JP6910563B1 - Air conditioning control device, power consumption prediction device, learning device and air conditioning control method - Google Patents

Abstract

The air conditioning control device (11) includes an information acquisition unit (111) that acquires event information, a computer resource prediction unit (112) that predicts the usage status of computer resources based on the event information, and a prediction usage status of computer resources. Also provided is an air conditioning intensity determining unit (113) that determines the air conditioning intensity of the air conditioner (20) based on the temperature detection data in the air conditioning area.

Description

The present disclosure relates to an air conditioning control device, a power consumption prediction device, a learning device, and an air conditioning control method.

Regarding the automatic control of air conditioning, for example, Patent Document 1 describes a device that uses reinforcement learning to determine the air conditioning intensity. This device strengthens and learns the temperature of the server room where the server, which is a computer, is located by using the weather information including the air temperature, and malfunctions or fails the server by using the detection data of the air temperature in the server room and the weather information. The temperature of the server room is automatically controlled so that it does not become so hot.

JP-A-2019-179518

In the device described in Patent Document 1, the temperature of the server room rises due to the heat generated from the server in which the computer resource is used, and after the detection data of the rised temperature is acquired, the air conditioning control according to the rise in the temperature. Is done. In air-conditioning control that follows such changes in temperature, for example, when the computer resources used increase and the temperature rises, the air-conditioning intensity in the direction of lowering the temperature is set in the air-conditioning device, and the temperature rises. When the temperature drops, the air conditioning intensity in the direction of raising the temperature is set in the air conditioner, so that the air conditioning intensity fluctuates greatly, and efficient air conditioning control with reduced power consumption cannot be performed.

The present disclosure solves the above problems, and an object of the present invention is to obtain an air conditioning control device, a power consumption prediction device, a learning device, and an air conditioning control method capable of performing efficient air conditioning control with reduced power consumption.

The air-conditioning control device according to the present disclosure is an air-conditioning control device that controls the air-conditioning operation of an air-conditioning area in which a computer is arranged by one or a plurality of air-conditioning devices, and uses physical or virtual computer resources in the computer. , Information acquisition unit that acquires event information related to pre-applied events, computer resource prediction unit that predicts the usage status of computer resources in the event based on the event information, and air conditioning that is predicted from the predicted usage status of computer resources. Based on the change in the temperature of the area and the detection data of the temperature in the air-conditioning area, the maximum air-conditioning intensity reached during the execution of the event is more than the air-conditioning control that follows the change in the temperature of the air-conditioning area accompanying the execution of the event. in so that falls below the target temperature changes in temperature of the air conditioning area is set to be relaxed and the air conditioning area value is small and prediction, setting the air conditioning system prior to a change in temperature of the air conditioning area is the predicted It is provided with an air conditioning strength determining unit that determines the air conditioning strength to be performed.

According to the present disclosure, event information regarding an event that uses a physical or virtual computer resource in a computer is acquired, the usage status of the computer resource in the event is predicted based on the event information, and the predicted usage status of the computer resource is predicted. And the air conditioning intensity is determined based on the temperature detection data in the air conditioning area. Before the temperature of the air-conditioning area changes due to the heat generated by the computer, the air-conditioning intensity that suppresses the temperature change can be determined based on the predicted usage status of the computer resources. Therefore, the air-conditioning control device according to the present disclosure suppresses power consumption. Efficient air conditioning control can be performed.

It is a block diagram which shows the structural example of the air-conditioning control device which concerns on Embodiment 1. FIG. It is a flowchart which shows the air-conditioning control method which concerns on Embodiment 1. FIG. 3A is a graph showing the relationship between the CPU load, the temperature and the air conditioning intensity in the conventional air conditioning control device, and FIG. 3B shows the relationship between the CPU load, the temperature and the air conditioning intensity in the air conditioning control device according to the first embodiment. It is a graph. FIG. 4A is a block diagram showing a hardware configuration that realizes the function of the air conditioning control device according to the first embodiment, and FIG. 4B is a block diagram that executes software that realizes the function of the air conditioning control device according to the first embodiment. It is a block diagram which shows the hardware configuration. It is a block diagram which shows the structure of the modification of the air-conditioning control device which concerns on Embodiment 1. FIG. It is a block diagram which shows the structural example of the power consumption prediction apparatus which concerns on Embodiment 2. FIG. It is a graph which shows the time-dependent change of the predicted power consumption and the computer resource usage charge in Embodiment 2. It is a block diagram which shows the structural example of the learning apparatus which concerns on Embodiment 3. It is explanatory drawing which shows the example of learning by a neural network. It is a flowchart which shows the learning process by the learning apparatus which concerns on Embodiment 3. It is a block diagram which shows the structural example of the inference apparatus which concerns on Embodiment 3. It is a flowchart which shows the inference processing by the inference apparatus which concerns on Embodiment 3.

Embodiment 1.
FIG. 1 is a block diagram showing the configuration of the air conditioning control device 11 according to the first embodiment. In FIG. 1, the air conditioning control device 11 controls the air conditioning operation of the air conditioning area by the air conditioning device 20. The air-conditioning area where the air-conditioning operation by the air-conditioning device 20 is performed is an air-conditioning target space in which one or a plurality of computers are arranged, for example, an indoor space of a data center in which a server which is a computer is arranged. In the following description, it is assumed that the air-conditioned area is, for example, a data center that provides a cloud service on the communication network 30.

The cloud service provides a server device managed by a cloud service operator as a service on the communication network 30. Cloud services can be set to automatically increase or decrease computer resources as communication traffic increases. The computer resources include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and a virtual machine included in the computer. A virtual machine is a virtual computer realized by software executed by the computer.

The number of servers arranged in the data center may be one, or may be multiple. That is, one or more computer resources can be operated in the data center, and the number of computer resources includes the number of virtual machines in addition to the number of computer hardware.

For example, when a sale is started in a mail-order service using the cloud, or an event to be viewed is started in a live video distribution service using the cloud, communication traffic between the service user's terminal and the server is generated. It is expected to increase rapidly. As communication traffic increases, so does the computer resources used for information communication and information processing.

If the increase in computer resources used for the service is expected due to the increase in communication traffic, the service user must apply in advance to the cloud service operator for the expected usage of computer resources for the service. There is. The cloud service operator can predict the computer resources used in the service based on the pre-application information from the service user. As the number of computer resources used increases, the amount of heat generated by the computer also increases, so the temperature of the data center where the server, which is the computer, is located also rises.

The air conditioner 20 performs air conditioning of the air conditioning intensity set by the air conditioner control device 11 on the data center. The number of air conditioners 20 arranged in the data center may be one, but may be multiple. The data center is provided with temperature sensors at one or more locations in the indoor space where the server is located, and the server is provided with temperature sensors that detect the temperature of the CPU or GPU. The temperature detection data in the air conditioning area detected by the temperature sensor is output to the air conditioning control device 11.

As shown in FIG. 1, the air conditioning control device 11 includes an information acquisition unit 111, a computer resource prediction unit 112, and an air conditioning intensity determination unit 113. The information acquisition unit 111 acquires event information. The information acquisition unit 111 acquires event information from an external device on the communication network 30. Event information is information about an event that uses physical or virtual computer resources on a computer. As an example of event information, there is the above-mentioned advance application information.

Pre-application information includes, for example, the number of requests for services per unit time during the expected peak time of server usage, the average size of request information and its response information, and the time period when computer resources are expected to increase ( Start time and end time), protocol type and protocol ratio used to send and receive requests and responses, communication traffic patterns, and factors that are expected to increase computer resources. Pre-application information is data set in the server of the data center.

The computer resource prediction unit 112 predicts the usage status of the computer resource in the event based on the event information. For example, the computer resource prediction unit 112 selects the computer resource used in the event of the service corresponding to the pre-application information from all the computer resources in the data center. Next, the computer resource prediction unit 112 predicts the usage status of the computer resource when the event of the service corresponding to the pre-application information is executed based on the content of the pre-application information. The predicted usage status of the computer resource generated by the computer resource prediction unit 112 is output to the air conditioning strength determination unit 113 as resource information.

When the computer resource used in the event is the CPU of the server, the computer resource prediction unit 112 predicts, for example, a change in the CPU usage rate during a time period in which the service corresponding to the advance application information is expected to be executed. .. The number of computer resources used in the event of the service corresponding to the pre-application information may be not limited to one, but may be plural.

The air-conditioning intensity determination unit 113 determines the air-conditioning intensity based on the predicted usage status of the computer resources and the temperature detection data in the air-conditioning area. The temperature detection data may be obtained at one place in the indoor space of the data center which is an air-conditioned area, or may be obtained at a plurality of places. Since the temperature detection data may be the temperature detection data in the air conditioning area, it is not limited to the temperature in the indoor space of the data center, but may be the temperature detection data of the server CPU or GPU, which is a computer resource. good.

For example, the air conditioning intensity determination unit 113 predicts the temperature change tendency of the data center based on the predicted usage status of computer resources. The temperature change trend of the data center tends to increase when the use of computer resources is concentrated, and tends to decrease as the use of computer resources decreases. In the air conditioning intensity determination unit 113, among the time changes of the target temperature in the operation plan, the temperature change tendency in the time zone which is different from the predicted temperature change tendency is in line with the time change of the target temperature in the operation plan. As described above, the air conditioning intensity (for example, set temperature, air volume) is sequentially determined based on the temperature detection data of the data center.

The operation plan is data showing the time change of the air conditioning state of the air conditioning area, and shows, for example, the time change of the target temperature of the air conditioning area. The target temperature is a temperature to be realized in the air conditioning area, and the air conditioning device 20 is set with an air conditioning intensity for controlling the air conditioning area to the target temperature. The air-conditioning intensity determination unit 113 may calculate the air-conditioning intensity by using a learned model in which event information and temperature detection data in a data center are input to infer the air-conditioning intensity.

The air conditioning control method according to the first embodiment is as follows.
FIG. 2 is a flowchart showing the air conditioning control method according to the first embodiment.
The information acquisition unit 111 acquires event information (step ST1). For example, the information acquisition unit 111 acquires event information from an external device on the communication network 30 via the network interface. For example, the external device is an external storage device that stores advance application information from a computer managed by a service user.

In mail-order services using the cloud, communication traffic between the service user terminal and the server increases significantly during the time period from the start time to the end time of the sale, and the computer resources used to provide the service also increase. Expected to increase. In addition, when the end time of the sale elapses, the communication traffic between the service user terminal and the server is significantly reduced, and the computer resources used for providing the service are also reduced. The pre-application information includes the time zone when the computer resources used are expected to increase and the processing contents in which the computer resources are used.

The computer resource prediction unit 112 predicts the usage status of the computer resource in the event based on the event information acquired by the information acquisition unit 111 (step ST2). For example, the computer resource prediction unit 112 identifies the computer resource used in the event of the service corresponding to the advance application information, and predicts the usage status of the computer resource when the service is provided. The information indicating the predicted usage status of the computer resource generated by the computer resource prediction unit 112 is output to the air conditioning strength determination unit 113.

The air-conditioning intensity determination unit 113 determines the air-conditioning intensity based on the predicted usage status of the computer resources and the temperature detection data in the air-conditioning area (step ST3). In a data center in which a plurality of air conditioners 20 are arranged and a plurality of servers are arranged, the air conditioning intensity determination unit 113 determines different air conditioning intensities for each air conditioner 20. For example, the air-conditioning intensity determination unit 113 determines the air-conditioning intensity for each air-conditioning device 20 according to the relationship between the physical position of each server in the data center and the position of the outlet of each air-conditioning device 20.

FIG. 3A is a graph showing the relationship between the CPU load, the temperature, and the air conditioning intensity in the conventional air conditioning control device. The time change of the CPU load of is indicated by reference numeral C. FIG. 3B is a graph showing the relationship between the CPU load, the temperature, and the air conditioning intensity in the air conditioning control device 11, in which the time change of the air conditioning intensity is indicated by the reference numeral A1 and the time change of the temperature of the data center is indicated by the reference numeral B1. The time change of the CPU load is indicated by reference numeral C1. Since FIGS. 3A and 3B show the case where the same event is executed, the time change of the CPU load used in the event is also the same.

The conventional air-conditioning control device performs air-conditioning control that follows changes in the air temperature of the data center based on the detection data of the air temperature of the data center. In FIG. 3A, the _{CPU is used from the event start time t 01} , and the CPU load starts to increase as indicated by reference numeral C. Heat is generated from the server as the CPU is used. When time _{elapses from time t 01} and time t ₀₂ , the temperature of the data center begins to rise due to the heat generated from the server, as shown by reference numeral B.

_{At time t03} , when the temperature raised by the heat generated by the server exceeds the threshold, the conventional air conditioning controller has a conventional air conditioning controller, as indicated by reference numeral A, so that the raised temperature becomes the target temperature set in the operation plan. Gradually set high air conditioning intensity to the air conditioner. _{At the time t 04} when a certain time elapses from the event start time t ₀₁ , the CPU load becomes a substantially constant value. In the conventional air conditioning control device, the air conditioning intensity is determined according to the change in air temperature. As shown in FIG. 3A, air temperature, during the final time t ₀₅ until an equilibrium state at the target temperature, the air conditioning strength moves up and down in response to the detection data of the temperature.

Further, the air conditioning control device 11 performs air conditioning control based on the predicted usage status of the computer resource at the event and the temperature detection data in the data center. As shown in FIG. 3B, the CPU is used from the event start time _{t 11,} CPU load as indicated by reference numeral C1 starts to rise. Heat is generated from the server as the CPU is used.

The air-conditioning controller 11 predicts the temperature rise due to the heat generated from the server based on the predicted usage status of the computer resources, so that the air-conditioning controller 11 starts from a _{time t 10} earlier than the _{event start time t 11 as indicated by reference numeral A1.} The air conditioning intensity in the direction of lowering the air temperature can be set in the air conditioner 20. Thus, in the event starting time t _11, as shown at B1, in a state where the temperature of the data center is below than the target temperature. Be continued CPU load increase from the event start time t ₁₁ to time t ₁₄ a certain time has elapsed, the temperature of the data center, an equilibrium state at time t ₁₅ remains below the set target temperature operation plan ..

When the change in temperature is predicted by the predicted usage status of the computer resource, the air conditioning control device 11 performs air conditioning control prior to the change in temperature so that the change in temperature at the event is mitigated. Therefore, the maximum value of the air conditioning intensity shown in FIG. 3B is smaller than the maximum value of the air conditioning intensity shown in FIG. 3A. Since the power consumption of the air conditioner 20 is proportional to the air conditioning intensity, the maximum value of the power consumption of the air conditioner control device 11 can be suppressed to be smaller than that of the conventional air conditioner control device, and the cost of the power contract is also reduced. In the air conditioning control device 11, the temperature does not rise sharply and the rise can be suppressed, so that stable operation of the server can be expected.

The hardware configuration that realizes the function of the air conditioning control device 11 is as follows.
FIG. 4A is a block diagram showing a hardware configuration that realizes the functions of the air conditioning control device 11. FIG. 4B is a block diagram showing a hardware configuration for executing software that realizes the functions of the air conditioning control device 11. In FIGS. 4A and 4B, the input interface 100 is an interface that relays the temperature detection data input from the temperature sensor by the air conditioning intensity determining unit 113. The output interface 101 is an interface that relays the air conditioning intensity output from the air conditioning intensity determining unit 113 to the air conditioning device 20. The network interface 102 is an interface that relays event information output from an external device to the air conditioning control device 11 via the communication network 30. The temperature detection data or the air conditioning intensity may be exchanged via the network interface.

The functions of the information acquisition unit 111, the computer resource prediction unit 112, and the air conditioning intensity determination unit 113 in the air conditioning control device 11 are realized by the processing circuit. That is, the air conditioning control device 11 includes a processing circuit for executing the processing from step ST1 to step ST3 shown in FIG. The processing circuit may be dedicated hardware, or may be a CPU that executes a program stored in the memory.

When the processing circuit is the processing circuit 103 of the dedicated hardware shown in FIG. 4A, the processing circuit 103 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, or an ASIC (Application Specific Integrated Circuitd). Circuit), FPGA (Field-Programmable Gate Array), or a combination thereof is applicable. The functions of the information acquisition unit 111, the computer resource prediction unit 112, and the air conditioning intensity determination unit 113 in the air conditioning control device 11 may be realized by separate processing circuits, and these functions are collectively realized by one processing circuit. May be good.

When the processing circuit is the processor 104 shown in FIG. 4B, the functions of the information acquisition unit 111, the computer resource prediction unit 112, and the air conditioning strength determination unit 113 in the air conditioning control device 11 are realized by software, firmware, or a combination of software and firmware. Will be done. The software or firmware is described as a program and stored in the memory 105.

The processor 104 realizes the functions of the information acquisition unit 111, the computer resource prediction unit 112, and the air conditioning strength determination unit 113 in the air conditioning control device 11 by reading and executing the program stored in the memory 105. For example, the air conditioning control device 11 includes a memory 105 that stores a program in which the processes from step ST1 to step ST3 shown in FIG. 2 are executed as a result when executed by the processor 104. These programs cause the computer to execute the procedure or method of the information acquisition unit 111, the computer resource prediction unit 112, and the air conditioning intensity determination unit 113. The memory 105 may be a computer-readable storage medium in which a program for making the computer function as an information acquisition unit 111, a computer resource prediction unit 112, and an air conditioning intensity determination unit 113 is stored.

The memory 105 is, for example, a non-volatile semiconductor such as a RAM (Random Access Memory), a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable Read Only Memory), an EEPROM (Electrolytically Magnetic), or an EPROM (Electrically-EPROM). This includes discs, flexible discs, optical discs, compact discs, mini discs, DVDs, and the like.

Some of the functions of the information acquisition unit 111, the computer resource prediction unit 112, and the air conditioning intensity determination unit 113 in the air conditioning control device 11 may be realized by dedicated hardware, and some may be realized by software or firmware. For example, the information acquisition unit 111 realizes a function by a processing circuit 103 which is dedicated hardware, and the computer resource prediction unit 112 and the air conditioning strength determination unit 113 read and execute a program stored in the memory 105 by the processor 104. The function is realized by doing. As described above, the processing circuit can realize the above-mentioned functions by hardware, software, firmware or a combination thereof.

A modification of the air conditioning control device 11 according to the first embodiment is as follows.
FIG. 5 is a block diagram showing a configuration of an air conditioning control device 12 which is a modification of the air conditioning control device 11. As shown in FIG. 5, the air conditioning control device 12 includes an information acquisition unit 121, a computer resource prediction unit 122, and an air conditioning intensity determination unit 123.

The information acquisition unit 121 acquires event information and weather information. The weather information is weather information including the temperature around the data center, which is an air-conditioned area. For example, the information acquisition unit 121 acquires event information and weather information from an external device on the communication network 30. Based on the event information, the computer resource prediction unit 122 predicts the usage status of the computer resource in the event in the same manner as the computer resource prediction unit 112.

The air-conditioning intensity determination unit 123 determines the air-conditioning intensity based on the predicted usage status of computer resources, temperature detection data in the air-conditioning area, and weather information acquired by the information acquisition unit 121. For example, the air conditioning intensity determination unit 123 predicts the temperature change tendency of the data center based on the predicted usage status of the computer resource and the weather information, and among the time changes of the target temperature in the operation plan, the predicted temperature change tendency. The air conditioning intensity (for example,) is determined based on the temperature detection data of the air conditioning area so that the temperature change tendency in the determined time zone is in line with the time change of the target temperature in the operation plan. , Set temperature, air volume) are determined in sequence.

The degree to which weather information is taken into account when predicting the temperature change trend of a data center is determined by the insulation performance of the data center. For example, when the value of the parameter that quantitatively expresses the heat insulation performance of the data center is equal to or less than the threshold value, the temperature of the indoor space of the data center changes due to the influence of the weather. In this case, the air-conditioning intensity determination unit 123 predicts the temperature change tendency of the data center in consideration of the change in temperature due to the weather in addition to the change in temperature due to the heat generated from the server. When the value of the parameter exceeds the threshold value, the indoor space of the data center is thermally isolated from the outside. In this case, the air-conditioning intensity determination unit 123 predicts the temperature change tendency of the data center in consideration of only the change in the air temperature due to the heat generated from the server, similarly to the air-conditioning intensity determination unit 113. Further, the air-conditioning intensity determination unit 123 may calculate the air-conditioning intensity by using a learned model in which event information, temperature detection data in a data center, and weather information are input to infer the air-conditioning intensity.

As described above, the air conditioning control device 11 according to the first embodiment includes an information acquisition unit 111 that acquires event information, a computer resource prediction unit 112 that predicts the usage status of computer resources based on the event information, and a computer. The air conditioning intensity determination unit 113 that determines the air conditioning intensity based on the predicted usage status of resources and the detection data of the temperature in the data center is provided. Since the air conditioning intensity that suppresses the temperature change can be determined based on the predicted usage status of the computer resources before the temperature of the data center changes due to the heat generated from the server, the air conditioning control device 11 is efficient with reduced power consumption. Air conditioning control can be performed.

In the air conditioning control device 12 according to the first embodiment, the information acquisition unit 121 acquires the event information and the weather information around the data center, and the air conditioning intensity determination unit 123 determines the predicted usage status of the computer resources and the data center. In addition to the temperature detection data, the air conditioning intensity is determined based on the weather information around the data center. Similar to the air conditioning controller 11, the air conditioning strength that suppresses the temperature change based on the predicted usage status of computer resources and the weather information around the data center before the temperature of the data center changes due to the heat generated from the server. Therefore, the air conditioning control device 12 can perform efficient air conditioning control with reduced power consumption.

Embodiment 2.
FIG. 6 is a block diagram showing a configuration example of the power consumption prediction device 31 according to the second embodiment. In FIG. 6, the power consumption prediction device 31 predicts the power consumption of the air conditioning operation of the air conditioning area by the air conditioning device 20 shown in FIG. 1, and sets the predicted power consumption for the air conditioning control device. As in the first embodiment, the air-conditioning area air-conditioned by the air-conditioning device 20 is an air-conditioning target space in which one or a plurality of computers are arranged. In the following description, it is assumed that the air-conditioned area is a data center that provides cloud services on the communication network 30.

The air conditioning control device in which the predicted power consumption is set by the power consumption prediction device 31 is a device that controls the air conditioning operation of the air conditioning device 20 according to the predicted power consumption, and is, for example, the air conditioning control device 11 or the air conditioning shown in the first embodiment. The control device 12. However, the air conditioning intensity determination unit 113 determines the air conditioning intensity based on the predicted usage status of the computer resources, the temperature detection data in the data center, and the predicted power consumption of the air conditioner 20 in the event. The air conditioning intensity determination unit 123 determines the air conditioning intensity based on the predicted usage status of the computer resources, the temperature detection data in the data center, the weather information around the data center, and the predicted power consumption of the air conditioner 20 in the event.

As shown in FIG. 6, the power consumption prediction device 31 includes an information acquisition unit 311, a computer resource prediction unit 312, an air conditioning intensity prediction unit 313, and a power consumption prediction unit 314. The information acquisition unit 311 acquires event information from an external device on the communication network 30. Event information is information about events that use physical or virtual computer resources on the server. As an example of the event information, there is the advance application information shown in the first embodiment. The information acquisition unit 311 acquires event information from an external device on the communication network 30 via the network interface. For example, the external device is an external storage device that stores advance application information from a computer managed by a service user.

The computer resource prediction unit 312 predicts the usage status of computer resources in an event based on the event information. For example, the computer resource prediction unit 312 selects the computer resource used in the event of the service corresponding to the pre-application information from all the computer resources in the data center. Next, the computer resource prediction unit 312 predicts the usage status of the computer resource when the event of the service corresponding to the pre-application information is executed based on the content of the pre-application information. The predicted usage status of the computer resource generated by the computer resource prediction unit 312 is output as resource information to the air conditioning intensity prediction unit 313 and the power consumption prediction unit 314.

The air-conditioning intensity prediction unit 313 predicts the air-conditioning intensity set in the air-conditioning device 20 based on the predicted usage status of computer resources. For example, the air-conditioning intensity prediction unit 313 predicts the temperature change tendency of the data center based on the predicted usage status of computer resources, similarly to the air-conditioning intensity determination unit 113. Subsequently, the air-conditioning intensity prediction unit 313 indicates that among the time changes of the target temperature in the operation plan, the temperature change tendency in the time zone which is different from the predicted temperature change tendency is the time of the target temperature in the operation plan. The air conditioning intensity is predicted in sequence to follow the changes.

Further, in a data center in which a plurality of air conditioners 20 are arranged and a plurality of servers are arranged, the air conditioning intensity prediction unit 313 may predict different air conditioning intensities for each air conditioner 20. For example, the air-conditioning intensity prediction unit 313 predicts the air-conditioning intensity set for each air-conditioning device 20 according to the relationship between the physical position of each server in the data center and the position of the outlet of each air-conditioning device 20. The air conditioning intensity prediction unit 313 may calculate the predicted air conditioning intensity by using a machine learning model that infers the air conditioning intensity by inputting the usage status of the computer resource in the event.

The power consumption prediction unit 314 calculates the predicted power consumption based on the predicted usage status of the computer resources and the predicted air conditioning intensity. For example, the power consumption prediction unit 314 predicts the power consumed by the server located in the data center at the event based on the predicted usage status of the computer resource. Subsequently, the power consumption prediction unit 314 sums the predicted value of the power consumed by the server and the predicted value of the power consumed by the air conditioner 20 calculated by the air conditioning intensity prediction unit 313 based on the operation plan. The predicted value of the power consumption of the entire data center is calculated, and the predicted power consumption is output to the air conditioning control device 11 or the air conditioning control device 12. When predicting the power consumption of the entire data center, power consumption other than air conditioning and a computer (for example, lighting) may be taken into consideration.

For example, the air-conditioning controller 11 or the air-conditioning controller 12 changes the temperature of the data center before the temperature of the data center changes due to the heat generated from the server, based on the predicted usage status of the computer resources and the temperature detection data in the data center. It is possible to determine the air conditioning strength that suppresses. Further, the air conditioning control device 11 or the air conditioning control device 12 can perform air conditioning control in consideration of the usage charge of the computer resource according to the power consumption by using the predicted power consumption calculated by the power consumption prediction unit 314.

FIG. 7 is a graph showing the time-dependent changes in the predicted power consumption and the computer resource usage charge. The time change in the computer resource usage charge is indicated by reference numeral D, and the time change in the predicted power consumption is indicated by reference numeral E. For example, it is assumed that the charge system of the cloud service that uses the server of the data center is a charge system in which the usage charge of the computer resource is significantly increased when the predicted power consumption exceeds the threshold value, as shown by the reference numeral E. The threshold value is, for example, the peak value of power consumption statistically calculated by the electric power company.

As the use of computer resources in a data center increases, the power consumption of the entire data center increases, and the temperature of the data center also rises due to the heat generated by the server. The air-conditioning intensity prediction unit 313 determines that the power consumption of the entire data center is the threshold when the usage of the computer resource is predicted to increase and the power consumption of the entire data center is predicted to exceed the threshold based on the predicted usage status of the computer resource. Before exceeding, predict the air conditioning intensity that suppresses the increase in power consumption of the entire data center.

The power consumption prediction unit 314 calculates the predicted power consumption based on the predicted usage status of the computer resource and the predicted air conditioning intensity, and sets the predicted power consumption in the air conditioning control device 11 or the air conditioning control device 12. The air-conditioning control device 11 or the air-conditioning control device 12 determines the air-conditioning intensity when the air-conditioning device 20 is operated with the predicted power consumption. As a result, the air conditioning control device 11 or the air conditioning control device 12 can perform efficient air conditioning control while suppressing the peak of power consumption, and can suppress an increase in the usage fee of computer resources.

Further, the information acquisition unit 311 may acquire weather information in addition to the event information. The weather information is weather information including the temperature around the data center, which is an air-conditioned area. For example, the information acquisition unit 311 acquires event information and weather information from an external device on the communication network 30. The air-conditioning intensity prediction unit 313 predicts the air-conditioning intensity based on the predicted usage status of computer resources, the temperature detection data in the air-conditioning area, and the weather information acquired by the information acquisition unit 311.

For example, the air conditioning intensity prediction unit 313 predicts the temperature change tendency of the data center based on the predicted usage status of the computer resource and the weather information around the data center, and predicts the time change of the target temperature in the operation plan. A time zone that is a change tendency different from the temperature change tendency is discriminated, and the air conditioning intensity (for example, set temperature) is sequentially predicted so that the temperature change tendency in the discriminated time zone is in line with the time change of the target temperature in the operation plan. do.

Efficient air-conditioning control with reduced power consumption is possible because the air-conditioning intensity at which temperature changes in the data center are suppressed is determined based on the predicted usage of computer resources and weather information around the data center. .. The air conditioning intensity prediction unit 313 may calculate the predicted air conditioning intensity by using a machine learning model that infers the air conditioning intensity by inputting the usage status of computer resources and the weather information around the data center.

The functions of the information acquisition unit 311, the computer resource prediction unit 312, the air conditioning intensity prediction unit 313, and the power consumption prediction unit 314 in the power consumption prediction device 31 are realized by the processing circuit in the same manner as the air conditioning control device 11 or the air conditioning control device 12. .. That is, the power consumption prediction device 31 includes a processing circuit for executing a series of processes for predicting the power consumption of the air conditioner 20. The processing circuit may be the hardware processing circuit 103 shown in FIG. 4A, or the processor 104 that executes the program stored in the memory 105 shown in FIG. 4B.

As described above, the power consumption prediction device 31 according to the second embodiment has an information acquisition unit 311 that acquires event information and a computer resource prediction unit 312 that predicts the usage status of computer resources in an event based on the event information. And the air conditioner strength prediction unit 313 that predicts the air conditioner intensity of the air conditioner 20 based on the predicted usage status of the computer resources, and the power consumption that calculates the predicted power consumption based on the predicted usage status of the computer resources and the predicted value of the air conditioner strength. A prediction unit 314 is provided. Before the temperature of the data center changes due to the heat generated from the server, the air conditioning control device 11 or the air conditioning control device 12 determines the air conditioning intensity that suppresses the temperature change based on the predicted power consumption set by the power consumption prediction device 31. Since it can be determined, efficient air conditioning control with reduced power consumption can be performed.

In the power consumption prediction device 31 according to the second embodiment, the information acquisition unit 311 acquires the event information and the weather information around the data center, and the air conditioning intensity prediction unit 313 predicts the usage status of the computer resources and the data center. In addition to the temperature detection data, the air conditioning intensity is predicted based on the weather information around the data center. Before the temperature of the data center changes due to the heat generated by the server, the air conditioning strength that suppresses the temperature change can be predicted based on the predicted usage of computer resources and the weather information around the data center, so power consumption can be reduced. Suppressed and efficient air conditioning control is possible.

Embodiment 3.
FIG. 8 is a block diagram showing a configuration example of the learning device 41 according to the third embodiment. The learning device 41 learns the air conditioning intensity set in the air conditioning device 20 shown in FIG. 1 to generate a learned model used by the air conditioning control device 11 or the air conditioning control device 12 shown in the first embodiment. Similar to the first embodiment, the air-conditioning area air-conditioned by the air-conditioning device 20 is an air-conditioning target space in which one or more computers are arranged. In the following description, it is assumed that the air-conditioned area is a data center that provides cloud services on the communication network 30. The learning device 41 may be built in the air conditioning intensity determining unit 113 or the air conditioning intensity determining unit 123, or may be provided in the server of the data center.

As shown in FIG. 8, the learning device 41 includes a data acquisition unit 411 and a model generation unit 412. The data acquisition unit 411 generates learning data including the usage status of computer resources based on event information, temperature detection data in the data center, and air conditioning intensity, and outputs the learning data to the model generation unit 412. The training data is data in which computer resource usage, temperature detection data, and air conditioning intensity are associated with each other.

The model generation unit 412 generates a trained model using the training data. The trained model infers the optimum air conditioning intensity by inputting the predicted usage of computer resources and the temperature detection data in the data center. The trained model generated by the model generation unit 412 is stored in the trained model storage unit 50. For example, the learned model storage unit 50 may be a storage unit included in the air conditioning control device 11 or the air conditioning control device 12, or is provided in an external storage device provided separately from the air conditioning control device 11 or the air conditioning control device 12. May be done.

Various learning algorithms such as supervised learning, unsupervised learning, and reinforcement learning can be used to generate the trained model by the model generation unit 412. For example, the model generation unit 412 learns the air conditioning intensity set in the air conditioning device 20 by so-called supervised learning according to the neural network model. Supervised learning is a learning method that uses a combination of input and result (label) data as learning data. The learning device 41 is a method of generating a trained model that infers a result from input data by learning a feature in a given learning data.

FIG. 9 is an explanatory diagram showing an example of learning by a neural network. In FIG. 9, the neural network consists of an input layer (X1, X2, X3) composed of a plurality of neurons, an intermediate layer (hidden layer) (Y1, Y2) composed of a plurality of neurons, and an output layer (Z1) composed of a plurality of neurons. , Z2, Z3). The intermediate layer shown in FIG. 9 is one layer, but may be two or more layers.

In the neural network shown in FIG. 9, when input data is input to each of the neurons X1, X2, and X3 in the input layer, the weights w11, w12, w13, w14, w15, and w16 are multiplied by the value of the input data. , The data of the multiplication result is input to the neural networks Y1 and Y2 in the intermediate layer. In the neurons Y1 and Y2, the weights w21, w22, w23, w24, w25, and w26 are multiplied by the value of the input data, and the multiplication result data is output from the neurons Z1, Z2, and Z3 in the output layer. The output result from the output layer changes depending on the weight value.

In the learning device 41, the neural network obtains the air conditioning intensity by so-called supervised learning using the learning data including the usage status of the computer resource acquired by the data acquisition unit 411, the temperature detection data, and the combination of the air conditioning intensity. learn. That is, the neural network learns by adjusting the weight so that the event information and the temperature detection data are input to the input layer and the value output from the output layer approaches the actual air conditioning intensity (correct answer).

The learning process by the learning device 41 is as follows.
FIG. 10 is a flowchart showing the learning process by the learning device 41.
The data acquisition unit 411 acquires learning data including the usage status of computer resources, temperature detection data in the data center, and air conditioning intensity (step ST1a). For example, the data acquisition unit 411 acquires the usage status of the computer resource while the event indicated by the event information is being executed, acquires the temperature detection data from the temperature sensor of the data center, and the air conditioner is operated by the air conditioner control device 11. Acquires the air conditioning strength set to 20.

The model generation unit 412 is set in the air conditioner 20 by performing supervised learning using learning data including the usage status of computer resources, temperature detection data, and air conditioning intensity acquired by the data acquisition unit 411. Generate a trained model that infers the air conditioning intensity (step ST2a). The trained model generated by the model generation unit 412 is stored in the trained model storage unit 50 (step ST3a).

The functions of the data acquisition unit 411 and the model generation unit 412 in the learning device 41 are realized by the processing circuit, similarly to the air conditioning control device 11 or the air conditioning control device 12. That is, the learning device 41 includes a processing circuit for executing the processing of steps ST1a to ST3a shown in FIG. The processing circuit may be the hardware processing circuit 103 shown in FIG. 4A, or the processor 104 that executes the program stored in the memory 105 shown in FIG. 4B.

Further, the data acquisition unit 411 acquires the weather information around the data center in addition to the usage status of computer resources, the temperature detection data in the data center, and the air conditioning intensity as learning data, and the model generation unit 412 uses the computer. You may generate a trained model that infers air conditioning intensity by inputting resource usage, temperature detection data in the data center, and weather information. By using this trained model, the air conditioning controller 11 or air conditioning controller 12 can determine the temperature of the data center based on the predicted usage of computer resources, the temperature detection data in the data center, and the weather information around the data center. The air conditioning strength that suppresses changes can be determined.

The configuration of the inference device according to the fourth embodiment is as follows.
FIG. 11 is a block diagram showing a configuration example of the inference device 42 according to the third embodiment. As shown in FIG. 11, the inference device 42 is, for example, the air conditioning intensity prediction unit 313 shown in FIG. The inference device 42 includes a data acquisition unit 421 and an inference unit 422. The data acquisition unit 421 acquires event information and temperature detection data in the data center.

In the inference unit 422, the learned model is read from the learned model storage unit 50, and the air conditioning intensity is inferred by the learned model in which the predicted usage status of the computer resource and the temperature detection data in the data center are input. The air conditioning intensity inferred by the trained model is set in the air conditioning device 20. The air conditioner 20 air-conditions the data center with the air-conditioning intensity set by the inference unit 422.

FIG. 12 is a flowchart showing the inference process by the inference device 42.
The data acquisition unit 421 acquires the predicted usage status of the computer resource and the temperature detection data in the data center (step ST1b). For example, the data acquisition unit 411 acquires the predicted usage status of the computer resource in the event calculated by the computer resource prediction unit 112, and detects the temperature detected by the temperature sensor in the data center during the execution of this event. To get.

The inference unit 422 reads the trained model from the trained model storage unit 50, and for the read trained model, predictively uses the computer resources acquired by the data acquisition unit 421 and detects the temperature in the data center. Input (step ST2b). The trained model in which the predicted usage status of the computer resource and the detection data of the temperature in the data center are input calculates the air conditioning intensity and outputs the calculated air conditioning intensity data (step ST3b).

The air conditioning intensity data calculated by the trained model is set in the air conditioning device 20 (step ST4b). The air conditioner 20 air-conditions the data center with the air-conditioning intensity set by the trained model. Since the air conditioning intensity set by the trained model is determined so that the temperature change of the data center is suppressed before the temperature of the data center is changed by the heat generated from the server, the air conditioning device 20 has a peak power consumption. Efficient air-conditioning operation is possible.

When the trained model stored in the trained model storage unit 50 is a model that infers the air conditioning intensity by inputting the predicted usage status of computer resources, the temperature detection data in the data center, and the weather information around the data center. , The data acquisition unit 421 acquires the predicted usage status of the computer resource, the temperature detection data in the data center, and the weather information around the data center.

The inference unit 422 calculates the air conditioning intensity using a trained model that infers the air conditioning intensity by inputting the predicted usage status of the computer resources acquired by the data acquisition unit 411, the temperature detection data in the data center, and the weather information. .. Since the air conditioner 20 operates with the air conditioning intensity calculated by the inference unit 422, efficient air conditioning operation with suppressed peak power consumption is possible.

The functions of the data acquisition unit 421 and the inference unit 422 in the inference device 42 are realized by the processing circuit, similarly to the air conditioning control device 11 or the air conditioning control device 12. That is, the inference device 42 includes a processing circuit for executing the processing of steps ST1b to ST4a shown in FIG. The processing circuit may be the hardware processing circuit 103 shown in FIG. 4A, or the processor 104 that executes the program stored in the memory 105 shown in FIG. 4B.

In the above description, the case where the model generation unit 412 performs supervised learning is shown, but the present invention is not limited to this. As the learning algorithm in the model generation unit 412, reinforcement learning, unsupervised learning, semi-supervised learning, or the like can be used in addition to supervised learning.

The model generation unit 412 may learn the air-conditioning intensity for each of the air-conditioning devices 20 by using the learning data corresponding to each of the plurality of air-conditioning devices 20. Further, the model generation unit 412 may acquire learning data from one data center, or may use learning data acquired from a plurality of air conditioners 20 that operate independently in a plurality of different data centers. You may learn the air conditioning strength.

Further, the data acquisition unit 411 may add learning data used for learning by the model generation unit 412, or may exclude arbitrary learning data from the acquired plurality of learning data. Further, the model generation unit 412 uses, for example, an air conditioner of a second air conditioner arranged in a data center different from the first air conditioner by using a learned model in which the air conditioner intensity of the first air conditioner is learned. By learning the intensity, the model may be updated to infer the air conditioning intensity of the second air conditioner.

In the above description, the case where the air-conditioned area is a data center has been shown, but the devices shown in the first to third embodiments can be applied as long as the air-conditioned area is where the computer is arranged. For example, the air conditioning control device 11 or the air conditioning control device 12 can be used in a train heating / cooling system automatically controlled by a computer.

As described above, the learning device 41 according to the third embodiment obtains the data acquisition unit 411 that acquires the learning data including the usage status of the computer resource, the temperature detection data in the data center, and the air conditioning intensity, and the learning data. It is provided with a model generation unit 412 that generates a trained model that infers the air conditioning intensity by inputting the predicted usage status of computer resources and the detection data of the temperature in the data center. By using this trained model, the learning device 41 can determine the air conditioning intensity that suppresses the temperature change in the data center based on the predicted usage status of the computer resources and the temperature detection data in the data center. As a result, the air conditioning control device 11 or the air conditioning control device 12 can perform efficient air conditioning control with reduced power consumption.

The inference device 42 according to the third embodiment inputs the data acquisition unit 421 that acquires the predicted usage status of the computer resource and the temperature detection data in the data center, and the predicted usage status of the computer resource and the temperature detection data in the data center. A reasoning unit 422 for calculating the air conditioning intensity is provided by using the trained model for inferring the air conditioning intensity. The inference device 42 can determine the air conditioning intensity that suppresses the temperature change in the data center based on the predicted usage status of the computer resource and the temperature detection data in the data center. As a result, the air conditioning control device 11 or the air conditioning control device 12 can perform efficient air conditioning control with reduced power consumption.

It should be noted that the combination of each embodiment, the modification of each arbitrary component of the embodiment, or the omission of any component in each of the embodiments is possible.

The air conditioning control device according to the present disclosure can be used, for example, for air conditioning control of a data center in which a plurality of computers are arranged.

11, 12 Air conditioning controller, 20 Air conditioning device, 30 Communication network, 31 Power consumption prediction device, 41 Learning device, 42 Inference device, 50 Learned model storage, 100 Input interface, 101 Output interface, 102 Network interface, 103 processing Circuit, 104 processor, 105 memory, 111 information acquisition unit, 112, 312 computer resource prediction unit, 113 air conditioning strength determination unit, 121 information acquisition unit, 122 computer resource prediction unit, 123 air conditioning strength determination unit, 311 information acquisition unit, 313 Air conditioning intensity prediction unit, 314 power consumption prediction unit, 411,421 data acquisition unit, 412 model generation unit, 422 inference unit.

Claims (10)

An air-conditioning controller that controls the air-conditioning operation of an air-conditioning area where a computer is located by one or more air-conditioning devices.
An information acquisition unit that acquires event information related to pre-applied events using physical or virtual computer resources in the computer, and an information acquisition unit.
Based on the event information, a computer resource prediction unit that predicts the usage status of the computer resource in the event, and a computer resource prediction unit.
Based on the change in the temperature of the air-conditioned area predicted from the predicted usage status of the computer resource and the detection data of the temperature in the air-conditioned area, the air-conditioning control that follows the change in the temperature of the air-conditioned area accompanying the execution of the event. in so that even fall below the target temperature changes in temperature of the air conditioning area in which the maximum value is small and the prediction is set to be relaxed and the air conditioning area of the air conditioning strength reached during which the event is running An air-conditioning control device comprising, an air-conditioning intensity determining unit for determining the air-conditioning intensity set in the air-conditioning device, prior to the predicted change in the temperature of the air-conditioning area. The information acquisition unit acquires the event information and the weather information around the air-conditioned area, and obtains the event information.
The air-conditioning intensity determining unit determines the air-conditioning intensity based on the weather information in addition to the change in the temperature of the air-conditioning area predicted from the predicted usage status of the computer resource and the detection data of the temperature in the air-conditioning area. The air conditioning control device according to claim 1, wherein the air conditioning control device is characterized by the above. The predicted power consumption is set for the air-conditioning control device that predicts the power consumption of the air-conditioning operation in the air-conditioning area where the computer is arranged by one or more air-conditioning devices and controls the air-conditioning operation by the air-conditioning device according to the predicted power consumption. It is a power consumption prediction device that
An information acquisition unit that acquires event information related to pre-applied events using physical or virtual computer resources in the computer, and an information acquisition unit.
Based on the event information, a computer resource prediction unit that predicts the usage status of the computer resource in the event, and a computer resource prediction unit.
While the event is being executed rather than the air conditioning control that follows the change in the temperature of the air conditioning area accompanying the execution of the event, based on the change in the temperature of the air conditioning area predicted from the predicted usage status of the computer resource. in so that falls below the target temperature changes in temperature of the air conditioning area in which the maximum value is small and the prediction is set to be relaxed and the air conditioning area of the air conditioning strength to reach, the air conditioning area is the predicted An air-conditioning intensity prediction unit that predicts the air-conditioning intensity set in the air-conditioning device prior to a change in temperature,
A power consumption prediction device including a power consumption prediction unit that calculates the predicted power consumption based on the predicted usage status of the computer resource and the predicted air conditioning intensity. The information acquisition unit acquires the event information and the weather information around the air-conditioned area, and obtains the event information.
The power consumption prediction device according to claim 3, wherein the air conditioning intensity prediction unit predicts the air conditioning intensity based on the weather information in addition to the predicted usage status of the computer resource. A learning device that generates a trained model used by the air conditioning control device according to claim 1.
A data acquisition unit that acquires learning data including the usage status of computer resources, temperature detection data in the air conditioning area, and air conditioning intensity.
Using the learning data, the temperature change in the air-conditioned area predicted from the predicted usage status of the computer resource and the temperature detection data in the air-conditioned area are input, and the air-conditioned area is subjected to the execution of the event. The maximum value of the air conditioning intensity reached during the execution of the event is smaller than that of the air conditioning control that follows the change in air temperature, the predicted change in air temperature is mitigated, and the target temperature set in the air conditioning area. in so that drops below the learning apparatus comprising: the model generation unit that generates the learned model prior to the change in temperature of the air conditioning area is the predicted inferring the air conditioning strength, .. A learning device that generates a trained model used by the air conditioning control device according to claim 2.
A data acquisition unit that acquires learning data including the usage status of the computer resource, temperature detection data in the air conditioning area, air conditioning intensity, and weather information.
By inputting the change in the temperature of the air-conditioned area predicted from the predicted usage status of the computer resource, the detection data of the temperature in the air-conditioned area, and the weather information, the change in the temperature of the air-conditioned area accompanying the execution of the event can be obtained. The maximum value of the air conditioning intensity reached during the execution of the event is smaller than that of the following air conditioning control, the predicted change in the temperature of the air conditioning area is mitigated, and the target temperature set in the air conditioning area is set. in so that drops below the learning apparatus comprising: the model generation unit that generates the learned model prior to the change in temperature of the air conditioning area is the predicted inferring the air conditioning strength, .. The air conditioning strength determining unit is an inference device and
The inference device
A data acquisition unit that acquires predicted usage of computer resources and temperature detection data in the air-conditioned area, and
Air-conditioning control that follows the change in the temperature of the air-conditioning area due to the execution of the event by inputting the change in the temperature of the air-conditioning area predicted from the predicted usage status of the computer resource and the detection data of the temperature in the air-conditioning area. that falls below the air conditioning strength target temperature change is set to be relaxed and the air conditioning area of temperature of the air conditioning area in which the maximum value is small and the prediction of reaching while the event is running than As described above, the claim is provided with a reasoning unit for calculating the air conditioning intensity by using a learned model for inferring the air conditioning intensity prior to the predicted change in the temperature of the air conditioning area. Item 1. The air conditioning control device according to item 1. The air conditioning strength determining unit is an inference device and
The inference device
A data acquisition unit that acquires the predicted usage status of the computer resource, the temperature detection data in the air-conditioned area, and the weather information.
The change in the temperature of the air-conditioned area predicted from the predicted usage status of the computer resource, the detection data of the temperature in the air-conditioned area, and the weather information are input to change the temperature of the air-conditioned area due to the execution of the event. The maximum value of the air conditioning intensity reached during the execution of the event is smaller than that of the following air conditioning control, the predicted change in the temperature of the air conditioning area is mitigated, and the target temperature set in the air conditioning area is set. in so that drops below the using the learned model prior to the change in temperature of the air conditioning area is the predicted inferring the air conditioning strength, further comprising a, an inference section for calculating the air-conditioning strength The air conditioning control device according to claim 2, wherein the air conditioning control device is characterized. An air-conditioning control method that controls the air-conditioning operation of an air-conditioning area where a computer is located by one or more air-conditioning devices.
A step in which the information acquisition unit acquires event information regarding a pre-requested event using physical or virtual computer resources in the computer.
A step in which the computer resource prediction unit predicts the usage status of the computer resource in the event based on the event information.
The air-conditioning intensity determination unit changes the temperature of the air-conditioning area with the execution of the event based on the change in the temperature of the air-conditioning area predicted from the predicted usage status of the computer resource and the detection data of the temperature in the air-conditioning area. The maximum value of the air-conditioning intensity reached during the execution of the event is smaller than that of the air-conditioning control following the above, the predicted change in the temperature of the air-conditioning area is mitigated, and the target temperature set in the air-conditioning area to so that drops below, the prior to the change in temperature of the air conditioning area is predicted, the air-conditioning control method characterized by comprising the steps of: determining the air conditioning strength to be set in the air conditioner. The information acquisition unit acquires the event information and the weather information around the air-conditioned area.
The air-conditioning intensity determination unit determines the temperature change in the air-conditioning area predicted from the predicted usage status of the computer resource, the temperature detection data in the air-conditioning area, and the weather information around the air-conditioning area. The air conditioning control method according to claim 9, wherein the air conditioning intensity is determined.

Images