Control device, control method, and program

Abstract

This control device comprises: an extraction unit that extracts, from time-series data obtained by measuring the indoor temperature of an air-conditioned space, an element included in a monitoring section stretching from a prescribed start condition for air-conditioning control by an air conditioner to an end condition therefor; an estimation unit that estimates, on the basis of the time-series data of the monitoring section, a time constant estimation value and a dead time estimation value of a primary delay response of the air-conditioning control; and a learning unit that learns a lookup table in which an explanatory variable, which is a condition affecting the air conditioning of the air-conditioned space, is associated with the dead time estimation value and the time constant estimation value.

Description

Control device, control method, and program

[0001] This application claims priority to Japanese Patent Application No. 2023-142859, filed on September 4, 2023, the contents of which are incorporated herein by reference.

[0002] For example, when an air conditioner is operated in cooling mode, it takes time for the room temperature to reach the target temperature set by the user after it is turned on, and the room temperature during that time is uncomfortable for the person in the room (user). For this reason, in order to make the room comfortable by the time the user arrives, people manually turn on the air conditioner (indoor unit) before the time they start to stay in the room, or use a function (warming-up operation) that automatically starts pre-cooling a certain time before (for example, 15 minutes or 30 minutes before).

[0003] The appropriate pre-cooling time not only differs for each air-conditioned space, such as a room, but also changes depending on conditions such as the outside temperature. Therefore, if the automatic pre-cooling start time is fixed, as in the warm-up operation described above, there is a possibility that the pre-cooling time may be too long or too short. If the air conditioner's pre-cooling time is too long, it will waste energy and ultimately lead to higher electricity bills. Also, if the pre-cooling time is too short, comfort will be reduced.

[0004] In addition, air conditioners are designed to be controlled in response to changes in conditions such as outside temperature, taking into account the dead time between when an operation is performed and when the result appears, and thermal response parameters including first-order lag (see, for example, Patent Documents 1 and 2).

[0005] Japanese Patent Application Laid-Open No. 2001-350503 Japanese Patent Application Laid-Open No. 2002-054837

[0006] Sufficient computing resources (memory and CPU) are required to calculate the thermal response parameters in real time according to constantly changing conditions and to calculate the pre-cooling time. However, most general air conditioner control devices (microcomputers) are equipped with only limited computing resources.

[0007] An object of the present disclosure is to provide a control device, a control method, and a program that can learn a lookup table that can predict thermal response parameters to a set temperature quickly and with low load, even when only limited computing resources are installed.

[0008] According to one aspect of the present disclosure, the control device includes an extraction unit that extracts elements included in a monitoring period from a predetermined start condition to an end condition of air conditioning control by an air conditioner from time series data measured of the indoor temperature of the air-conditioned space, an estimation unit that estimates a time constant estimate and a dead time estimate of a first-order lag response of the air conditioning control based on the time series data of the monitoring period, and a learning unit that learns a lookup table that associates explanatory variables, which are conditions that affect the air conditioning of the air-conditioned space, with the dead time estimate and the time constant estimate.

[0009] According to one aspect of the present disclosure, a control method includes the steps of: extracting elements included in a monitoring interval from a predetermined start condition to an end condition of air conditioning control by an air conditioner from time series data measuring the indoor temperature of the air-conditioned space; estimating a time constant estimate and a dead time estimate of a first-order lag response of the air conditioning control based on the time series data of the monitoring interval; and learning a lookup table that associates explanatory variables, which are conditions that affect the air conditioning of the air-conditioned space, with the dead time estimate and the time constant estimate.

[0010] According to one aspect of the present disclosure, the program causes the control device to execute the following steps: extracting elements included in a monitoring period from a predetermined start condition to an end condition of air conditioning control by an air conditioner from time series data measuring the indoor temperature of the air-conditioned space; estimating a time constant estimate and a dead time estimate of a first-order lag response of the air conditioning control based on the time series data of the monitoring period; and learning a lookup table that associates explanatory variables, which are conditions that affect the air conditioning of the air-conditioned space, with the dead time estimate and the time constant estimate.

[0011] According to the above aspect, even when only limited computational resources are installed, it is possible to learn a lookup table that can predict thermal response parameters to a set temperature quickly and with low load.

[0012] 1 is a diagram showing a functional configuration of an air conditioning system according to a first embodiment. FIG. 2 is a diagram showing an example of a prediction model (lookup table) according to the first embodiment. FIG. 3 is a diagram showing an example of an explanatory variable and level category assignment according to the first embodiment. FIG. 4 is a flowchart showing an example of a thermal response parameter estimation process according to the first embodiment. FIG. 5 is a flowchart showing an example of a prediction model learning process according to the first embodiment. FIG. 6 is a block diagram showing an example of a lookup table update process according to the first embodiment. FIG. 7 is a flowchart showing an example of a predicted arrival time prediction process according to the first embodiment. FIG. 8 is a block diagram showing an example of a predicted arrival time prediction process according to the first embodiment. FIG. 9 is a diagram showing an example of a display of a terminal device according to the first embodiment. FIG. 10 is a diagram showing an example of a lookup table update process according to the first embodiment.

[0013] <First embodiment> A control method according to the first embodiment will be described below with reference to the drawings. The control method according to this embodiment can be applied to the control of various heat source devices such as air conditioners, water heaters, and refrigerators, but the following description will be given taking the case where the control method is applied to an air conditioner as an example.

[0014] (Overall Configuration of Air Conditioning System) FIG. 1 is a diagram showing the functional configuration of an air conditioning system according to a first embodiment. As shown in FIG. 1, the air conditioning system 100 includes an air conditioner 1 and a terminal device 20. The air conditioner 1 includes an outdoor air temperature sensor 2, an indoor temperature sensor 3, a memory 4, an indoor unit 5, and an outdoor unit 6. The outdoor air temperature sensor 2 and the indoor temperature sensor 3 are connected to a control device 10, and the control device 10 acquires the temperatures measured by the outdoor air temperature sensor 2 and the indoor temperature sensor 3. The indoor unit 5 and the outdoor unit 6 are connected to the control device 10, and the control device 10 controls the indoor unit 5 and the outdoor unit 6 and detects their operating states. The control device 10 and the terminal device 20 are connected to each other via a network NW so as to be able to communicate with each other. The terminal device 20 is, for example, a smartphone or tablet equipped with an application that enables remote operation of the air conditioner 1.

[0015] (Functional Configuration of Control Device) The control device 10 is a computer including a CPU (Central Processing Unit) such as a microcomputer and an MPU (Micro Processing Unit). The control device 10 includes an information acquisition unit 11, an extraction unit 12, an estimation unit 14, a learning unit 15, a prediction unit 16, a control unit 17, and a communication unit 18.

[0016] The information acquisition unit 11 acquires the outdoor temperature measured by the outdoor temperature sensor 2 and the indoor temperature measured by the indoor temperature sensor 3 .

[0017] The extraction unit 12 extracts elements included in the monitoring period from a specified start condition to an end condition of the air conditioning control by the air conditioner 1 from time series data measuring the indoor temperature in the room (air-conditioned space) where the air conditioner 1 is installed.

[0018] The estimation unit 14 estimates a time constant estimate value τ and a dead time estimate value L of the first-order lag response of air conditioning control based on the time-series data of the monitoring section.

[0019] The learning unit 15 learns a prediction model (lookup table) that associates explanatory variables, which are conditions that affect the air conditioning of the air-conditioned space, with the estimated dead time value L and the estimated time constant value τ. The learned lookup table is stored in the memory 4.

[0020] The prediction unit 16 predicts a predicted time D until the air-conditioned space reaches the target temperature, based on the trained prediction model (lookup table) read from the memory 4 and the current explanatory variables.

[0021] The control unit 17 controls the indoor unit 5 and the outdoor unit 6. Based on the set temperature and set time settings received from the terminal device 20 via the communication unit 18 and the predicted arrival time D predicted by the prediction unit 16, the control unit 17 operates the indoor unit 5 and the outdoor unit 6 so that the air-conditioned space approaches the set temperature at the set time. For example, the control unit 17 starts air conditioning at a time that is the predicted arrival time D before the set time or slightly before that time, and performs cooling operation or heating operation so that the set temperature is achieved by the set time.

[0022] The communication unit 18 communicates with the terminal device 20. For example, the communication unit 18 transmits the predicted arrival time D predicted by the prediction unit 16 to the terminal device 20. The communication unit 18 may also receive setting conditions such as the set temperature and set time, and operation instructions such as turning on the power of the air conditioner 1 from the terminal device 20.

[0023] (Regarding the Prediction Model) FIG. 2 is a diagram illustrating an example of a prediction model (lookup table) according to the first embodiment. FIG. 3 is a diagram illustrating an example of the assignment of explanatory variables and level categories according to the first embodiment. As shown in FIG. 2 , in this embodiment, the prediction model of the thermal response parameter is represented by a lookup table LUT of the time constant estimate value τ and the dead time estimate value L for each operation mode. In the example of FIG. 2 , the table of the time constant estimate value τ and the table of the dead time estimate value L each have a table for each operation mode (heating operation and cooling operation). Therefore, the lookup table LUT according to this embodiment has four tables: a time constant table LUT1 for cooling operation, a dead time table LUT2 for cooling operation, a time constant table LUT3 for heating operation, and a dead time table LUT4 for heating operation. While the example of FIG. 2 illustrates a table using two explanatory variables, this is not limiting. In other embodiments, the number of explanatory variables may be one or more. Each explanatory variable has multiple level categories. The level categories of the explanatory variables are used as indices to identify the cells of each table.

[0024] For example, as shown in Fig. 3, assume that explanatory variable 1 is air volume and explanatory variable 2 is outdoor temperature. For air volume, a level category is assigned for each air volume setting value of the fan of the indoor unit 5. For outdoor temperature, a level category is assigned for each range of measurement values ​​of the outdoor temperature sensor 2. Note that, as in the example of Fig. 3, different temperature ranges may be set for the outdoor air temperature level category during cooling operation and heating operation.

[0025] 2 and 3, each table stores in each cell an estimated value of the thermal response parameter (time constant τ or dead time L) corresponding to the combination of air volume and outdoor temperature for each operation mode (heating operation or cooling operation). This allows different thermal response parameters to be updated and referenced depending on the settings of the air conditioner 1, environmental conditions, etc. Each cell also stores the number of samples n (the number of times the thermal response parameter is recorded). The process of updating the lookup table (learning process) and the process of referencing it (prediction process) will be described later.

[0026] (Process for Estimating Thermal Response Parameters) FIG. 4 is a flowchart showing an example of a process for estimating thermal response parameters according to the first embodiment. Here, the flow of the process by which the control device 10 estimates thermal response parameters will be described with reference to FIG. 4. When the air conditioner 1 is operating, the control device 10 repeatedly performs the series of processes shown in FIG. 2 at predetermined control intervals (for example, every 50 ms). During this time, the information acquisition unit 11 sequentially acquires sensor measurement values ​​and set values ​​such as the outdoor air temperature, indoor temperature, and air volume from the outdoor air temperature sensor 2, indoor temperature sensor 3, indoor unit 5, etc.

[0027] First, the extraction unit 12 determines whether the start condition of the monitoring period has been met (step S100). The monitoring period is a period for monitoring (extracting and storing) data to estimate thermal response parameters in the air conditioning control of the air conditioner 1. For example, the extraction unit 12 determines that the start condition of the monitoring period has been met when the indoor unit 5 is switched on or the set temperature changes.

[0028] If the start condition is not met (step S100; NO), the extraction unit 12 ends the process. On the other hand, if the start condition is met (step S100; YES), the extraction unit 12 starts extracting and storing necessary information (step S101). Thereafter, the extraction unit 12 records the measured indoor temperature values ​​(time-series data) in a buffer secured in the memory 4 at each predetermined control period. In addition, only when the extraction unit 12 starts recording the indoor temperature, it records the operating mode of the air conditioner (heating operation or cooling operation) and explanatory variables used in the learning process described below in the memory 4. The explanatory variables include, for example, the time when the measurement value storage started, the outdoor air temperature, the airflow rate of the indoor unit 5, and the temperature gap between the indoor temperature and the set temperature.

[0029] Next, the extractor 12 determines whether a predetermined termination condition is met (step S102). For example, the extractor 12 determines that the termination condition is met when the end of the buffer (the upper limit of the number or amount of time-series data) is reached, when the indoor unit 5 is switched off, when the set temperature is changed, etc.

[0030] If the termination condition is not met (step S102; NO), the extraction unit 12 temporarily terminates the process. At this time, the time-series data of the indoor temperature continues to be stored. Therefore, until the extraction unit 12 determines that the termination condition is met, the indoor temperature at each step of the control cycle is extracted and stored in a buffer. On the other hand, if the termination condition is met (step S102; YES), the extraction unit 12 further determines whether the monitoring interval is a valid interval (step S103).

[0031] For example, if the length of the data (time-series data of indoor temperature) recorded in the buffer is less than a threshold value, or if the indoor temperature does not reach the set temperature by the time the buffer reaches its end, the extraction unit 12 determines that the monitoring section is invalid (step S103; NO) if the time-series data does not meet a predetermined standard (i.e., it is difficult to estimate the thermal response parameters based on the time-series data). In this case, the extraction unit 12 clears the buffer (step S106) and ends the process. On the other hand, if the time-series data meets the standard, the extraction unit 12 determines that the monitoring section is valid (step S103; YES).

[0032] Next, the estimator 14 estimates thermal response parameters (step S105). The estimator 14 according to the present embodiment estimates the length of dead time and the time constant of the first-order lag response as thermal response parameters. Known techniques may be used to estimate the dead time estimate L and the time constant estimate τ of the first-order lag response.

[0033] The estimation unit 14 stores the thermal response parameters, including the dead time estimated value L and the time constant estimated value τ, in the memory 4 in association with the previously recorded operation mode and explanatory variables (step S106). Thereafter, the estimation unit 14 clears the buffer (step S107) and ends the process.

[0034] (Prediction Model Learning Process) Fig. 5 is a flowchart showing an example of a prediction model learning process according to the first embodiment. Fig. 6 is a block diagram showing an example of a lookup table update process according to the first embodiment. Here, the flow of the process by which the control device 10 learns a prediction model will be described with reference to Figs. 5 and 6. Note that the control device 10 executes the series of processes shown in Fig. 5 each time the estimator 14 estimates a thermal response parameter.

[0035] First, the learning unit 15 reads from the memory 4 the estimated value of the thermal response parameter newly estimated by the estimation unit 14 and the explanatory variables recorded together with the estimated value (step S200).

[0036] Next, the learning unit 15 determines the index of the cell in which the newly estimated value of the thermal response parameter is to be stored (step S201).

[0037] First, the learning unit 15 selects either the heating operation table or the cooling operation table as the storage destination based on the operation mode read from the memory 4. Furthermore, the learning unit 15 determines the index of the storage destination cell based on the values ​​of the explanatory variables read from the memory 4. For example, assume that the operation mode is heating operation, explanatory variable 1 (air volume) is Middle, and explanatory variable 2 (outdoor temperature) is 6°C. According to the example of FIGS. 2 and 3, explanatory variable 1 corresponds to level 2, and explanatory variable 2 corresponds to level 3. Therefore, in this case, the learning unit 15 selects the cell where the row of level 2 of explanatory variable 1 intersects with the column of level 3 of explanatory variable 2 in the time constant table LUT3 for heating operation and the dead time table LUT4 for heating operation shown in FIG. 2 as the storage destination cell for the time constant estimate τ(tau_est) and the dead time estimate L(lag_est), respectively.

[0038] Next, the learning unit 15 calculates the average value of the new estimated value and the past estimated value (step S202). First, the learning unit 15 reads out the values ​​recorded in the storage destination cells of each table as the past average values ​​(tau_avg_old, lag_avg_old).

[0039] The learning unit 15 also calculates an average value (tau_avg_new) between the new time constant estimate (tau_est) and the average value (tau_avg_old) of the past time constant estimates read from the storage cell. Similarly, the learning unit 15 calculates an average value (lag_avg_new) between the new time delay estimate (lag_est) and the average value (lag_avg_old) of the past time delay estimates read from the storage cell. The learning unit 15 may calculate the new average value by weighting the average value with a weight α, as shown in the following equations (1) and (2). The weight α may be set to any value within the range of 0<α≦1. For example, by setting the weight α to a value greater than 0.5, the influence of the new estimate is significantly reflected, and the effect of forgetting the past estimate can be achieved. This allows the estimated value of the thermal response parameter to track the actual value, for example, when the thermal response parameter gradually changes due to aging or other reasons.

[0040] tau_avg_new=α*tau_est+(1-α)*tau_avg_old…(1)

[0041] lag_avg_new=α*lag_est+(1-α)*lag_avg_old…(2)

[0042] The learning unit 15 inputs the calculated new average values ​​(tau_avg_new, lag_avg_new) into the storage destination cells of each table to update the lookup table LUT (step S203). At this time, the learning unit 15 increments the number of samples n in these storage destination cells by +1.

[0043] If the number of samples n of the storage cell is 0, i.e., if this is the first time that thermal response parameters have been input to this storage cell, the learning unit 15 does not calculate the average value, but instead inputs the thermal response parameter estimates (tau_est, lag_est) read from the memory 4 directly into the storage cell, thereby updating the lookup table LUT (step S202).

[0044] (Prediction process of predicted arrival time) FIG. 7 is a flowchart showing an example of the prediction process of predicted arrival time according to the first embodiment. FIG. 8 is a block diagram showing an example of the prediction process of predicted arrival time according to the first embodiment. Here, the flow of the process of predicting the predicted arrival time at which the indoor temperature will reach the set temperature will be described with reference to FIGS. 7 and 8. Note that the control device 10 executes the series of processes shown in FIG. 7 when, for example, an application of the terminal device 20 is started or operated.

[0045] First, the information acquisition unit 11 acquires explanatory variables to be used for prediction (step S300). For example, if explanatory variable 1 is air volume and explanatory variable 2 is outdoor temperature, the information acquisition unit 11 acquires the air volume setting value of the air conditioner 1 (indoor unit 5) and also acquires the current outdoor temperature from the outdoor temperature sensor 2. The information acquisition unit 11 may also simultaneously acquire the current set temperature (or, if the power is OFF, the set temperature at the time of the previous operation).

[0046] Next, the prediction unit 16 reads out the thermal response parameters corresponding to the obtained explanatory variables (step S301). For example, if the current operation mode (or the previous operation mode, if the power is OFF) is heating operation, the prediction unit 16 reads out the parameters (tau_avg, lag_avg) corresponding to the explanatory variables from the time constant table LUT3 and the dead time table LUT4 for heating operation. The method of determining the index of the storage cell for each parameter is the same as step S201 in FIG. 6.

[0047] The prediction unit 16 also predicts a predicted time D until the indoor temperature of the air-conditioned space reaches the target temperature based on the read thermal response parameters (step S302). The predicted time D is calculated, for example, using the following formula (3). The target temperature is expressed, for example, as x% (any value, such as 90%, 95%, etc.) of the current set temperature. The coefficient C of the time constant τ (tau_avg) is a value equal to or greater than 1 and can be changed depending on the target temperature. For example, if the target temperature is 95% of the set temperature, C=3, and if the target temperature is 90% of the set temperature, C=2.3.

[0048] D=C×tau_avg+lag_avg…(3)

[0049] Next, the communication unit 18 transmits the predicted arrival time D predicted by the prediction unit 16 to the terminal device 20 (step S303).

[0050] 10 is a diagram showing an example of a display of a terminal device according to the first embodiment. For example, as shown in FIG. 10, the application screen of the terminal device 20 displays information such as current information C201 and predicted information C202, as well as an operation reception section C203 that receives remote operation of the air conditioner 1 by the user. The current information C201 includes information such as the current date and time, the indoor temperature, the outdoor temperature, and the set temperature. The predicted information C202 includes a predicted arrival time D received from the control device 10 of the air conditioner 1. The predicted information C202 may also include a predicted time obtained by adding the predicted arrival time D to the current time. The operation reception section C203 may also include a "Now ON / OFF button," a "Timer setting ON / OFF button," a "Timer date and time setting button," and the like.

[0051] The user remotely controls the air conditioner 1 via the operation reception unit C203 while referring to the current time and the predicted time D until the set temperature is reached. For example, in the example of FIG. 10 , the predicted time D until the set temperature is reached is approximately 13 minutes, and the user plans to be in the air-conditioned space for about 15 minutes. In this case, the user operates the "ON / OFF NOW button" to power on the air conditioner 1. The control unit 17 of the control device 10 then receives an operation instruction (power ON) from the terminal device 20 via the communication unit 18 (step S304) and controls the air conditioner 1 (indoor unit 5) to power on (step S305).

[0052] Furthermore, for example, if there is time before the user enters the air-conditioned space, the user may operate the "Timer Setting ON / OFF" button in the operation reception unit C203 of the terminal device 20 to turn on the timer setting. The user may also operate the timer date and time setting to set the indoor temperature to the set temperature at any set time. In this case, the control unit 17 of the control device 10 receives an operation instruction (timer setting ON and set time) from the terminal device 20 via the communication unit 18 (step S304). The control unit 17 automatically starts air conditioning at a time that is the predicted arrival time D before the set time or slightly before that time, so that the air-conditioned space reaches the set temperature at the set time set by the user, and performs cooling or heating operation to achieve the set temperature by the set time (step S305).

[0053] The user may also change the current set temperature. For example, the user may be able to change the set temperature by tapping the set temperature portion of the current information C201. In response, the control device 10 receives an operation instruction (an instruction to change the set temperature) from the terminal device 20 via the communication unit 18. In this case, the control device 10 may return to step S302 and re-predict the predicted arrival time.

[0054] (Actions and Effects) As described above, the control device 10 according to this embodiment includes an extraction unit 12 that extracts elements included in a monitoring period from a predetermined start condition to an end condition from time series data obtained by measuring the indoor temperature of the space to be air-conditioned, an estimation unit 14 that estimates a time constant estimate τ and a dead time estimate L of the first-order lag response of the air-conditioning control based on the time series data of the monitoring period, and a learning unit 15 that learns a lookup table LUT that associates explanatory variables, which are conditions that affect the air-conditioning of the space to be air-conditioned, with the dead time estimate L and the time constant estimate τ.

[0055] In this way, the control device 10 can learn a lookup table that serves as a substitute for a prediction model using a simpler and lighter algorithm than learning a prediction model using a statistical method, machine learning, etc. In other words, even if the control device 10 is equipped with only limited computational resources, it can learn a lookup table LUT that can predict thermal response parameters to a set temperature quickly and with low load.

[0056] In addition, the learning unit 15 determines a storage cell corresponding to the explanatory variable from the lookup table LUT, and updates the dead time estimate value and the time constant estimate value in the lookup table LUT by taking a weighted average of the new estimated value by the estimation unit 14 and the past estimated value recorded in the storage cell.

[0057] In this way, the control device 10 can update the lookup table LUT to reflect new estimated values ​​and gradually forget past estimated values, thereby allowing the influence of, for example, aging on the thermal response parameters to be reflected in the lookup table LUT.

[0058] The control device 10 further includes a prediction unit 16 that predicts a predicted time D required for the air conditioned space to reach the target temperature based on the lookup table LUT and the current explanatory variables.

[0059] In this way, the control device 10 can use the thermal response parameters in the lookup table LUT to accurately predict the predicted time D until the temperature of the air-conditioned space reaches the target temperature.

[0060] Second Embodiment Next, a second embodiment will be described with reference to Fig. 10. Components common to the above-described embodiment will be assigned the same reference numerals, and detailed description thereof will be omitted.

[0061] In the first embodiment, the learning unit 15 updates the new thermal response parameter estimates (tau_est, lag_est) by reflecting them in only one specific storage cell. In the present embodiment, the learning unit 15 updates the new thermal response parameter estimates in the storage cell and its surrounding cells by reflecting them.

[0062] FIG. 10 is a diagram illustrating an example of a lookup table update process according to the first embodiment. Here, an example of updating the time constant table LUT1 for cooling operation will be described. For example, as shown in FIG. 10 , the learning unit 15 determines cell P1 as the storage cell for the new time constant estimate value tau_est (step S201 in FIG. 5 ). In this embodiment, in step S201, the learning unit 15 determines at least one cell surrounding the storage cell P1 as a target cell in which the new time constant estimate value tau_est will be reflected. The surrounding target cell is, for example, cell P2 adjacent to the storage cell P1 on the top, bottom, left, and right. The surrounding target cells may also include cell P3, which is diagonally adjacent to the storage cell P1.

[0063] Next, as in the first embodiment, the learning unit 15 inputs the weighted average of the new time constant estimate tau_est and the previous time constant estimate tau_avg_old to the storage cell P1 (step S203 in FIG. 5). Furthermore, the learning unit 15 according to this embodiment applies the new time constant estimate tau_est to the target cells P2 and P3 surrounding the storage cell P1, in addition to the storage cell P1.

[0064] Specifically, if the target cells P2 and P3 do not have a previously recorded average value tau_avg_old of the time constant estimates (i.e., if the value is null), the learning unit 15 directly assigns the new time constant estimate tau_est to the target cells P2 and P3. On the other hand, if the target cells P2 and P3 have a previously recorded average value tau_avg_old of the time constant estimates (i.e., if the value is non-null), the learning unit 15 calculates a new time constant average value tau_avg_new using equation (1) (step S202 in FIG. 5 ). At this time, the learning unit 15 calculates the average value for the target cells P2 and P3 by using a smaller weight α than that for the storage cell P1. The learning unit 15 also inputs the calculated new time constant average value tau_avg_new into the target cells P2 and P3 to update the table (step S203 in FIG. 5 ).

[0065] In the example of FIG. 10 , among the target cells P2 and P3, cells P2a, P2c, P3a, P3b, and P3c are null, so the learning unit 15 directly assigns the new time constant estimate tau_est to these cells. Furthermore, since cells P2b and P3d are non-null, the learning unit 15 uses equation (1) to calculate the weighted average of the new time constant estimate tau_est and the average value tau_avg_old of past time constant estimates, and assigns the new average value tau_avg_new of the time constant estimate to cells P2b and P3d, respectively. Note that the value of the weight α may be different for each of the target cells P2 and P3, so that the value of the weight α decreases as the distance between the storage destination cell P1 and the target cell increases. In the example of FIG. 10 , the value of the weight α of the target cell P2 may be greater than the value of the weight α of the target cell P3.

[0066] The learning unit 15 also performs the same processing on the estimated dead time value.

[0067] (Action, effect) As described above, in the control device 10 according to the second embodiment, the learning unit 15 determines at least one target cell P2, P3 surrounding the storage cell P1, and updates the dead time estimate and the time constant estimate in the lookup table LUT by taking a weighted average of the new estimate by the estimation unit 14 and the past estimates recorded in the target cells P2, P3.

[0068] For example, depending on the operating mode (cooling or heating), a bias may occur between cells with a large number of samples and cells with a small number of samples (or no samples at all). On the other hand, adjacent cells have similar level categories of explanatory variables, so the estimated values ​​of the thermal response parameters are likely to be similar. Therefore, as described above, the learning unit 15 according to this embodiment can compensate for the lack of samples by applying new estimated values ​​of the thermal response parameters to the target cells P2 and P3 surrounding the storage destination cell P1 identified by the explanatory variables. Furthermore, for example, in the early stages of learning, the contents of the lookup table LUT can be quickly enriched with a small number of samples. This reduces the possibility that the predicted arrival time D will be difficult to predict due to a lack of samples (null).

[0069] Other Embodiments Although one embodiment has been described in detail above with reference to the drawings, the specific configuration is not limited to the above, and various design modifications are possible. That is, in other embodiments, the order of the above-described processes may be changed as appropriate. Furthermore, some processes may be executed in parallel.

[0070] For example, in the above-described embodiment, an example has been described in which the control device 10 (learning unit 15) of the air conditioner 1 learns a prediction model (lookup table LUT). However, this is not limiting. In another embodiment, the thermal response parameters estimated by the control devices 10 (estimation units 14) of multiple air conditioners 1 may be transmitted to a server for aggregation, and the server may learn a prediction model. In this case, the server performs statistical processing or machine learning on the multiple thermal response parameters to generate and update the prediction model. The server also periodically predicts the thermal response parameters (time constant τ, dead time L) for each air conditioner 1 using the prediction model and transmits them to each air conditioner 1. The control device 10 (learning unit 15) of the air conditioner 1 overwrites the lookup table LUT in the memory 4 with the thermal response parameters received from the server. In yet another embodiment, an application on the terminal device 20 may acquire the thermal response parameters from the control device 10 and learn the prediction model. The application of the terminal device 20 periodically predicts the thermal response parameters (time constant τ, dead time L) of the air conditioner 1 using a prediction model and transmits the predicted values ​​to the air conditioner 1. The control device 10 (learning unit 15) of the air conditioner 1 overwrites the lookup table LUT in the memory 4 with the thermal response parameters received from the terminal device 20. This makes it possible to reduce the processing load of the learning unit 15.

[0071] In the above-described embodiment, the control device 10 (prediction unit 16) of the air conditioner 1 determines a storage cell corresponding to an explanatory variable from the lookup table LUT and predicts the predicted arrival time D based on the thermal response parameter read from the storage cell. However, this is not limiting. In other embodiments, the prediction unit 16 may predict the predicted arrival time D using the average value of the estimated thermal response parameters of the cells surrounding the storage cell when, for example, the storage cell does not have an estimated thermal response parameter (the value is null). For example, as shown in the example of FIG. 10 , when the storage cell P1 is null, the prediction unit 16 predicts the predicted arrival time D using the average value of the estimated values ​​read from the surrounding target cells P2 and P3. In this way, even when the lookup table LUT contains many null cells, such as in the early stages of learning, predictions can be made using the estimated values ​​of the surrounding target cells P2 and P3.

[0072] <Additional Notes> The control device, control method, and program described in the above-described embodiment can be understood, for example, as follows.

[0073] (1) According to the first aspect, the control device 10 includes an extraction unit that extracts elements included in a monitoring period from a predetermined start condition to an end condition of air conditioning control by an air conditioner from time series data obtained by measuring the indoor temperature of the air-conditioned space; an estimation unit that estimates a time constant estimate and a dead time estimate of a first-order lag response of the air conditioning control based on the time series data of the monitoring period; and a learning unit that learns a lookup table that associates explanatory variables, which are conditions that affect the air conditioning of the air-conditioned space, with the dead time estimate and the time constant estimate.

[0074] In this way, the control device 10 can learn a lookup table that serves as a substitute for a prediction model using a simpler and lighter algorithm than learning a prediction model using a statistical method, machine learning, etc. In other words, even if the control device 10 is equipped with only limited computational resources, it can learn a lookup table LUT that can predict thermal response parameters to a set temperature quickly and with low load.

[0075] (2) According to the second aspect, in the control device 10 relating to the first aspect, the learning unit determines a storage cell corresponding to the explanatory variable from the lookup table, and updates the dead time estimate and the time constant estimate in the lookup table by taking a weighted average of the new estimate by the estimation unit and the past estimate recorded in the storage cell.

[0076] In this way, the control device 10 can update the lookup table LUT to reflect new estimated values ​​and gradually forget past estimated values, thereby allowing the influence of, for example, aging on the thermal response parameters to be reflected in the lookup table LUT.

[0077] (3) According to a third aspect, in the control device 10 relating to the first or second aspect, the learning unit determines at least one target cell surrounding the storage cell, and updates the dead time estimate and the time constant estimate in the lookup table by taking a weighted average of the new estimate by the estimation unit and the past estimate recorded in the target cell.

[0078] For example, depending on the operating mode (cooling or heating), a bias may occur between cells with a large number of samples and cells with a small number of samples (or no samples at all). On the other hand, adjacent cells have similar level categories of explanatory variables, so the estimated values ​​of the thermal response parameters are likely to be similar. Therefore, as described above, the learning unit 15 according to this embodiment can compensate for the lack of samples by reflecting new estimated values ​​of the thermal response parameters in the target cells P2 and P3 surrounding the storage destination cell P1 identified by the explanatory variables. Furthermore, for example, in the early stages of learning, the contents of the lookup table LUT can be quickly enriched with a small number of samples.

[0079] (4) According to a fourth aspect, the control device 10 according to any one of the first to third aspects further includes a prediction unit that predicts the predicted time until the air-conditioned space reaches the target temperature based on the lookup table and the current explanatory variables.

[0080] In this way, the control device 10 can use the thermal response parameters in the lookup table LUT to accurately predict the predicted time D until the temperature of the air-conditioned space reaches the target temperature.

[0081] (5) According to a fifth aspect, the control method includes the steps of: extracting elements included in a monitoring period from a predetermined start condition to an end condition of air conditioning control by an air conditioner from time series data obtained by measuring the indoor temperature of the air-conditioned space; estimating a time constant estimate and a dead time estimate of a first-order lag response of the air conditioning control based on the time series data of the monitoring period; and learning a lookup table that associates explanatory variables, which are conditions that affect the air conditioning of the air-conditioned space, with the dead time estimate and the time constant estimate.

[0082] (6) According to the sixth aspect, the program causes the control device to execute the following steps: extracting elements included in a monitoring period from a predetermined start condition to an end condition of air conditioning control by an air conditioner from time series data measuring the indoor temperature of the air-conditioned space; estimating a time constant estimate and a dead time estimate of a first-order lag response of the air conditioning control based on the time series data of the monitoring period; and learning a lookup table that associates explanatory variables, which are conditions that affect the air conditioning of the air-conditioned space, with the dead time estimate and the time constant estimate.

[0083] According to the above aspect, even when only limited computational resources are installed, it is possible to learn a lookup table that can predict thermal response parameters to a set temperature quickly and with low load.

[0084] REFERENCE SIGNS LIST 1 Air conditioner 2 Outdoor temperature sensor 3 Indoor temperature sensor 4 Memory 5 Indoor unit 6 Outdoor unit 10 Control device 11 Information acquisition unit 12 Extraction unit 14 Estimation unit 15 Learning unit 16 Prediction unit 17 Control unit 18 Communication unit 20 Terminal device 100 Air conditioning system LUT Look-up table

Claims

1. A control device comprising: an extraction unit that extracts elements included in a monitoring period from a specified start condition to an end condition of air conditioning control by an air conditioner from time series data measuring the indoor temperature in a space to be air conditioned; an estimation unit that estimates a time constant estimate and a dead time estimate of a first-order lag response of the air conditioning control based on the time series data of the monitoring period; and a learning unit that learns a lookup table that associates explanatory variables, which are conditions that affect the air conditioning of the space to be air conditioned, with the dead time estimates and the time constant estimates.

2. The control device as described in claim 1, wherein the learning unit determines a storage cell corresponding to the explanatory variable from the lookup table, and updates the dead time estimate and the time constant estimate in the lookup table by taking a weighted average of a new estimate by the estimation unit and a past estimate recorded in the storage cell.

3. The control device described in claim 2, wherein the learning unit determines at least one target cell surrounding the storage cell, and updates the dead time estimate and the time constant estimate in the lookup table by taking a weighted average of a new estimate by the estimation unit and a past estimate recorded in the target cell.

4. The control device according to any one of claims 1 to 3, further comprising a prediction unit that predicts a predicted time until the temperature of the air-conditioned space reaches a target temperature based on the lookup table and the current explanatory variables.

5. A control method comprising the steps of: extracting elements included in a monitoring period from a predetermined start condition to an end condition of air conditioning control by an air conditioner from time series data measuring the indoor temperature in the space to be air conditioned; estimating a time constant estimate and a dead time estimate of a first-order lag response of the air conditioning control based on the time series data of the monitoring period; and learning a lookup table that associates explanatory variables, which are conditions that affect the air conditioning of the space to be air conditioned, with the dead time estimates and the time constant estimates.

6. A program that causes a control device to execute the steps of: extracting elements included in a monitoring period from a specified start condition to an end condition of air conditioning control by an air conditioner from time series data measuring the indoor temperature in the space to be air conditioned; estimating a time constant estimate and a dead time estimate of a first-order lag response of the air conditioning control based on the time series data of the monitoring period; and learning a lookup table that associates explanatory variables, which are conditions that affect the air conditioning of the space to be air conditioned, with the dead time estimate and the time constant estimate.

Images