An AI energy-saving control method and centralized control cabinet for a data center computer room

Abstract

The present application discloses an AI energy-saving control method and a centralized control cabinet for a data center computer room, and relates to the field of computer room control. The AI ​​energy-saving control method for a data center computer room includes: obtaining environmental status data in the data center computer room; constructing an objective function with the goal of minimizing the power consumption of the IT system and the air-conditioning system and minimizing the difference between the ambient temperature and the target ambient temperature; determining the optimal control parameters of each execution device in the data center computer room with the goal of optimizing the objective function; and controlling each execution device according to the optimal control parameters of each execution device. The present application determines the optimal control parameters and performs control with the goal of minimizing the power consumption of the IT system and the air-conditioning system and minimizing the difference between the ambient temperature and the target ambient temperature, thereby achieving energy-saving effects while realizing precise control of the ambient temperature.

Description

Technical Field

[0001] The present application relates to the field of computer room control, and in particular to an AI energy-saving control method and a centralized control cabinet for a data center computer room. Background Art

[0002] The ambient temperature in the data center room is crucial to the operation of each device in the data center room. The current control method is usually based on the temperature data monitored by the sensor, and the staff sends control instructions, or mechanically increases or decreases the power of the air-conditioning system and related equipment based on the monitored temperature data, resulting in a waste of electricity. Summary of the invention

[0003] The purpose of this application is to provide an AI energy-saving control method and a centralized control cabinet for a data center computer room, which can achieve precise control of the ambient temperature in the data center computer room and achieve energy-saving effects.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides an AI energy-saving control method for a data center computer room, and the AI ​​energy-saving control method for a data center computer room includes:

[0006] Obtaining environmental status data in the data center computer room; the environmental status data includes IT system power consumption data, air conditioning system power consumption data, ambient temperature data, air conditioning system operation data, air supply system operation data and air conditioning auxiliary system operation data;

[0007] The objective function is constructed with the goal of minimizing the power consumption of the IT system and the air conditioning system and minimizing the difference between the ambient temperature and the target ambient temperature;

[0008] Taking the optimization of the objective function as the goal, determine the optimal control parameters of each execution device in the data center computer room; each execution device includes: each air conditioning device in the air conditioning system, each air supply device in the air supply system and each auxiliary device in the air conditioning auxiliary system, each auxiliary device includes a frequency conversion control device, atomization spray device and a natural cold source heat pipe system;

[0009] Each execution device is controlled according to its optimal control parameters.

[0010] In a second aspect, the present application provides a centralized control cabinet, which applies the above-mentioned data center computer room AI energy-saving control method;

[0011] The centralized control cabinet is respectively connected to the header cabinet, the control host and the temperature sensor in the computer room of the data center. The centralized control cabinet is used to obtain the power consumption data of the IT system from the header cabinet, obtain the power consumption data of the air conditioning system, the operation data of the air conditioning system, the operation data of the air supply system and the operation data of the air conditioning auxiliary system from the control host, and obtain the ambient temperature data collected by the temperature sensor;

[0012] The centralized control cabinet is also connected to the control terminals of various execution devices in the computer room of the data center.

[0013] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0014] The present application provides an AI energy-saving control method for a data center computer room and a centralized control cabinet, the AI ​​energy-saving control method for a data center computer room comprising: obtaining environmental status data in the data center computer room; constructing an objective function with the goal of minimizing the power consumption of the IT system and the air-conditioning system and minimizing the difference between the ambient temperature and the target ambient temperature; determining the optimal control parameters of each execution device in the data center computer room with the goal of optimizing the objective function; and controlling each execution device according to the optimal control parameters of each execution device. The present application determines the optimal control parameters and performs control with the goal of minimizing the power consumption of the IT system and the air-conditioning system and minimizing the difference between the ambient temperature and the target ambient temperature, thereby achieving energy-saving effects while realizing precise control of the ambient temperature. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A flowchart of an AI energy-saving control method for a data center computer room provided in one embodiment of the present application;

[0017] Figure 2 A schematic diagram of the connection relationship between two master and one slave devices and an intermediate relay provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0020] In an exemplary embodiment, Figure 1 As shown, an AI energy-saving control method for a data center computer room is provided, including the following steps 101 to 104.

[0021] Step 101, obtaining environmental status data in a data center computer room; the environmental status data includes IT system power consumption data, air conditioning system power consumption data, ambient temperature data, air conditioning system operation data, air supply system operation data and air conditioning auxiliary system operation data.

[0022] Step 102 , constructing an objective function with the goal of minimizing the power consumption of the IT system and the air conditioning system and minimizing the difference between the ambient temperature and the target ambient temperature.

[0023] Step 103, with the objective of optimizing the objective function, determine the optimal control parameters of each execution device in the data center computer room; each execution device includes: each air-conditioning device in the air-conditioning system, each air-supply device in the air-supply system and each auxiliary device in the air-conditioning auxiliary system, each auxiliary device includes a frequency conversion control device, an atomizing spray device and a natural cold source heat pipe system.

[0024] Step 104: Control each execution device according to the optimal control parameter of each execution device.

[0025] By implementing the above steps 101 to 104, the precise control of the ambient temperature is achieved while achieving energy saving.

[0026] In another exemplary embodiment of the present application, in order to determine the optimal control parameters, the above step 103 is replaced by the following steps 201 to 205.

[0027] Step 201, initialization.

[0028] Generate a random position vector of each search operator in the search operator group.

[0029] Set the maximum number of iterations, the number of search operators, the dimension, and the upper and lower bounds of the search space.

[0030] Initialize the deep neural network model (DNN) and train the deep neural network model.

[0031] Step 202, location update.

[0032] Calculate the fitness value of each search operator in the current search operator group.

[0033] Determine the location of the current optimal search operator and its fitness value.

[0034] According to different search behavior probabilities, one of the following three behaviors is selected to update the position.

[0035] Step 203: Dynamic parameter adjustment.

[0036] Dynamically adjust parameters based on the number of iterations .

[0037] Step 204, iterative update.

[0038] The position update step is repeated until the maximum number of iterations is reached or the convergence condition is met.

[0039] Step 205: output the optimal solution and the optimal fitness value.

[0040] In the above step 202, according to different optimization behavior probabilities , select one of the following three location update methods.

[0041] 1. Optimal point shrinkage and encirclement mechanism (optimal behavior probability ).

[0042] .

[0043] .

[0044] .

[0045] .

[0046] .

[0047] in, is the position vector of the search operator for the next iteration, is the position vector of the search operator of the current iteration, t is the number of current iterations, is the position vector of the current optimal search operator, , and are the first intermediate variable, the second intermediate variable and the third intermediate variable respectively, To adjust the parameters, As the number of iterations decreases, is the maximum number of iterations, and is a random vector; each time you save and The value of is used for deep neural network model training.

[0048] The optimal point of evaluation is the fitness function y=F(X), X is the position vector of the search operator, that is, the vector composed of the return air temperature set values ​​of all non-faulty air conditioners.

[0049] X =[x1,x2,x3…xn], that is, [return air temperature setting value of air conditioner 1, return air temperature setting value of air conditioner 2, return air temperature setting value of air conditioner 3,…, return air temperature setting value of air conditioner n].

[0050] The parameter t is the number of iterations. The maximum number of iterations max_iter is defined when the parameter is initialized. t increases from 0 to max_iter.

[0051] Parameter t=0, the first search is a random search.

[0052] 2. Spiral update mechanism ( ).

[0053] .

[0054] in, is the fourth intermediate variable, , is the natural base, is a constant, is a random number. and The value of is used for deep neural network model training.

[0055] 3. Estimation mechanism based on deep learning AI ( And t 10).

[0056] .

[0057] in, It is a deep neural network model.

[0058] The deep neural network model includes an input layer, a hidden layer, and an output layer. The input layer includes n neurons, and the input data is X =[x1,x2,x3……xn], XAfter normalization, the input layer is input, the hidden layer is 2n neurons, the activation function from the input layer to the hidden layer is sigmoid, the number of neurons in the output layer is n, and the prediction of the next step is .

[0059] Formation and application of deep neural network model: When t 10 means that there are 10 historical data, and the model is trained, saved and called based on the historical data. Before calling the model, the historical data is input into the model for model training to form a deep neural network model.

[0060] In an exemplary embodiment, a centralized control cabinet is provided, which applies the AI ​​energy-saving control method for the data center computer room of the above embodiment to output optimal control parameters or cut off control instructions. The centralized control cabinet is respectively connected to the header cabinet, the control host and the temperature sensor in the data center computer room, and the centralized control cabinet is used to obtain the power consumption data of the IT system from the header cabinet, the power consumption data of the air conditioning system, the operation data of the air conditioning system, the operation data of the air supply system and the operation data of the air conditioning auxiliary system from the control host, and obtain the ambient temperature data collected by the temperature sensor; the centralized control cabinet is also connected to the power supply circuit of each execution device in the data center computer room.

[0061] The centralized control cabinet in this embodiment faces the interior of the data center computer room. On the basis of the original acquisition and control system, artificial intelligence technology is used to optimize the operation process of the entire computer room system to achieve energy saving. The centralized control cabinet realizes AI calculation, data acquisition, control instruction issuance, user operation interface, system operation status display, etc., all in one control cabinet.

[0062] Control objects of the centralized control cabinet: The ambient temperature in the computer room is the key to the stable operation of the IT system. This embodiment does not consider the specific form of the air-conditioning system, that is, whether the outdoor heat dissipation form is an air-cooled or water-cooled system. It only cares about the refrigeration terminal equipment that affects the ambient temperature, such as the air-conditioning system, and related equipment that affects the airflow organization in the computer room, such as the air supply system.

[0063] Related equipment that affects indoor temperature and airflow and is connected to indoor equipment, such as outdoor condensers, dry coolers, water-cooled hosts, water pumps and other related equipment, are organized by independent control systems and connected to the centralized control cabinet through the network or control bus to achieve overall system operation control.

[0064] So far, the specific parameter types monitored by the centralized control cabinet include IT system power consumption data, air conditioning system power consumption data, air conditioning system operation data, ambient temperature data, air supply system operation data, and other system operation data. Among them, the IT system power consumption data can be collected through the head cabinet or the upper-level UPS system or the upper-level low-voltage distribution cabinet of the UPS. The air conditioning system power consumption data must realize the independent monitoring of the power of each air conditioning device. The air conditioning system operation data is realized by connecting the data bus with the air conditioning indoor host PLC or single-chip microcomputer. The ambient temperature data is obtained by connecting the temperature sensors that have been deployed or newly added indoors. The air supply system operation data is realized by connecting the data bus with the existing PLC. Other system monitoring data refers to the newly added frequency conversion control, atomization spray, natural cold source heat pipe system, etc. of the air conditioner, which are all realized by connecting the data bus with the existing PLC.

[0065] Exemplarily, the centralized control cabinet system includes: a cabinet body, a touch screen, a start / stop button, a power switch, a field monitoring unit, two master and one slave units, an intermediate relay, a fault control switch, an industrial control host, a grounding terminal block, a wiring terminal block, an electric meter and a network switch.

[0066] Cabinet: Can carry the industrial control host and all related components.

[0067] Touch screen: Inlaid on the front of the cabinet.

[0068] Start / Stop button: for two-master-one-slave fault signal, internal power switch signal.

[0069] Power supply switching power supply: for 24V components and 15V touch screen.

[0070] On-site monitoring unit: The collected data is converted from 485 signals to TCP / IP signals.

[0071] Two-master-one-slave unit: connect the original control signal and the newly added control signal to the device at the same time.

[0072] Intermediate relay: ensures the original communication line is connected when the two masters and one slave are powered off.

[0073] Fault control switch: In the event of a two-master-one-slave signal failure, the intermediate relay is de-energized.

[0074] Industrial control host: provides specific delivery strategies and collection interfaces.

[0075] Grounding terminal block: 485 grounding, cabinet grounding, circuit grounding.

[0076] Terminal block: The entire wiring node.

[0077] Electricity meter: electricity statistics inside the cabinet.

[0078] Network switch: networking equipment, connected to industrial control host, FSU, existing dynamic ring, etc.

[0079] Industrial control host: According to the size of the project, in small projects, the AI ​​model (i.e. the AI ​​energy-saving control method of the data center room mentioned above) and computing power requirements are relatively small, and local AI model deployment can be implemented. The AI ​​model is deployed in the industrial control host as an independent node. In large projects, the AI ​​energy-saving control component is deployed in the industrial control host as an edge node, and the actual control strategy is obtained through data interaction with the central node, and cached in the local edge system. In the event of a communication failure between the edge node and the central node, the environment can be loaded with monitoring data to choose to perform a fallback operation or an emergency operation, respectively for the cache strategy, initialization strategy, and full load strategy.

[0080] In the embodiment of the present application, all equipment components in the centralized control cabinet can be generally divided into a power supply system, a signal system and a branch system. Among them, the signal system is centered on the field monitoring unit, and all remote devices are connected to the field monitoring unit through 485 communication. However, since the device itself is connected to the dynamic environment or automatic control system, Figure 2 As shown, before connecting to the on-site monitoring unit, the original equipment wiring needs to be divided into two by two master and one slave devices, and the original dynamic ring or automatic control system is connected to one line, for example, Figure 2 In the process, the dynamic ring FSU and the air conditioner are connected through an intermediate relay, and the centralized control cabinet of this application is connected to a line, such as Figure 2 The two master and one slave devices are used to connect the dynamic environment FSU and the air conditioner, and at the same time, the air conditioner is connected to the energy efficiency field collection unit. The energy efficiency field collection unit sends the collected information to the edge computing server through the network cable. Figure 2 The switching power supply in the device is used to power the intermediate relay and the two master and one slave device. Figure 2 N and L in represent the neutral line access terminal and the live line access terminal of the switching power supply respectively, V- and V+ represent the negative output terminal and the positive output terminal of the switching power supply, A1, A2, A, B, 11, 21, 14, 24, 12, 22, slave A, slave B, master 1A, master 1B, master 2A, master 2B are all connection terminals, and port represents the port. Since the line communication can be connected only after the two-master and one-slave devices are powered on, the original communication line will be affected in the event of a power failure. Therefore, the original communication line is accessed through the normally closed contact of the intermediate relay. When the two-master and one-slave devices are powered on, ensure that the intermediate relay is powered on at the same time, disconnect the original communication line, and communicate with the original dynamic ring or automatic control system through the two-master and one-slave, and open the communication link with the field control unit in the centralized control cabinet of this application.

[0081] The fault control switch is connected to the on-site control unit to issue on-off commands to each branch. Each branch connects the power supply circuit to the two master-one slave device and the intermediate relay respectively (it should be noted that the power supply circuit here is the power supply circuit of the control end of the two master-one slave device and the intermediate relay), realizing remote power supply control of the equipment (that is, realizing remote control of the two master-one slave device and the intermediate relay). After a communication failure occurs, the power supply of the corresponding link is cut off, and the normally closed contacts of the intermediate relay are closed to ensure normal communication of the original link.

[0082] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0083] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. An AI energy-saving control method for a data center computer room, characterized in that: The data center computer room AI energy-saving control method includes: Obtaining environmental status data in the data center computer room; the environmental status data includes IT system power consumption data, air conditioning system power consumption data, ambient temperature data, air conditioning system operation data, air supply system operation data and air conditioning auxiliary system operation data; The objective function is constructed with the goal of minimizing the power consumption of the IT system and the air conditioning system and minimizing the difference between the ambient temperature and the target ambient temperature; Taking the optimization of the objective function as the goal, determine the optimal control parameters of each execution device in the data center computer room; each execution device includes: each air conditioning device in the air conditioning system, each air supply device in the air supply system and each auxiliary device in the air conditioning auxiliary system, each auxiliary device includes a frequency conversion control device, atomization spray device and a natural cold source heat pipe system; Control each execution device according to the optimal control parameters of each execution device; With the optimization of the objective function as the goal, determine the optimal control parameters of each execution device in the data center room, including: Initializing a position vector of each search operator in the search operator group; the position vector includes control parameters of each execution device; Taking the objective function as the fitness function, calculating the fitness value of each search operator; According to the fitness values ​​of each search operator, the position vector of the current optimal search operator is determined; Update the position vector of each search operator; Return to the step of "using the objective function as the fitness function to calculate the fitness value of each search operator" until the maximum number of iterations is reached, and output the control parameters in the position vector of the current optimal search operator as the optimal control parameters; Update the position vector of each search operator, including: Use random generation function to generate search behavior probability ; when When , the optimal point shrinkage and encirclement mechanism is used to update the position vector of each search operator; when When , the spiral update mechanism is used to update the position vector of each search operator; when When , a deep neural network model is used to update the position vectors of each search operator; the deep neural network model is obtained by training the position vectors updated based on the optimal point shrinking and surrounding mechanism or the spiral updating mechanism; wherein t is the current iteration number; when When , the optimal point shrinkage and encirclement mechanism or spiral update mechanism is used to update the position vector of each search operator; The formula for updating the position vector of each search operator using the optimal point shrinkage and encirclement mechanism is: ； ； ； ； ； in, is the position vector of the search operator for the next iteration, is the position vector of the search operator for the current iteration, is the position vector of the current optimal search operator, , and are the first intermediate variable, the second intermediate variable and the third intermediate variable respectively, To adjust the parameters, is the maximum number of iterations, and is a random vector; The formula for updating the position vector of each search operator using the spiral update mechanism is: ； in, is the fourth intermediate variable, , is the natural base, is a constant, is a random number; The formula for updating the position vector of each search operator using the deep neural network model is: ； in, It is a deep neural network model.

2. The AI ​​energy-saving control method for a data center computer room according to claim 1 is characterized in that: Calculate the fitness value of each search operator, including: Input the control parameters of each execution device in the environmental state data and the position vector of the search operator into the digital twin model of the data center computer room, and calculate the predicted environmental state data corresponding to the search operator; Calculate the sum of the IT system power consumption data and the air conditioning system power consumption data in the predicted environmental status data, and the absolute value of the difference between the ambient temperature data and the target temperature data in the predicted environmental status data; The fitness value of the search operator is calculated according to the sum of the IT system power consumption data and the air conditioning system power consumption data in the predicted environmental state data, and the absolute value of the difference between the ambient temperature data and the target temperature data in the predicted environmental state data.

3. A centralized control cabinet, characterized in that: The centralized control cabinet applies the AI ​​energy-saving control method for a data center computer room according to any one of claims 1-2; The centralized control cabinet is respectively connected to the header cabinet, the control host and the temperature sensor in the computer room of the data center. The centralized control cabinet is used to obtain the power consumption data of the IT system from the header cabinet, obtain the power consumption data of the air conditioning system, the operation data of the air conditioning system, the operation data of the air supply system and the operation data of the air conditioning auxiliary system from the control host, and obtain the ambient temperature data collected by the temperature sensor; The centralized control cabinet is also connected to the control terminals of various execution devices in the computer room of the data center.

4. The centralized control cabinet according to claim 3, characterized in that: The centralized control cabinet includes an industrial control host and a connecting device, wherein the connecting device is used to connect the industrial control host to a cabinet, a control host or a temperature sensor in a data center computer room, or to connect the industrial control host to the control end of each execution device in a data center computer room.

5. The centralized control cabinet according to claim 4, characterized in that: The connection device comprises: two master and one slave devices and an intermediate relay; The intermediate relay is connected to the original communication link, which is the communication link between the cabinet, the control host, the temperature sensor or each execution device and the monitoring center of the data center computer room; The two masters of the two-master-one-slave device are respectively connected to the industrial control host and the monitoring center of the data center computer room, and one slave of the two-master-one-slave device is connected to the head cabinet, the control host, the temperature sensor or each execution device; The two master-one slave devices and the intermediate relay are controlled by the same control signal; When the control signal is at a low level, the intermediate relay is closed to connect the original communication link; When the control signal is at a high level, the two master and one slave devices are closed and the intermediate relay is disconnected to simultaneously connect the original communication link and the centralized control cabinet link; the centralized control cabinet link is the communication link between the queue cabinet, control host, temperature sensor or each execution device and the industrial control host.

6. The centralized control cabinet according to claim 5, characterized in that: The connection device also includes: a fault control switch, which is serially connected to the connection line between the control end of the intermediate relay and the industrial control host. The control end of the fault control switch is connected to the industrial control host and is used to close the intermediate relay when the two master and one slave devices fail.

Images