JP6384971B2 - Training data collection system, air conditioning control system, training data collection method and program - Google Patents

Description

  The present invention relates to a training data collection system, an air conditioning control system, a training data collection method, and a program.

  There is a technique called ADR (Automated Demand Response) that adjusts the balance between power demand and supply by, for example, raising or lowering the electricity rate and suppressing the amount of power supply during peak hours. For example, in the United States, realization of power charges by OpenADR has been developed in the general market, and operation such as selecting a cheap power supply company in real time is possible. In Japan, there is a movement to develop a domestic version of OpenADR2.0 in the United States as an ADR standard led by the government.

  By the way, in an air conditioner, control which does not impair a user's comfort is requested | required, suppressing electric power consumption and suppressing an electric power charge cheaply. In the conventional air conditioner, a control method has been pursued on the premise that the electric power charge does not change for a certain period, and the comfort is not impaired under the condition and the electric power charge is suppressed. On the other hand, the applicant of the present application relates to a control method for an air conditioner that reduces the power charge while achieving a desired comfortable temperature under the condition that the power charge changes in real time in anticipation of the spread of the ADR. An application has already been filed (Japanese Patent Application No. 2006-173173).

  As a related technique, Patent Document 1 describes a device management system that predicts the comfort level, power consumption, and the like when the operation state of a device is changed and displays a prediction result for a device such as an air conditioner. There is.

Japanese Patent No. 5491891

  In the technology according to the previous application (Japanese Patent Application No. 2006-173173), the future indoor temperature and power consumption are predicted based on the prediction model, and the future air conditioning is based on the schedule information of the electric power unit price that changes each time. The operation state of the machine and the electricity charge are simulated, and the simulation results are evaluated in terms of both comfort and electricity charge to determine the operation method of the air conditioner that optimizes the trade-off between electricity charge and comfort. In this technique, the prediction accuracy of the prediction model is important, but training data is required to construct the prediction model. In general, it is considered that training data is often collected by conducting a step response test or the like. However, since the response characteristics of the air conditioner differ depending on the nominal capacity of the air conditioner, the structure of the building where the air conditioner is installed, etc., build a prediction model with high prediction accuracy for the air conditioner installed in a room In order to do this, it is necessary to conduct step response tests in the installed environment and collect training data. However, it is not realistic to perform the step response test at the installation site (customer) of the air conditioner, and a technique for collecting training data during normal operation has been desired.

  Accordingly, an object of the present invention is to provide a training data collection system, an air conditioning control system, a training data collection method, and a program that can solve the above-described problems.

The first aspect of the present invention is a data collection for acquiring control data for collecting training data that is output to the air conditioner in superposition with the normal operation during normal operation of the air conditioner, and response data for the control data. parts, Bei give a control data for said training data acquisition includes a power limit value, the response data includes indoor temperature and power consumption of the result of the operation the air conditioner under the constraint of the power limit value, the training It is a data collection system.

The training data collection system according to the second aspect of the present invention further includes a power limit value determining unit that varies the magnitude of the power limit value within a predetermined range.

The training data collection system according to the third aspect of the present invention monitors the room temperature as a result of operating the air conditioner based on the power limit value determined by the power limit value determination unit, and the room temperature is a set target. A temperature monitoring unit that instructs to change the power limit value when the temperature does not fall within a predetermined range based on the temperature.

The data collection unit in the fourth aspect of the present invention further collects operating environment information of the air conditioner.

The training data collection system according to the fifth aspect of the present invention further includes a prediction model construction unit that constructs a prediction model that predicts a response to certain control data based on the control data and the response data.

A sixth aspect of the present invention is an air conditioning control system comprising one or more air conditioners, a control device that controls normal operation of the air conditioners, and the training data collection system according to any one of the above. is there.

According to a seventh aspect of the present invention, control data for training data collection that is output to the air conditioner in a superimposed manner with the normal operation of the air conditioner, and response data for the control data are acquired during the normal operation , The control data for training data collection includes a power limit value, and the response data includes a room temperature and power consumption as a result of operating the air conditioner under the restriction of the power limit value .

According to an eighth aspect of the present invention, the power limit value included in the control data for training data collection that is output to the air conditioner by superimposing the computer of the training data collection system on the air conditioner during normal operation of the air conditioner. It is a program for functioning as means for changing the size within a predetermined range.

  According to the present invention, training data for constructing a prediction model used for optimizing the operation of an air conditioner can be collected during normal operation without performing a step response test.

It is a figure which shows an example of the air-conditioning control system in one Embodiment of this invention. It is a block diagram which shows the structural example of the training data collection system in one Embodiment of this invention. It is a figure explaining the control method of the training data collection system in one embodiment of the present invention. It is a 1st flowchart which shows an example of a process of the training data collection system in one Embodiment of this invention. It is a 2nd flowchart which shows an example of a process of the training data collection system in one Embodiment of this invention.

<Embodiment>
Hereinafter, a training data collection system according to an embodiment of the present invention will be described with reference to FIGS.
FIG. 1 is a diagram illustrating an example of an air conditioning control system according to an embodiment of the present invention.
The air conditioning control system 100 is a control system for operating the air conditioner 31 and the like provided in the building 4 and the like at a low cost while maintaining the user's comfort. The air conditioning control system 100 includes a server device 1, a client device 2, a gateway 3, an air conditioner 31, an air conditioner 32,..., An air conditioner 3m, and a training data collection system 10. . The server device 1 is, for example, a server device DRAS operated by an electric power company. The electric power company changes the power rate unit price at a predetermined time interval (for example, every 10 minutes). For example, if a power shortage due to power demand is expected, the power company will increase the unit price of the electricity to encourage consumers to save electricity, and reduce the unit price of the electricity charge when the demand for electricity is expected to be relatively low. To encourage power consumption. The server apparatus 1 and the client apparatus 2 are communicably connected, and perform communication based on a publicly disclosed protocol (for example, OpenADR2.0). For example, the server device 1 transmits schedule information of power charges for a predetermined period in the future to the client device 2.

  The client device 2 is an air conditioning management device operated by a company that operates the air conditioner 31, for example. The client device 2 is an operation that indicates what kind of operation should be performed based on the schedule information of the power rate acquired from the server device 1 and that the air conditioner can be operated at a low power rate while realizing the target temperature set by the user. Control information is calculated. The client device 2 is communicably connected to the gateway 3 via a network such as the Internet. The client device 2 transmits the calculated operation control information to the gateway 3. Incidentally, the client device 2 uses a certain prediction model in order to calculate the operation control information to be transmitted to the gateway 3. This prediction model predicts how the room temperature will change in the future (for example, 5 minutes ahead) when the air conditioner is operated under certain conditions, and how much electricity charge will be charged during the operation. Model (for example, a neural network model). As will be described later, the present embodiment relates to a training data (learning data) collection method for constructing this prediction model.

The gateway 3 is a communication device provided in the building 4. The building 4 is provided with multi-type air conditioners 31, 32, 3m having a plurality of indoor units. The air conditioner 31 includes an outdoor unit 301, indoor units 311, 31n, and the like. The air conditioner 32 includes an outdoor unit 302, indoor units 321, 32n, and the like. The air conditioner 3m includes an outdoor unit 30m, indoor units 3m1, 3mn, and the like. For example, the outdoor unit 301 includes a compressor, a heat exchanger, and the like, and sends the refrigerant to the indoor units 311 and 31n. The indoor unit 311 and the like are provided in a room of the building 4 and perform air conditioning so that the temperature of the room becomes a temperature desired by the user.
The gateway 3 outputs the operation control information for each air conditioner acquired from the client device 2 to the corresponding air conditioner 31 and the like. The air conditioner 31 controls the operation of the own device based on the acquired operation control information.

  The training data collection system 10 is, for example, one or a plurality of air conditioning management devices operated by a company that operates the air conditioner 31 or the like. The training data collection system 10 causes the air conditioner 31 and the like to execute the training data collection operation superimposed on the normal operation. The training data collection system 10 is communicably connected to the client device 2, and transmits training data collection control data (power limit command) to each air conditioner 31 and the like via the client device 2 to collect training data. The operation is executed, and training data is acquired from each air conditioner 31 and the like via the client device 2. The training data collection system 10 will be described in detail with reference to FIG.

FIG. 2 is a block diagram illustrating a configuration example of a training data collection system according to an embodiment of the present invention.
The training data collection system 10 superimposes a small-amplitude power limit command during normal operation of the air conditioner while monitoring the temperature of the space targeted by the air conditioner so as not to deviate significantly from the set target temperature set by the resident. To collect the power limit command and response data (indoor temperature, power consumption) indicating the influence thereof. The collected data is used as training data for constructing a prediction model. As shown in FIG. 2, the training data collection system 10 includes an online data collection device 11, a power limit modulator device 12, a temperature monitoring device 13, and a prediction model construction device 14.

The online data collection device 11 collects training data for constructing a prediction model. As illustrated, the online data collection device 11 includes a data collection unit 111 and a storage unit 112.
The data collection unit 111 air-conditions time-series data (training data) of the power limit value, the indoor temperature as a result of operating the air conditioner 31 and the like under the restriction of the power limit value, and the power consumption consumed in the operation. Acquire every machine. In addition, the data collection unit 111 acquires an outdoor temperature during operation of the air conditioner 31 or the like (for example, a temperature measured around the outdoor unit 301).
The storage unit 112 stores the training data acquired by the data collection unit 111.

The power limit modulator device 12 determines a small amplitude power limit value that is instructed to each air conditioner 31 or the like when the air conditioner 31 or the like is operating normally. As illustrated, the power limit modulator device 12 includes an input / output unit 121 and a power limit value determination unit 122.
The input / output unit 121 inputs and outputs data with other devices. For example, the input / output unit 121 acquires the current power consumption for each air conditioner from the client device 2. Further, the input / output unit 121 outputs a power limit value for each air conditioner to the client device 2.
The power limit value determination unit 122 varies the magnitude of the power limit value instructed to each air conditioner within a predetermined range. For example, the power limit value determination unit 122 randomly varies within a predetermined range (for example, about ± 20%) of the current power consumption every 10 minutes.

The temperature monitoring device 13 monitors the temperature of the space in which each air conditioner 31 performs air conditioning under the restriction by the small amplitude power limit value determined by the power limit modulator device 12. As illustrated, the temperature monitoring device 13 includes an input / output unit 131 and a room temperature change determination unit 132.
The input / output unit 131 inputs and outputs data with other devices. For example, the input / output unit 131 receives information on the indoor temperature of the room or the like that the air conditioner (air conditioner 31 or the like) is in charge of for each air conditioner from the client device 2 (in each room where the indoor unit 311 or the like is installed). Temperature). In addition, the input / output unit 131 outputs power limit value change instruction information that is output when the room temperature is no longer an appropriate temperature, to the power limit modulator device 12.
The room temperature change determination unit 132 determines whether the room temperature acquired by the input / output unit 131 is within a predetermined range based on the set target temperature set for each air conditioner. In addition, the room temperature change determination unit 132 generates power limit value change instruction information when the room temperature does not fall within a predetermined range.

The prediction model construction device 14 uses the training data collected by the online data collection device 11 to construct a prediction model that predicts a change in room temperature until the future and power consumption during that time. As illustrated, the prediction model construction device 14 includes an input / output unit 141 and a prediction model construction unit 142.
The input / output unit 141 performs data input / output with other devices. For example, the input / output unit 141 acquires training data from the online data collection device 11. For example, the input / output unit 141 outputs the constructed prediction model to the client device 2.
The prediction model construction unit 142 constructs a prediction model by learning / analyzing training data using a modeling technique such as an AR model or a neural network.

Next, the training data collection operation by the training data collection system 10 will be described with reference to FIG.
FIG. 3 is a diagram for explaining a control method of the training data collection system in one embodiment of the present invention.
The upper and lower graphs of FIG. 3 show the control data during the training data collection operation for one of the air conditioners (the air conditioner 31) that the training data collection system 10 collects training data. The transition of the transition and the response to the control data are shown. The control data is the power limit value, and the response data is the room temperature (the room temperature of the room where the indoor unit 311 under the air conditioner 31 is installed) and the actual power consumption.
The upper diagram of FIG. 3 is a diagram illustrating an example of a time transition of a small amplitude power limit command value commanded by the power limit modulator device 12 and actual power consumption. In FIG. 3, the vertical axis represents power (kW), and the horizontal axis represents time (minutes).
The lower diagram of FIG. 3 is a diagram illustrating an example of a temporal transition of the room temperature monitored by the temperature monitoring device 13. The vertical axis in the lower diagram of FIG. 3 indicates the room temperature (° C.), and the horizontal axis indicates time (minutes). The time indicated by a position on the horizontal axis in the upper diagram of FIG. 3 corresponds to the time at the same position in the lower diagram of FIG.

  The graph PW in the upper diagram of FIG. 3 shows the transition of the power limit command value for the air conditioner 31 determined by the power limit value determination unit 122. The graph PR shows the transition of the power actually consumed when the air conditioner 31 is operated under the restriction of the power limitation shown in the graph PW. As illustrated, the air conditioner 31 is operated so as not to exceed the power limit value of the graph PW. On the other hand, the graph TR in the lower diagram of FIG. 3 shows the transition of the indoor temperature achieved by the operation of the air conditioner 31 under the above-described restrictions. For the space where the air conditioner 31 performs air conditioning, a set target temperature (for example, 25 ° C.) that is a target indicated by the graph TB is set. Further, for this set target temperature, a fluctuation range a ° C. (for example, −2 ° C. to + 2 ° C.) that prescribes an allowable range in which a deviation from the set target temperature is not noticed by a resident in the space is determined in advance. The temperature monitoring device 13 monitors whether or not the room temperature of the space to be air-conditioned falls within this allowable range (23 ° C. to 27 ° C. in this example).

  Next, control on the training data collection system 10 side during training data collection operation will be described. First, the power limit value determination unit 122 determines a value within a predetermined range (for example, −20% to + 20%) based on the current power consumption of the air conditioner 31 as the power limit value. That is, if the current power consumption is “10”, the power limit value determination unit 122 determines the power limit value in the range of 8-12. Within this range, the power limit value determination unit 122 can randomly determine the magnitude of the power limit value. For example, the power limit value determination unit 122 may determine the magnitude of the power limit value completely arbitrarily, such as 8.5, 9, 10, 11.5, and 12. Here, the air conditioner 31 is in the cooling operation, and the power limit value determination unit 122 determines the power limit value to be “12” with respect to the current power consumption “10”, and the air conditioner 31 is the first 10 shown in FIG. Suppose that you drive for this minute under this constraint. Then, based on the fact that a power limit value larger than the current power consumption is set and that the current room temperature (at 0 minutes) is higher than the set target temperature, the air conditioner 31 increases the operation intensity. . The actual power consumption (graph PR) by the air conditioner 31 follows the power limit value. Further, as the operating intensity increases, the room temperature (graph TR) approaches the set target temperature.

  Here, the data collection unit 111 acquires the determined power limit value “12”. Further, as response data to the addition of the power limit value “12”, when a predetermined time (for example, 5 minutes) has elapsed since the start of operation under the restriction of the power limit value, the air conditioner 31 Power consumption and the indoor temperature achieved by the operation of the air conditioner 31 are acquired. For example, in the case of the above example, the data collection unit 111 acquires the power limit value, the actual power consumption, and the room temperature at the time point of 5 minutes in FIG. For example, this data set is one piece of training data.

  When 10 minutes have elapsed since the start of the operation with the power limit value “12”, the power limit value determination unit 122 determines the power limit value for the next 10 minutes. As in the previous time, the power limit value determination unit 122 determines a value within a predetermined range based on the current power consumption of the air conditioner 31 as the next power limit value. For example, the current power consumption of the air conditioner 31 is “11.5”, and the power limit value determining unit 122 sets the lower limit value “9.2 (11.5 × 0) of the predetermined range based on this power consumption. .8) "is determined as the next power limit value. Then, the air conditioner 31 performs a cooling operation to achieve the set target temperature under the restriction of the power limit value (10 minutes to 20 minutes). The data collection unit 111 acquires training data (power limit value, power consumption, room temperature) when 5 minutes have elapsed after the application of the new power limit value.

  Thereafter, the training data collection system 10 repeats this control. That is, the power limit value determination unit 122 changes the power limit value every 10 minutes, and the data collection unit 111 collects training data for this change when 5 minutes have passed after the change. Here, the power limit value is changed after another 5 minutes from the collection of the training data, for example, the response data (room temperature for 5 minutes) from when the training data is collected until the next power limit value is determined. And power consumption) are the training data under the old power limit value, the response data until the training data is collected 5 minutes after the next power limit value is determined as the training data under the new power limit value. To train a prediction model that realizes such an input / output relationship using the training data and the new power limit value under the power limit value and the old power limit value as input parameters, and the response data under the new power limit value as output parameters. It is.

  If it describes in more detail about an example of the training data which the data collection part 111 acquires, for example, the data collection part 111 will be at the time of the last collection at the time of training data collection (after 5 minutes from the determination of the current power limit value). 10 minutes before), the power limit value, power consumption, and room temperature at that time are acquired along with the time. Then, the power limit value determined last time, the power consumption and the room temperature at each of five points from one minute after the previous collection to the time when the current power limit value is determined, and the power limit value determined this time are predicted. It is assumed that the input parameters for model learning (12 in total). Furthermore, the data collection unit 111 acquires the outdoor temperature (operation environment information) from the previous training data collection time to the current training data collection time as an input parameter. The outdoor temperature may be a representative value (that is, one value) for 10 minutes. The representative value of the outdoor temperature is, for example, an average value or a median value of outdoor temperatures for 10 minutes, or an outdoor temperature measured at a certain timing. For example, the data collection unit 111 acquires these 13 values as input parameters.

  Further, the data collection unit 111 acquires, as output parameters, the room temperature and power consumption at each of the five points in total from 1 minute after the current power limit value determination to 5 minutes in 1 minute increments. The data collection unit 111 writes and stores these input / output parameters (training data) in the storage unit 112.

  As described above, the power limit value determination unit 122 superimposes the small amplitude modulation of the power limit command during the normal operation of the air conditioner 31, but for the purpose of collecting the training data for constructing the prediction model, it is as diverse as possible. Preferably data can be collected. Therefore, for the determination of the power limit value by the power limit value determination unit 122, for example, an algorithm that generates random numbers without bias may be used. For example, in cooperation with the temperature monitoring device 13, if a certain power limit value is given and the room temperature is within the allowable range but shows a tendency to fall below the lower limit value, the power for the next 10 minutes As the limit value, the power limit value determination algorithm may be configured to randomly determine the next power limit value from the range of the power limit value that causes the room temperature to recover to the set target temperature.

  Next, control of the temperature monitoring device 13 will be described. In the temperature monitoring device 13, the input / output unit 131 acquires the indoor temperature of the space that is air-conditioned by the air conditioner 31, and the room temperature change determination unit 132 determines whether the indoor temperature falls within an allowable range (23 ° C. to 27 ° C.). judge. In the example of FIG. 3, after being within the allowable range immediately after the start of the output of the power limit value, it is always within the range of ± a ° C. The room temperature change determination unit 132 monitors the room temperature so that the room temperature changes within the allowable range as in the example of FIG. When the room temperature exceeds the allowable range, the room temperature change determination unit 132 generates power limit value change instruction information and outputs the instruction information to the power limit modulator device 12 via the input / output unit 131. Here, the power limit value change instruction information is, for example, information that instructs the power limit modulator device 12 to release the power limit value. The power limit modulator device 12 that has acquired the power limit value change instruction information stops outputting the power limit value. Then, the air conditioner 31 resumes the operation that brings the room temperature close to the set target temperature without any restriction. Then, the room temperature approaches the set target temperature again.

  Further, the power limit value change instruction information is information for instructing to increase the power limit value, for example, when the room temperature is higher than the allowable range during the cooling operation. Receiving this instruction information, the air conditioner 31 increases the operating strength (for example, increases the rotational speed of the compressor). Then, the room temperature falls to a temperature within the allowable range. Further, for example, the power limit value change instruction information is information for instructing to reduce the size of the power limit value when the room temperature becomes lower than the allowable range during the cooling operation. Receiving this instruction information, the air conditioner 31 reduces the operating strength. Then, the room temperature rises to a temperature within the allowable range.

  In addition, when 50 minutes in FIG. 3 are observed, the room temperature rises to the vicinity of the upper limit value of the allowable range. As described above, when the room temperature approaches the upper and lower limit values of the allowable range, the room temperature change determination unit 132 may output warning information to the power limit modulator device 12. Further, as exemplified in about 60 minutes in FIG. 3, the power limit modulator device 12 that has acquired the warning information has a power limit value that approaches the set target temperature at the next timing (in this example, cooling operation). Since the room temperature has increased, a higher power limit value is determined).

  In this way, the temperature monitoring device 13 restricts the operation of the power limiting modulator device 12 to keep the room temperature during the training data collection operation within a predetermined range so that the occupant does not feel the change in the room temperature. To control. Thereby, the air conditioner 31 can continue the parallel execution of the normal operation and the training data collection operation without making the resident feel uncomfortable. According to this embodiment, since training data can be automatically collected in parallel with normal operation of the air conditioner 31, it is not necessary to perform a step response test for training data collection at the customer. Since it is not necessary to perform a step response test, the time and labor required for collecting training data can be greatly reduced. Since data is collected during normal operation of the air conditioner 31, training data corresponding to the actual operating environment such as the size of the space where the air conditioner 31 is installed and the structure of the building (for the construction of a prediction model with high prediction accuracy) Necessary high quality training data) can be collected. Accordingly, it is possible to construct a prediction model necessary for realizing an operation for reducing the power charge necessary for the operation of the air conditioner while achieving the user-desired set target temperature as much as possible and maintaining the comfort.

FIG. 4 is a first flowchart showing an example of processing of the training data collection system in one embodiment of the present invention.
The flow of training data collection processing will be described with reference to FIG.
As a premise, it is assumed that each of the air conditioners 31, 32,..., 3m is performing a normal operation, and an operator or the like instructs the start of the training data collection operation.
The client device 2 has predetermined operating state information (air temperature, indoor temperature for each indoor unit, power consumption, set target temperature set by the user, etc.) for each air conditioner 31, 32,. Time intervals (eg, every minute). Further, when the power limit modulator device 12 outputs the power limit command value to the client device 2, the operation limit command for each air conditioner is supplied from the client device 2 to each of the air conditioners 31, 32,. The value is notified, and each air conditioner 31 or the like performs normal operation by controlling the indoor units under each air conditioner under the restriction of the operation restriction command value.

First, the training data collection system 10 acquires operating condition information (step S10). More specifically, the input / output unit 131 of the temperature monitoring device 13 sets the set target temperature for each air conditioner 31, 32,..., 3m (more precisely, for each indoor unit under each air conditioner). Obtained from the client device 2.
Next, the power limit modulator device 12 determines a power limit value (step S11). More specifically, the input / output unit 121 acquires the current power consumption for each of the air conditioners 31, 32,..., 3 m from the client device 2. The power limit value determination unit 122 calculates a power limit value within a predetermined range (for example, ± 20%) based on the power consumption of the air conditioner 31. The power limit value determination unit 122 similarly calculates a power limit value based on the respective current power consumption for the other air conditioners 32. The input / output unit 121 outputs the power limit value for each air conditioner 31 and the like calculated by the power limit value determination unit 122 to the client device 2.

  In the online data collection device 11, the data collection unit 111 acquires training data at a predetermined interval (for example, every minute) (step S12). More specifically, the data collection unit 111 acquires the indoor temperature and the outdoor temperature for each air conditioner from the client device 2. The data collection unit 111 records the acquired training data in the storage unit 112 together with the acquired time. As described with reference to FIG. 3, the data collection unit 111 may collect training data such as room temperature and power consumption in units of 1 minute from the client device 2 every 10 minutes.

  In the temperature monitoring device 13, the input / output unit 131 acquires the room temperature for each air conditioner 31 and the like from the client device 2. The room temperature change determination unit 132 determines whether the room temperature acquired for the air conditioner 31 is within a predetermined range (for example, ± 2 ° C.) based on the set target temperature of the air conditioner 31 (step S13). If not within the predetermined range (step S13; No), the room temperature change determination unit 132 generates, for example, power limit value change instruction information for instructing cancellation of the power limit value, and the power limit modulator via the input / output unit 131. Output to the device 12. In the power limit modulator device 12, when the input / output unit 121 acquires this instruction information, the power limit value determination unit 122 cancels the power limit value (step S16). For example, the power limit value determination unit 122 notifies each air conditioner 31 and the like via the client device 2 and the gateway 3 of the cancellation of the power limit value. Each air conditioner 31 or the like performs a normal operation without restriction by the power limit value. Then, the determination in step S13 is repeated. If the room temperature returns to within a predetermined range due to the cancellation of the power limit value, the determination in step S13 is Yes, and the process proceeds to step S14. The processing in steps S13 to S14 is also performed for the other air conditioners 32 and the like.

  On the other hand, when the room temperature is within the predetermined range (step S13; Yes), the end of the training data collection operation is determined (step S14). For example, when an end instruction is input to the training data collection system 10 from an operator or the like, the training data collection operation is terminated (step S14; Yes). The online data collection device 11, the power limiting modulator device 12, and the temperature monitoring device 13 stop operating, and this flowchart ends. If not finished (step S14; No), the power limit value determination unit 122 determines whether 10 minutes have elapsed since the previous determination of the power limit value (step S11) (step S15). When 10 minutes have not elapsed (step S15; No), the processing from step S11 is repeated. That is, for each of all the air conditioners 31..., 3 m, etc., the collection of training data by the data collecting unit 111 is continued while the room temperature is monitored by the room temperature change determining unit 132. On the other hand, when 10 minutes have elapsed (step S15; Yes), the processing from step S12 is repeated. That is, for each of the air conditioners 31..., 3 m, etc., the power limit value determination unit 122 determines the power limit value for the next 10 minutes, and superimposes the normal operation on each air conditioner 31 etc. Make a command.

Next, the flow of automatic training data collection processing and prediction model construction processing will be described.
FIG. 5 is a second flowchart showing an example of processing of the training data collection system in one embodiment of the present invention.
As a premise, the prediction model is recorded in a storage unit (not shown) of the client device 2, and the client device 2 calculates operation control information for realizing the optimized operation of the air conditioner 31 and the like using this prediction model. doing.
The training data collection system 10 commands a power limit value during normal operation. The air conditioner 31 and the like execute the training data collection operation superimposed on the normal operation (step S20), and then determine whether the training data collection operation is finished (step S20). For example, when an instruction to end the training data collection operation is input from the operator, it is determined that the training data collection operation is to be ended. When it determines with not complete | finishing (step S21; No), the air conditioner 31 grade | etc., Continues execution of training data collection driving | operation (step S21). When it determines with complete | finishing (step S21; Yes), the training data collection system 10 stops instruction | command of an electric power limit value, and collection of training data, and the air conditioner 31 grade | etc., Ends training data collection driving | operation.

Further, in the training data collection system 10, when an instruction to end the training data collection operation is input, the prediction model construction device 14 constructs a prediction model (step S22). For example, the input / output unit 141 acquires time-series training data stored in units of air conditioners from the storage unit 112 of the online data collection device 11. Next, the prediction model construction unit 142 constructs a prediction model for each air conditioner 31 and the like by a neural network, deep learning such as AR modeling, and a statistical modeling method. For example, the prediction model construction unit 142 inputs the power limit value, the room temperature, the power consumption, the outdoor temperature, the new power limit value for the past 5 minutes, the new power limit value, and the room temperature for 5 minutes after commanding the new power limit value. Then, a neural network is constructed that uses power consumption as an output parameter and outputs a value given as an output parameter when an input parameter value is input. When the construction of the prediction model is completed, the prediction model constructed by the input / output unit 141 is output to the client device 2. The client device 2 updates the existing prediction model with a new prediction model.
Note that the update timing of the prediction model is not limited to this. For example, the existing prediction model may be updated or improved at predetermined time intervals during execution of the training data collection operation.

  Conventionally, training data for constructing the above prediction model has been collected by a step response test. In the step response test, it is necessary to design hundreds to thousands of operation patterns in advance, and to actually execute the operation patterns to collect data, which takes a lot of labor and time. In addition, because it is not possible to conduct step response tests using the space where the resident lives, it is necessary to prepare a building for the test or to build a prediction model corresponding to each of the various existing living spaces In order to do this, there were problems such as the need for step response tests using various buildings. On the other hand, according to the training data collection system 10 of the present embodiment, the process from the creation of test data (power limit command value) to the collection of training data is automatically performed, so the labor and time required for training data collection are reduced. It can be greatly reduced. In addition, since data can be collected in parallel with normal operation without being understood by the resident, it is possible to acquire high-quality training data that reflects the characteristics specific to the installation environment of the air conditioner. Also, since training data can be collected automatically during normal operation after customer installation, the step response test process before delivery of the air conditioner can be reduced, or training data can be collected at any time after installation. Therefore, it is possible to obtain training data in accordance with changes such as aging of air conditioners and renovation of residential space, so that high practicality can be obtained.

In addition, using the training data collected by the training data collection system 10, even in an environment where the unit price of power charges fluctuates, a prediction model necessary for operating an air conditioner with reduced power charges while satisfying user comfort is constructed. can do.
In the above example, the case where the function for calculating the operation control information is implemented in the client device 2 has been described as an example, but the present invention is not limited to this. For example, these functions may be implemented in the gateway 3.
Moreover, although outdoor temperature was illustrated as driving environment information of an input parameter, you may include weather information, the number of people detected by the human sensor, etc. in the input parameter.

  In addition, the process of each process in the training data collection system 10 mentioned above is memorize | stored in the computer-readable recording medium in the format of a program, and when the computer with which the training data collection system 10 is equipped reads and runs this program The above processing is performed. Here, the computer-readable recording medium means a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like. Alternatively, the computer program may be distributed to the computer via a communication line, and the computer that has received the distribution may execute the program. The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, what is called a difference file (difference program) may be sufficient.

  In addition, it is possible to appropriately replace the components in the above-described embodiments with known components without departing from the spirit of the present invention. The technical scope of the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit of the present invention. For example, in the present embodiment, an example in which the online data collection device 11, the power limit modulator device 12, the temperature monitoring device 13, and the prediction model construction device 14 are mounted on separate computers has been described. A plurality of devices may be mounted on one computer. The client device 2 is an example of a control device. The room temperature change determination unit 132 is an example of a temperature monitoring unit.

DESCRIPTION OF SYMBOLS 100 ... Air-conditioning control system 1 ... Server apparatus 2, 2a ... Client apparatus 3 ... Gateway 31, 32, 3m ... Air conditioner 4 ... Building 301, 302, 30m ... Outdoor Machine 311, 31n, 321, 32n, 3m1, 3mn ... Indoor unit 10 ... Training data collection system 10
DESCRIPTION OF SYMBOLS 11 ... Online data collection device 111 ... Data collection part 112 ... Memory | storage part 12 ... Power limit modulator apparatus 121 ... Input / output part 122 ... Power limit value determination part 13 ... Temperature Monitoring device 131... I / O unit 132... Room temperature change determination unit 14... Prediction model construction device 141.

Claims (8)

A data collection unit that acquires control data for training data collection that is output to the air conditioner in superposition with the normal operation during normal operation of the air conditioner, and response data for the control data,
Bei to give a,
The control data for collecting training data includes a power limit value, and the response data includes a room temperature and power consumption as a result of operating the air conditioner under the restriction of the power limit value.
Training data collection system. A power limit value determining unit that varies the magnitude of the power limit value within a predetermined range;
The training data collection system according to claim 1 , further comprising: When the room temperature is monitored as a result of operating the air conditioner based on the power limit value determined by the power limit value determination unit, and the room temperature does not fall within a predetermined range based on a set target temperature, A temperature monitoring unit for instructing to change the power limit value;
The training data collection system according to claim 2 , further comprising: The data collection unit further collects operating environment information of the air conditioner;
The training data collection system according to any one of claims 1 to 3 . A prediction model construction unit for constructing a prediction model for predicting a response to certain control data based on the control data and the response data;
Training data acquisition system according to any one of claims 1 to 4, further comprising a. One or more air conditioners;
A control device for controlling the normal operation of the air conditioner;
The training data collection system according to any one of claims 1 to 5 ,
An air conditioning control system. Control data for training data collection to be output to the air conditioner superimposed on the normal operation of the air conditioner, and response data for the control data are acquired during the normal operation ,
The control data for collecting training data includes a power limit value, and the response data includes a room temperature and power consumption as a result of operating the air conditioner under the restriction of the power limit value.
Training data collection method. Training data collection system computer,
Means for varying the magnitude of the power limit value included in the control data for training data collection to be output to the air conditioner superimposed on the normal operation during the normal operation of the air conditioner, within a predetermined range;
Program to function as.

Images