Environmental parameter control equipment cluster control method, device, equipment and storage medium

Abstract

An embodiment of the present invention provides a method, apparatus, device and storage medium for controlling a cluster of environmental parameter control devices. The method includes: collecting environmental parameter samples and energy efficiency samples every time a control cycle is reached, and updating them to an environmental parameter sample set and an energy efficiency sample set accordingly; using the energy efficiency sample set to train a context-sensitive (context-related) UCB model for model training, and using the environmental parameter sample set to train a regression model; in the recommendation stage, determining multiple candidate environmental parameter setting points and candidate startup numbers, based on the state parameters and candidate startup numbers collected when the current control cycle is reached, using the UCB model to predict the corresponding energy efficiency score prediction value, based on the state parameters and candidate environmental parameter setting points, using the regression model to predict the corresponding environmental parameter prediction value, and deciding the startup number and environmental parameter setting point based on the energy efficiency score prediction value and the environmental parameter prediction value and running the decision.

Description

Technical Field

[0001] The present invention relates to the field of automation technology, and in particular to a method, device, equipment and storage medium for controlling a cluster of environmental parameter control equipment. Background Art

[0002] With the rapid development of big data and the ever-increasing amount of data processed, data centers are rapidly proliferating. Data centers typically consist of multiple cabinets, servers, uninterruptible power supplies (UPS), air conditioners, temperature and humidity sensors, and more. The data center environment is constantly affected by factors such as heat generated by servers and environmental fluctuations. Air conditioners are used to regulate the data center's temperature, ensuring a constant temperature. As high-power electrical appliances, air conditioners consume significant amounts of electricity during use, and the efficiency of converting electricity into cooling varies at different operating powers. Therefore, controlling multiple air conditioners to reduce the power usage effectiveness (PUE) of data centers has become a major trend.

[0003] The traditional group control method of data center air conditioning is relatively simple, and most of them are based on rules designed according to manual experience to control the air conditioners in groups, which is not ideal.

[0004] In recent years, research institutions and researchers have been studying machine learning-based methods for air conditioner group control, focusing primarily on reinforcement learning, neural network, and regression models. However, single regression models have limited accuracy and relatively weak online learning capabilities. While reinforcement learning and neural network models offer superior learning capabilities, they are difficult to adjust parameters and exhibit slow convergence. Consequently, in practice, these models have been ineffective in achieving equipment safety and energy consumption control in air conditioner group control. Summary of the Invention

[0005] The embodiments of the present invention provide a method, apparatus, device and storage medium for controlling a cluster of environmental parameter control devices, which are used to solve the problem in the prior art that the existing model is used to perform a single air conditioning group control, resulting in poor equipment safety and energy consumption control.

[0006] An embodiment of the present invention provides a method for controlling a cluster of environmental parameter control devices, which is applied to energy-saving and environmental parameter adjustment in a data center, including:

[0007] Every time a control cycle is reached, environmental parameter samples and energy efficiency samples are collected and updated to the environmental parameter sample set and energy efficiency sample set accordingly;

[0008] When triggering model training, the sample features in the energy efficiency sample set are input into the UCB model with context, and the model training is performed with the goal of outputting the sample labels in the energy efficiency sample set; the sample features in the environmental parameter sample set are input into the regression model, and the model training is performed with the goal of outputting the sample labels in the environmental parameter sample set;

[0009] When it is determined that the current stage is in the recommendation stage, each time a control cycle is reached, multiple candidate environmental parameter set points and multiple candidate startup numbers are determined, based on the state parameters and candidate startup numbers collected when the current control cycle is reached, the corresponding energy efficiency score prediction value is predicted using the UCB model, based on the state parameters and candidate environmental parameter set points, the corresponding environmental parameter prediction value is predicted using the regression model, and the startup number and environmental parameter set point are determined based on the energy efficiency score prediction value and the environmental parameter prediction value, and the operation is carried out;

[0010] Among them, the sample features in the energy efficiency sample include the status parameters and the number of startups of the data center obtained when the previous control cycle arrives, and the sample label is the energy efficiency score calculated when the current control cycle arrives. The sample features in the environmental parameter sample include the status parameters and the environmental parameter setting points of the data center obtained when the previous control cycle arrives, and the sample label includes the environmental parameter measurement values ​​collected when the current control cycle arrives.

[0011] Optionally, the energy efficiency score is a score determined according to an energy efficiency index of the data center;

[0012] The energy efficiency index includes the total power of the data center or the total power of the environmental parameter control device cluster or the power usage efficiency (PUE) of the data center; wherein:

[0013] If the environmental parameter measurement value collected when the current control cycle arrives is less than or equal to the preset environmental parameter deviation from the environmental parameter set point of the previous control cycle, the energy efficiency score is calculated using the first formula based on the energy efficiency index of the data center;

[0014] If the difference between the environmental parameter measurement value collected at the time of the current control cycle and the environmental parameter set point of the previous control cycle is greater than the preset environmental parameter deviation, the energy efficiency score is calculated using the second formula based on the energy efficiency index of the data center;

[0015] Among them, when the energy efficiency indicators are the same, the energy efficiency score calculated using the first formula is greater than the energy efficiency score calculated using the second formula.

[0016] Optionally, the sample features in the environmental parameter sample further include: the number of startups in the previous control cycle;

[0017] Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including:

[0018] Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively;

[0019] Determining the number of startups according to the energy efficiency score prediction value;

[0020] Input the startup number, the state parameters collected when the current control cycle arrives, and the candidate environmental parameter set points into the regression model to predict the corresponding environmental parameter prediction value;

[0021] The environmental parameter set point is determined according to the environmental parameter prediction value.

[0022] Optionally, the sample features in the energy efficiency sample further include: an environmental parameter set point of the previous control cycle;

[0023] Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including:

[0024] Inputting the state parameters and candidate environmental parameter set points collected when the current control cycle arrives into the regression model to predict the corresponding environmental parameter prediction values;

[0025] Determine the environmental parameter set point according to the environmental parameter prediction value;

[0026] Input the candidate startup number, the environmental parameter set point, and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction value respectively;

[0027] The number of startups is determined according to the energy efficiency score prediction value.

[0028] Optionally, based on the state parameters and candidate startup numbers collected when the current control cycle arrives, using the UCB model to predict a corresponding energy efficiency score prediction value, based on the state parameters and candidate environmental parameter set points, using a regression model to predict a corresponding environmental parameter prediction value, and deciding the startup number and the environmental parameter set point based on the energy efficiency score prediction value and the environmental parameter prediction value, including:

[0029] Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively;

[0030] Determining the number of startups according to the energy efficiency score prediction value;

[0031] Inputting the state parameters collected when the current control cycle arrives and the candidate environmental parameter set points into the regression model, and predicting the corresponding environmental parameter prediction values ​​respectively;

[0032] The environmental parameter set point is determined according to the environmental parameter prediction value.

[0033] Optionally, if the greater the energy efficiency score, the higher the energy consumption efficiency of the environmental parameter control device cluster, then determining the number of startup devices according to the predicted value of the energy efficiency score includes:

[0034] Determine the candidate startup number corresponding to the largest energy efficiency score prediction value as the startup number;

[0035] Determining the environmental parameter set point according to the environmental parameter prediction value includes:

[0036] The candidate environmental parameter setting point corresponding to the environmental parameter prediction value closest to the target environmental parameter is decided as the environmental parameter setting point.

[0037] Optionally, when a control period is reached, model training is triggered;

[0038] Inputting sample features in the energy efficiency index sample set into the upper confidence UCB model with context, and performing model training with the goal of outputting sample labels in the energy efficiency sample set, including:

[0039] Inputting sample features of the energy efficiency samples in the current control cycle in the energy efficiency sample set into the UCB model, and performing model training with the goal of outputting sample labels of the energy efficiency samples in the current control cycle; or

[0040] The sample features of all energy efficiency samples in the energy efficiency sample set are sequentially input into the upper confidence UCB model, and the model training is performed with the goal of outputting the corresponding sample labels;

[0041] Inputting sample features in the environmental parameter sample set into a regression model, and performing model training with the goal of outputting labels in the environmental parameter sample set, including:

[0042] The sample features of all environmental parameter samples in the environmental parameter sample set are sequentially input into the regression model, and the model training is performed with the goal of outputting corresponding sample labels.

[0043] Optionally, the method further includes:

[0044] When it is determined that the model decision condition is not satisfied, each time a control cycle is reached, the environmental parameter measurement values ​​collected when the current control cycle of the data center environment arrives are obtained;

[0045] Adjust the startup number and environmental parameter set point according to the difference between the environmental parameter measurement value and the target environmental parameter and run;

[0046] When it is determined that the model decision conditions are met, it is determined to enter the recommendation stage.

[0047] Optionally, when it is determined that the model decision condition is not satisfied, each time a control cycle is reached, an environmental parameter measurement value of the data center environment is obtained, and the number of startups and the environmental parameter set point are adjusted according to the difference between the environmental parameter measurement value and the target environmental parameter, including:

[0048] When it is determined that the model decision condition is not met, the difference ΔE = EnvirMeasure-EnvirTarget is calculated every time a control cycle is reached;

[0049] If the difference ΔE is less than the first threshold, the environmental parameter set point of the current control period is determined to be the environmental parameter set point of the previous control period increased by the preset environmental parameter adjustment value, and the number of startups in the current control period is determined to be the number of startups in the previous control period decreased by the startup number adjustment value;

[0050] If the difference ΔE is greater than a second threshold, determining that the environmental parameter set point of the current control cycle is the environmental parameter set point of the previous control cycle minus the preset environmental parameter adjustment value, and the number of startups in the current control cycle is the number of startups in the previous control cycle plus the preset startup number adjustment value;

[0051] If the difference ΔE is greater than or equal to the first threshold and less than or equal to the second threshold, determining that the environmental parameter set point of the current control period is the environmental parameter set point of the previous control period and is randomly increased or decreased by a preset environmental parameter adjustment value, and the number of startups in the current control period is the number of startups in the previous control period and is randomly increased or decreased by a preset startup number adjustment value;

[0052] Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, and EnvirTarget is the target environmental parameter.

[0053] Optionally, determining a plurality of candidate environmental parameter set points and a plurality of candidate startup numbers comprises:

[0054] determining a plurality of candidate environmental parameter set points centered around an environmental parameter set point of a previous control cycle;

[0055] determining, among the plurality of candidate environmental parameter set points, an alternative environmental parameter set point that meets the environmental parameter set point range and the environmental parameter set point of the previous control cycle as candidate environmental parameter set points;

[0056] Determine multiple candidate startup numbers centered on the startup number of the previous control cycle;

[0057] Determine the candidate startup numbers that meet the startup number range among the multiple candidate startup numbers and the startup number in the previous control cycle as candidate startup numbers.

[0058] Optionally, the method further includes:

[0059] At each data sampling time, the state parameters of the data center are collected, and the environmental parameters of the data center are triggered to be detected. When the safe environmental parameters are exceeded, the number of startups and the environmental parameter set points are adjusted according to the set adjustment range;

[0060] The length of one control cycle is equal to an integer multiple of the interval between two adjacent state parameter sampling moments.

[0061] Optionally, each time a data sampling time is reached, the state parameters of the data center are collected, and the environmental parameters of the data center are triggered to be detected. When the safe environmental parameter range is exceeded, the number of startups and the environmental parameter set point are adjusted according to a set adjustment range, including:

[0062] Every time the state parameter sampling time is reached, the environmental parameter EnvirMeasure of the data center is triggered;

[0063] Calculate ΔE = EnvirMeasure - EnvirTarget;

[0064] If ΔE<-DeadLine, the environmental parameter set point of the current control cycle is increased by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is reduced by the preset startup number adjustment value;

[0065] If ΔE>DeadLine, the environmental parameter set point of the current control cycle is reduced by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is increased by the preset startup number adjustment value;

[0066] Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, EnvirTarget is the target environmental parameter, and DeadLine is the preset environmental parameter difference and is a positive number.

[0067] Optionally, the UCB model is a linear upper confidence LinUCB model or a Gaussian UCB model.

[0068] Optionally, the regression model is any one of the following:

[0069] xgboost model, random forest RF model, support vector machine SVM model, neural network model.

[0070] Optionally, the state parameter includes at least one of the following: load power, average supply air environment parameter, average return air environment parameter, average hot channel side environment parameter, and average cold channel side environment parameter.

[0071] Optionally, the environmental parameter is temperature or humidity.

[0072] Based on the same inventive concept, an embodiment of the present invention further provides an environmental parameter control device cluster control device, which is applied to energy-saving adjustment and environmental parameter adjustment in a data center, including:

[0073] The sample collection module is used to collect environmental parameter samples and energy efficiency samples every time a control cycle is reached, and update them to the environmental parameter sample set and energy efficiency sample set accordingly;

[0074] A model training module is configured to, when triggering model training, input the sample features in the energy efficiency sample set into the upper confidence UCB model with context, and perform model training with the goal of outputting the sample labels in the energy efficiency sample set; and input the sample features in the environmental parameter sample set into the regression model, and perform model training with the goal of outputting the sample labels in the environmental parameter sample set;

[0075] A recommendation module is configured to determine, when the current recommendation phase is in progress, a plurality of candidate environmental parameter set points and a plurality of candidate startup numbers for each control cycle reached, predict a corresponding energy efficiency score prediction value using the UCB model based on the state parameters and candidate startup numbers collected when the current control cycle reaches the end, predict a corresponding environmental parameter prediction value using a regression model based on the state parameters and candidate environmental parameter set points, and decide on the startup number and environmental parameter set point based on the energy efficiency score prediction value and the environmental parameter prediction value, and then execute the operation;

[0076] Among them, the sample features in the energy efficiency sample include the status parameters and the number of startups of the data center obtained when the previous control cycle arrives, and the sample label is the energy efficiency score calculated when the current control cycle arrives. The sample features in the environmental parameter sample include the status parameters and the environmental parameter setting points of the data center obtained when the previous control cycle arrives, and the sample label includes the environmental parameter measurement values ​​collected when the current control cycle arrives.

[0077] Optionally, the sample features in the environmental parameter sample further include: the number of startups in the previous control cycle;

[0078] Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including:

[0079] Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively;

[0080] Determining the number of startups according to the energy efficiency score prediction value;

[0081] Input the startup number, the state parameters collected when the current control cycle arrives, and the candidate environmental parameter set points into the regression model to predict the corresponding environmental parameter prediction value;

[0082] The environmental parameter set point is determined according to the environmental parameter prediction value.

[0083] Optionally, the sample features in the energy efficiency sample further include: an environmental parameter set point of the previous control cycle;

[0084] Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including:

[0085] Inputting the candidate environmental parameter set point and the state parameter collected when the current control cycle arrives into the regression model to predict the corresponding environmental parameter prediction value;

[0086] Determine the environmental parameter set point according to the environmental parameter prediction value;

[0087] Input the candidate startup number, the state parameters collected when the current control cycle arrives, and the environmental parameter set point into the UCB model to predict the corresponding energy efficiency score prediction values ​​respectively;

[0088] The number of startups is determined according to the energy efficiency score prediction value.

[0089] Optionally, based on the state parameters and candidate startup numbers collected when the current control cycle arrives, using the UCB model to predict a corresponding energy efficiency score prediction value, based on the state parameters and candidate environmental parameter set points, using a regression model to predict a corresponding environmental parameter prediction value, and deciding the startup number and the environmental parameter set point based on the energy efficiency score prediction value and the environmental parameter prediction value, including:

[0090] Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively;

[0091] Determining the number of startups according to the energy efficiency score prediction value;

[0092] Inputting the candidate environmental parameter set point and the state parameter collected when the current control cycle arrives into the regression model, and predicting the corresponding environmental parameter prediction values ​​respectively;

[0093] The environmental parameter set point is determined according to the environmental parameter prediction value.

[0094] Optionally, the device further comprises:

[0095] The initial control module is used to obtain the environmental parameter measurement values ​​collected when the current control cycle of the data center environment arrives each time a control cycle is reached when it is determined that the model decision condition is not met;

[0096] Adjust the startup number and environmental parameter set point according to the difference between the environmental parameter measurement value and the target environmental parameter and run;

[0097] When it is determined that the model decision conditions are met, it is determined to enter the recommendation stage.

[0098] Optionally, when it is determined that the model decision condition is not satisfied, each time a control cycle is reached, an environmental parameter measurement value of the data center environment is obtained, and the number of startups and the environmental parameter set point are adjusted according to the difference between the environmental parameter measurement value and the target environmental parameter, including:

[0099] When it is determined that the model decision condition is not met, the difference ΔE = EnvirMeasure-EnvirTarget is calculated every time a control cycle is reached;

[0100] If the difference ΔE is less than the first threshold, the environmental parameter set point of the current control period is determined to be the environmental parameter set point of the previous control period increased by the preset environmental parameter adjustment value, and the number of startups in the current control period is determined to be the number of startups in the previous control period decreased by the startup number adjustment value;

[0101] If the difference ΔE is greater than a second threshold, determining that the environmental parameter set point of the current control cycle is the environmental parameter set point of the previous control cycle minus the preset environmental parameter adjustment value, and the number of startups in the current control cycle is the number of startups in the previous control cycle plus the preset startup number adjustment value;

[0102] If the difference ΔE is greater than or equal to the first threshold and less than or equal to the second threshold, determining that the environmental parameter set point of the current control period is the environmental parameter set point of the previous control period and is randomly increased or decreased by a preset environmental parameter adjustment value, and the number of startups in the current control period is the number of startups in the previous control period and is randomly increased or decreased by a preset startup number adjustment value;

[0103] Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, and EnvirTarget is the target environmental parameter.

[0104] Optionally, the device further comprises:

[0105] A safety maintenance module is used to collect the status parameters of the data center at each data sampling time, trigger the detection of the environmental parameters of the data center, and adjust the number of startups and the environmental parameter set points according to the set adjustment range when the safety environmental parameter range is exceeded;

[0106] The length of one of the control cycles is an integer multiple of the time interval between adjacent data samplings.

[0107] Optionally, each time a data sampling time is reached, the state parameters of the data center are collected, and the environmental parameters of the data center are triggered to be detected. When the safe environmental parameter range is exceeded, the number of startups and the environmental parameter set point are adjusted according to a set adjustment range, including:

[0108] Every time the data sampling time arrives, the state parameters of the data center are collected, and the environmental parameters EnvirMeasure of the data center are triggered;

[0109] Calculate ΔE = EnvirMeasure - EnvirTarget;

[0110] If ΔE<-DeadLine, the environmental parameter set point of the current control cycle is increased by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is reduced by the preset startup number adjustment value;

[0111] If ΔE>DeadLine, the environmental parameter set point of the current control cycle is reduced by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is increased by the preset startup number adjustment value;

[0112] Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, EnvirTarget is the target environmental parameter, and DeadLine is the preset environmental parameter difference and is a positive number.

[0113] Based on the same inventive concept, an embodiment of the present invention further provides an electronic device, comprising: a processor and a memory for storing instructions executable by the processor;

[0114] The processor is configured to execute the instructions to implement the environmental parameter control device cluster control method.

[0115] Based on the same inventive concept, an embodiment of the present invention further provides a storage medium, wherein the computer storage medium stores a computer program, and the computer program is used to implement the environmental parameter control device cluster control method.

[0116] The beneficial effects of the present invention are as follows:

[0117] The environmental parameter control device cluster control method, apparatus, device, and storage medium provided by embodiments of the present invention combine a context-sensitive Unified Combination (UCB) model algorithm with a regression model algorithm to recommend data center configurations. This combined model decouples the action space for air conditioning configuration decisions, reducing the action space by nearly tenfold and accelerating learning convergence. The main model, using the UCB algorithm, also exhibits rapid convergence and online learning capabilities, while also achieving high accuracy. The number of air conditioners on and the environmental parameter set points are adjusted within a specific control cycle, thereby improving the data center's environmental parameters and reducing PUE (Power Usage Effectiveness) in a safe and reliable manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0118] Figure 1 One of the flow charts of the environmental parameter control device cluster control method provided by an embodiment of the present invention;

[0119] Figure 2 This is a schematic diagram of the effect of training the UCB model and the regression model in an embodiment of the present invention;

[0120] Figure 3 Flowchart 2 of the method for controlling a cluster of environmental parameter control devices provided by an embodiment of the present invention;

[0121] Figure 4-1 This is one of the input and output diagrams for recommendation using the UCB model and regression model in an embodiment of the present invention;

[0122] Figure 4-2 This is the second input and output diagram of using the UCB model and regression model for recommendation in an embodiment of the present invention;

[0123] Figure 4-3 This is the third input and output diagram of using the UCB model and regression model for recommendation in an embodiment of the present invention;

[0124] Figure 5 Schematic diagram of the effect of the environmental parameter control device cluster control method in an embodiment of the present invention;

[0125] Figure 6 Flowchart 3 of the method for controlling a cluster of environmental parameter control devices provided by an embodiment of the present invention;

[0126] Figure 7 Flowchart 4 of the method for controlling a cluster of environmental parameter control devices provided by an embodiment of the present invention;

[0127] Figure 8-1 for Figure 4-1 Schematic diagram of the specific input and output of the UCB model;

[0128] Figure 8-2 for Figure 4-1 Schematic diagram of the specific input and output of the regression model;

[0129] Figure 9 Flowchart 5 of the method for controlling a cluster of environmental parameter control devices provided by an embodiment of the present invention;

[0130] Figure 10 A schematic diagram of the structure of an environmental parameter control device cluster control device provided by an embodiment of the present invention;

[0131] Figure 11 A schematic diagram of the electronic structure provided by an embodiment of the present invention;

[0132] Figure 12 For application Figure 11 Schematic diagram of the structure of a data center control system for electronic equipment. DETAILED DESCRIPTION

[0133] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention will be further described below with reference to the accompanying drawings and examples. However, the example embodiments can be implemented in various forms and should not be understood as being limited to the embodiments described herein; on the contrary, these embodiments are provided to make the present invention more comprehensive and complete, and to fully convey the concepts of the example embodiments to those skilled in the art. The same figure marks in the figures represent the same or similar structures, and their repeated descriptions will be omitted. The words expressing position and direction described in the present invention are all explained with reference to the accompanying drawings as examples, but changes can be made as needed, and the changes made are all included in the scope of protection of the present invention. The drawings of the present invention are only used to illustrate the relative position relationship and do not represent the true proportion.

[0134] It should be noted that specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can be implemented in a variety of ways different from those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. The subsequent description of the specification is a preferred embodiment of the present application, but the description is for the purpose of illustrating the general principles of the present application and is not intended to limit the scope of the present application. The scope of protection of the present application shall be determined as defined by the appended claims.

[0135] In the embodiment of the present invention, the environmental parameter is temperature or humidity. The following description will be made by taking the environmental parameter being temperature as an example. The implementation of the environmental parameter being humidity is substantially the same as that of temperature, so the embodiment of temperature can be referred to and will not be repeated here.

[0136] Before introducing the embodiments of the present invention, the terms that will appear below are first explained.

[0137] Temperature Setpoint: The temperature set during operation of a temperature control device cluster. For example, if the temperature control device is an air conditioner, the temperature setpoint is the air outlet temperature of the air conditioner.

[0138] Temperature measurement value: The ambient temperature of the data center to be controlled by the temperature control device cluster, collected by the temperature sensor.

[0139] Target temperature: the desired ambient temperature ultimately achieved by controlling the data center through the temperature control device cluster.

[0140] The following describes in detail the environmental parameter control device cluster control method, apparatus, device, and storage medium provided by the embodiments of the present invention in conjunction with the accompanying drawings.

[0141] First aspect:

[0142] The embodiment of the present invention provides a method for controlling a cluster of environmental parameter control devices, which is applied to energy saving and temperature regulation in a data center. Figure 1 As shown, including:

[0143] When the cluster control of the environmental parameter control device starts, the steps of the decision control part and the model training part are performed respectively; first, step S110 of the decision control part and step S210 of the model training part are performed;

[0144] S110, determining whether the control period has been reached;

[0145] If the result of step S110 is yes, execute step S120;

[0146] like Figure 2As shown, S120, every time a control cycle is reached, temperature samples and energy efficiency samples are collected and updated to the temperature sample set and energy efficiency sample set accordingly;

[0147] The sample features of the energy efficiency sample include the state parameters and the number of powered-on machines of the data center obtained at the arrival of the previous control cycle, and the sample label is the energy efficiency score calculated at the arrival of the current control cycle. The sample features of the temperature sample include the state parameters and the temperature set point of the data center obtained at the arrival of the previous control cycle, and the sample label includes the temperature measurement value collected at the arrival of the current control cycle.

[0148] S130, determining whether the model decision condition is met;

[0149] If the result of step S130 is yes, it is determined that the current stage is the recommendation stage, and step S150 is executed;

[0150] S150. Each time a control cycle is reached, a plurality of candidate temperature set points and a plurality of candidate startup numbers are determined. Based on the state parameters and candidate startup numbers collected when the current control cycle is reached, the UCB model is used to predict the corresponding energy efficiency score prediction value. Based on the state parameters and candidate temperature set points, the regression model is used to predict the corresponding ambient temperature prediction value. Based on the energy efficiency score prediction value and the ambient temperature prediction value, the startup number and temperature set point are decided and the operation is carried out.

[0151] S210, determining whether the model training conditions are met;

[0152] If the result of step S210 is yes, execute step S220; if the result of step S210 is no, continue waiting until the result of step S210 is yes;

[0153] S220. Input the sample features in the energy efficiency sample set into an upper confidence bound (UCB) model with context, and perform model training with the goal of outputting the sample labels in the energy efficiency sample set; input the sample features in the temperature sample set into a regression model, and perform model training with the goal of outputting the sample labels in the temperature sample set.

[0154] In a specific implementation, the UCB model is a UCB model with context. Optionally, the UCB model is a Linear Upper Confidence Bound (LinUCB) model or a Gaussian Process Upper Confidence Bound (GPUCB) model.

[0155] In a specific implementation process, optionally, the regression model is any one of the following:

[0156] xgboost model, Random Forest (RF) model, Support Vector Machine (SVM) model, and neural network model.

[0157] The regression model may also be other models not mentioned above, which can be selected according to actual needs and is not limited here.

[0158] During the specific implementation process, the model training condition in step S210 can trigger model training once for each control cycle, or can trigger model training once after multiple control cycles, or can trigger model training once when certain conditions are met (for example, when multiple consecutive control cycles arrive, the temperature set point and the number of startups obtained by the decision in step S150 are used to control the environmental parameter control device cluster, and the state parameters of the data center obtained when the next control cycle arrives do not meet the preset state parameter range). There is no limitation here.

[0159] As an optional implementation, when the control cycle is reached, the model training is triggered. That is, the model training is performed once in each control cycle. The implementation method of triggering the model training once in each control cycle can be found in Figure 3 Schematic flow chart, where Figure 3 The steps shown in Figure 1 Basically the same, please refer to the above content, so I will not repeat it here.

[0160] In a specific implementation process, optionally, in step S120, corresponding updates to the temperature sample set and the energy efficiency sample set include:

[0161] If the number of samples in the temperature sample set and the energy efficiency sample set is equal to the sample set capacity, the temperature samples and energy efficiency samples corresponding to the earliest control period in the temperature sample set and the energy efficiency sample set are deleted, and the temperature samples and energy efficiency samples corresponding to the current control period are updated to the temperature sample set and the energy efficiency sample set.

[0162] For example, if the control cycle time is 1 hour, and the sample set capacity of the temperature sample set and the energy efficiency sample set is set to the number of samples for 30 days, then the sample set capacity is 24×30=720 samples. When the environmental parameter control device cluster runs to the 721st hour, the temperature samples and energy efficiency samples corresponding to the 1st control cycle in the two sample sets will be deleted, and the temperature samples and energy efficiency samples corresponding to the 721st control cycle will be updated to the two sample sets. In the subsequent control cycles, the temperature samples and energy efficiency samples corresponding to the 2nd control cycle will be deleted in the 722nd control cycle, and the temperature samples and energy efficiency samples corresponding to the 722nd control cycle will be updated to the two sample sets, and so on.

[0163] This application combines the context-based UCB model algorithm with the regression model algorithm to recommend data center temperature and energy consumption configurations, with high accuracy and convergence speed.

[0164] During implementation, the energy consumption sample set and the temperature sample set need to be used to train the UCB model and the regression model respectively.

[0165] The training process of the regression model may optionally include inputting sample features in the temperature sample set into the regression model and performing model training with the goal of outputting labels in the temperature sample set, including:

[0166] The sample features of all temperature samples in the temperature sample set are sequentially input into the regression model, and the model training is performed with the goal of outputting corresponding sample labels.

[0167] In the training process of the UCB model, the sample features in the energy efficiency index sample set are input into the upper confidence UCB model with context, and the model training is performed with the goal of outputting the sample labels in the energy efficiency sample set. Any of the following implementation methods can be adopted:

[0168] Method A: The sample features of all energy efficiency samples in the energy efficiency sample set are sequentially input into the UCB model, and the model training is performed with the goal of outputting corresponding sample labels.

[0169] Mode B: When the control cycle is reached, model training is triggered. Sample features of the energy efficiency samples in the current control cycle in the energy efficiency sample set are input into the UCB model, and model training is performed with the goal of outputting sample labels of the energy efficiency samples in the current control cycle.

[0170] In this way, the UCB model is trained using method B, and only updated energy efficiency samples are used for training each time a control cycle is reached, which can reduce the amount of training data and speed up the training.

[0171] Alternatively, as Figure 1 and Figure 3 As shown, the method further includes:

[0172] If the result of step S130 is no, it is determined that the current stage is the initialization control stage, and step S141 is executed;

[0173] S141. Every time a control cycle is reached, obtaining a temperature measurement value collected when the current control cycle of the data center environment is reached;

[0174] S142: Adjust the number of startups and the temperature set point according to the difference between the temperature measurement value and the target temperature, and then run.

[0175] During the specific implementation process, due to the fact that some data centers do not have the characteristics of customized data centers, the environment in which the data centers are located is different (for example, the temperature change patterns in different regions are different, which makes the temperature and energy efficiency change patterns of the data centers also different accordingly), and other factors, a unified training sample is pre-set to train the UCB model and the regression model. The model trained in this way is used to recommend control of the environmental parameter control equipment cluster, which may have problems with its safety and energy saving effect. Then, in the absence of data for model training in advance, the number of startups and the temperature set point are first adjusted according to the difference between the temperature measurement value and the target temperature in the above implementation method, and the temperature sample and the energy efficiency sample are collected when each control cycle arrives, thereby accumulating training samples for the UCB model and the regression model, which is convenient for the subsequent use of the trained model for recommended control.

[0176] Optionally, in step S142, adjusting the number of startups and the temperature set point according to the difference between the temperature measurement value and the target temperature includes:

[0177] When it is determined that the model decision condition is not met, the difference ΔT = TempMeasure-TempTarget is calculated every time a control cycle is reached;

[0178] If the difference ΔT is less than the first threshold, the temperature setting point of the current control cycle is determined to be the temperature setting point of the previous control cycle increased by the preset temperature adjustment value, and the number of startups in the current control cycle is determined to be the number of startups in the previous control cycle decreased by the startup number adjustment value;

[0179] If the difference ΔT is greater than a second threshold, the temperature setting point of the current control cycle is determined to be the temperature setting point of the previous control cycle minus the preset temperature adjustment value, and the number of startups in the current control cycle is determined to be the number of startups in the previous control cycle plus the preset startup number adjustment value;

[0180] If the difference ΔT is greater than or equal to the first threshold and less than or equal to the second threshold, determine that the temperature set point of the current control cycle is the temperature set point of the previous control cycle and is randomly increased or decreased by a preset temperature adjustment value, and the number of startups in the current control cycle is the number of startups in the previous control cycle and is randomly increased or decreased by a preset startup number adjustment value;

[0181] Among them, TempMeasure is the temperature measurement value collected when the current control cycle arrives, and TempTarget is the target temperature.

[0182] For example, the preset temperature adjustment value is 1°C, the preset power-on count adjustment value is 1, the first threshold is -2°C, and the second threshold is 2°C. Then, when the difference ΔT is less than -2°C, the environmental parameter control device cluster increases the temperature set point by 1°C and decreases the power-on count by 1; when the difference ΔT is greater than 2°C, the environmental parameter control device cluster decreases the temperature set point by 1°C and increases the power-on count by 1; and when the difference -2°C ≤ ΔT ≤ 2°C, the environmental parameter control device cluster randomly increases or decreases the temperature set point by 1°C and randomly increases or decreases the power-on count by 1.

[0183] In this way, when the temperature measurement value is significantly lower than the target temperature, the temperature set point is raised and the number of powered-on machines is reduced, thereby controlling the ambient temperature of the data center to rise. When the temperature measurement value is significantly higher than the target temperature, the temperature set point is lowered and the number of powered-on machines is increased, thereby controlling the ambient temperature of the data center to fall, thereby adjusting the ambient temperature of the data center to the target temperature as quickly as possible. Furthermore, when the temperature measurement value is close to the target temperature, randomly controlling the temperature set point and the number of powered-on machines can enrich the training samples for the UCB model and the regression model, thereby improving the reliability of the model recommendations obtained after training.

[0184] Optionally, the energy efficiency score is a score determined according to an energy efficiency index of the data center;

[0185] The energy efficiency index includes the total power of the data center or the total power of the environmental parameter control device cluster or the PUE of the data center; wherein:

[0186] If |TempMeasure-Temp|≤DeadLine, the energy efficiency score is calculated using the first formula according to the energy efficiency index of the data center;

[0187] If |TempMeasure-Temp|>DeadLine, the energy efficiency score is calculated using the second formula according to the energy efficiency index of the data center;

[0188] Wherein, when the energy efficiency indicators are the same, the energy efficiency score calculated using the first formula is greater than the energy efficiency score calculated using the second formula;

[0189] TempMeasure is the temperature measurement value collected when the current control cycle arrives, Temp is the temperature set point of the previous control cycle, and DeadLine is the preset temperature difference and is a positive number.

[0190] In the specific implementation process, the energy efficiency indicator is taken as PUE as an example. Optionally, the first formula is:

[0191]

[0192] The second formula is:

[0193]

[0194] Where score is the energy efficiency score, A>B. For example, A=1, B=0.8.

[0195] Since the total power of the data center and the total power of the environmental parameter control equipment cluster have similar changing patterns to the PUE of the data center, the PUE in the above first and second formulas can also be replaced by the total power of the data center and the total power of the environmental parameter control equipment cluster, and the values ​​of A and B can be adjusted as needed, which will not be repeated here.

[0196] Of course, the first formula and the second formula are not limited to the above-mentioned inverse proportional relationship, and can also be other types of formulas, which are not limited here.

[0197] In this way, by calculating the energy efficiency scores of the temperature measurement value and the temperature set point of the previous control cycle through different formulas, a lower energy efficiency score can be given when the difference between the two values ​​is large, so that the UCB model can take the influence of temperature into account when making recommendation decisions.

[0198] Optionally, in step S150, determining a plurality of candidate temperature setting points and a plurality of candidate startup numbers includes:

[0199] determining a plurality of candidate temperature set points centered around a temperature set point of a previous control cycle;

[0200] determining an alternative temperature set point that meets the temperature set point range among the plurality of alternative temperature set points and the temperature set point of the previous control cycle as a candidate temperature set point;

[0201] Determine multiple candidate startup numbers centered on the startup number of the previous control cycle;

[0202] Determine the candidate startup numbers that meet the startup number range among the multiple candidate startup numbers and the startup number in the previous control cycle as candidate startup numbers.

[0203] During the specific implementation process, the temperature set point range is a pre-set range, such as 10°C-30°C. When the temperature set point of the previous control cycle is 30°C, the alternative temperature set point will have values ​​greater than 30°C and less than 30°C, but the alternative temperature set point greater than 30°C does not conform to the temperature set point range and will be discarded. The candidate temperature set point is finally determined to be the alternative temperature set point less than 30°C and 30°C. For the power-on number range, it can be directly determined as 0 to the number of devices in the environmental parameter control device cluster, or it can be further set to a subset thereof (for example, half to all of the number of devices in the environmental parameter control device cluster). If the range from 0 to the number of devices in the environmental parameter control device cluster is used directly, when the alternative power-on number is less than 0 or greater than the number of devices in the environmental parameter control device cluster, this alternative power-on number will be discarded.

[0204] During specific implementation, the number n1 of the alternative temperature set points, the numerical difference Δ1 between two adjacent alternative temperature set points, the number n2 of the alternative power-on numbers, and the numerical difference Δ2 between two adjacent alternative power-on numbers (n1 and n2 are both positive even numbers) can be determined as needed. For example, if the number n1 of the alternative temperature set points is determined to be 4 and the numerical difference Δ1 between the alternative temperature set points is determined to be 2°C, when the temperature set point of the previous control cycle is 20°C, the alternative temperature set points are determined to be 16°C, 18°C, 22°C, and 24°C. The method for setting the number of alternative power-on numbers is similar, so it will not be repeated here.

[0205] In this way, when the recommended decision adjustment is made in each control cycle, the temperature setting point and the startup number finally determined from the candidate temperature setting points and the candidate startup numbers are close to the temperature setting point and the startup number in the previous control cycle, avoiding drastic changes in these two setting parameters causing abnormal operation of the temperature control device.

[0206] As a preferred embodiment, determining multiple candidate temperature set points centered around the temperature set point of the previous control cycle includes:

[0207] Two candidate temperature set points centered around the temperature set point of the previous control cycle are determined, and the candidate temperature set points differ from the temperature set point of the previous control cycle by a minimum temperature set point change value.

[0208] For example, the minimum temperature set point change value is 1°C. When the temperature set point of the previous control cycle is 20°C, the alternative temperature set points are determined to be 19°C and 21°C.

[0209] Determine multiple candidate startup numbers centered on the startup number of the previous control cycle, including:

[0210] Two candidate startup numbers centered around the startup number of the previous control cycle are determined, and the candidate startup numbers differ from the startup number of the previous control cycle by 1.

[0211] In this way, the temperature changes in the data center are more balanced by changing the number of startups and the temperature set point by at most one unit each time.

[0212] In the recommendation stage, after completing the training of the UCB model and the regression model, when using the UCB model and the regression model to predict the energy efficiency score and the ambient temperature respectively, the UCB model and the regression model can be used respectively for energy efficiency score prediction and ambient temperature prediction, or the two can be used in combination, and the parameters obtained by using the prediction results of one model to recommend the decision are input into the other model to affect the prediction results and recommendation decisions of the other parameters.

[0213] Optionally, based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score, based on the state parameters and candidate temperature set points, the regression model is used to predict the corresponding ambient temperature prediction value, and the startup number and temperature set point are determined based on the energy efficiency score and the ambient temperature prediction value, including any of the following implementations:

[0214] Method 1:

[0215] In the first embodiment, the sample characteristics of the temperature sample further include: the number of power-on times in the previous control cycle.

[0216] like Figure 4-1 As shown, the candidate startup number and the state parameters collected when the current control cycle arrives are input into the UCB model to predict the corresponding energy efficiency score prediction values ​​respectively;

[0217] Determining the number of startups according to the energy efficiency score prediction value;

[0218] Input the number of startups, the state parameters collected when the current control cycle arrives, and the candidate temperature set point into the regression model to predict the corresponding ambient temperature prediction value;

[0219] The temperature set point is determined according to the predicted ambient temperature value.

[0220] Method 2:

[0221] In the second approach, the sample characteristics in the energy efficiency sample further include: a temperature set point in the previous control cycle.

[0222] like Figure 4-2 As shown, the candidate temperature set point and the state parameters collected when the current control cycle arrives are input into the regression model to predict the corresponding ambient temperature prediction value;

[0223] Determine the temperature set point according to the ambient temperature prediction value;

[0224] Input the candidate startup number, the state parameters collected when the current control cycle arrives, and the temperature set point into the UCB model to predict the corresponding energy efficiency score prediction values ​​respectively;

[0225] The number of startups is determined according to the energy efficiency score prediction value.

[0226] Method 3:

[0227] like Figure 4-3 As shown, the candidate startup number and the state parameters collected when the current control cycle arrives are input into the UCB model to predict the corresponding energy efficiency score prediction values ​​respectively;

[0228] Determining the number of startups according to the energy efficiency score prediction value;

[0229] Inputting the candidate temperature set point and the state parameters collected when the current control cycle arrives into the regression model, and predicting the corresponding ambient temperature prediction values ​​respectively;

[0230] The temperature set point is determined according to the predicted ambient temperature value.

[0231] In a specific implementation, the temperature set point closest to the target temperature among the predicted ambient temperature values ​​is determined as the temperature set point for the current control cycle. If the energy efficiency score calculated based on the energy efficiency index is higher, the higher the energy efficiency index, the higher the energy efficiency score, then the candidate startup number corresponding to the maximum value among the predicted energy efficiency scores is determined as the startup number for the current control cycle.

[0232] In this way, Methods 1 and 2 use two models to couple the number of startups and temperature set points, reducing the recommendation space and significantly improving the model convergence speed. Method 3 uses two models to determine the number of startups and temperature set points separately, which is a simpler solution.

[0233] At each control cycle, in addition to adjusting the startup number and temperature set point of the environmental parameter control device cluster using the above-mentioned implementation method, in order to ensure the safety of device operation, Figure 5 As shown, it is also possible to set a periodic state parameter sampling moment for detection and adjustment.

[0234] Alternatively, as Figure 6 and Figure 7 As shown, except for Figure 1 、 Figure 3 In addition to the same steps as those in the method shown, the method further comprises:

[0235] S160: Determine whether a data sampling time has arrived, and collect the status parameters of the data center once;

[0236] If the result of step S160 is yes, execute step S170; if the result of step S160 is no, return to step S110;

[0237] S170: Detect the ambient temperature of the data center to determine whether it exceeds a safe temperature range;

[0238] If the result of step S170 is yes, execute step S180; if the result of step S170 is no, return to step S110;

[0239] S180, adjusting the number of startups and the temperature set point according to the set adjustment range; returning to step S110;

[0240] The length of one of the control cycles is an integer multiple of the time interval between adjacent data samplings.

[0241] In a specific implementation process, the duration of the data sampling time can be set according to actual needs. For example, if the control period is 1 hour, the data sampling time is 5 minutes.

[0242] In this way, potential safety hazards in the data center can be avoided.

[0243] Optionally, the step S170 of detecting the ambient temperature of the data center to determine whether it exceeds a safe temperature range includes:

[0244] TempMeasure, which detects the ambient temperature of the data center;

[0245] Calculate ΔT = TempMeasure - TempTarget;

[0246] Determine whether ΔT<-DeadLine, or ΔT>DeadLine.

[0247] The step S180, adjusting the number of startups and the temperature set point according to the set adjustment range, includes:

[0248] If ΔT < -DeadLine, the temperature set point of the current control cycle is increased by the preset temperature adjustment value, and the number of startups in the current control cycle is reduced by the preset startup number adjustment value;

[0249] If ΔT>DeadLine, the temperature set point of the current control cycle is reduced by the preset temperature adjustment value, and the number of startups in the current control cycle is increased by the preset startup number adjustment value;

[0250] Among them, TempMeasure is the temperature measurement value collected when the current control cycle arrives, TempTarget is the target temperature, and DeadLine is the preset temperature difference and is a positive number.

[0251] For example, the Deadline is a preset temperature difference of 3° C., the preset temperature adjustment value is 1° C., and the preset power-on number adjustment value is 1.

[0252] Optionally, the state parameter includes at least one of the following: load power, average supply air temperature, average return air temperature, average temperature on the hot aisle side, and average temperature on the cold aisle side.

[0253] If the technical solution of method 1 is used in model decision making, the input and output parameters of the UCB model and the regression model will be as follows: Figure 8-1 and Figure 8-2 shown.

[0254] A specific example is given below to illustrate the environmental parameter control device cluster control method provided by the present invention.

[0255] In this example, the temperature control device is an air conditioner, the control cycle is 1 hour, and the interval between adjacent data sampling is 5 minutes. The UCB model is the LinUCB model or the GPUCB model, and the regression model is the xgboost model. The target temperature TempTarget = 22°C, the preset temperature difference DeadLine = 3°C, the number of devices in the air conditioner cluster ranges from half to the full number of devices in the air conditioner cluster, the temperature setpoint range is [15°C, 20°C], and the sample set capacity of the temperature sample set and the energy efficiency sample set is 720. The recommendation phase begins at the 241st control cycle, with the first threshold being -2°C and the second threshold being 2°C.

[0256] like Figure 9 As shown, the control device cluster control method includes:

[0257] S300, CallTime=0.

[0258] S310: Determine whether the control cycle has arrived.

[0259] If the result of step S310 is yes, execute step S320; if the result of step S310 is no, execute step S380.

[0260] S320: Collect temperature samples and energy efficiency samples every time a control cycle is reached.

[0261] Among them, the sample features in the energy efficiency sample include the status parameters and the number of startups n of the data center obtained when the previous control cycle arrives, and the sample label is the energy efficiency score calculated when the current control cycle arrives. The sample features in the temperature sample include the status parameters, temperature set point AcTempSet, and the number of startups n of the data center obtained when the previous control cycle arrives, and the sample label includes the temperature measurement value TempMeasure collected when the current control cycle arrives.

[0262] The calculation method of energy efficiency score is:

[0263] If |TempMeasure-Temp|≤DeadLine,

[0264] If |TempMeasure-Temp|>DeadLine,

[0265] S321. Determine whether the number of samples in the temperature sample set and the energy efficiency sample set is equal to the sample set capacity.

[0266] If the result of step S321 is yes, execute step S322; if the result of step S320 is no, execute step S323.

[0267] S322: Delete the temperature sample and energy efficiency sample corresponding to the earliest control period in the temperature sample set and energy consumption sample set.

[0268] S323: Update the collected temperature samples and energy efficiency samples into the temperature sample set and energy efficiency sample set accordingly.

[0269] S330. Input the sample features in the energy efficiency sample set into the UCB model, and perform model training with the goal of outputting the sample labels in the energy efficiency sample set; input the sample features in the temperature sample set into the xgboost model, and perform model training with the goal of outputting the sample labels in the temperature sample set.

[0270] S340. Determine whether CallTime>InitTimeTh; wherein InitTimeTh=240.

[0271] If the result of step S340 is yes, execute step S360; if the result of step S340 is no, execute step S350.

[0272] S350 : Obtain a temperature measurement value TempMeasure collected when the current control cycle of the data center environment arrives, and calculate a difference ΔT=TempMeasure−TempTarget.

[0273] If ΔT<-2°C, execute step S351; if ΔT>2°C, execute step S352; if -2°C≤ΔT≤2°C, execute step S353.

[0274] S351 , n=n−1, AcTempSet=AcTempSet+1° C. Execute step S370 .

[0275] S352, n=n+1, AcTempSet=AcTempSet-1°C. Execute step S370.

[0276] S353 , n is randomly increased or decreased by 1, and AcTempSet is randomly increased or decreased by 1° C. Execute step S370 .

[0277] S360. Increase and decrease the temperature set point AcTempSet of the previous control cycle by 1°C and 1°C respectively to obtain two alternative temperature set points. After removing the alternative temperature set points that are not in [15°C, 20°C], the two alternative temperature set points are formed together with the temperature set point AcTempSet of the previous control cycle as the candidate temperature set point. Increase and decrease the number of power-on devices n of the previous control cycle by 1 and 1 respectively to obtain two alternative power-on numbers. After removing the alternative power-on numbers that are not in the range of half to all the number of air-conditioning cluster devices, the two alternative power-on numbers are formed together with the power-on number n of the previous control cycle as the candidate power-on number.

[0278] S361: Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively.

[0279] S362: Determine the candidate startup number corresponding to the maximum value of the energy efficiency score prediction value as the startup number n of the current control cycle.

[0280] S363: Input the number of startups in the current control cycle, the state parameters collected when the current control cycle is reached, and the candidate temperature set point into the xgboost model to predict the corresponding ambient temperature prediction value;

[0281] S364: Determine the temperature set point closest to the target temperature TempTarget among the predicted ambient temperature values ​​as the temperature set point AcTempSet of the current control cycle.

[0282] S370. CallTime=CallTime+1.

[0283] S380: Determine whether the data sampling time has arrived, and collect the status parameters of the data center once.

[0284] If the result of step S380 is yes, execute step S381; if the result of step S380 is no, execute step S310.

[0285] S381. Detect the ambient temperature TempMeasure of the data center; calculate the difference ΔT=TempMeasure-TempTarget; and determine whether ΔT<-DeadLine or ΔT>DeadLine.

[0286] If ΔT<-DeadLine, execute step S382; if ΔT>DeadLine, execute step S383; if -DeadLine≤ΔT≤DeadLine, return to step S310.

[0287] S382, AcTempSet=AcTempSet+1°C, n=n-1. Return to step S310.

[0288] S383, AcTempSet = AcTempSet - 1°C, n = n + 1. Return to step S310.

[0289] In the above example implementation, the parameter Alpha in the LinUCB model or Delta in the GPUCB model must be pre-set. Alpha / Delta are parameters that adjust the prediction strategy in the LinUCB / GPUCB model, controlling whether the model tends to exploit or explore. Specifically, the UCB model learns underlying patterns from historical samples and predicts the expected value for the working condition for which a recommendation decision is to be made. Obviously, for a constant environment, the model's selection of the condition with the highest expected value is the most appropriate and yields the greatest benefit. However, when the environment changes, the new state may not be included in the historical samples, so the model has not learned about it, and the model's predictions are likely to be inaccurate. Therefore, when the environment changes, appropriate exploration attempts are beneficial for the model to find the maximum expected value in the new environment. Thus, an exploitative bias means that the model tends to select the condition with the highest expected value, while an exploration bias means that the model tends to select the condition with a lower expected value. The more exploitative a model is, the more stable it is, but its adaptability to environmental changes is less. Conversely, a model with a more exploration bias is more adaptable to environmental changes, but excessive exploration can lead to poor stability.

[0290] At the same time, the parameters max_depth and learning_rate in the xgboost model also need to be pre-set. Among them, max_depth represents the maximum depth of the tree in the xgboost model. The larger the value, the stronger the fitting ability of the sample, but if it is too large, it is easy to fit to the noise and cause overfitting. Therefore, the value should be neither too large nor too small. When modeling, the data set needs to be divided and selected according to the test results. Preferably, max_depth = 5. learning_rate represents the learning rate in the xgboost model, also known as the learning step size. The smaller the value, the more iterations of weak learners are required, and the better the generalization. However, if the value is too small, the fitting effect may be reduced. When modeling, the data set also needs to be divided and selected according to the test results.

[0291] Second aspect:

[0292] Based on the same inventive concept, the embodiment of the present invention also provides an environmental parameter control device cluster control device, which is applied to energy-saving adjustment and environmental parameter adjustment of a data center, such as Figure 10 Shown, including:

[0293] The sample collection module M101 is used to collect environmental parameter samples and energy efficiency samples every time a control cycle is reached, and update them to the environmental parameter sample set and energy efficiency sample set accordingly;

[0294] The model training module M102 is used to input the sample features in the energy efficiency sample set into the upper confidence UCB model with context when triggering model training, and perform model training with the goal of outputting the sample labels in the energy efficiency sample set; and input the sample features in the environmental parameter sample set into the regression model, and perform model training with the goal of outputting the sample labels in the environmental parameter sample set;

[0295] The recommendation module M104 is configured to determine, when the current recommendation phase is in progress, a plurality of candidate environmental parameter set points and a plurality of candidate startup numbers for each control cycle, predict a corresponding energy efficiency score prediction value using the UCB model based on the state parameters and candidate startup numbers collected when the current control cycle arrives, predict a corresponding environmental parameter prediction value using a regression model based on the state parameters and candidate environmental parameter set points, and decide on the startup number and environmental parameter set points based on the energy efficiency score prediction value and the environmental parameter prediction value, and then execute the operation;

[0296] Among them, the sample features in the energy efficiency sample include the status parameters and the number of startups of the data center obtained when the previous control cycle arrives, and the sample label is the energy efficiency score calculated when the current control cycle arrives. The sample features in the environmental parameter sample include the status parameters and the environmental parameter setting points of the data center obtained when the previous control cycle arrives, and the sample label includes the environmental parameter measurement values ​​collected when the current control cycle arrives.

[0297] Optionally, the energy efficiency score is a score determined according to an energy efficiency index of the data center;

[0298] The energy efficiency index includes the total power of the data center or the total power of the environmental parameter control device cluster or the power usage efficiency (PUE) of the data center; wherein:

[0299] If the environmental parameter measurement value collected when the current control cycle arrives is less than or equal to the preset environmental parameter deviation from the environmental parameter set point of the previous control cycle, the energy efficiency score is calculated using the first formula based on the energy efficiency index of the data center;

[0300] If the difference between the environmental parameter measurement value collected at the time of the current control cycle and the environmental parameter set point of the previous control cycle is greater than the preset environmental parameter deviation, the energy efficiency score is calculated using the second formula based on the energy efficiency index of the data center;

[0301] Among them, when the energy efficiency indicators are the same, the energy efficiency score calculated using the first formula is greater than the energy efficiency score calculated using the second formula.

[0302] Optionally, the sample features in the environmental parameter sample further include: the number of startups in the previous control cycle; based on the state parameters and candidate startup numbers collected when the current control cycle arrives, using the UCB model to predict the corresponding energy efficiency score; based on the state parameters and candidate environmental parameter set points, using a regression model to predict the corresponding environmental parameter predicted value; and deciding the startup number and environmental parameter set point based on the energy efficiency score and the environmental parameter predicted value, including:

[0303] Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively;

[0304] Determining the number of startups according to the energy efficiency score prediction value;

[0305] Input the startup number, the state parameters collected when the current control cycle arrives, and the candidate environmental parameter set points into the regression model to predict the corresponding environmental parameter prediction value;

[0306] The environmental parameter set point is determined according to the environmental parameter prediction value.

[0307] Optionally, the sample features in the energy efficiency sample further include: an environmental parameter set point of the previous control cycle;

[0308] Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; based on the energy efficiency score and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including:

[0309] Inputting the candidate environmental parameter set point and the state parameter collected when the current control cycle arrives into the regression model to predict the corresponding environmental parameter prediction value;

[0310] Determine the environmental parameter set point according to the environmental parameter prediction value;

[0311] Input the candidate startup number, the state parameters collected when the current control cycle arrives, and the environmental parameter set point into the UCB model to predict the corresponding energy efficiency score prediction value;

[0312] The number of startups is determined according to the energy efficiency score prediction value.

[0313] Optionally, based on the state parameters and candidate startup numbers collected when the current control period arrives, the UCB model is used to predict a corresponding energy efficiency score; based on the state parameters and candidate environmental parameter set points, a regression model is used to predict a corresponding environmental parameter predicted value; and based on the energy efficiency score and the environmental parameter predicted value, the startup number and the environmental parameter set point are decided, including:

[0314] Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively;

[0315] Determining the number of startups according to the energy efficiency score prediction value;

[0316] Inputting the candidate environmental parameter set point and the state parameter collected when the current control cycle arrives into the regression model, and predicting the corresponding environmental parameter prediction values ​​respectively;

[0317] The environmental parameter set point is determined according to the environmental parameter predicted value.

[0318] Optionally, if the greater the energy efficiency score, the higher the energy consumption efficiency of the environmental parameter control device cluster, then determining the number of startup devices according to the predicted value of the energy efficiency score includes:

[0319] Determine the candidate startup number corresponding to the largest energy efficiency score prediction value as the startup number;

[0320] Determining the environmental parameter set point according to the environmental parameter prediction value includes:

[0321] The candidate environmental parameter setting point corresponding to the environmental parameter prediction value closest to the target environmental parameter is decided as the environmental parameter setting point.

[0322] Optionally, when a control period is reached, model training is triggered;

[0323] Inputting sample features in the energy efficiency index sample set into the upper confidence UCB model with context, and performing model training with the goal of outputting sample labels in the energy efficiency sample set, including:

[0324] Inputting sample features of the energy efficiency samples in the current control cycle in the energy efficiency sample set into the UCB model, and performing model training with the goal of outputting sample labels of the energy efficiency samples in the current control cycle; or

[0325] The sample features of all energy efficiency samples in the energy efficiency sample set are sequentially input into the upper confidence UCB model, and the model training is performed with the goal of outputting the corresponding sample labels;

[0326] Inputting sample features in the environmental parameter sample set into a regression model, and performing model training with the goal of outputting labels in the environmental parameter sample set, including:

[0327] The sample features of all environmental parameter samples in the environmental parameter sample set are sequentially input into the regression model, and the model training is performed with the goal of outputting corresponding sample labels.

[0328] Optionally, the device further comprises:

[0329] The initial control module M103 is configured to obtain, when it is determined that the model decision condition is not satisfied, the environmental parameter measurement values ​​collected when the current control cycle of the data center environment arrives each time a control cycle is reached;

[0330] Adjust the startup number and environmental parameter set point according to the difference between the environmental parameter measurement value and the target environmental parameter and run;

[0331] When it is determined that the model decision conditions are met, it is determined to enter the recommendation stage.

[0332] Optionally, when it is determined that the model decision condition is not satisfied, each time a control cycle is reached, an environmental parameter measurement value of the data center environment is obtained, and the number of startups and the environmental parameter set point are adjusted according to the difference between the environmental parameter measurement value and the target environmental parameter, including:

[0333] When it is determined that the model decision condition is not met, the difference ΔE = EnvirMeasure-EnvirTarget is calculated every time a control cycle is reached;

[0334] If the difference ΔE is less than the first threshold, the environmental parameter set point of the current control period is determined to be the environmental parameter set point of the previous control period increased by the preset environmental parameter adjustment value, and the number of startups in the current control period is determined to be the number of startups in the previous control period decreased by the startup number adjustment value;

[0335] If the difference ΔE is greater than a second threshold, determining that the environmental parameter set point of the current control cycle is the environmental parameter set point of the previous control cycle minus the preset environmental parameter adjustment value, and the number of startups in the current control cycle is the number of startups in the previous control cycle plus the preset startup number adjustment value;

[0336] If the difference ΔE is greater than or equal to the first threshold and less than or equal to the second threshold, determining that the environmental parameter set point of the current control period is the environmental parameter set point of the previous control period and is randomly increased or decreased by a preset environmental parameter adjustment value, and the number of startups in the current control period is the number of startups in the previous control period and is randomly increased or decreased by a preset startup number adjustment value;

[0337] Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, and EnvirTarget is the target environmental parameter.

[0338] Optionally, determining a plurality of candidate environmental parameter set points and a plurality of candidate startup numbers comprises:

[0339] determining a plurality of candidate environmental parameter set points centered around an environmental parameter set point of a previous control cycle;

[0340] determining, among the plurality of candidate environmental parameter set points, an alternative environmental parameter set point that meets the environmental parameter set point range and the environmental parameter set point of the previous control cycle as candidate environmental parameter set points;

[0341] Determine multiple candidate startup numbers centered on the startup number of the previous control cycle;

[0342] Determine the candidate startup numbers that meet the startup number range among the multiple candidate startup numbers and the startup number in the previous control cycle as candidate startup numbers.

[0343] Optionally, the device further comprises:

[0344] The safety maintenance module M105 is used to collect the status parameters of the data center at each data sampling time, trigger the detection of the environmental parameters of the data center, and adjust the number of startups and the environmental parameter set points according to the set adjustment range when the safety environmental parameter range is exceeded;

[0345] The length of one of the control cycles is an integer multiple of the time interval between adjacent data samplings.

[0346] Optionally, each time a data sampling time is reached, the state parameters of the data center are collected, and the environmental parameters of the data center are triggered to be detected. When the safe environmental parameter range is exceeded, the number of startups and the environmental parameter set point are adjusted according to a set adjustment range, including:

[0347] Every time the data sampling time arrives, the state parameters of the data center are collected, and the environmental parameters EnvirMeasure of the data center are triggered;

[0348] Calculate ΔE = EnvirMeasure - EnvirTarget;

[0349] If ΔE<-DeadLine, the environmental parameter set point of the current control cycle is increased by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is reduced by the preset startup number adjustment value;

[0350] If ΔE>DeadLine, the environmental parameter set point of the current control cycle is reduced by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is increased by the preset startup number adjustment value;

[0351] Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, EnvirTarget is the target environmental parameter, and DeadLine is the preset environmental parameter difference and is a positive number.

[0352] Optionally, the UCB model is a linear upper confidence LinUCB model or a Gaussian UCB model.

[0353] Optionally, the regression model is any one of the following:

[0354] xgboost model, random forest RF model, support vector machine SVM model, neural network model.

[0355] Optionally, the state parameter includes at least one of the following: load power, average supply air environment parameter, average return air environment parameter, average hot channel side environment parameter, and average cold channel side environment parameter.

[0356] Optionally, the environmental parameter is temperature or humidity.

[0357] During the specific implementation process, the specific working principles of the environmental parameter control equipment cluster control device and the environmental parameter control equipment cluster control method are similar, so the specific implementation method of the environmental parameter control equipment cluster control method can be referred to for corresponding implementation, and no further details will be given here.

[0358] The third aspect:

[0359] Based on the same inventive concept, an embodiment of the present invention further provides an electronic device 100, such as Figure 11 As shown, it includes: a processor 110 and a memory 120 for storing executable instructions of the processor 110; wherein the processor 110 is configured to execute the instructions to implement the environmental parameter control device cluster control method.

[0360] In a specific implementation, the device 100 may have relatively large differences due to different configurations or performances, and may include one or more processors 110 and memories 120, and one or more storage media 130 storing applications 131 or data 132. The memories 120 and storage media 130 may be temporary storage or permanent storage. The application 131 stored in the storage medium 130 may include one or more of the above units ( Figure 11 (not shown), each module may include a series of instruction operations in the environmental parameter control device cluster control device. Further, the processor 110 may be configured to communicate with the storage medium 130 and execute a series of instruction operations in the storage medium 130 on the device 100. The device 100 may also include one or more power supplies ( Figure 11 one or more network interfaces 140, wherein the network interface 140 includes a wired network interface 141 or a wireless network interface 142; one or more input and output interfaces 143; and / or one or more operating systems 133, such as Windows, Mac OS, Linux, IOS, Android, Unix, FreeBSD, etc.

[0361] Figure 12 The data center control system composed of the electronic device 100 provided by the embodiment of the present invention is illustrated. Figure 12 As shown, the data center control system includes the electronic device 100, data center equipment 200, and monitoring system equipment 300. The data center equipment 200 includes a temperature control device cluster and / or a humidity control device cluster, a cabinet containing data center server equipment, temperature sensors and / or humidity sensors, and the like. There is at least one monitoring system device 300, which is used to control the operating status of the data center equipment. The electronic device 100 receives state parameters collected by the data center and processed and forwarded by the monitoring system device 300, and determines environmental parameter set points and the number of startups based on these state parameters. The monitoring system device 300 then controls the corresponding data center equipment 200 based on the decision made by the electronic device 100.

[0362] Fourth aspect:

[0363] Based on the same inventive concept, an embodiment of the present invention further provides a computer storage medium, wherein the computer storage medium stores a computer program, and the computer program is used to implement the environmental parameter control device cluster control method.

[0364] The temperature and humidity control device cluster control method, apparatus, device, and storage medium provided by the present invention combine a contextualized Unified Combination (UCB) model algorithm with a regression model algorithm to recommend data center configurations, achieving high accuracy and rapid convergence. The method adjusts the number of air conditioners on and the temperature set point within a specific control cycle, thereby improving data center temperature and reducing PUE in a safe and reliable manner.

[0365] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0366] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0367] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0368] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0369] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A method for controlling a cluster of environmental parameter control devices, applied to energy-saving and environmental parameter regulation in a data center, characterized in that: include: Every time a control cycle is reached, environmental parameter samples and energy efficiency samples are collected and updated to the environmental parameter sample set and energy efficiency sample set accordingly; When triggering model training, the sample features in the energy efficiency sample set are input into the upper confidence UCB model with context, and the model training is performed with the goal of outputting the sample labels in the energy efficiency sample set; the sample features in the environmental parameter sample set are input into the regression model, and the model training is performed with the goal of outputting the sample labels in the environmental parameter sample set; When it is determined that the current stage is in the recommendation stage, each time a control cycle is reached, multiple candidate environmental parameter set points and multiple candidate startup numbers are determined, based on the state parameters and candidate startup numbers collected when the current control cycle is reached, the UCB model is used to predict the corresponding energy efficiency score prediction value, based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value, and the startup number and environmental parameter set point are respectively decided based on the energy efficiency score prediction value and the environmental parameter prediction value and run; wherein, when it is determined that the UCB model and the regression model meet the model decision conditions, it is determined to enter the recommendation stage; the environmental parameter set point is a parameter value for setting temperature or humidity, and the state parameter is a state parameter of the environmental parameter control device; Otherwise, each time a control cycle is reached, the environmental parameter measurement value collected when the current control cycle of the data center environment arrives is obtained; the number of startups and the environmental parameter set point are adjusted according to the difference between the environmental parameter measurement value and the target environmental parameter, and the operation is carried out; Among them, the sample features in the energy efficiency sample include the status parameters and the number of startups of the data center obtained when the previous control cycle arrives, and the sample label is the energy efficiency score calculated when the current control cycle arrives. The sample features in the environmental parameter sample include the status parameters and the environmental parameter setting points of the data center obtained when the previous control cycle arrives, and the sample label includes the environmental parameter measurement values ​​collected when the current control cycle arrives.

2. The method according to claim 1, wherein The energy efficiency score is a score determined according to the energy efficiency index of the data center; The energy efficiency index includes the total power of the data center or the total power of the environmental parameter control device cluster or the power usage efficiency (PUE) of the data center; wherein: If the environmental parameter measurement value collected when the current control cycle arrives is less than or equal to the preset environmental parameter deviation from the environmental parameter set point of the previous control cycle, the energy efficiency score is calculated using the first formula based on the energy efficiency index of the data center; If the difference between the environmental parameter measurement value collected at the time of the current control cycle and the environmental parameter set point of the previous control cycle is greater than the preset environmental parameter deviation, the energy efficiency score is calculated using the second formula based on the energy efficiency index of the data center; Wherein, when the energy efficiency indicators are the same, the energy efficiency score calculated using the first formula is greater than the energy efficiency score calculated using the second formula; The first formula is: The second formula is: Wherein score is the energy efficiency score, PUE is the energy efficiency index, and A>B.

3. The method according to claim 1, wherein The sample features in the environmental parameter sample also include: the number of startups in the previous control cycle; Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including: Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively; Determining the number of startups according to the energy efficiency score prediction value; Inputting the number of startups determined according to the energy efficiency score prediction value, the state parameters collected when the current control cycle arrives, and the candidate environmental parameter set points into the regression model to predict the corresponding environmental parameter prediction value; The environmental parameter set point is determined according to the environmental parameter predicted value.

4. The method according to claim 1, wherein The sample characteristics in the energy efficiency sample also include: the environmental parameter set point of the previous control cycle; Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including: Inputting the candidate environmental parameter set point and the state parameter collected when the current control cycle arrives into the regression model to predict the corresponding environmental parameter prediction value; Determine the environmental parameter set point according to the environmental parameter prediction value; Input the candidate startup number, the state parameters collected when the current control cycle arrives, and the environmental parameter set point determined according to the environmental parameter prediction value into the UCB model, and predict the corresponding energy efficiency score prediction value respectively; The number of startups is determined according to the energy efficiency score prediction value.

5. The method according to claim 1, wherein Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including: Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively; Determining the number of startups according to the energy efficiency score prediction value; Inputting the candidate environmental parameter set point and the state parameter collected when the current control cycle arrives into the regression model, and predicting the corresponding environmental parameter prediction values ​​respectively; The environmental parameter set point is determined according to the environmental parameter predicted value.

6. The method according to any one of claims 3 to 5, characterized in that If the energy efficiency score is larger, the energy consumption efficiency of the environmental parameter control device cluster is higher, and the number of startup devices is determined according to the predicted value of the energy efficiency score, including: Determine the candidate startup number corresponding to the largest energy efficiency score prediction value as the startup number; Determining the environmental parameter set point according to the environmental parameter prediction value includes: The candidate environmental parameter setting point corresponding to the environmental parameter prediction value closest to the target environmental parameter is decided as the environmental parameter setting point.

7. The method according to claim 1, wherein When the control period is reached, model training is triggered; Inputting sample features in the energy efficiency index sample set into the context-based UCB model, and performing model training with the goal of outputting sample labels in the energy efficiency sample set, including: Inputting sample features of the energy efficiency samples in the current control cycle in the energy efficiency sample set into the UCB model, and performing model training with the goal of outputting sample labels of the energy efficiency samples in the current control cycle; or Input the sample features of all energy efficiency samples in the energy efficiency sample set into the UCB model in sequence, and perform model training with the goal of outputting corresponding sample labels; Inputting sample features in the environmental parameter sample set into a regression model, and performing model training with the goal of outputting labels in the environmental parameter sample set, including: The sample features of all environmental parameter samples in the environmental parameter sample set are sequentially input into the regression model, and the model training is performed with the goal of outputting corresponding sample labels.

8. The method according to claim 1, wherein When it is determined that the model decision conditions are not met, each time a control cycle is reached, the environmental parameter measurement values ​​of the data center environment are obtained, and the number of startups and the environmental parameter set points are adjusted according to the difference between the environmental parameter measurement values ​​and the target environmental parameters, including: When it is determined that the model decision condition is not met, the difference ΔE = EnvirMeasure-EnvirTarget is calculated every time a control cycle is reached; If the difference ΔE is less than the first threshold, the environmental parameter set point of the current control period is determined to be the environmental parameter set point of the previous control period increased by the preset environmental parameter adjustment value, and the number of startups in the current control period is determined to be the number of startups in the previous control period decreased by the startup number adjustment value; If the difference ΔE is greater than a second threshold, determining that the environmental parameter set point of the current control cycle is the environmental parameter set point of the previous control cycle minus the preset environmental parameter adjustment value, and the number of startups in the current control cycle is the number of startups in the previous control cycle plus the preset startup number adjustment value; If the difference ΔE is greater than or equal to the first threshold and less than or equal to the second threshold, determining that the environmental parameter set point of the current control period is the environmental parameter set point of the previous control period and is randomly increased or decreased by a preset environmental parameter adjustment value, and the number of startups in the current control period is the number of startups in the previous control period and is randomly increased or decreased by a preset startup number adjustment value; Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, and EnvirTarget is the target environmental parameter.

9. The method according to claim 1, wherein Determine multiple candidate environmental parameter set points and multiple candidate start-up numbers, including: determining a plurality of candidate environmental parameter set points centered around an environmental parameter set point of a previous control cycle; determining, among the plurality of candidate environmental parameter set points, an alternative environmental parameter set point that meets the environmental parameter set point range and the environmental parameter set point of the previous control cycle as candidate environmental parameter set points; Determine multiple candidate startup numbers centered on the startup number of the previous control cycle; Determine the candidate startup numbers that meet the startup number range among the multiple candidate startup numbers and the startup number in the previous control cycle as candidate startup numbers.

10. The method according to claim 1, wherein Also includes: At each data sampling time, the state parameters of the data center are collected, and the environmental parameters of the data center are triggered to be detected. When the safe environmental parameters are exceeded, the number of startups and the environmental parameter set points are adjusted according to the set adjustment range; The length of one of the control cycles is an integer multiple of the time interval between adjacent data samplings.

11. The method according to claim 10, wherein At each data sampling time, the data center's status parameters are collected and the data center's environmental parameters are triggered to detect. When the safe environmental parameter range is exceeded, the number of startups and the environmental parameter set point are adjusted according to the set adjustment range, including: Every time the data sampling time arrives, the state parameters of the data center are collected, and the environmental parameters EnvirMeasure of the data center are triggered; Calculate ΔE = EnvirMeasure - EnvirTarget; If ΔE<-DeadLine, the environmental parameter set point of the current control cycle is increased by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is reduced by the preset startup number adjustment value; If ΔE>DeadLine, the environmental parameter set point of the current control cycle is reduced by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is increased by the preset startup number adjustment value; Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, EnvirTarget is the target environmental parameter, and DeadLine is the preset environmental parameter difference and is a positive number.

12. The method according to claim 1, wherein The UCB model is a linear upper confidence LinUCB model or a Gaussian upper confidence GPUCB model.

13. The method according to claim 1, wherein The regression model is any of the following: xgboost model, random forest RF model, support vector machine SVM model, neural network model.

14. The method according to claim 1, wherein The state parameters include at least one of the following: Load power, average supply air environmental parameters, average return air environmental parameters, average hot aisle side environmental parameters, and average cold aisle side environmental parameters.

15. The method according to claim 1, wherein The environmental parameter is temperature or humidity.

16. An environmental parameter control device cluster control device, used for energy saving and environmental parameter adjustment in a data center, characterized in that: include: The sample collection module is used to collect environmental parameter samples and energy efficiency samples every time a control cycle is reached, and update them to the environmental parameter sample set and energy efficiency sample set accordingly; A model training module is configured to, when triggering model training, input the sample features in the energy efficiency sample set into a UCB model with context, and perform model training with the goal of outputting the sample labels in the energy efficiency sample set; and input the sample features in the environmental parameter sample set into a regression model, and perform model training with the goal of outputting the sample labels in the environmental parameter sample set; a recommendation module, configured to determine, when currently in a recommendation phase, a plurality of candidate environmental parameter set points and a plurality of candidate startup numbers for each control cycle reached, predict a corresponding energy efficiency score prediction value using the UCB model based on the state parameters and candidate startup numbers collected when the current control cycle reaches the end, predict a corresponding environmental parameter prediction value using a regression model based on the state parameters and candidate environmental parameter set points, and decide on the startup number and environmental parameter set point based on the energy efficiency score prediction value and the environmental parameter prediction value, respectively, and then operate; wherein the environmental parameter set point is a parameter value for setting temperature or humidity, and the state parameter is a state parameter of an environmental parameter control device; The initial control module is configured to obtain, when determining that the model decision conditions are not met, the environmental parameter measurement values ​​collected when the current control cycle of the data center environment arrives at each control cycle; adjust the number of startups and the environmental parameter set points based on the difference between the environmental parameter measurement values ​​and the target environmental parameters and operate; and determine to enter the recommendation phase when determining that the UCB model and the regression model meet the model decision conditions; Among them, the sample features in the energy efficiency sample include the status parameters and the number of startups of the data center obtained when the previous control cycle arrives, and the sample label is the energy efficiency score calculated when the current control cycle arrives. The sample features in the environmental parameter sample include the status parameters and the environmental parameter setting points of the data center obtained when the previous control cycle arrives, and the sample label includes the environmental parameter measurement values ​​collected when the current control cycle arrives.

17. The device according to claim 16, characterized in that The sample features in the environmental parameter sample also include: the number of startups in the previous control cycle; Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including: Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively; Determining the number of startups according to the energy efficiency score prediction value; Inputting the number of startups determined according to the energy efficiency score prediction value, the state parameters collected when the current control cycle arrives, and the candidate environmental parameter set points into the regression model to predict the corresponding environmental parameter prediction value; The environmental parameter set point is determined according to the environmental parameter prediction value.

18. The device according to claim 16, wherein The sample characteristics in the energy efficiency sample also include: the environmental parameter set point of the previous control cycle; Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including: Inputting the candidate environmental parameter set point and the state parameter collected when the current control cycle arrives into the regression model to predict the corresponding environmental parameter prediction value; Determine the environmental parameter set point according to the environmental parameter prediction value; Input the candidate startup number, the state parameters collected when the current control cycle arrives, and the environmental parameter set point determined according to the environmental parameter prediction value into the UCB model, and predict the corresponding energy efficiency score prediction value respectively; The number of startups is determined according to the energy efficiency score prediction value.

19. The device according to claim 16, wherein Based on the state parameters and candidate startup numbers collected when the current control cycle arrives, the UCB model is used to predict the corresponding energy efficiency score prediction value; based on the state parameters and candidate environmental parameter set points, the regression model is used to predict the corresponding environmental parameter prediction value; and based on the energy efficiency score prediction value and the environmental parameter prediction value, the startup number and the environmental parameter set point are determined, including: Input the candidate startup number and the state parameters collected when the current control cycle arrives into the UCB model, and predict the corresponding energy efficiency score prediction values ​​respectively; Determining the number of startups according to the energy efficiency score prediction value; Inputting the candidate environmental parameter set point and the state parameter collected when the current control cycle arrives into the regression model, and predicting the corresponding environmental parameter prediction values ​​respectively; The environmental parameter set point is determined according to the environmental parameter predicted value.

20. The device according to claim 16, wherein When it is determined that the model decision conditions are not met, each time a control cycle is reached, the environmental parameter measurement values ​​of the data center environment are obtained, and the number of startups and the environmental parameter set points are adjusted according to the difference between the environmental parameter measurement values ​​and the target environmental parameters, including: When it is determined that the model decision condition is not met, the difference ΔE = EnvirMeasure-EnvirTarget is calculated every time a control cycle is reached; If the difference ΔE is less than the first threshold, the environmental parameter set point of the current control period is determined to be the environmental parameter set point of the previous control period increased by the preset environmental parameter adjustment value, and the number of startups in the current control period is determined to be the number of startups in the previous control period decreased by the startup number adjustment value; If the difference ΔE is greater than a second threshold, determining that the environmental parameter set point of the current control cycle is the environmental parameter set point of the previous control cycle minus the preset environmental parameter adjustment value, and the number of startups in the current control cycle is the number of startups in the previous control cycle plus the preset startup number adjustment value; If the difference ΔE is greater than or equal to the first threshold and less than or equal to the second threshold, determining that the environmental parameter set point of the current control period is the environmental parameter set point of the previous control period and is randomly increased or decreased by a preset environmental parameter adjustment value, and the number of startups in the current control period is the number of startups in the previous control period and is randomly increased or decreased by a preset startup number adjustment value; Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, and EnvirTarget is the target environmental parameter.

21. The device according to claim 16, wherein Also includes: A safety maintenance module is used to collect the status parameters of the data center at each data sampling time, trigger the detection of the environmental parameters of the data center, and adjust the number of startups and the environmental parameter set points according to the set adjustment range when the safety environmental parameter range is exceeded; The length of one of the control cycles is an integer multiple of the time interval between adjacent data samplings.

22. The device according to claim 21, wherein At each data sampling time, the data center's status parameters are collected and the data center's environmental parameters are triggered to detect. When the safe environmental parameter range is exceeded, the number of startups and the environmental parameter set point are adjusted according to the set adjustment range, including: Every time the data sampling time arrives, the state parameters of the data center are collected, and the environmental parameters EnvirMeasure of the data center are triggered; Calculate ΔE = EnvirMeasure - EnvirTarget; If ΔE<-DeadLine, the environmental parameter set point of the current control cycle is increased by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is reduced by the preset startup number adjustment value; If ΔE>DeadLine, the environmental parameter set point of the current control cycle is reduced by the preset environmental parameter adjustment value, and the number of startups in the current control cycle is increased by the preset startup number adjustment value; Among them, EnvirMeasure is the environmental parameter measurement value collected when the current control cycle arrives, EnvirTarget is the target environmental parameter, and DeadLine is the preset environmental parameter difference and is a positive number.

23. An electronic device, characterized in that: include: a processor and a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the environmental parameter control device cluster control method according to any one of claims 1 to 15.

24. A storage medium, characterized in that The computer storage medium stores a computer program, which is used to implement the environmental parameter control device cluster control method according to any one of claims 1 to 15.

Images