditions of the heat source machine, determining the optimal operating conditions, and then operating the machine accordingly. However, as the number of controllable parameters increases, the computational load increases, making it increasingly difficult to determine the optimal operating conditions under all possible operating conditions. For this reason, practical solutions include narrowing the number of variable parameters, reducing the amount of computation by making various assumptions, or lengthening the computation cycle. With such limited control, it is difficult to operate the machine under truly optimal operating conditions. Therefore, the heat source machine system 1 according to this embodiment utilizes a control model 80 that has undergone machine learning processing, as described below, to achieve the optimal operating conditions as much as possible. The procedure for generating the control model 80 is outlined as follows: first, training data is created, and then machine learning processing is performed on a computer using the created training data.

[0052] <Creating training data> FIG. 2 is a flowchart showing the procedure for creating training data. In the following description, when referring to the configuration of the heat source equipment system 1, reference will be made to FIG. 1 as appropriate. The heat source equipment 11, 12, and 13 may be collectively referred to as heat source equipment. The hot and cold water pumps 21, 22, and 23, the cooling towers 31, 32, and 33, and the cooling water pumps 41, 42, and 43 may be collectively referred to as heat source auxiliary equipment. In this embodiment, a simulation is used to create training data. The simulation in this embodiment is typically performed at a location different from the installation location of the heat source equipment system 1. The simulation in this embodiment is used to estimate predetermined indicators under given operating conditions and assumed operating states. Here, the predetermined indicators are criteria that a user of the heat source equipment system 1 focuses on when operating the heat source equipment system 1. Examples of the predetermined indicators include the power consumption, operating costs, and carbon dioxide emissions of the heat source equipment and heat source auxiliary equipment when the heat source equipment system 1 is operated. Simulation calculations for the heat source equipment system 1 generally require recursive calculations, which take a certain amount of time. Conventional heat source equipment systems use simulation results directly to control the equipment, but due to issues such as installation space, power consumption, noise environment, and cost, it is difficult to adopt high-performance calculation devices, and calculation times tend to be long. However, the calculation device used for the simulation to create the training data shown in Figure 2 does not need to be installed at the site where the heat source equipment system 1 is constructed, so a high-performance calculation device can be used. Another advantage is that sufficient time is available for simulation calculations, allowing for the possibility of assuming a wide range of operating conditions to find the optimal operating condition.

[0053] As shown in Figure 2, when creating training data, first, fixed conditions are prepared (S1). Fixed conditions are unchanging elements such as the energy consumption characteristics and pressure loss coefficient (different from the pressure loss coefficient of the heat demand equipment 99) of each heat source equipment and heat source auxiliary equipment, and are specific to the heat source equipment and heat source auxiliary equipment. Fixed conditions can generally be prepared using characteristic data for each equipment and fluid equations (e.g., the relationship between flow rate and pressure loss). Once the fixed conditions are prepared, operating conditions are assumed (S2). Operating conditions can be assumed within the expected range of application. The expected range of application (expected operating conditions) is intended to exclude conditions such as an outside air temperature of 70°C, which is unlikely to occur in reality, for example. Specific examples of assumed operating conditions include the heat demand of the heat demand facility 99 being 6000 kW, the outdoor air temperature being 25°C, the target temperature of the discharged chilled / hot water CH being 7°C, and the pressure loss coefficient on the heat demand facility 99 side being 200 kPa / 1000 LPM, as shown in operating condition "1" in Fig. 3. In addition, specific examples of assumed operating conditions include an electricity rate of 15 yen / kWh, a water rate of 150 yen / m 3 Although not shown in the figure, the electricity carbon dioxide emission coefficient and the water carbon dioxide emission coefficient may also be included.

[0054] Once the operating conditions have been assumed, the operating state is assumed (S3). The operating state can also be assumed within the range of expected application. The range of expected application (expected operating state) is intended to exclude, for example, an unrealizable state such as operating five heat source units when there are three. A specific example of an assumed operating state is shown in operating state "1" in FIG. 4, where all three heat source units 11, 12, and 13 are operating and the flow rate of chilled or hot water CH from the three chilled or hot water pumps 21, 22, and 23 is 100%. In operating state "1," it is also assumed that all three cooling towers 31, 32, and 33 are operating and the flow rate of cooling water CD from the three cooling water pumps 41, 42, and 43 is 100%. Once the operating state has been assumed, predetermined indicators under the assumed operating conditions and the assumed operating state are calculated by simulation (S4). The calculation of these predetermined indicators can be performed, for example, as follows.

[0055] First, the total flow rate of chilled / hot water CH can be calculated from the number of operating heat source units 11, 12, and 13 under assumed operating conditions and the flow rate of chilled / hot water CH from each chilled / hot water pump 21, 22, and 23. Then, the return temperature of the chilled / hot water CH (the temperature of the chilled / hot water CH entering the heat source unit) can be calculated from the total flow rate of the chilled / hot water CH and the target temperature and heat demand of the chilled / hot water CH under assumed operating conditions. Next, the load heat quantity of each heat source unit 11, 12, and 13 can be calculated from the temperature difference between the inlet and outlet of the chilled / hot water CH for each heat source unit 11, 12, and 13 and the flow rate of the chilled / hot water CH. Then, the efficiency (COP) of each heat source unit 11, 12, and 13 is assumed to calculate the amount of heat to be processed. Then, the temperature of the chilled water CD entering each heat source unit 11, 12, and 13 can be calculated from the amount of heat to be processed, the flow rate of the chilled water CD in each chilled water pump 41, 42, and 43, and the outside air temperature. The efficiency of each heat source unit 11, 12, 13 is calculated from the inlet temperature of the cooling water CD, the flow rate of the cooling water CD, and the conditions on the chilled / hot water CH side. Here, iterative calculations (convergence calculations) are performed for each heat source unit 11, 12, 13 so that the calculated efficiency matches the efficiency previously assumed. This determines the efficiency of each heat source unit 11, 12, 13, and thereby makes it possible to determine the power consumption and makeup water amount of each heat source unit 11, 12, 13.

[0056] Regarding the power consumption of the heat source auxiliary equipment, for example, the power consumption of each chilled / hot water pump 21, 22, 23 can be calculated as follows. First, the pressure loss on the heat demand equipment 99 side is calculated from the flow rate of the chilled / hot water CH and the pressure loss coefficient on the heat demand equipment 99 side. Then, the pressure loss of each heat source equipment 11, 12, 13 is calculated from the pressure loss data of each heat source equipment 11, 12, 13 defined as a fixed condition. The required head (pressure) of each chilled / hot water pump 21, 22, 23 is calculated by adding up both pressure losses. The power consumption of each chilled / hot water pump 21, 22, 23 can be estimated from this and the flow rate of the chilled / hot water CH. Similarly, the power consumption of the other heat source auxiliary equipment can be estimated and added up to calculate the power consumption of the entire heat source equipment system 1. Furthermore, the water consumption can be calculated from the amount of evaporated water, which is calculated by dividing the heat amount processed in each heat source equipment 11, 12, 13 by the latent heat of evaporation of water, and the target concentration ratio. Alternatively, as is commonly done, water consumption can be calculated by multiplying the flow rate of cooling water CD by a certain coefficient (about 5%). If power consumption and water consumption can be calculated in this way, operating costs can be calculated by multiplying these by the cost per unit quantity and adding them up, and carbon dioxide emissions can be calculated by multiplying these by the carbon dioxide emission coefficient and adding them up.

[0057] After calculating the predetermined indexes for the assumed operating conditions and the assumed operating states as described above, it is determined whether the calculated number of predetermined indexes for the operating conditions is sufficient (S5). Here, "whether the calculated number of predetermined indexes is sufficient" refers to whether there are sufficient predetermined indexes to determine the desired operating state, which is determined by comparing and considering multiple predetermined indexes to determine the operating state under the operating conditions. This means that the above-described simulation is performed multiple times under the operating conditions to calculate the predetermined indexes by changing the operating state. In the description of the procedure for creating training data up to this point, the predetermined indexes are calculated for only one operating state under the operating conditions, so the calculated number of predetermined indexes is not sufficient. If the calculated number of predetermined indexes for the operating conditions is not sufficient in the step (S5), an unused operating state under the operating conditions is assumed (S6).

[0058] An example of an unused operating state is, as shown in operating state "2" in Fig. 4, changing the flow rate of chilled / hot water CH in the first chilled / hot water pump 21 and the second chilled / hot water pump 22 to 50% without changing the operating states of the heat source units 11, 12, and 13 and the third chilled / hot water pump 23. In addition, as shown in operating state "2" in Fig. 4, changing the flow rate of chilled water CD in the first chilled water pump 41 and the second chilled water pump 42 to 50% without changing the operating states of the cooling towers 31, 32, and 33 and the third chilled water pump 43. Alternatively, as another example of an unused operating state, as shown in operating state "3" in Fig. 4, in comparison with operating state "1", stopping the third heat source unit 13 and the third cooling tower 33 and stopping the third chilled / hot water pump 23 and the third chilled water pump 43 (flow rate 0). It is preferable that the flow rates of the chilled / hot water CH and cooling water CD flowing through the first heat source unit 11 and the second heat source unit 12, which are variable speed turbo chillers, are the same for both heat source units 11, 12. The reason for this is to reduce the calculation load, since, in consideration of symmetry, it is extremely unlikely that energy savings can be achieved by operating the two heat source units 11, 12 under different conditions.

[0059] As described above, when changing the operating state under the operating conditions, it is preferable to apply each operating state in a combination (brute force) manner to avoid overlooking certain operating states. In this embodiment, the heat source units 11, 12, and 13 and the cooling towers 31, 32, and 33 are either on or off, limiting the number of combinations. On the other hand, the flow rates of the chilled / hot water CH from the chilled / hot water pumps 21, 22, and 23 and the cooling water CD from the cooling water pumps 41, 42, and 43 can be varied continuously, allowing for an extremely large number of combinations. In this case, for example, the range of change for each flow rate, other than when it is off (0%), can be set between 50% and 100% in 10% or 5% increments, thereby limiting the number of combinations. However, it is advisable to exclude clearly inappropriate operating states before or during the simulation. For example, if the total maximum output of the operating heat source units does not meet the heat demand for the operating conditions, the simulation need not be performed. In addition, cases can be excluded where the return temperature of the chilled water CH (outlet temperature of the heat demand equipment 99) derived from the demanded heat quantity, the supply temperature of the chilled water CH (temperature of the chilled water CH leaving the heat source equipment), and the flow rate of the chilled water CH exceeds the specified temperature.In addition, cases can be excluded where the temperature of the chilled water CD is above a certain level and variable flow rate control of the flow rate of the chilled water CD is not necessary.

[0060] Once the unused operating state is assumed as described above, the process returns to step S4 of calculating the predetermined index and follows the flow described above. Then, multiple predetermined indexes are calculated. In step S5, it is determined whether the calculated number of predetermined indexes for the operating conditions is sufficient. If the number is sufficient, the operating state in which the predetermined index satisfies the condition is identified (S7). Here, the value of the predetermined index that satisfies the condition is a value that is reasonably worth selecting, and is typically an optimal value. For example, if the predetermined index is operating cost, the lower the better; if it is carbon dioxide emissions, the lower the better; and if it is energy efficiency, the higher the better. There are several ways to determine the optimal value of the predetermined index. For example, if the predetermined index is operating cost, the minimum value may be simply selected. Alternatively, other considerations, such as a low number of operating heat source equipment and cooling towers, may be taken into account among operating states within a range of approximately 1-2% from the minimum (the difference is considered to be within the margin of error). Alternatively, a weighted average of multiple values ​​within an appropriate range of the predetermined index may be used to determine the operating state with the specified index value. Alternatively, the average of driving conditions corresponding to a plurality of values ​​(for example, the top three) within an appropriate range of a predetermined index may be adopted as the driving condition.

[0061] Once the combinations of operating conditions and operating states that satisfy the predetermined indexes are identified, it is determined whether the number of combinations satisfies the required number (S8). Here, the required number of combinations corresponds to the number of training data required to generate the control model 80, and although it depends on the model configuration, it is generally about several hundred to several thousand combinations. If the number of combinations does not satisfy the required number in the step (S8) of determining whether the number of combinations satisfies the required number, unused operating conditions are assumed (S9). Unused operating conditions are operating conditions that have not been used to identify combinations of operating conditions and operating states that satisfy the predetermined indexes. In this embodiment, several hundred to several thousand operating conditions are ultimately assumed in order to ultimately identify several hundred to several thousand combinations of operating conditions and operating states.

[0062] When assuming a relatively large number of operating conditions, it would be ideal if all of the operating conditions could be assumed. However, as the number of operating condition items increases, the number of operating conditions increases exponentially, resulting in a huge number of operating conditions. Furthermore, because the operating conditions are analog values, the number of operating conditions generated is determined by the dividing step. Therefore, if sufficient machine learning cannot be performed with the previously generated training data, it is tedious to create new operating conditions for machine learning. Therefore, in this embodiment, the computer performing the simulation randomly determines the operating conditions (creates them using random numbers). That is, each operating condition is set using random numbers within a variable range. This is repeated each time the step (S9) of assuming an unused operating condition is performed. In this way, even though the number of operating conditions is smaller than when identifying them by combination (brute force), each item (parameter) is evenly distributed. Therefore, when setting the operating conditions using random numbers, training data capable of creating a highly accurate control model 80 can be obtained, even if less data (data creation time) is required than when identifying them by combination (brute force). Furthermore, when new operating conditions are assumed to add training data after the fact, the necessary number of data sets can be created again using random numbers.

[0063] Once unused operating conditions have been assumed in the above manner, the process returns to the step (S3) of assuming an operating state, and thereafter follows the flow described above. Then, multiple pairs of operating conditions and operating states that meet the predetermined indexes are identified, and in the step (S8) of determining whether the number of pairs satisfies the required number, if it does, the creation of training data is terminated.

[0064] Although teacher data can be created in this way, if data used in another system exists and can be used, the creation of teacher data can be simplified, for example, by doing the following. The gist of this is that if the difference between teacher data already created for an existing heat source machine system and teacher data to be created for a new heat source machine system 1 is a partial difference in the range of operating conditions, the remaining common parts can be reused. A specific example will be explained below.

[0065] FIG. 5 is a flowchart illustrating the procedure for creating training data by reusing some existing data. In this specific example, the newly created training data (hereinafter referred to as "new data") has a maximum number of operating heat source machines of 3, a demand heat amount of 200 to 6000 kW, and 9000 pairs of operating conditions and operating states that meet the specified index (hereinafter referred to as "group data"). The existing training data (hereinafter referred to as "existing data") has a maximum number of operating heat source machines of 5, a demand heat amount of 200 to 10000 kW, and 9000 pairs of data. First, from the existing data, data with a demand heat amount that matches the content of the new data is extracted (S11). As a result, data with a maximum number of operating heat source machines of 5 and a demand heat amount of 200 to 6000 kW is extracted from the existing data, and in this example, the number of such data is 6000. Next, the existing data extracted by demand heat amount is sorted by the number of operating heat source machines (S12). As a result, the heat demand for both cases is between 200 and 6,000 kW, but these are divided into those with three or fewer operating heat source machines and those with more than three operating heat source machines. In this example, there are 5,000 cases with three or fewer operating heat source machines and 1,000 cases with more than three operating heat source machines. Of these, those with more than three operating heat source machines do not match the new data for three operating heat source machines, so they cannot be used as new data as is. However, in this example, rather than unconditionally excluding them, the operating conditions are retained and the operating state is changed to values ​​that match the new data, and then a simulation is performed to create a set of data (S13). As a result, 1,000 sets of data for a maximum of three operating heat source machines and a heat demand of 200 to 6,000 kW are created using the existing data as a reference. Next, the 5,000 sets of data extracted from the existing data and the 1,000 sets of data created using the existing data as a reference are added together (S14). This makes it possible to obtain 6,000 pieces of data for a maximum of three operating heat source machines and a heat demand of 200 to 6,000 kW. At this point, there are 3,000 pieces of data that are not sufficient for the new data that is being sought. Therefore, 3,000 new pieces of data are created for a maximum of three operating heat source machines and a heat demand of 200 to 6,000 kW (S15).When creating these 3,000 new data items, it is advisable to set the operating conditions randomly, as described above. Finally, these newly created 3,000 data items are added to the 6,000 data items obtained based on the existing data (S16). This results in 9,000 sets of data with a maximum number of operating heat source machines of 3 and a heat demand of 200 to 6,000 kW. These sets of data can be used as training data to be applied to the control model 80 of the new heat source machine system 1. In this way, the burden of creating new data can be reduced.

[0066] It goes without saying that the creation of the training data described above can be done automatically by creating a program on a general-purpose computer.

[0067] <Generation of control model (trained model)> Once the training data is created, a computer performs machine learning processing using the training data to create a control model 80. In this embodiment, for each of the numerous pairs (several hundred to several thousand pairs) of operating conditions and operating states that satisfy a predetermined index, created as training data, the operating conditions are input and the operating states are output, and machine learning processing is performed on the computer to create the control model 80. The control model 80 can be of any type or machine learning method, and any of a wide variety of proposed methods can be used. However, in this embodiment, a neural network is used. A neural network generally has perceptrons (operators) in an input layer, an intermediate layer, and an output layer. The output of the previous perceptron is multiplied by a weighting coefficient in the subsequent perceptron, summed, and processed through an activation function to generate a new output. The control model 80 is a model that derives a large number of numerical values ​​from a large number of numerical inputs. Here, the output layer of a neural network is generally expressed as a probability value, and therefore is output as a real number (decimal). As a result, the number of operating units, which should be an integer among the output operating states, is output as a real number (decimal) instead of an integer. Therefore, to determine the actual operating state, integer conversion processing such as rounding up or down is required. However, this results in the number of operating units differing from the output of control model 80, making it impossible to say with certainty that the state is optimal. Therefore, in this embodiment, control model 80 is divided into a first control model that outputs the number of operating units as an integer value among the output operating states, and a second control model that inputs the integer value of the number of operating units output by the first control model as one of the operating conditions.

[0068] FIG. 6 is a schematic diagram of the first control model (hereinafter referred to as "first control model 81"). FIG. 7 is a schematic diagram of the second control model (hereinafter referred to as "second control model 82"). As shown in FIG. 6, the first control model 81 includes an input layer, an intermediate layer, and an output layer, and is configured so that operating conditions are input to the input layer and operating states are output from the output layer. The first control model 81 includes, as operating conditions input to the input layer, the items exemplified in FIG. 3 (calorie demand, outside air temperature, target temperature of chilled or hot water CH, etc.), as well as an electricity carbon dioxide emission coefficient and a water supply carbon dioxide emission coefficient. The electricity carbon dioxide emission coefficient is the amount of carbon dioxide emission per unit of power consumed, and the water supply carbon dioxide emission coefficient is the amount of carbon dioxide emission per unit of water consumed. Including the unit cost of consumable resources and the carbon dioxide emission per unit consumption in the operating conditions input to the input layer has the following advantages. As a premise, even if these variables fluctuate, they generally fluctuate over a period of several months to several years, unlike heat demand or outdoor temperature, which fluctuate over a period of several minutes to a few hours at most. Therefore, in conventional control using simulations, it has been reasonable to treat these variables as fixed conditions and change them when they fluctuate. However, if these variables were treated as fixed conditions in this embodiment, the control model 80 would have to be rebuilt (machine learning processing would be performed again) when they fluctuated. Therefore, in this embodiment, these variables are treated as operating conditions, so that the control model 80 does not have to be rebuilt even when they fluctuate. This is of great benefit to users who experience daily changes in energy costs due to the introduction of a variable electricity pricing system, as has been considered in recent years, or the diversification of electricity procurement sources.

[0069] As shown in FIG. 6 , the first control model 81 outputs the operating status to the output layer, including the number of operating fixed-speed centrifugal chillers and the number of operating variable-speed centrifugal chillers, as well as the flow rates of chilled water CH and cooling water CD for each of these chillers. As mentioned above, the fixed-speed centrifugal chiller corresponds to the third heat source unit 13 in FIG. 1 , and the variable-speed centrifugal chillers correspond to the first heat source unit 11 and the second heat source unit 12. Therefore, the number of operating fixed-speed centrifugal chillers is 0 or 1, and the number of operating variable-speed centrifugal chillers is 0, 1, or 2, all of which are integer values. Furthermore, because the number of operating cooling towers 31, 32, and 33 matches the number of operating heat source units 11, 12, and 13, the cooling towers are not shown in FIG. 6. In other words, the number of operating heat source units 11, 12, and 13 and the number of operating cooling towers 31, 32, and 33 are the same integer values. 6 is the flow rate of the third chilled / hot water pump 23, and in this embodiment, is either 0% or a value (continuous value) between 50 and 100%. Similarly, the variable speed centrifugal chiller operation chilled / hot water flow rate is the total flow rate of the first chilled / hot water pump 21 and the second chilled / hot water pump 22. The fixed speed centrifugal chiller operation cooling water flow rate is the flow rate of the third cooling water pump 43, and the variable speed centrifugal chiller operation cooling water flow rate is the total flow rate of the first cooling water pump 41 and the second cooling water pump 42. In this embodiment, the range of the flow rate of each of these pumps is either 0% or a value (continuous value) between 50 and 100%.

[0070] In light of the above-mentioned purpose, the output layer of the first control model 81 only needs to output the number of operating units, which is not suitable for output as a decimal value. Therefore, other factors such as flow rate may not be included in the output to the output layer. However, in reality, the number of operating heat source units and the chilled / hot water flow rate (i.e., the load on each heat source unit) are closely related. Therefore, adding the chilled / hot water flow rate to the output may result in a perceptron in the middle layer that is closely related to this, and learning may proceed more efficiently. In this case, multiple control models can be trained by providing the same training data, and the optimal model can be determined from the training results. Note that adding the flow rate and other factors to the output here may shorten the training time of the second control model 82, as described below. In the case of a neural network, model training means adjusting the weight coefficients of the perceptron in the middle layer so that the input data and output data of the training data match. The general procedure is to first separate the training data into "training" and "verification," then use the training data to adjust the weighting coefficients using a technique called backpropagation, check the accuracy using the verification data, and if it is insufficient, adjust the weighting coefficients again.If the required accuracy cannot be achieved even after repeated training, as mentioned above, new training data can be generated by randomly assuming operating conditions, and additional training can be performed using the new training data.

[0071] As shown in FIG. 7 , the second control model 82 also includes an input layer, an intermediate layer, and an output layer, and is configured so that operating conditions are input to the input layer and operating states are output from the output layer. The second control model 82 includes, as operating conditions input to the input layer, the respective items input to the input layer of the first control model 81 (see FIG. 6 ) as well as the number of operating heat source machines (and interlocking cooling towers) output to the output layer of the first control model 81. The value of the number of operating heat source machines, etc. input to the input layer of this second control model 82 is an integer value. On the other hand, the operating state output to the output layer of the second control model 82 does not include the number of operating fixed-speed centrifugal chillers or the number of operating variable-speed centrifugal chillers, but is the flow rate of chilled or hot water CH and the flow rate of cooling water CD for each fixed-speed centrifugal chiller and variable-speed centrifugal chiller. In this embodiment, these values ​​output to the output layer of the second control model 82 are also output to the output layer of the first control model 81, but may differ from the corresponding values ​​output to the output layer of the first control model 81. The reason for this is that, as described above, the second control model 82 limits the value of the number of operating units of each heat source machine, etc., input to the input layer to an integer value. Therefore, the second control model 82 can output operating states such as the flow rate of chilled or heated water CH and the flow rate of cooling water CD, in which predetermined indicators more closely match the conditions, in a more realistic state than when the number of operating units of each heat source machine, etc., is output as a real number (including decimals).

[0072] As described above, when the output of the second control model 82 is included in the output of the first control model 81, it is advisable to configure the intermediate layers of both models 81 and 82 to be identical and use the intermediate layer coefficients of the first control model 81 as the initial values ​​of the intermediate layer coefficients of the second control model 82. This is a type of so-called "transfer learning," in which the only difference between the first control model 81 and the second control model 82 is whether or not the input includes the number of operating vehicles. Therefore, as described above, by configuring the intermediate layers of both models 81 and 82 to be identical and using the intermediate layer coefficients of the first control model 81 as the initial values ​​of the intermediate layer coefficients of the second control model 82, it may be possible to significantly reduce the learning time required to generate the second control model 82. Note that the first control model 81 and the second control model 82 are conceptually distinct and may be physically configured as separate entities or as an integrated entity.

[0073] It is preferable to create a control model 80 (including the case where the control model 80 is divided into a first control model 81 and a second control model 82) for each predetermined index to be optimized. For example, it is preferable to create a model that minimizes operating costs, a model that minimizes carbon dioxide emissions, etc. In this way, it becomes possible to switch between control models 80 depending on the situation, enabling more appropriate operation of the heat source machine system 1.

[0074] The control model 80 configured (generated) as described above (including the case where it is divided into a first control model 81 and a second control model 82) is mounted on the control device 70 of the heat source machine system 1. The control device 70 mounted with the control model 80 is configured to control the operation of the heat source equipment and heat source auxiliary equipment that make up the heat source machine system 1 so as to achieve the operating state output by the control model 80. The operation of the heat source machine system 1, including the control of the control device 70, will be described below.

[0075] <Heat source system operation> Figure 8 is a block diagram showing the calculation flow of the control device 70 of the heat source equipment system 1. Below, the operation of the heat source equipment system 1 will be explained mainly with reference to Figures 1 and 8, and also with reference to Figures 2 to 7 as appropriate. When the heat source equipment system 1 is in operation, the necessary ones of the heat source equipment 11, 12, 13 operate, and in conjunction with this, the necessary ones of the chilled / hot water pumps 21, 22, 23, the cooling towers 31, 32, 33, and the cooling water pumps 41, 42, 43 operate. For the time being, to avoid complicating the situation, the explanation will be given assuming that the heat source equipment 11, 12, 13 and their associated heat source auxiliary equipment are all operating.

[0076] By operating each chilled / hot water pump 21, 22, 23, chilled / hot water CH flows from the return header 29 into each heat source unit 11, 12, 13 via each chilled / hot water return pipe 25, 26, 27. The chilled / hot water CH flowing into each heat source unit 11, 12, 13 is cooled (during cooling) or heated (during heating), and its temperature is typically adjusted so that the difference with the ambient temperature increases. The chilled / hot water CH whose temperature has been adjusted in each heat source unit 11, 12, 13 is transported to the forward header 19 via each chilled / hot water forward pipe 15, 16, 17. The chilled / hot water CH flowing into the forward header 19 is supplied to the heat demand facility 99 by a secondary pump (not shown) through a supply pipe 91, and then flows through a recovery pipe 92 and into the return header 29. The chilled / hot water CH supplied to the heat demand facility 99 is used for heat load treatment, and its temperature changes so that the difference with the ambient temperature decreases.

[0077] On the other hand, operation of each cooling water pump 41, 42, 43 causes cooling water CD to flow from each cooling tower 31, 32, 33 through each cooling water supply pipe 35, 36, 37 into each heat source unit 11, 12, 13. The cooling water CD that flows into each heat source unit 11, 12, 13 exchanges heat with the refrigerant in each heat source unit 11, 12, 13, and then returns to each cooling tower 31, 32, 33 through each cooling water return pipe 45, 46, 47. As described above, the control device 70 controls the start / stop of each heat source unit 11, 12, 13. Based on the operation of each heat source unit 11, 12, 13, the auxiliary power panel 75 controls the start / stop of each cooling tower 31, 32, 33 and the discharge flow rate of each chilled / hot water pump 21, 22, 23 and each cooling water pump 41, 42, 43. When the heat source machine system 1 is operating in this manner, the control device 70 controls the system in the following manner so that the operating state becomes appropriate (typically, optimal).

[0078] The control device 70 receives temperature information from the outdoor air thermometer 61. The control device 70 also receives temperature and pressure information from each heat source unit 11, 12, and 13, and measures the heat demand of the heat demand facility 99, the flow rate of the chilled or hot water CH, and the pressure loss of the heat demand facility 99. The control device 70 then calculates the pressure loss coefficient of the heat demand facility 99 from the pressure loss of the heat demand facility 99 and the flow rate of the chilled or hot water CH. The control device 70 also receives information regarding the target temperature of the chilled or hot water CH from the heat demand facility 99. Meanwhile, the control device 70 stores the electricity unit price, water rate, electricity carbon dioxide emission coefficient, and water carbon dioxide emission coefficient as set values. As mentioned above, these set values ​​generally do not change in a short period of time, and therefore, it is often sufficient to correct them when they change. The control device 70 inputs these identified values ​​into the control model 80 as operating conditions.

[0079] When operating conditions are input, the control model 80 performs processing based on a trained algorithm and outputs an appropriate operating state. Generally, learning (backpropagation calculation) of a mathematical model, particularly using a neural network, requires extremely high computing power, such as a computing element like a GPU. However, inference (forward propagation calculation) using a trained model does not require such high computing power. Therefore, even when using a dedicated computing element, inference using a trained model can be performed sufficiently with a relatively inexpensive computing device that consumes little power. In light of this, the control model 80 that performs inference based on a trained algorithm has a smaller computational load than performing a simulation when creating training data, and is suitable for installation on site.

[0080] When the operating conditions are input, the control model 80 first inputs the operating conditions to the first control model 81 (see FIG. 6). The first control model 81, to which the operating conditions have been input, performs processing based on a trained algorithm, converts the output into integers, and determines the number of operating heat source machines 11, 12, and 13. At this time, in order to improve the accuracy of the output, it is preferable to also output the flow rate of the chilled / hot water CH and the flow rate of the cooling water CD, as described above. However, the flow rate of the chilled / hot water CH and the flow rate of the cooling water CD output by the first control model 81 are not intended to be directly used for controlling the heat source machine system 1. After obtaining the output of the first control model 81, the control model 80 inputs the operating conditions input to the first control model 81 and the integer value of the number of operating heat source machines 11, 12, and 13 output from the first control model 81 to the second control model 82 (see FIG. 7). The second control model 82, to which the integer value of the number of operating units and the operating conditions have been input, performs processing based on the learned algorithm and outputs the flow rate of chilled / heated water CH and the flow rate of cooling water CD. In this embodiment, the first control model 81 outputs the number of operating units as an integer value, so that the output of the integer value can be applied directly to actual control. Furthermore, because the integer value of the number of operating units is included in the input of the second control model 82, the accuracy of the output of the second control model 82 can be improved.

[0081] The control device 70 controls the operation of each heat source unit 11, 12, and 13 so that the number of operating units is equal to the number output by the first control model 81. The control device 70 also controls the discharge flow rates of each chilled / hot water pump 21, 22, and 23 and each cooling water pump 41, 42, and 43 so that the flow rates of chilled / hot water CH and cooling water CD are equal to the flow rates output by the second control model 82. By controlling each device constituting the heat source unit system 1 according to the output of the control model 80, it is possible to operate the heat source unit system 1 so that predetermined indicators set by the user, such as power consumption, operating costs, and carbon dioxide emissions, meet the conditions (e.g., minimize them). The calculations using the control model 80 and the control of each device based on the output may, for example, change the number of units in stages or gradually change the command value for the flow rate in order to avoid sudden changes in the operating state. The calculations in the control model 80 and the adjustment of the number of operating units and the flow rate based on the output of the control model 80 may be performed at predetermined intervals. The predetermined interval can be determined appropriately depending on the situation, taking into consideration the calculation time (or calculation load) in the control model 80 and the control accuracy of the heat source machine system 1. Examples of the predetermined interval include 3 minutes, 5 minutes, 10 minutes, 15 minutes, 30 minutes, etc. However, for example, if the demand heat amount changes by more than a set amount, or if other operating conditions change significantly, calculations using the control model 80 and control calculations for each device based on the output of the calculations may be performed at timings other than the predetermined intervals that have been set in advance.

[0082] The control device 70 may be equipped with a plurality of control models 80 corresponding to the types of predetermined indexes, and may switch to an appropriate control model 80 as appropriate based on user settings or instructions. For example, a plurality of control models 80 may be provided, each with a minimum power consumption, a minimum operating cost, and a minimum amount of carbon dioxide emissions as predetermined indexes that meet the conditions, and these may be switched according to the situation. The predetermined index used to generate one control model 80 is not limited to one type; for example, the control device 70 may have a control model 80 with the minimum amount of carbon dioxide emissions within a 5% range from the minimum operating cost. In this case, the operating cost corresponds to the first predetermined index (the index with the highest priority), and the carbon dioxide emissions correspond to the second predetermined index (the index with the second highest priority).

[0083] Furthermore, instead of obtaining all of the operating states for controlling the equipment constituting the heat source equipment system 1 from the output of the control model 80, the control device 70 may control some of the operating states based on the output of the control model 80. In this case, the remaining operating states may be controlled based on the results of a simulation performed by the control device 70 or based on a rule base predefined in the control device 70. For example, the number of operating heat source equipment 11, 12, 13 and the flow rate of chilled or hot water CH in each chilled or hot water pump 21, 22, 23 may be controlled based on the output of the control model 80, and the flow rate of cooling water CD in each cooling water pump 41, 42, 43 may be controlled based on the rule base.

[0084] As described above, according to the heat source equipment system 1 of this embodiment, the constituent devices are controlled based on the output of the control model 80, so that operation with good response can be performed in an appropriate operating state according to changes in the operating conditions. Furthermore, since the first control model 81 that outputs the number of operating units as an integer value and the second control model 82 that includes this integer value as an input are provided, an appropriate operating state that conforms to actual operation can be maintained. Furthermore, if the operating conditions include the pressure loss coefficient on the heat demand facility 99 side, appropriate operation can be performed even if the flow rate of chilled or hot water CH supplied to the heat demand facility 99 changes.

[0085] <Other> In the above description, the first heat source unit 11 and the second heat source unit 12 are variable-speed turbo chillers, and the third heat source unit 13 is a fixed-speed turbo chiller. However, various types of heat source units other than turbo chillers can be used depending on the application, such as absorption chillers, chilled / hot water generators, and heat pumps. In the above description, the first heat source unit 11 and the second heat source unit 12 are the same type of units having the same characteristics, but they may be the same type of units having different characteristics, or they may be different types of units. In the above description, three heat source units 11, 12, and 13 are provided as heat source units, but the total number of heat source units is not limited to three and may be more or less than three depending on the application. For example, more than three units of different types of units may be provided, or multiple types of units of the same type (e.g., two or three units) may be provided.

[0086] In the above explanation, it has been assumed that each of the heat source machines 11, 12, and 13 is capable of detecting the temperature difference and pressure difference of the inflowing and outflowing chilled or hot water CH. However, instead of each of the heat source machines 11, 12, and 13 being provided with a meter for detecting temperature and pressure, a meter for detecting temperature and pressure may be provided on a nearby pipe.

[0087] In the above explanation, cooling towers 31, 32, and 33 are provided as the heat-source fluid supply device, and cooling water CD is used as the heat-source fluid. However, the heat-source fluid supply device may be an air-cooled heat pump chiller, and the heat-source fluid may be air. In this case, the air-cooled heat pump chiller serves as both the heat source device and the heat-source fluid supply device; in other words, the heat source device and the heat-source fluid supply device are typically configured as a single physical unit (housed in a single housing).

[0088] In the above description, the control device 70 indirectly controls the heat source auxiliary machines via the auxiliary machine power panel 75, but the control device 70 may also directly control the heat source auxiliary machines.

[0089] In the above description, it has been described that when creating training data, it is preferable to combine (try all possible combinations) the driving states to be changed for the assumed driving conditions, but it is also possible to set the driving states to be changed randomly. However, as described above, it is preferable to set the driving states to be changed in a combined (try all possible combinations) in order to avoid overlooking specific driving states.

[0090] In the above explanation, fixed conditions are prepared before the operating conditions are assumed, but the items listed as examples of fixed conditions may be treated as operating conditions. In this case, the step (S1) of preparing fixed conditions in the flowchart shown in Figure 2 is omitted.

[0091] In the above explanation, the items included in the operating conditions input to the input layer are the heat demand, outside air temperature, target temperature of the chilled / hot water CH, pressure loss coefficient on the heat demand equipment 99 side, unit cost of consumed resources, and unit carbon dioxide emission amount. However, the number of input items may be increased or decreased as appropriate. For example, if a heat source device (e.g., an absorption chiller) that obtains heat by burning fuel such as oil or gas is installed, the fuel unit cost may be included as an input item. Conversely, the pressure loss coefficient on the heat demand equipment 99 side and other items may be excluded from the input items. However, including the pressure loss coefficient on the heat demand equipment 99 side as an input item has the advantage that the power of each chilled / hot water pump 21, 22, 23 can be accurately estimated even if the flow rate of the chilled / hot water CH on the heat demand equipment 99 side changes. Furthermore, instead of the heat demand value itself being input as an operating condition, a physical quantity correlated with the heat demand may be input as an operating condition. Examples of physical quantities correlated with the demanded heat quantity include the return temperature of the chilled / hot water CH (the temperature of the chilled / hot water CH flowing into each heat source unit 11, 12, 13). Furthermore, instead of the outside air temperature being input as an operating condition, a physical quantity correlated with the outside air temperature may be input as an operating condition. Examples of physical quantities correlated with the outside air temperature include the inlet temperature of the cooling water CD (the temperature of the cooling water CD flowing into each heat source unit 11, 12, 13) and the temperature of the lower water tank of the cooling towers 31, 32, 33.

[0092] In the above explanation, the integer value output by the first control model 81 is used to control the number of operating heat source units 11, 12, and 13. However, if each of the chilled / hot water pumps 21, 22, and 23 and / or each of the cooling water pumps 41, 42, and 43 is configured to perform unit number control and an integer value output is required as their operating state, the output of the first control model 81 may be used as their operating state. On the other hand, if each of the heat source units 11, 12, and 13 is configured to perform capacity control with continuously adjustable output (for example, capable of operating at 50% to 100% of maximum output), the first control model 81 may be omitted and the output of the second control model 82 may be used for control. [Explanation of symbols]

[0093] 1 Heat source system 11, 12, 13 Heat source equipment (heat source equipment) 21, 22, 23 Chilled and hot water pump (heat source auxiliary equipment, heat medium pump) 31, 32, 33 Cooling tower (heat source auxiliary equipment, heat source fluid supply device) 41, 42, 43 Cooling water pump (heat source auxiliary equipment, heat source fluid supply device) 70 Control device 80 Control Model (Trained Model) 81 First control model (trained model) 82 Second control model (trained model) 99 Heat demand equipment CD Cooling water (heat source fluid) CH Cold and hot water (heat medium)

Claims

1. a heat source device that cools or heats a heat medium to be supplied to a heat demand facility; a heat source auxiliary machine that operates in conjunction with the operation of the heat source equipment; A control device that adjusts the operating states of the heat source equipment and the heat source auxiliary equipment and has a learned control model, The heat source auxiliary equipment includes a heat medium pump that causes the heat medium to flow through the heat source equipment, and a heat source fluid supply device that supplies the heat source equipment with a heat source fluid that directly or indirectly exchanges heat with the heat medium in the heat source equipment, the operating state includes at least one of an operating status of the heat source equipment, a flow rate of the heat medium discharged by the heat medium pump, and a flow rate of the heat source fluid supplied by the heat source fluid supply device; the control model is subjected to machine learning processing using training data so as to output the operating state in which a predetermined index becomes a value that satisfies the condition when an operating condition that affects the operating state is input; the operating conditions include at least one of a heat demand of the heat demanding facility or a physical quantity correlated thereto, and an outside air temperature or a physical quantity correlated thereto; the predetermined index includes at least one of the power consumption of the heat source equipment and the heat source auxiliary equipment, the operating costs of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emissions of the heat source equipment and the heat source auxiliary equipment; the control device controls the heat source equipment and the heat source auxiliary equipment so as to achieve the operating state output by the control model; At least one of the heat source equipment, the heat medium pump, and the heat source fluid supply device is configured by a plurality of units, the operating state includes the number of operating units among the heat source equipment, the heat medium pump, and the heat source fluid supply device; the control models include a first control model that outputs the number of operating units as an integer value among the operating states to be output, and a second control model that inputs the integer value of the number of operating units output by the first control model as one of the operating conditions. Heat source machine system.

2. the predetermined indicators include two or more of the power consumption of the heat source equipment and the heat source auxiliary equipment, the operating costs of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emissions of the heat source equipment and the heat source auxiliary equipment, The control model is configured to select a plurality of the operating states in which a first predetermined index among the plurality of predetermined indexes is a value that satisfies a condition, and to output the operating state in which a second predetermined index different from the first predetermined index is a value that satisfies a condition from the selected plurality of the operating states. The heat source system according to claim 1 .

3. The control device uses the output of the control model for some of the operating states among the heat processing amount of the heat source equipment, the flow rate of the heat medium discharged by the heat medium pump, and the flow rate of the heat source fluid supplied by the heat source fluid supply device, and determines the remaining operating states by simulation or rule-based. The heat source machine system according to claim 1 or 2.

4. The operating conditions include a pressure loss coefficient of the heat demand equipment. The heat source machine system according to any one of claims 1 to 3.

5. The operating conditions include at least one of a cost per unit of power consumption of the heat source equipment and the heat source auxiliary equipment, and an amount of carbon dioxide emission per unit of power consumption of the heat source equipment and the heat source auxiliary equipment. The heat source machine system according to any one of claims 1 to 4.

6. the control model includes a plurality of control models in which the predetermined index is one or more of: power consumption of the heat source equipment and the heat source auxiliary equipment; operating costs of the heat source equipment and the heat source auxiliary equipment; and carbon dioxide emissions of the heat source equipment and the heat source auxiliary equipment; the control device uses an appropriate control model from among the plurality of control models in accordance with the predetermined index desired by a user. The heat source machine system according to any one of claims 1 to 5.

7. A method for generating a trained model used to control a heat source equipment system including a heat source equipment that cools or heats a heat medium supplied to a heat demand facility and a heat source auxiliary equipment that operates in conjunction with the operation of the heat source equipment, A process of generating teacher data by obtaining a plurality of predetermined indices for an assumed operating state of the heat source machine system under assumed operating conditions while changing the operating state, and then defining the operating state when the predetermined indices meet the conditions as a relationship with the operating condition, and performing this for a plurality of operating conditions to obtain a plurality of pairs of relationships between the operating state when the predetermined indices meet the conditions and the operating condition; and performing machine learning processing using the training data to generate a trained model in which the operating conditions are input and the operating state is output, The heat source auxiliary equipment includes a heat medium pump that causes the heat medium to flow through the heat source equipment, and a heat source fluid supply device that supplies the heat source equipment with a heat source fluid that directly or indirectly exchanges heat with the heat medium in the heat source equipment, the operating conditions include at least one of a heat demand of the heat demanding facility or a physical quantity correlated thereto, and an outside air temperature or a physical quantity correlated thereto; the operating state includes at least one of an operating status of the heat source equipment, a flow rate of the heat medium discharged by the heat medium pump, and a flow rate of the heat source fluid supplied by the heat source fluid supply device; the predetermined index includes at least one of the power consumption of the heat source equipment and the heat source auxiliary equipment, the operating costs of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emissions of the heat source equipment and the heat source auxiliary equipment; At least one of the heat source equipment, the heat medium pump, and the heat source fluid supply device is configured by a plurality of units, the operating state includes the number of operating units among the heat source equipment, the heat medium pump, and the heat source fluid supply device; The step of generating the trained model includes a step of generating a first trained model including an integer value of the number of operating vehicles in the output operating state, and a step of generating a second trained model that inputs the integer value of the number of operating vehicles as one of the operating conditions. How to generate a trained model.

8. The first trained model and the second trained model use a neural network, and the first trained model and the second trained model have a common intermediate layer; The driving state item that is an output of the first trained model includes the driving state item that is an output of the second trained model, The step of generating the trained model includes first performing a machine learning process on a first trained model to generate the first trained model, and then performing a machine learning process on a second trained model in which initial values ​​of weight coefficients of the intermediate layer of the second trained model are set to the weight coefficients of the intermediate layer of the first trained model to generate the second trained model. The method for generating a trained model according to claim 7.

9. In the step of generating the teacher data, at least one of the operating conditions and the operating state when performing the simulation is determined randomly. A method for generating a trained model according to claim 7 or claim 8.

10. In the step of generating the teacher data, when data of the predetermined index for the operating state in a wider range than the operating state assumed under the operating conditions in a wider range than the assumed operating conditions already exists, the predetermined index for the operating state assumed under the assumed operating conditions is extracted from the already existing data and used as the teacher data. A method for generating a trained model according to any one of claims 7 to 9.

11. The step of generating the teacher data includes changing values ​​of items in the unextracted data that are incompatible with the expected operating conditions and operating states among the operating conditions and operating states to compatible values, and then performing a simulation to generate the teacher data. The method for generating a trained model according to claim 10.

12. The operating conditions include a pressure loss coefficient of the heat demand facility, the step of generating the teacher data includes calculating a pressure loss in the heat demanding facility based on the pressure loss coefficient and the flow rate of the heat medium in the simulation, and then obtaining the predetermined index. A method for generating a trained model according to any one of claims 7 to 11.

13. A trained model installed in a computer used to control a heat source system including a heat source device that cools or heats a heat medium supplied to a heat demand facility and a heat source auxiliary device that operates in conjunction with the operation of the heat source device, a first control model; a second control model; The first control model and the second control model each include: an input layer to which operating conditions of the heat source machine system are input; an output layer that outputs the operating status of the heat source machine system; an intermediate layer in which parameters are learned using training data in which the operating conditions are input and the operating state in which a predetermined index has a value that satisfies the condition is output; The heat source auxiliary equipment includes a heat medium pump that causes the heat medium to flow through the heat source equipment, and a heat source fluid supply device that supplies the heat source equipment with a heat source fluid that directly or indirectly exchanges heat with the heat medium in the heat source equipment, At least one of the heat source equipment, the heat medium pump, and the heat source fluid supply device is configured by a plurality of units, the operating conditions include at least one of a heat demand of the heat demanding facility or a physical quantity correlated thereto, and an outside air temperature or a physical quantity correlated thereto; the operating state includes at least one of an operating status of the heat source equipment, a flow rate of the heat medium discharged by the heat medium pump, and a flow rate of the heat source fluid supplied by the heat source fluid supply device; the operating state output from the first control model includes the number of operating units among the heat source equipment, the heat medium pump, and the heat-source fluid supply device; the predetermined index includes at least one of the power consumption of the heat source equipment and the heat source auxiliary equipment, the operating costs of the heat source equipment and the heat source auxiliary equipment, and the carbon dioxide emissions of the heat source equipment and the heat source auxiliary equipment; When the computer is caused to function to input the operating conditions to the input layer, perform calculations in the intermediate layer, and output the operating states from the output layer, the first control model outputs the number of operating units as an integer value among the operating states to be output, and the second control model inputs the integer value of the number of operating units output by the first control model as one of the operating conditions. Trained model.

14. A control device having the trained model according to claim 13; The heat source equipment; The heat source auxiliary machine, The control device controls the heat source equipment and the heat source auxiliary equipment so as to achieve the operating state output by the learned model. Heat source machine system.

Images