Air-conditioning controller

Abstract

To improve estimation accuracy of room temperature when the room temperature is estimated from measured temperatures by temperature sensors built in an air-conditioning controller.SOLUTION: An air-conditioning controller 23 includes a control board 50 and an enclosure 30 housing the control board 50. Electronic parts mounted on the control board 50 includes a P / S 55, an LDO 56, an MCU 61 and a graphic IC 63. These electronic parts act as a heat source to increase temperature in the enclosure 30 during operation. A first region 91 of the control board 50 where the P / S 55, the LDO 56, the MCU 61 and the graphic IC 63 are arranged becomes higher in temperature than a second region 92 that is a peripheral region thereof. A temperature sensor 67 is arranged in the first region 91, and a temperature sensor 68 is arranged in the second region 92. Measured temperatures by these temperature sensors 67, 68 are inputted in a learning model generated by machine learning, and thereby a room temperature is estimated.SELECTED DRAWING: Figure 8

Description

[Technical field]

[0001] The present invention relates to an air conditioning controller. [Background technology]

[0002] For example, some air conditioning equipment installed in buildings such as houses includes an air conditioning controller that allows a user to perform various operations such as temperature setting, and an air conditioning device that performs air conditioning based on instructions from the air conditioning controller. In recent years, a technology has been proposed in which a temperature sensor that measures the internal temperature of the air conditioning controller is housed in the housing of the air conditioning controller, and air conditioning control is performed based on the measurement value of this temperature sensor.

[0003] Many of the electrical components mounted on the control board generate heat during operation. If the temperature inside the housing rises due to heat from the electrical components, the temperature sensor's measurement value will be higher than the actual room temperature. In other words, the measured internal temperature may not match the room temperature. In consideration of this situation, Patent Document 1 takes measures such as adding a heat source temperature sensor that measures the heat source temperature near the heat source (controller) and estimating the room temperature taking the heat source temperature into consideration. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2022-150688 A Summary of the Invention [Problem to be solved by the invention]

[0005] Here, depending on the positional relationship between the air conditioning controller and the air outlet of the air conditioner, the wind direction of the air conditioner, the air volume, etc., the wind from the air outlet may hit the air conditioning controller. The temperature (internal temperature) of the air conditioning controller may drop due to the influence of this wind. In this regard, the air conditioning controller described in the above Patent Document 1 uses a temperature sensor for measuring the internal temperature in combination with a temperature sensor for measuring the heat source temperature, and calculates the room temperature simply by subtracting the difference between the heat source temperature and the internal temperature multiplied by an influence coefficient (a constant calculated by experiment) from the heat source temperature. The influence coefficient is based on the premise that the heat source temperature, the internal temperature, and the actual room temperature have a predetermined temperature gradient relationship, and that this temperature gradient is constant. It is described that by using this influence coefficient, even if the temperature of the air conditioning controller drops due to the wind, the deviation between the room temperature and the estimated temperature can be suppressed. However, while this method can easily improve the estimation accuracy, it can also be said that it is difficult to further improve the estimation accuracy due to its characteristics. In a configuration in which the room temperature is estimated using a temperature sensor built into the air conditioning controller, there is still room for improvement in terms of being able to accurately estimate the room temperature even in a situation where it is affected by a heat source or wind. In a configuration in which air conditioning control is performed according to the estimated temperature, improving the estimation accuracy can contribute to optimizing air conditioning control, so there is great technical significance in improving the estimation accuracy.

[0006] The present invention has been made in consideration of the problems exemplified above, and its main object is to improve the estimation accuracy when estimating room temperature from the temperature measured by a temperature sensor built into an air conditioning controller. [Means for solving the problem]

[0007] Means for solving the above problems will be described below.

[0008] First means: An air conditioning controller having a control board that controls an air conditioning unit and a housing that houses the control board, wherein the control board is equipped with heat-generating components that serve as a heat source that raises the temperature inside the housing when electricity is passed through them, and wherein temperature sensors are respectively arranged at multiple positions on the control board where temperature differences occur due to heat from the heat-generating components, and wherein the controller has a learning model for estimating room temperature generated by machine learning, and wherein each of the temperature data from each of the temperature sensors is input to the learning model, and an estimation unit uses the learning model to estimate the temperature of the room in which the air conditioning controller is installed.

[0009] As shown in this means, if a configuration is adopted in which temperature data from a plurality of temperature sensors is input into a learning model for estimating room temperature to estimate the room temperature, it becomes possible to estimate the room temperature with high accuracy. [Brief description of the drawings]

[0010] [Figure 1] 1 is a schematic diagram showing a building and an air conditioning system according to a first embodiment. [Diagram 2] FIG. 2 is a perspective view of the air conditioning controller as viewed from the front side. [Diagram 3] FIG. 2 is a block diagram showing the electrical configuration of the air conditioning controller. [Figure 4] FIG. 4 is a schematic diagram showing types of electrical components. [Diagram 5] FIG. 1 is a schematic diagram for explaining a problem. [Figure 6] FIG. 1 is a schematic diagram showing a configuration related to generation of a learning model. [Figure 7] Schematic diagram showing the difference between temperature change due to the influence of a heat source and temperature change due to the influence of wind. [Figure 8] FIG. 4A is a front view of the control board, FIG. 4B is a distribution diagram showing temperature distribution, and FIG. 4C is a schematic diagram comparing the first region and the second region. [Figure 9] FIG. 2 is a schematic diagram showing a configuration related to room temperature estimation. [Figure 10] 11 is a schematic diagram illustrating an example of a difference between an actual room temperature and an estimated temperature. [Figure 11]FIG. 13A is a schematic diagram showing a configuration related to generation of a learning model in the second embodiment; FIG. 13B is a schematic diagram showing a configuration related to estimation of room temperature. [Figure 12] 11 is a schematic diagram illustrating an example of a difference between an actual room temperature and an estimated temperature. [Figure 13] FIG. [Figure 14] FIG. [Figure 15] FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] <First embodiment> DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will now be described with reference to the drawings. In this embodiment, the present invention is embodied in an air conditioning system installed in a building.

[0012] 1, an air conditioning system 20 provided in a building 10 includes an air conditioner 21a installed indoors (machine room 11), an outdoor unit 21b installed outdoors, heat exchange piping 24 connecting the air conditioner 21a and the outdoor unit 21b, intake ducts 25, 26 that take in air for air conditioning from rooms 15, 16, etc. of the building 10, and supply ducts 27, 28 that supply the conditioned air to the rooms 15, 16, etc. Note that the building 10 shown in this embodiment is a highly airtight and highly insulated house, and the air conditioning system 20 air-conditions the entire building (so-called whole-building air-conditioning).

[0013] Air outlets 27a, 28a of supply ducts 27, 28 are provided in each of the rooms 15, 16 (for example, in the ceiling). Movable louvers are provided in these air outlets 27a, 28a, and the air direction can be changed by changing the position of these louvers.

[0014] An air conditioning controller 23 that controls the air conditioning device 21 (air conditioner 21a and outdoor unit 21b) is attached to the partitions (walls 17, 18) of each of the rooms 15, 16. The air conditioning controller 23 is connected to the air conditioning device 21 etc. so as to be able to communicate with them. The air conditioning controller 23 is capable of performing various setting operations such as setting the operation mode of the air conditioning device 21 (cooling mode, heating mode, dehumidification mode, fan mode), setting the temperature, setting the air volume, setting the air direction, etc., and when these settings are changed, it transmits a command to the air conditioning device 21 indicating that. The air conditioning device 21 switches the operation mode etc. based on the received command.

[0015] 2, the air conditioning controller 23 comprises a touch panel display 40 and a housing 30 that houses the display 40, with the display 40 exposed through an opening formed in the front portion 31 of the housing 30. It can be said that the housing 30 and the display 40 form the outer shell of the air conditioning controller 23. The housing 30 has a thin rectangular parallelepiped shape, and is sized so that when air flows around the air conditioning controller 23, the air conditioning controller 23 (excluding the parts of the housing 30 that face the walls 17 and 18) is assumed to come into contact with the air almost evenly.

[0016] A horizontally long rectangular control board 50 is housed in the housing 30. The control board 50 is disposed so as to face the rear surface of the display 40, and although details will be described later, one of the plate surfaces of the control board 50 (the plate surface facing the display 40) is provided with various electric components and wiring patterns connecting these electric components.

[0017] The housing 30 has a peripheral portion 32 formed along the edge of the control board 50, and a plurality of ventilation holes 34 penetrating from the inside to the outside are formed in the peripheral portion 32. Some of the electrical components mounted on the control board 50 generate heat when electricity is applied, and when the temperature inside the housing 30 rises due to the heat generated by these electrical components, the warmed air inside the housing 30 is exhausted through the ventilation holes 34, and indoor air flows into the housing 30 through these ventilation holes 34.

[0018] Next, the electrical configuration of the air conditioning controller 23 will be described with reference to FIG. 3. The control board 50 of the air conditioning controller 23 is provided with a power supply unit 51 that supplies power from an external power source to the display 40 and various electrical components mounted on the control board 50. The power supply unit 51 includes a connector 52 that connects the control board 50 to the external power source, an RSC 54 (Rectify-Smooth Circuit) that is a conversion unit that converts the power supplied from the external power source from AC power to DC power, a polyswitch 53 provided between the connector 52 and the RSC 54, a P / S 55 (Power Supply) that is a stable supply unit for realizing a stable supply of the DC power converted by the RSC 54, and an LDO 56 (Low Drop Out) that is a regulator that reduces the voltage of the power from the P / S 55 to a predetermined voltage. Note that the power (operating power) required for the display 40 and various electrical components during operation differs, so a plurality of LDOs 56 are provided corresponding to each operating power.

[0019] The control board 50 is also provided with an MCU 61 as a control unit that performs main control of the air conditioning, a WD 62 (WATCHDOG) as a monitoring unit that monitors whether the MCU 61 is operating normally, a graphic IC 63 that controls the display of the display 40 based on commands from the MCU 61, a touch IC 64 that detects a change in current (change in capacitance) at a part touched by a user's finger when the touch panel of the display 40 is operated and transmits the detected information to the MCU 61, a sound output unit 65 that outputs operation sounds and messages, a communication unit 66 that communicates with the air conditioner 21, and temperature sensors 67 and 68 that measure the temperature inside the housing 30. The temperature sensors 67 and 68 are thermistor types, but the specific configuration is arbitrary as long as they can be accommodated in the housing 30. For example, they may be a resistance temperature detector type or a linear resistor type.

[0020] The electrical components mounted on the control board 50 are roughly classified into "first type electrical components" that become a heat source that increases the temperature inside the housing 30 when electricity is applied, and "second type electrical components" that generate a negligible amount of heat even when electricity is applied, and do not actually become a heat source that increases the temperature inside the housing 30. In other words, if the amount of heat that can increase the temperature inside the housing 30 is defined as a reference heat value, electrical components that generate a heat value greater than the reference heat value are "first type electrical components," and electrical components that generate a heat value less than the reference heat value are "second type electrical components." In this embodiment, the "first type electrical components" correspond to "heat-generating components."

[0021] As shown in Figure 4, the P / S 55, LDO 56, and MCU 61 have high heat generation temperatures and the amount of heat generated far exceeds the reference amount of heat generation for temperature rise. The graphic IC 63 has a lower heat generation temperature than the MCU 61 and other components, but the amount of heat generated exceeds the reference amount of heat generation. These P / S 55, LDO 56, MCU 61, and graphic IC 63 are classified as "Type 1 electrical components."

[0022] In contrast, the heat generation temperatures of the connector 52, polyswitch 53, RSC 54, WD 62, touch IC 64, audio output unit 65, communication unit 66, and temperature sensors 67 and 68 are all low, so the heat generation amount is below the reference heat generation amount. The connector 52, polyswitch 53, RSC 54, WD 62, touch IC 64, audio output unit 65, communication unit 66, and temperature sensors 67 and 68 correspond to the "second type electrical components." Incidentally, in addition to the various electrical components described above, the control board 50 also includes a connector, a humidity sensor, a memory, and the like, which also correspond to the "second type electrical components."

[0023] The air conditioning controller 23 shown in this embodiment is capable of measuring the temperature inside the housing 30 (hereinafter also referred to as the internal temperature) using the temperature sensors 67, 68 built into the housing 30. Then, the room temperature is estimated from the measured internal temperature, and air conditioning control is performed based on the estimated temperature. As described above, the first-class electric components built into the housing 30 together with the temperature sensors 67, 68 become a heat source that increases the internal temperature of the housing 30. Specifically, when the first-class electric components generate heat, the heat is transferred from the first-class electric components to the control board 50, and the temperature increases not only in the first-class electric components but also in the surrounding areas. In other words, the temperature sensors 67, 68 mounted on the control board 50 are also influenced by the heat sources to a certain extent, and the temperatures measured by the temperature sensors 67, 68 become higher than the actual room temperature (hereinafter also referred to as the actual room temperature).

[0024] Here, in addition to heat generation inside the housing 30, wind from the air conditioning controller 23 is one of the factors that affect the internal temperature. That is, when wind from the air conditioning device 21 hits the air conditioning controller 23, the influence of the wind can cool the housing 30 and the display 40, causing a drop in the internal temperature of the housing 30. In the following explanation, a distinction is made between a state in which the wind from the air conditioning device 21 hits the air conditioning controller 23 or a state in which the wind does not hit the air conditioning controller 23 directly but creates an air flow around the air conditioning controller 23 that causes a drop in the internal temperature, and a state in which the wind blows out from the air conditioning device 21 but does not hit the air conditioning controller 23 and does not create an air flow around the air conditioning controller 23 that causes a drop in the internal temperature, or a state in which the temperature becomes appropriate and blowing is temporarily stopped, as a "no wind state."

[0025] For example, as illustrated in FIG. 5, when the difference between the measured temperature and the room temperature is maintained substantially constant in a windless state, the estimated temperature can be easily adjusted to the room temperature by correcting the measured temperature using a predetermined correction value. Here, when the windless state is switched to the blowing state, heat is taken from the air conditioning controller, and the internal temperature drops. This causes the temperature measured by the temperature sensor to drop. On the other hand, the actual room temperature does not change as much as the internal temperature. Therefore, when the same correction as in the windless state is performed, the corrected temperature (estimated temperature) and the actual room temperature will deviate by the amount of the drop in the internal temperature. In other words, the correction is excessive. In addition, the influence of the wind varies depending on the positional relationship between the air conditioning controller 23 and the air outlets (e.g., air outlets 27a, 28a) of the air conditioner 21, the operating conditions of the air conditioner 21 (wind direction, air volume, set temperature), etc. Therefore, it is difficult to deal with the influence of the heat source and the influence of the wind by the simple correction as described above.

[0026] In recent years, as described in the above Patent Document 1, a technology has been proposed that can deal with the influence of heat sources and wind in a simple manner, but there is still room for improvement in the configuration related to room temperature estimation in order to improve the estimation accuracy under various conditions. One of the features of this embodiment is that the room temperature can be accurately estimated from the internal temperature of the housing 30 by using machine learning. The characteristic configuration of this embodiment will be described below. Note that the machine learning in this embodiment is performed in a factory laboratory, and the product is shipped in a state in which a machine learning model for room temperature estimation (hereinafter referred to as a learning model) generated by machine learning is stored in the MPU 61.

[0027] The MPU 61 is provided with a model generation unit 61b that generates a learning model (see FIG. 3). The model generation unit 61b applies so-called supervised learning as a learning algorithm executed by the model generation unit 61b. Supervised learning is a method of learning a model for estimating results for new conditions by identifying features that suggest a correlation between conditions and results from a data set of conditions and their corresponding results (so-called teacher data). In this supervised learning, a neural network is used to construct an operation setting model.

[0028] As shown in FIG. 6, in the model generation unit 61b shown in this embodiment, the measured temperatures (learning data) of the temperature sensor 67, the measured temperatures (learning data) of the temperature sensor 68, and the measured temperatures (teaching data) of the room temperature sensor 100 that measures the room temperature of the laboratory in which the air conditioning controller 23 is installed are used as input data, and a multi-layered calculation according to a neural network is performed to generate a learning model capable of outputting an appropriate estimated temperature as output data.

[0029] When generating a learning model, learning is performed in the above-mentioned windless state and in the blowing state using an air conditioner installed in a lab. In particular, in learning in the blowing state, the control parameters of the air conditioner's wind direction and air volume are appropriately changed to vary the influence of wind in various ways. This creates a pseudo environment similar to the state in which the air conditioner is installed in a house or the like (an environment close to actual use).

[0030] To generate a learning model appropriately, it is necessary to use data that is correlated with the two feature quantities, the effect of the heat source and the effect of the wind, as input data. Here, we will provide additional explanation on these feature quantities with reference to Figure 7. Figure 7(a) is a schematic diagram showing how heat is transmitted, and Figure 7(b) is a schematic diagram showing how heat is dissipated.

[0031] In FIG. 7(a), points P1 and P2 on the control board are at different distances from the heat source. Therefore, the heat from the heat source is transferred to points P1 and P2 through the control board, but the way in which the heat is transferred differs depending on the distance from the heat source. If the temperature inside the housing 30 rises due to the influence of the heat source, and a difference occurs between the temperature of point P1 and the temperature of point P2, it can be said that the influence of the heat source is reflected in the difference. If the influence of the heat source is considered as a characteristic amount, the larger the temperature difference, the better. And if the temperature is high due to the influence of the heat source, the range by which the temperature can be lowered by cooling will also be large. In other words, it can be said that the larger the temperature difference, the better, even when the influence of wind is considered as a characteristic amount.

[0032] Also, at point P1, the temperature rises rapidly due to the influence of the heat source, while at point P2 the temperature rise is slower than at point P1. In other words, there is a difference between point P1 and point P2 not only in the final reached temperature but also in the rate at which the temperature rises.

[0033] In contrast, as shown in Figure 7(b), when wind hits the housing (air conditioning controller), the temperature of the housing drops more or less evenly. The heat of the control board escapes through the air and the contact points with the housing, but there is little difference in the heat dissipation rate between points P1 and P2. In other words, the difference in temperature between points P1 and P2 is unlikely to change.

[0034] For the above reasons, the relationship between the temperature of point P1 and the temperature of point P2 differs between when the temperature rises due to the influence of a heat source and when the temperature drops due to the influence of wind. In other words, even if the temperature of point P1 is a predetermined value, the temperature of point P2 is likely to differ between when the temperature rises due to the influence of a heat source (no wind state) and when the temperature drops due to the influence of wind (blowing state).

[0035] If we assume that points P1 and P2 are close to each other, the behavior when the temperature rises and the behavior when the temperature falls will be close to each other. This is because the temperature difference between points P1 and P2 is hidden by noise, measurement errors, etc., making it difficult to distinguish between the effects of the heat source and the effects of wind.

[0036] In this embodiment, taking these circumstances into consideration, the measured temperatures of the two built-in temperature sensors 67, 68 are used as input data, and the temperature sensors 67, 68 are arranged in separate areas where the temperature difference becomes large when affected by a heat source. The arrangement of the temperature sensors 67, 68 will be further explained below with reference to Fig. 8. Note that Fig. 8(a) is a plan view of the control board 50, but the electrical components and wiring patterns mounted on the control board 50 are omitted for convenience.

[0037] The first-class electric components, namely, the P / S 55, the LDO 56, the MCU 61, and the graphic IC 63, are arranged together in a specific location (hereinafter referred to as the first region 91) on the control board 50. By gathering the first-class electric components, which are heat sources, the first region 91 is configured to have a higher temperature than the surrounding region (hereinafter referred to as the second region 92) on the control board 50. The first region 91 also has second-class electric components arranged therein, and has a denser wiring pattern than the second region 92. The wiring patterns easily transmit heat from the heat source, and the heat is released from these wiring patterns. In other words, the densely arranged wiring patterns are also one of the factors that cause the temperature of the first region 91 to be higher.

[0038] The first-class electric components are not all maintained in a constantly energized state, and some of the first-class electric components may stop operating depending on the situation. Since the first-class electric components are concentrated in the first region 91, even if some of the first-class electric components stop operating, the range of temperature fluctuation in the first region 91 is small.

[0039] The temperature sensor 67 is disposed in this first region 91. Specifically, although the temperature sensor 67 is disposed slightly away from the first-class electric components, the first region 91 is filled with the first-class electric components that act as heat sources, and therefore the temperature at the measurement point is sufficiently high. In other words, in order to properly generate a learning model, it is preferable that the difference in temperature measured by the temperature sensors 67 and 68 is large, but a configuration has been realized in which it is not necessary to forcibly dispose the temperature sensor 67 close to the first-class electric components in order to ensure the temperature difference. This is also preferable in terms of achieving coexistence of the temperature sensor 67 and other electric components on the control board 50.

[0040] Heat from the heat source is transmitted mainly through the control board 50 or through the air in the housing 30, but the influence of the mode of transmission through the control board 50 is greater than the influence of the mode of transmission through the air in the housing 30. Here, to provide a supplementary explanation on the arrangement of the temperature sensor 67, the temperature sensor 67 is located in a range where heat from the P / S 55, LDO 56, MCU 61, and graphic IC 63 is transmitted through the control board 50 (including the wiring pattern), so it can be said that the temperature is easily increased by the influence of the heat source. Note that the temperature sensor 68 is located in a position where little heat is transmitted through the control board 50, but is also affected by heat transmitted through the air, so it can be said that it is located in a position where the temperature is at least higher than room temperature.

[0041] The second region 92 is larger than the first region 91, and is a region where second-class electric components are mainly arranged. Since the second region 92 and the first region 91 are connected, heat generated in the first region 91 is transferred to the second region 92 via the control board 50, but the temperature in the second region 92 is basically lower than that in the first region 91, and the temperature difference with the first region 91 becomes more significant as the region moves away from the first region 91. Electrical components with low heat resistance are arranged in the second region 92 to suppress the effects of heat.

[0042] A linear slit 75 is formed in the control board 50 so as to surround a part of the second region 92, specifically the lower right corner 72. By forming the slit 75 and partially interrupting the heat transfer path, the transfer of heat to the corner 72, i.e., the influence of the heat source, is suppressed. As a result, the temperature of the corner 72 surrounded by the slit 75 becomes lower than that of other parts of the second region 92. The above-mentioned ventilation hole 34 is located near the corner 72.

[0043] The temperature sensor 68 is disposed in this corner 72. To further explain the relationship between the temperature sensor 68, the first type electric components, and the slits 75, the slits 75 extend so as to intersect with an imaginary straight line FL connecting the temperature sensor 68 and each of the first type electric components when viewed from the front of the control board 50. This is preferable in terms of efficiently suppressing the effects of the heat source.

[0044] By arranging the temperature sensor 67 in the first region 91 and the temperature sensor 68 in the second region 92 (particularly the corner 72), under conditions where there is an influence of a heat source, the temperature measured by the temperature sensor 67 differs greatly from the temperature measured by the temperature sensor 68. This is advantageous in terms of appropriately generating a learning model and improving the estimation accuracy of the room temperature.

[0045] 9, a learning model 81 generated using input data from the temperature sensors 67, 68 is stored in the MCU 61 (estimation unit 61a). In the estimation unit 61a, the measured temperatures of the temperature sensors 67, 68 are input to this learning model 81, and an estimated temperature is obtained as output data from the learning model 81. The MCU 61 performs air conditioning control based on this estimated temperature.

[0046] FIG. 10 shows an example of the results of estimating room temperature using the learning model 81. According to this result, the difference between the estimated temperature and the actual room temperature is within the range of ±2°F except when the range is significantly exceeded at the timing when switching from a no-wind state to an air blowing state, and most of the difference is within the target range of ±1°F in both the no-wind state and the air blowing state. Similar results are obtained even when conditions such as the position of the air conditioning controller 23 are changed. Since high estimation accuracy can be ensured under various conditions, it can contribute to optimizing air conditioning control.

[0047] The temperature sensors 67, 68 built into the air conditioning controller 23 may have different measured temperatures due to differences in the influence of heat sources. In addition, if there is a difference in temperature depending on the positions of the temperature sensors 67, 68, there may also be a difference in temperature or the amount of change when the temperature inside the housing 30 drops due to the wind from the air conditioner 21 hitting the air conditioning controller 23. That is, the influence of the heat source and the influence of the wind are reflected in the measured temperatures of the temperature sensors 67, 68. Here, the way the wind hits the air conditioning controller 23 varies, and the temperature of the heat source is not necessarily stable. That is, in the environment in which the air conditioning controller 23 is actually used, there may be differences in the influence of the heat source and the influence of the wind due to various factors. In this regard, as shown in this embodiment, if a learning model 81 is generated based on learning data indicating the temperature measured by the temperature sensors 67, 68 and teacher data indicating the room temperature, and the temperature data from the temperature sensors 67, 68 is input to the learning model 81 for room temperature estimation to estimate the room temperature, the room temperature can be accurately estimated.

[0048] In generating the learning model 81, it is preferable that the influence of the heat source and the influence of the wind are significantly different depending on the placement locations of the temperature sensors 67, 68. Therefore, when multiple heat-generating components are mounted, the temperature sensors 67, 68 are placed so that the distance from each heat-generating component to the temperature sensor 67 is smaller than the distance from each heat-generating component to the temperature sensor 68, thereby suitably improving the accuracy of the learning model 81 generated from the learning data related to each temperature sensor 67, 68.

[0049] In the air conditioning controller 23, heat may be generated by the components constituting the power supply unit 51 and the MCU 61, which is the control unit. Therefore, by arranging the temperature sensors 67, 68 based on these heat-generating components, a practically preferable configuration can be realized.

[0050] In addition to the heat-generating components (first-class electric components), various electric components (second-class electric components) are mounted on the control board 50, and furthermore, wiring patterns are formed to connect these components. Therefore, if an attempt is made to arrange the temperature sensor 67 close to a specific heat-generating component, it may be difficult to coexist with these components and wiring patterns. In this regard, according to the configuration shown in the present embodiment in which a plurality of heat-generating components are concentrated so that the temperature of the first region 91 is high, the degree of freedom in the arrangement of the temperature sensor 67 within the first region 91 is increased. This makes it possible to suppress stricter restrictions on the arrangement of other components and wiring patterns.

[0051] In a small control board 50 as shown in this embodiment, the distance from the heat source to the low-temperature side temperature sensor 68 is small, making it difficult to ensure a significant temperature difference between the temperature sensors 67, 68. In this regard, if the configuration is such that the slits 75 are used to suppress the transfer of heat to the temperature sensor 68 as shown in this embodiment, a significant temperature difference can be ensured even in cases where it is difficult to ensure a sufficient distance from the heat source to the temperature sensor 68. In other words, the slits 75 shown in this embodiment are technically significant in realizing a miniaturized control board 50 (air conditioning controller 23).

[0052] <Second embodiment> In the first embodiment, the learning model 81 is generated using the temperature measured by the temperature sensor 67 and the temperature measured by the temperature sensor 68 as learning data. The temperature sensors 67 and 68 are thermistors whose resistance changes with temperature, and are configured to obtain analog output. Due to the nature of analog output, the output changes in small increments. Therefore, the temperature measured by the temperature sensors 67 and 68 includes unnecessary variable factors. When a learning model is generated using learning data including such variable factors, the reliability of the model may be reduced by the variable factors. In particular, it is advantageous to use multiple temperature sensors 67 and 68 in combination to generate a learning model that corresponds to the influence of a heat source and the influence of wind, but the learning variation increases due to the overlap of variable factors for each learning data. One of the features of the present embodiment is that a design that takes such circumstances into consideration is made. Hereinafter, the design will be described with reference to FIG. 11. Note that in FIG. 11(a), the illustration of an air conditioner (see FIG. 7) for model generation is omitted.

[0053] In this embodiment, the temperatures measured by the temperature sensors 67 and 68 are not input as learning data as they are, but are smoothed by the moving average filter (filtering or smoothing) of the MCU 61 before being input. In other words, the smoothed data becomes learning data. The temperature sensors 67 and 68 measure temperatures at a period of 1 second, and the moving average filter calculates the average value from the past 60 temperature measurements.

[0054] The room temperature sensor 100 installed in the lab is also a thermistor type, like the temperature sensors 67 and 68. The temperature measured by the room temperature sensor 100 is not input as training data as is, but is smoothed by the moving average filter (filtering or smoothing) of the MCU 61 before being input. In other words, the smoothed data becomes training data. The measurement cycle of the room temperature sensor 100 is 1 second, and the moving average filter calculates the average value from the past 60 temperature measurements.

[0055] In this way, by making both the learning data and the teacher data smoothed data, the above-mentioned fluctuation factors are prevented from being reflected in the learning model 81X.

[0056] Also, when estimating the room temperature using the learning model 81X, the measured values ​​by the temperature sensors 67 and 68 are not input directly to the learning model 81X, but are smoothed by a moving average filter (filtering or smoothing process) before being input to the learning model 81X. The specific manner of smoothing is the same as that when generating the learning model, and therefore will not be described.

[0057] According to the second embodiment described above in detail, as illustrated in Fig. 12 (experimental results), the difference between the actual room temperature and the estimated temperature is significantly reduced. Specifically, not only is the difference within ±1°F except for the timing of switching from the no-wind state to the blowing state, but the variation is approximately half that of the experimental results in the first embodiment (see Fig. 10). In other words, most of the difference is within ±0.5°F. For these reasons, it can be said that the accuracy of estimating the room temperature using the learning model 81X can be further improved by smoothing the learning data and teacher data using a moving average.

[0058] <Other embodiments> The present invention is not limited to the contents of the above-mentioned embodiments, and may be implemented, for example, as follows. Each of the following configurations may be applied individually to each of the above-mentioned embodiments, or may be applied in combination with some or all of the following configurations to each of the above-mentioned embodiments. It is also possible to arbitrarily combine all or some of the various configurations shown in each of the above-mentioned embodiments. In this case, it is preferable that the technical significance (effects exerted) of each of the configurations to be combined is guaranteed. Each of the following configurations may be applied individually to a new configuration consisting of a combination of the embodiments, or may be applied in combination with some or all of the following configurations.

[0059] In each of the above embodiments, when estimating room temperature using the learning model 81, the temperatures measured by the temperature sensors 67, 68 are each smoothed by a moving average before being input to the learning model 81, but this is not limited to the above. As shown in FIG. 13, the output value from the learning model 81 may be smoothed by a moving average, that is, a moving average filter may be disposed on the output side instead of the input side, and the smoothed data may be used as the estimated temperature. According to the configuration shown in this modified example, the configuration related to smoothing can be simplified. However, if emphasis is placed on improving the estimation accuracy, it is preferable to provide a moving average filter on the input side.

[0060] In the above embodiments, the learning model 81 is generated and the room temperature is estimated using the temperature sensor 67 arranged in the first region 91 and the temperature sensor 68 arranged in the second region 92. Although the temperature of the entire first region 91 increases due to the concentration of heat sources, a certain degree of temperature difference occurs depending on the distance from the heat source, and the heat may be transmitted differently. Therefore, multiple temperature sensors 67 may be individually arranged in the first region 91 at locations where the influence of the heat source differs, that is, at locations where the temperature during heat generation differs. For example, in the example shown in FIG. 14(a), the temperature sensor 67A is arranged at a position where the temperature is relatively high in the first region 91, and the temperature sensor 67B is arranged at a position where the temperature is relatively low in the first region 91, specifically, at a position closer to the second region 92 in the first region 91. If the learning model is generated using these temperature sensors 67A, 67B and the temperature sensor 68, it can contribute to further optimization of the learning model. In addition, when the configuration shown in this modified example is used, the temperature data to be input to the learning model during temperature estimation should be temperature data related to temperature sensor 67A, temperature data related to temperature sensor 67B, and temperature data related to temperature sensor 68.

[0061] Also, a significant temperature difference occurs between the corner 72 and other parts in the second region 92. In consideration of this situation, as shown in Fig. 14(b), a temperature sensor 69 may be added to a part of the second region 92 other than the corner 72, and a learning model may be generated based on learning data from three temperature sensors 67 to 69.

[0062] Incidentally, in both of the modified examples shown in Figures 14(a) and 14(b), a configuration is described in which a learning model is generated based on learning data from three temperature sensors, but the number of temperature sensors may be four, and it is not intended to deny the use of five or more temperature sensors.

[0063] In the above-mentioned embodiments, the learning data related to the temperature sensor 67 and the learning data related to the temperature sensor 68 arranged in the first region 91 are used to generate the learning model 81. However, the arrangement of the temperature sensors may be changed as desired as long as a significant difference can be ensured in at least both the influence of the heat source and the influence of the wind. For example, as shown in FIG. 15(a), instead of the temperature sensor 67 in the first region 91, the temperature sensor 69 may be arranged in the second region 92 (excluding the corner 72), and the learning data related to the temperature sensor 68 and the learning data related to the temperature sensor 69 may be used to generate the learning model. In particular, in a configuration in which the slit 75 formed so as to surround the corner 72 suppresses the transfer of heat to the corner 72, the temperature difference between the corner 72 and other parts in the second region 92 can be increased. Therefore, even when the temperature sensors 68 and 69 are arranged in the second region 92, the significant difference described above can be easily ensured by using the slit 75 (a configuration that suppresses the transfer of heat) in combination. This is also preferable in terms of improving the degree of freedom in the arrangement of the temperature sensors.

[0064] In each of the above embodiments, the temperature sensor 68 is disposed in the second region 92 of the control board 50 at a position (corner 72) near the ventilation hole 34. Considering that a significant temperature difference occurs in the second region 92 from the first region 91 as described above, it is also possible to dispose the temperature sensor 68 in a portion of the second region 92 other than the corner 72. For example, as shown in Fig. 15(b), it is preferable to dispose the temperature sensor 68 in the corner 73 of the control board 50 that is farthest from the first region 91 (a position where the influence of the heat source is small).

[0065] In each of the above embodiments, the slits 75 are used to suppress the transfer of heat to the corners 72, but the specific configuration for suppressing the transfer of heat is arbitrary. For example, a groove may be provided instead of the slits 75, or a heat dissipation member that dissipates heat more easily than the control board 50 may be provided instead of the slits.

[0066] In the above embodiments, when generating the learning model 81, the test air conditioner is operated while changing the control parameters of the airflow direction and the airflow volume, but this is not necessarily limited to this. For example, the control parameter to be changed may be only the airflow direction, or only the airflow volume. Also, in addition to the airflow direction and the airflow volume, the set temperature may be changed.

[0067] In each of the above embodiments, the learning model 81 for room temperature estimation is generated by the model generation unit 61b of the air conditioning controller 23, but this is not limited to the above. It is also possible to generate a learning model in a device (e.g., a learning device) other than the air conditioning controller 23, and store the generated learning model in the air conditioning controller 23.

[0068] <Inventions extracted from the above embodiments> The following describes the features of the inventions extracted from the above-mentioned embodiments, while indicating, as necessary, their effects, etc. In the following, for ease of understanding, the corresponding configurations in the above-mentioned embodiments are appropriately indicated in parentheses, but the present invention is not limited to the specific configurations indicated in parentheses.

[0069] Feature 1. An air conditioning controller (air conditioning controller 23) having a control board (control board 50) that controls an air conditioning device (air conditioning device 21) and a housing (housing 30) that houses the control board, Heat-generating components (MCU 61, P / S 55, LDO 56, etc.) that become a heat source that increases the temperature inside the housing when current is applied are mounted on the control board, and temperature sensors (temperature sensors 67, 68, etc.) are disposed at a plurality of positions on the control board where a temperature difference occurs due to heat from the heat-generating components, An air conditioning controller having a learning model (learning model 81) for estimating room temperature generated by machine learning, and an estimation unit (estimation unit 61a) that inputs each of the temperature data from each of the temperature sensors to the learning model and uses the learning model to estimate the temperature of the room in which the air conditioning controller is installed.

[0070] The multiple temperature sensors built into the air conditioning controller may have different measured temperatures due to differences in the influence of heat sources. In addition, if there is a difference in temperature depending on the position of the temperature sensor, there may also be a difference in temperature or the amount of change when the temperature inside the housing drops due to wind from the air conditioner hitting the air conditioning controller. In other words, the influence of the heat source and the influence of the wind are reflected in the measured temperature of each temperature sensor. Here, the way the wind hits the air conditioning controller varies, and the temperature of the heat source is not necessarily stable. In other words, in the environment in which the air conditioning controller is actually used, there may be differences in the influence of the heat source and the influence of the wind due to various factors. In this regard, as shown in this feature, if the room temperature is estimated by inputting temperature data from multiple temperature sensors into a learning model for room temperature estimation, the room temperature can be accurately estimated.

[0071] Feature 2. An air conditioning controller according to Feature 1, wherein the learning model is generated by machine learning based on learning data indicating temperatures detected by each of the temperature sensors and teacher data indicating room temperature.

[0072] As shown in Feature 1, the influence of heat sources and wind appears in the temperature detected by each temperature sensor. Therefore, as shown in this feature, if a learning model is generated based on learning data indicating the detected temperature and teacher data indicating the room temperature, it is possible to accurately estimate the room temperature even when it is influenced by a heat source or wind.

[0073] Feature 3. A plurality of the heat-generating components are mounted on the control board, The temperature sensors include a first temperature sensor (temperature sensor 67) and a second temperature sensor (temperature sensor 68), 3. The air conditioning controller according to feature 1 or 2, wherein a distance from each of the heat generating components to the first temperature sensor is shorter than a distance from each of the heat generating components to the second temperature sensor.

[0074] In generating a learning model, it is preferable that the influence of heat sources and wind is significantly different for each temperature sensor. Therefore, when multiple heat-generating components are mounted, the temperature sensors are arranged so that the distance from each heat-generating component to the first temperature sensor is shorter than the distance from each heat-generating component to the second temperature sensor, thereby suitably improving the accuracy of the learning model generated from the learning data indicating the temperatures detected by each temperature sensor.

[0075] Feature 4: The air conditioning controller according to Feature 3, wherein the heat-generating components include components (e.g., P / S 55 and LDO 56) that constitute a power supply unit (power supply unit 51) on the control board, and a control unit (MCU 61) that controls the air conditioning device.

[0076] In an air conditioning controller, the components constituting the power supply unit and the control unit can generate a lot of heat. Therefore, by arranging the temperature sensors based on these heat-generating components, a practically preferable configuration can be realized.

[0077] Feature 5. The control board has a plurality of heat-generating components mounted thereon; The air conditioning controller according to feature 1 or 2 has, as the temperature sensors, a first temperature sensor (temperature sensor 67) arranged in a specified area (first area 91) where the heat-generating components are concentrated, and a second temperature sensor (temperature sensor 68) arranged in another area (second area 92) outside the specified area.

[0078] By gathering multiple heat generating components in a specific area, it is possible to not only increase the temperature in the specific area but also reduce the temperature fluctuation. If a first temperature sensor is placed in such an area and a second temperature sensor is placed in another area, the effects of the heat source and wind will be more easily reflected in the temperatures detected by each temperature sensor.

[0079] In addition to the heat-generating components, various electric components are mounted on the control board, and wiring patterns are formed to connect these components. Therefore, if the first temperature sensor is placed close to a specific heat-generating component, it may be difficult to coexist with these components and wiring patterns. In this regard, according to the configuration in which multiple heat-generating components are concentrated so that the temperature of the entire predetermined area is high, as shown in this feature, the degree of freedom in placing the temperature sensor in the predetermined area is increased. This makes it possible to suppress stricter restrictions on the placement of other components and wiring patterns.

[0080] Feature 6. The control board has a rectangular shape, An air conditioning controller as described in feature 5, wherein the specified area is located on one side of a pair of sides of the control board, and the second temperature sensor is arranged in a portion of the other area that is closer to the other of the pair of sides.

[0081] With this arrangement, the difference between the temperature detected by the first temperature sensor and the temperature detected by the second temperature sensor can be made significant.

[0082] Feature 7: An air conditioning controller as described in Feature 5 or Feature 6, wherein the heat-generating components arranged in the specified area include components (e.g., P / S 55 and LDO 56) that constitute a power supply unit (power supply unit 51) on the control board, and a control unit (MCU 61) that controls the air conditioning device.

[0083] In an air conditioning controller, the components constituting the power supply unit and the control unit can generate a lot of heat. Therefore, by arranging the temperature sensors based on these heat-generating components, a practically preferable configuration can be realized.

[0084] Feature 8. The air conditioning controller according to any one of Features 3 to 7, wherein the first temperature sensor is disposed at a position where heat is transferred from the plurality of heat-generating components through the control board.

[0085] By placing the first temperature sensor on the higher temperature side in the range where heat is transmitted from multiple heat-generating components through the control board, the difference between the temperature detected by the first temperature sensor and the temperature detected by the second temperature sensor can be made significant.

[0086] Feature 9 (3 temperature sensors). The first temperature sensor is provided in plurality. The air conditioning controller according to any one of features 3 to 8, wherein the first temperature sensors are arranged at positions that are differently affected by heat generated by the heat-generating component.

[0087] If multiple first temperature sensors on the high temperature side are prepared and the first temperature sensors are arranged in positions where the influence of heat generation is different, it can contribute to further improving the estimation accuracy by the learning model. Note that, when generating the learning model, it is preferable to configure it so that the temperatures detected by each first temperature sensor are input as learning data.

[0088] Feature 10 (Board shape): An air conditioning controller described in any one of Features 3 to 8, in which a suppression section (slit 75) is provided in a portion of the control board between the heat-generating component and the second temperature sensor, for suppressing the transfer of heat from the heat-generating component toward the position of the second temperature sensor.

[0089] If the control board is large, it is easy to increase the distance from the heat source to the second temperature sensor. However, if an attempt is made to miniaturize the control board, the distance from the heat source to the second temperature sensor becomes smaller, making it difficult to ensure a significant temperature difference for each temperature sensor. In this regard, if a configuration is adopted in which the suppression unit suppresses the transfer of heat to the second temperature sensor as shown in this feature, a significant temperature difference can be ensured even in cases where it is difficult to ensure the distance from the heat source to the second temperature sensor. In other words, the configuration shown in this feature has technical significance in realizing the miniaturization of the control board (air conditioning controller).

[0090] Feature 11 (moving average). The learning model is generated by machine learning based on learning data indicating temperatures detected by each of the temperature sensors and teacher data indicating room temperature; 2. The air conditioning controller according to feature 1, wherein the learning data is data obtained by smoothing the temperatures detected by the respective temperature sensors using a moving average.

[0091] If the temperature detected by the temperature sensor contains unnecessary variables and is input as learning data, the variables may become a factor that hinders the optimization of the learning model. In particular, when learning data is provided from multiple temperature sensors, proper learning becomes difficult due to the variables remaining in each learning data. Therefore, by using data smoothed by a moving average as shown in this feature, such concerns can be eliminated and the accuracy of room temperature estimation can be further improved.

[0092] Feature 12 (Double Moving Average): The temperature data that is input to the learning model during the estimation is data obtained by smoothing the temperatures detected by each of the temperature sensors using a moving average. An air conditioning controller as described in Feature 11.

[0093] By smoothing the detected temperature using a moving average as the temperature data input to the learning model when estimating room temperature, it is possible to contribute to improving the accuracy of room temperature estimation. [Explanation of symbols]

[0094] 20...air conditioning equipment, 21...air conditioning device, 23...air conditioning controller, 30...housing, 50...control board, 51...power supply unit, 55...P / S, 56...LDO, 61...MCU, 61a...estimation unit, 61b...model generation unit, 63...graphics IC, 67...temperature sensor, 68...temperature sensor, 81...learning model, 91...first area, 92...second area, FL...virtual line.

Claims

1. An air conditioning controller having a control board for controlling an air conditioner and a housing for accommodating the control board, a heat generating component that becomes a heat source that increases a temperature inside the housing when current is applied is mounted on the control board, and temperature sensors are disposed at a plurality of positions on the control board where a temperature difference occurs due to heat from the heat generating component, An air conditioning controller having a learning model for estimating room temperature generated by machine learning, each of the temperature data from each of the temperature sensors being input to the learning model, and having an estimation unit that uses the learning model to estimate the temperature of the room in which the air conditioning controller is installed.

2. A plurality of the heat generating components are mounted on the control board, The temperature sensors include a first temperature sensor and a second temperature sensor, The air conditioning controller according to claim 1 , wherein a distance from each of the heat generating components to the first temperature sensor is smaller than a distance from each of the heat generating components to the second temperature sensor.

3. A plurality of the heat generating components are mounted on the control board, 2. The air conditioning controller according to claim 1, wherein the temperature sensors include a first temperature sensor arranged in a predetermined area where the heat-generating components are concentrated, and a second temperature sensor arranged in another area outside the predetermined area.

4. The air conditioning controller according to claim 2 , wherein the first temperature sensor is disposed at a position where heat is transferred from the plurality of heat-generating components through the control board.

5. 4. The air conditioning controller according to claim 1, wherein the learning model is generated by machine learning based on learning data indicating temperatures detected by each of the temperature sensors and teacher data indicating room temperature.

6. The learning model is generated by machine learning based on learning data indicating temperatures detected by each of the temperature sensors and teacher data indicating room temperature, The air conditioning controller according to claim 5 , wherein the learning data is data obtained by smoothing the temperatures detected by the respective temperature sensors using a moving average.

Images