Method for collecting learning data for controlling air conditioning equipment, air conditioning system

Abstract

To provide a learning data collection method for controlling an air conditioner that allows learning data of a prediction model in a control object region to be collected reliably in a short time without imposing a burden on the device.SOLUTION: A learning data collection method includes steps of region setting, learning start point setting, state point moving, and data collection. In the region setting step, a region R1 is set for creating a prediction model on a psychrometric chart. In the learning start point setting step, a plurality of learning start points S1-S9, which are starting points at the time of random vibration, are set in the region R1. In the state point moving step, an air conditioning air state point K1 in the same position as the position of an outside air state point G1 is moved to the initial learning start point S1. At this time, air conditioner instruments 22-25 are operated for executing temperature and humidity control. In the data collection step, random vibration of the air conditioner instruments 22-25 is executed while moving the air conditioning air state point K1 between learning start points, and learning data is collected.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates to a learning data collection method for controlling an air conditioner, and an air conditioning system. [Background technology]

[0002] Generally, a painting facility includes a painting booth for applying paint to objects such as automobile bodies, and a drying oven for drying the paint on the objects after it has passed through the painting booth. In such painting facilities, air whose temperature and humidity have been adjusted by a paint booth air conditioner is sent into the painting booth to perform painting. The paint booth air conditioner is composed of devices such as a heating device, a cooling device, a humidifying device (washer), and a blower fan, and these devices are operated in combination to control the conditioned air to a target temperature and humidity (see, for example, Patent Documents 1 and 2). PID control has traditionally been used for this purpose.

[0003] Incidentally, booth air conditioners in automotive paint booths require strict temperature and humidity control to maintain the quality of the paint job on the product. Recently, a new method for achieving strict temperature and humidity control has been proposed: model predictive control (MPC), which controls while predicting future reactions. Model predictive control is a control method that appropriately captures the dynamics of air conditioning equipment and uses a modeled predictive model. Furthermore, the predictive model used in model predictive control for air conditioning uses machine learning for its identification. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 3993358 [Patent Document 2] Japanese Patent Application Laid-Open No. 2010-119901 Summary of the Invention [Problem to be solved by the invention]

[0005] Generally, identification data for learning is obtained by randomly changing the operation amount of each air conditioning device (i.e., applying random vibration) for several hours, and repeating measurement and recording.

[0006] However, such random vibrations place a burden on each air conditioning device, causing it to malfunction. Furthermore, randomly changing the manipulated variable can unintentionally activate the device's anomaly detection device, interrupting measurement and recording. Furthermore, because random vibrations are applied starting from the outside air at that time, it is difficult to reliably collect data on the area targeted for temperature and humidity control in a short period of time.

[0007] The present invention has been made in consideration of the above-mentioned problems, and its purpose is to provide a learning data collection method for air conditioning device control, and an air conditioning system, which can reliably collect learning data for a predictive model in a controlled area in a short time without placing a burden on the device. [Means for solving the problem]

[0008] In order to solve the above problem, the invention described in Means 1 is a method for collecting learning data for a predictive model used in model predictive control of an air conditioning device that adjusts the temperature and humidity of taken-in outdoor air using multiple types of air conditioning equipment, the method comprising: an area setting step for setting an area on a psychrometric chart for creating the predictive model; a learning start point setting step for setting multiple learning start points within the set area that serve as starting points for performing random excitation of the air conditioning equipment; a state point moving step for moving an air-conditioned air state point located at the same position as the outdoor air state point on the psychrometric chart to one of the multiple learning start points by operating the air conditioning equipment to control the temperature and humidity; and a data collection step for collecting the learning data by performing random excitation of the air conditioning equipment while moving the air-conditioned air state point between the multiple learning start points.

[0009] Therefore, according to the invention described in means 1, even if there is no outdoor air state point within the set region, learning data collection is started after moving the air-conditioned air state point to one of the learning start points within the region. In other words, the starting point for data collection is not the outdoor air state point at that time, but a state point within the region. This makes it possible to reliably collect data for the region to be targeted for temperature and humidity control. Furthermore, because random excitation of the air conditioning equipment is performed while the air-conditioned air state point is moved between multiple learning start points within the region, learning data within the region can be reliably collected.

[0010] The invention described in means 2 is characterized in that in means 1, in the data collection step, the air-conditioned air state point displaced from the learning start point by the random vibration is controlled to return to the learning start point by PID control.

[0011] Therefore, according to the invention described in means 2, even if random vibration is applied, correction by PID control works, so data collection can be performed without placing a burden on each air conditioning device.

[0012] The invention described in Means 3 is characterized in that in Means 1 or 2, in the state point moving step, when the outside air state point is not within the region, the air-conditioned air state point located at the same position as the outside air state point is moved to one of the plurality of learning start points by controlling the temperature and humidity of the air-conditioning equipment using PID control.

[0013] Therefore, according to the invention described in means 3, it is possible to quickly and efficiently move an air-conditioned air state point that is not within the region to one of the learning start points, and it is also possible to reduce the burden on each air-conditioning device at that time.

[0014] The invention described in means 4 is characterized in that, in any one of means 1 to 3, in the data collection step, the temperature and humidity of the air conditioning equipment are controlled by PID control, thereby moving the air-conditioned air state point from the current learning start point to the next learning start point.

[0015] Therefore, according to the invention described in means 4, movement between learning start points for the conditioned air state points can be performed quickly and efficiently.

[0016] The invention described in means 5 is characterized in that, in any one of means 1 to 4, in the data collection step, upper and lower limit values ​​are set for the operating amount of the air conditioning equipment when the random vibration is performed.

[0017] Therefore, according to the invention described in means 5, by setting upper and lower limit values ​​for the manipulated variable, it is possible to prevent the air conditioning equipment from being operated with an excessive manipulated variable during random vibration. As a result, the occurrence of abnormalities in the air conditioning equipment due to excessive operation is avoided, and learning data can be collected efficiently without interruption.

[0018] The invention described in means 6 is characterized in that, in any one of means 1 to 5, the air conditioning device is an air conditioning device for a paint booth that includes a preheating device, a humidifying device, a cooling device, and a reheating device as the air conditioning equipment.

[0019] The invention described in Means 7 is an air conditioning system comprising an air conditioning device that adjusts the temperature and humidity of taken-in outside air using multiple types of air conditioning equipment, and an air conditioning control device that controls the operation amount of the air conditioning equipment, wherein the air conditioning control device is equipped with a learning data collection device that collects learning data of a predictive model used for model predictive control of the air conditioning device, and the learning data collection device is characterized by including: an area setting unit that sets an area on a psychrometric chart for creating the predictive model; a learning start point setting unit that sets multiple learning start points within the set area that serve as starting points for performing random excitation of the air conditioning equipment; a state point moving unit that moves an air-conditioned air state point located at the same position as the outside air state point on the psychrometric chart to one of the multiple learning start points by operating the air conditioning equipment to control the temperature and humidity; and a data collection unit that collects the learning data by performing random excitation of the air conditioning equipment while moving the air-conditioned air state point between the multiple learning start points.

[0020] Therefore, according to the invention described in means 7, even if there is no outdoor air state point within the set region, the state point moving unit moves the air-conditioned air state point to any learning start point within the region, and then the data collecting unit starts collecting learning data. In other words, the starting point for data collection is not the outdoor air state point at that time, but any learning start point within the region. This makes it possible to reliably collect data for the region to be targeted for temperature and humidity control. Furthermore, the data collecting unit randomly vibrates the air conditioning equipment while moving the air-conditioned air state point between multiple learning start points within the region. This makes it possible to reliably collect learning data within the region. [Effects of the Invention]

[0021] As described above in detail, according to the inventions described in claims 1 to 3, it is possible to provide a learning data collection method for air conditioning device control and an air conditioning system that can reliably collect learning data for a predictive model in a controlled area in a short time without placing a burden on the device. [Brief explanation of the drawings]

[0022] [Figure 1] 1 is a block diagram illustrating an air conditioning system for a painting facility according to an embodiment of the present invention; [Figure 2] 4 is a flowchart for explaining a learning data collection method in the air conditioning system for the painting equipment. [Figure 3] 4 is a psychrometric chart for explaining a learning data collection method in the above-mentioned air conditioning system for painting equipment. [Figure 4] 4 is a psychrometric chart for explaining a learning data collection method in the above-mentioned air conditioning system for painting equipment. [Figure 5] 4 is a psychrometric chart for explaining a learning data collection method in the above-mentioned air conditioning system for painting equipment. [Figure 6] 4 is a psychrometric chart for explaining a learning data collection method in the above-mentioned air conditioning system for painting equipment. [Figure 7] 1A is a graph for explaining random vibration in the learning data collection method of the embodiment, and FIG. 1B is a graph for explaining random vibration in the learning data collection method of the comparative example. [Figure 8] FIG. 10 is a block diagram illustrating an air conditioning system for a painting facility according to another embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0023] An air conditioning system 11 for a painting facility according to one embodiment of the present invention will be described in detail below with reference to FIGS.

[0024] 1 is a painting equipment air conditioning system 11, which includes an air conditioner 21 that adjusts the temperature and humidity of the taken-in outside air using multiple types of air conditioning equipment, and an air conditioning control device 31 that controls the operation amount of the air conditioning equipment. The air conditioner 21 in this embodiment is a painting booth air conditioner 21 that includes a preheating device, a humidifying device, a cooling device, and a reheating device as air conditioning equipment.

[0025] The paint booth, to which the conditioned air generated by the paint equipment air conditioning system 11 is supplied, is generally installed in an area of ​​a paint workpiece transport line where paint is applied to the workpieces. The paint booth comprises a paint chamber, an air supply chamber located above the paint chamber for supplying air in a downflow (a fixed direction from above to below) to the paint chamber, and an exhaust chamber located below the paint chamber for exhausting the air from the paint chamber. In the paint booth of this embodiment, conditioned air discharged from the paint booth air conditioner 21 is supplied into the paint chamber in a downflow manner from the air supply chamber.

[0026] In the paint booth's paint chamber, paint is applied to objects by spraying paint mist from a paint sprayer (not shown). During this process, paint mist that is oversprayed from the paint sprayer and scattered is exhausted from the chamber to an exhaust chamber by the downflow of conditioned air acting within the chamber. In the exhaust chamber, the paint mist contained in the air is captured using booth circulating water, and the paint is recovered. Furthermore, the air exhausted from the exhaust chamber is released into the atmosphere by a blower fan.

[0027] As shown in Fig. 1, the paint booth air conditioner 21 (air conditioning device) in this embodiment is configured to include multiple types of air conditioning equipment. This paint booth air conditioner 21 is a device that adjusts air taken in from outside the device to a predetermined temperature (e.g., approximately 23°C) and a predetermined humidity (e.g., approximately 70% RH) and sends the air to the paint booth. Specifically, this paint booth air conditioner 21 is equipped with a preheater 22 (preheating device), a washer 23 (humidifying device), a cooling coil 24 (cooling device), a reheater (reheating device) 25, and a blower fan 26.

[0028] The preheater 22 is a type of temperature control means for adjusting the temperature of the taken-in air and is a device for heating the air to preheat it. The washer 23 is a type of humidity control means for adjusting the humidity of the taken-in air and is a device for increasing the humidity of the air by injecting water into the air that has passed through the preheater 22. The cooling coil 24 is a type of temperature control means for adjusting the temperature of the taken-in air and is a cooling device for cooling the air that has passed through the washer 23 to lower its temperature. The reheater 25 is a type of temperature control means for adjusting the temperature of the taken-in air and is a reheating device for reheating the air that has passed through the cooling coil 24 to raise its temperature. The blower fan 26 is an air compression device for compressing and sending the temperature- and humidity-controlled air (i.e., conditioned air) to the painting booth.

[0029] Sensing devices are provided at multiple locations in the paint booth air conditioner 21. Specifically, the paint booth air conditioner 21 is equipped with a first sensor 27a for measuring temperature and humidity, a second sensor 27b for measuring temperature and humidity, and a third sensor 27c for measuring temperature. The first sensor 27a is used to measure the temperature and humidity of the outside air before conditioning and is located near the outside air intake port of the paint booth air conditioner 21. The second sensor 27b is used to measure the temperature and humidity of the outside air after conditioning and is located on the outlet side of the blower fan 26 from which the conditioned air is sent out. The third sensor 27c is used to measure the temperature of the outside air that has passed through the preheater 22 and is located upstream of the washer 23.

[0030] The air conditioning control device 31 for a painting facility in this embodiment is a device for controlling the manipulated variables of air conditioning equipment and is composed of one or more well-known computers including a CPU, memory means (ROM, RAM), etc. As shown in the block diagram of FIG. 1, the air conditioning control device 31 includes an air supply target input unit 32, an air supply setting calculation unit 33, and an equipment control unit 34. The equipment control unit 34 includes a PID controller 35 that controls the manipulated variables of the air conditioning equipment using PID control. The memory means of the air conditioning control device 31 stores a program for controlling temperature and humidity, and the CPU of the air conditioning control device 31 reads and sequentially executes the program from the memory means. In addition to the program, the memory means also stores data related to a psychrometric chart (psychrometric chart table), which represents air state values ​​on a coordinate system. Incidentally, enthalpy increases toward the upper right of the psychrometric chart and decreases toward the lower left.

[0031] PID control (Proportional-Integral-Differential Control) is a type of feedback control that controls input values ​​using three elements: the deviation between an output value and a target value, and its integral and derivative. The PID controller 35 of this embodiment has the same number of PID loops as the number of controlled objects (specifically, four). The PID controller 35 and each air conditioning device (i.e., the preheater 22, the washer 23, the cooling coil 24, and the reheater 25) are electrically connected via driver circuits (not shown). The PID controller 35 is also electrically connected to each of the sensors 27a-27c. Therefore, the PID controller 35 outputs drive control signals to each controlled object, thereby PID-controlling the manipulated variables of each air conditioning device. As a result, the outdoor air temperature and humidity are adjusted to reach the target temperature and humidity. At this time, the sensors 27a-27c input temperature and humidity measurement results. Therefore, the PID controller 35 can perform feedback control based on the measurement results.

[0032] The air supply target input unit 32 is electrically connected to the equipment control unit 34 via the air supply setting calculation unit 33. The air supply target input unit 32 is used to input target values ​​for the temperature and humidity of the conditioned air to be supplied to the painting booth, and is configured to include means such as a keyboard, a touch panel, etc. An output signal from the air supply target input unit 32 is input to the air supply setting calculation unit 33.

[0033] The air supply setting calculation unit 33 is electrically connected to each of the sensors 27a to 27c. Therefore, the measured values ​​of temperature and humidity output from each of the sensors 27a to 27c are input to the air supply setting calculation unit 33. The air supply setting calculation unit 33 performs calculations based on the input target values ​​of temperature and humidity and the measured values, and calculates the minimum enthalpy required to reach the target temperature and humidity. Then, the air supply setting calculation unit 33 sets target values ​​for the operation amounts of each air conditioning device based on the calculation results, and outputs the target values ​​to the PID controller 35.

[0034] Furthermore, the air conditioning control device 31 is equipped with a learning data collection device 41 that collects learning data of the prediction model used for model predictive control (MPC) of the paint booth air conditioner 21. The learning data collection operation performed by the learning data collection device 41 of the air conditioning control device 31 will be described below. The program for collecting learning data is stored in the storage means of the air conditioning control device 31. The CPU in the air conditioning control device 31 reads out the program from the storage means as needed and executes it sequentially.

[0035] As shown in FIG. 1, the learning data collection device 41 in this air conditioning control device 31 includes a region setting unit 42, a learning start point setting unit 43, a state point moving unit 44, a data collection unit 45, and the like.

[0036] As shown in FIG. 3 and other figures, the region setting unit 42 sets a region R1 on the psychrometric chart for which a prediction model is to be created, i.e., a region to be targeted for temperature and humidity control (control region R1). In this embodiment, for example, a CPU in the air conditioning control device 31 functions as the region setting unit 42. The control region R1 can also be described as a region for which a high-quality prediction model is required to achieve highly accurate temperature and humidity control. For example, in FIG. 3, the control region R1 is shown as a region surrounded by four line segments.

[0037] The learning start point setting unit 43 also sets multiple learning start points S1 to S9 within the set control target region R1, which serve as starting points for random vibration of the air conditioning equipment. Specifically, the number and positions of the learning start points S1 to S9, as well as the order in which they are moved, are also set (see FIG. 3). In this embodiment, for example, a CPU in the air conditioning control device 31 functions as the learning start point setting unit 43. The number of learning start points S1 to S9 must be multiple, but this number is not limited and can be set arbitrarily, for example, five or more points, and in this embodiment, nine points. The number of learning start points S1 to S9 is preferably ten or more points. The positions of the learning start points S1 to S9 are also not limited and can be set arbitrarily, for example, at positions spaced apart from each other on a psychrometric chart. In this case, the multiple learning start points S1 may be arranged along a line segment defining the control target region R1. Furthermore, the learning start point setting unit 43 may set any one of the multiple set learning start points S1 to S9 as the "first learning start point S1" that serves as a starting point when random vibration is first performed.

[0038] The state point moving unit 44 moves the conditioned air state point K1, which is located at the same position as the outside air state point G1 on the psychrometric chart, to one of the plurality of learning start points S1 to S9 by operating the air conditioning equipment to control the temperature and humidity. In this embodiment, for example, a CPU in the air conditioning control device 31 functions as the state point moving unit 44. The state point moving unit 44 in this embodiment moves the conditioned air state point K1 to one of the plurality of learning start points S1 to S9 by controlling the temperature and humidity of the air conditioning equipment using PID control.

[0039] The data collection unit 45 collects learning data by randomly exciting the air conditioning equipment while moving the conditioned air state point K1 among multiple learning start points S1-S9 within the control target region R1. In this embodiment, the data collection unit 45 uses PID control to return the conditioned air state point K1, which has been displaced from the learning start points S1-S9 due to the random exciting, back to the learning start point S1-S9. PID control is also performed when moving the conditioned air state point K1 from the current learning start point to the next learning start point. Furthermore, the data collection unit 45 performs random exciting within a range that does not exceed preset upper and lower limit values ​​for the manipulated variable of each air conditioning equipment. These upper and lower limit values ​​are set for each air conditioning equipment so that the manipulated variable does not place a heavy burden on the equipment. Note that FIG. 7(a) is a graph illustrating random exciting in the learning data collection method of this embodiment, and FIG. 7(b) is a graph illustrating random exciting in the learning data collection method of the comparative example. In both graphs, curves L1, L2, L3, and L4, which increase and decrease stepwise, represent the manipulated variables of the preheater 22, the washer 23, the cooling coil 24, and the reheater 25, respectively. Two dashed lines sandwiching each of the curves L1 to L4 represent upper and lower limit values. Although upper and lower limit values ​​are not set in the comparative example, these are depicted in FIG. 7(b) for ease of explanation. It can be seen that in the present embodiment shown in FIG. 7(a), the manipulated variable is not set beyond the set upper and lower limit values. On the other hand, in the comparative example shown in FIG. 7(b), since upper and lower limit values ​​are not specifically set, the manipulated variable may be set beyond these limits (see the point indicated by T1 in the figure).

[0040] Next, the learning data collection method of this embodiment will be described based on the flowchart of FIG. 2 and with reference to the psychrometric charts of FIGS.

[0041] First, the region setting unit 42 operates to set a control region R1 for which a prediction model is to be created on a psychrometric chart (region setting step: S110). Here, the control region R1 may be set by the operator defining a plurality of line segments, or region data previously set and stored in the device may be read and used.

[0042] Next, the learning start point setting unit 43 operates to set a plurality of learning start points S1 to S9, which will be the starting points when randomly exciting the air conditioning equipment, at predetermined positions within the set control target region R1 (learning start point setting step: step S120).The learning start point setting unit 43 also sets the order and route for moving the air-conditioned air state point K1.

[0043] Next, the state point moving unit 44 is activated to confirm the position of the outside air state point G1 (step S130). Next, the air conditioning equipment is operated using PID control to control the temperature and humidity of the air-conditioned air state point K1, which is located at the same position as the outside air state point G1 on the psychrometric chart. This causes the air conditioning air state point K1 to be moved to one of the multiple learning start points S1 to S9 (step S140: state point moving step). Specifically, for example, in the psychrometric chart of FIG. 3, the outside air state point G1 is not within the control target region R1 but is located away from the control target region R1. Therefore, PID control is performed to move the air-conditioned air state point K1 from the outside air state point G1 to the first learning start point S1 (see FIG. 4). In FIG. 3, the first learning start point S1 is the closest of the multiple learning start points S1 to S9.

[0044] Next, the data collection unit 45 operates to repeatedly control the air-conditioned air state point K1, which has been displaced from the initial learning start point S1 due to random vibration, to return it to the learning start point S1 using PID control, and collects data (step S150: data collection step, see Figure 5).

[0045] After collecting a predetermined number of data points for one learning start point S1, the data collection unit 45 proceeds to the next step S160. In step S160, it is determined whether data collection has been completed for all learning start points S1 to S9. If the determination result is NO, the data collection unit 45 controls the temperature and humidity of the air conditioner using PID control to move the conditioned air state point K1 from the current learning start point S1 to the next learning start point S2 (step S170, see FIG. 6). After executing step S170, the process returns to step S150, where the conditioned air state point K1, which has been displaced from the second learning start point S2 by the random excitation, is returned to the learning start point S2 using PID control, thereby collecting data. This process is repeated until data collection for all learning start points S1 to S9 is completed. If the determination result in step S160 is YES, i.e., if data collection for all learning start points S1 to S9 is completed, the data collection unit 45 ends the series of processes. After this, the learning data collection device 41 in the air conditioning control device 31 performs machine learning based on the identification data collected as described above, creates a prediction model to be used for model predictive control for air conditioning control, and stores this in a memory means.

[0046] Therefore, according to this embodiment, the following effects can be obtained.

[0047] (1) The air conditioning system 11 of this embodiment includes an air conditioning control device 31, which in turn includes a learning data collection device 41 that collects learning data for a predictive model used in model predictive control of the paint booth air conditioner 21. As described above, the learning data collection device 41 includes a region setting unit 42, a learning start point setting unit 43, a state point moving unit 44, and a data collection unit 45. Therefore, even if the outdoor air state point G1 is not within the set control target region R1, the state point moving unit 44 moves the conditioned air state point K1 to one of the learning start points S1-S9 within the control target region R1. After the movement, the data collection unit 45 starts collecting learning data. In other words, the starting point for data collection is not the outdoor air state point G1 at that time, but one of the learning start points S1-S9 within the control target region R1. This ensures reliable collection of data for the control target region R1, which is the target of temperature and humidity control. Furthermore, the data collection unit 45 performs random vibration on the air conditioning equipment while moving the conditioned air state point K1 among a plurality of learning start points S1 to S9 within the control target region R1, thereby making it possible to reliably collect learning data within the control target region R1.

[0048] (2) In this embodiment, in the data collection step performed by the data collection unit 45, the air-conditioned air state point K1, which has been displaced from the learning start points S1 to S9 due to random excitation, is controlled to return to the learning start point S1 to S9 by PID control. Therefore, even if random excitation is performed, correction by PID control works, so data collection can be performed without placing a burden on each air-conditioning device.

[0049] (3) In this embodiment, in the state point movement step performed by the state point movement unit 44, when the outside air state point G1 is not within the control target region R1, the conditioned air state point K1, which is located at the same position as the outside air state point G1, is moved to one of the plurality of learning start points S1 to S9. The movement in this case is performed by controlling the temperature and humidity of the air conditioning equipment using PID control. Therefore, the conditioned air state point K1, which is not within the control target region R1, can be moved quickly and efficiently to one of the learning start points S1 to S9, and the burden on each air conditioning equipment at that time can be reduced.

[0050] (4) In this embodiment, in the data collection step performed by the data collection unit 45, the temperature and humidity of the air conditioning equipment are controlled by PID control, thereby moving the conditioned air state point K1 from the current learning start point S1 to the next learning start point S2. That is, in this embodiment, the learning start point is moved sequentially starting from S1, then S2, S3, S4, and so on. This allows the conditioned air state point K1 to be moved between learning start points quickly and efficiently.

[0051] (5) In this embodiment, in the data collection step performed by the data collection unit 45, upper and lower limits for the operation amount of the air conditioning equipment when random vibration is applied are set in advance. This prevents the air conditioning equipment from being operated with an excessive operation amount when random vibration is applied. As a result, the occurrence of abnormalities in the air conditioning equipment due to excessive operation is avoided, and learning data can be collected efficiently without interruption.

[0052] Each embodiment of the present invention may be modified as follows.

[0053] In the above embodiment, four line segments are defined on the psychrometric chart to define the control region R1 for which a forecast model is to be created. However, this is not limited to this method. For example, a range defined by three or five or more line segments may be defined as the control region R1. Alternatively, multiple points may be defined and the range including these points may be defined as the control region R1.

[0054] In the above embodiment, the initial learning start point S1 is set near the outer periphery of the control target region R1, and a movement path is set that starts from the initial learning start point S1, moves to multiple learning start points S2 to S7 located near the outer periphery, and then moves to multiple learning start points S8 and S9 located in the center of the control target region R1. However, this is not limited to this. For example, the movement path may be set oppositely, such that the initial learning start point is set near the center of the control target region R1, and starts from the initial learning start point, moves to multiple learning start points located near the center, and then moves to multiple learning start points located on the outer periphery of the control target region R1.

[0055] In the above embodiment, learning data for a prediction model used in model predictive control of the paint booth air conditioner 21 is collected, and a prediction model for use in another device is created based on that data. However, the present invention is not limited to this. For example, the equipment control unit 34 may include a model predictive control unit (MPC controller) in addition to the PID controller 35, and a prediction model may be created based on the collected learning data and used within the same device.

[0056] In the above embodiment, the paint booth air conditioner 21 includes a preheater 22 (preheating device), a washer 23 (humidifying device), a cooling coil 24 (cooling device), a reheater (reheating device) 25, and a blower fan 26. However, this is not limiting and other configurations may be used. For example, the cooling coil 24 may be configured as two stages rather than one stage, or may be omitted if unnecessary. The reheater 25 may also be omitted if unnecessary. In other words, the air conditioner is not limited to one that has the functions of heating, humidifying, and cooling the outside air it takes in, but may have heating and humidifying functions but not cooling functions, or cooling and humidifying functions but not heating functions.

[0057] In the above embodiment, the air conditioning system 11 of the present invention is embodied in an air conditioning system for a painting facility that includes an air conditioner 21 for a painting booth, but it may of course be embodied in an air conditioning system that includes an air conditioner for other purposes.

[0058] In the above embodiment, the outside air state point G1, which is not within region R1, is moved to the initial learning start point S1 in the state point moving step, but this is not limited to this. For example, the closest learning start point (to which the outside air state point G1 can be moved with the smallest enthalpy) may be calculated, redefined as the initial learning start point S1, and the outside air state point G1 may be moved to the redefined initial learning start point S1.

[0059] In the above embodiment, the air supply setting calculation unit 33 calculates the minimum enthalpy based on the temperature and humidity measurement results from the three sensing means (first, second, and third sensors 27a, 27b, and 27c) provided in the paint booth air conditioner 21, and the PID controller 35 performs feedback control. However, this is not limited to this. For example, measurement results from sensing means provided in equipment other than the paint booth air conditioner 21 may be used. In another embodiment of the air conditioning system 11 shown in FIG. 8, a heat pump (HP) supplies a heat source and chilled water to the paint booth air conditioner 21. The paint booth air conditioner 21 and the heat pump are connected via a first path 61 that supplies the heat source and a second path 62 that supplies chilled water. A fourth sensor 27d is provided on the first path 61 that supplies the heat source to the preheater 22. A fifth sensor 27e is provided on the first path 61 that supplies the heat source to the reheater 25. A sixth sensor 27f is provided on the second path 62 that supplies chilled water to the cooling coil 24. Examples of the fourth sensor 27d and the fifth sensor 27e include a gas flow sensor, a steam flow sensor, and a hot water temperature flow sensor. Examples of the sixth sensor 27f include a chilled water flow sensor and a chilled water temperature sensor. The air supply setting calculation unit 33 may calculate the minimum enthalpy based on sensing information from sensors 27d to 27f in addition to sensing information from sensors 27a to 27c, and the PID controller 35 may perform feedback control. Using sensing information from energy-related sensors in this way can more accurately calculate the minimum enthalpy using the air supply setting calculation unit 33, thereby enabling a more reliable reduction in the amount of energy consumed by air conditioning. Of course, the operating power of the heat pump may be sensed and the sensing information may be used in the above-described minimum enthalpy calculation.

[0060] Next, in addition to the technical ideas set forth in the claims, the technical ideas grasped by the above-described embodiments will be listed below.

[0061] (1) In any one of the above-mentioned means 2 to 6, the PID control is performed based on sensing information from a temperature and humidity sensor provided in the air conditioner.

[0062] (2) In any of the above means 2 to 6, the PID control is performed based on sensing information from an energy-related sensor installed on a path that fluidly connects the air conditioner and a heat pump that supplies a heat source and chilled water to the air conditioner. [Explanation of symbols]

[0063] 11:Air conditioning system 21: Paint booth air conditioner as an air conditioning device 22: Preheater as air conditioning equipment 23: Washer as air conditioning equipment 24: Cooler as air conditioning equipment 25: Reheater as air conditioning equipment 31: Air conditioning control device 41: Learning data collection device 42: Area setting section 43: Learning starting point setting section 44: State point movement section 45: Data collection section R1: Control area as a region K1: Conditioned air state point G1: Outside air state point S1~S9: Starting point of learning

Claims

1. A method for collecting learning data of a prediction model used in model predictive control of an air conditioning device that adjusts the temperature and humidity of taken-in outdoor air using multiple types of air conditioning equipment, comprising: a region setting step of setting a region on a psychrometric chart for creating the prediction model; a learning start point setting step of setting a plurality of learning start points that serve as starting points when performing random vibration on the air conditioning equipment within the set region; a state point moving step of moving an air-conditioned air state point that is at the same position as the outdoor air state point on the psychrometric chart to one of the plurality of learning start points by operating the air-conditioning equipment to control temperature and humidity; a data collection step of collecting the learning data by performing random vibration on the air conditioning equipment while moving the air-conditioned air state point among the plurality of learning start points; A learning data collection method for controlling an air conditioner, comprising:

2. 2. The learning data collection method for controlling an air conditioning device according to claim 1, wherein in the data collection step, the air-conditioned air state point displaced from the learning start point by the random vibration is controlled to return to the learning start point by PID control.

3. 3. The learning data collection method for controlling an air conditioning device according to claim 1, wherein in the state point moving step, the air-conditioned air state point that is located at the same position as the outside air state point when the outside air state point is not within the region is moved to one of the plurality of learning start points by controlling the temperature and humidity of the air-conditioning equipment using PID control.

4. 4. The learning data collection method for controlling an air conditioning device according to claim 1, wherein in the data collection step, the temperature and humidity of the air conditioning equipment are controlled by PID control, thereby moving the air-conditioned air state point from the current learning start point to the next learning start point.

5. 5. The learning data collection method for controlling an air conditioning device according to claim 1, wherein in the data collection step, upper and lower limit values ​​of the operation amount of the air conditioning equipment when the random vibration is performed are set.

6. 6. The learning data collection method for air conditioning device control according to claim 1, wherein the air conditioning device is an air conditioning device for a paint booth, the air conditioning device including a preheating device, a humidifying device, a cooling device, and a reheating device.

7. An air conditioning system including an air conditioning device that adjusts the temperature and humidity of taken-in outside air using a plurality of types of air conditioning equipment, and an air conditioning control device that controls the operation amount of the air conditioning equipment, The air conditioning control device includes a learning data collection device that collects learning data of a prediction model used in model predictive control of the air conditioning device, and the learning data collection device an area setting unit that sets an area on a psychrometric chart for creating the prediction model; a learning start point setting unit that sets a plurality of learning start points that serve as starting points when random vibration is applied to the air conditioning device within the set region; a state point moving unit that moves an air-conditioned air state point that is at the same position as the outdoor air state point on the psychrometric chart to one of the plurality of learning start points by operating the air-conditioning equipment to control temperature and humidity; a data collection unit that collects the learning data by performing random vibration on the air conditioning equipment while moving the air-conditioned air state point between the plurality of learning start points; An air conditioning system comprising:

Images