Method and apparatus for predicting return water temperature of data center cooling system

Abstract

The application relates to the technical field of time series prediction, in particular to a data center cooling system return water temperature prediction method and device, wherein the method comprises the following steps: collecting initial operation data of a cooling system in a data center; pre-processing the initial operation data to obtain final operation data meeting certain conditions; inputting the final operation data into a pre-trained time series prediction model to obtain a cooling system return water temperature prediction result, wherein the time series prediction model has a Kronecker product layer, a kernel matrix layer and a multilayer perception layer. Thus, the problems that, in the related art, model parameters are large, blackboxing is caused by expressing nonlinear characteristics through a complex activation function, it is difficult to directly express cross-variable relationships, the effectiveness and interpretability of the model in actual application are limited, and the understanding and trust of a user for a model prediction result are affected are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of time series prediction, in particular to a method and device for predicting return water temperature of a data center cooling system. BACKGROUND

[0002] A data center is an infrastructure for guaranteeing the operation of servers and network devices, wherein a cooling system is an important component of the operation of the data center, and if not properly controlled, it will lead to an increase in server failure rate and even service interruption.

[0003] In the related art, an original database can be constructed by acquiring thermodynamic parameters of a data center cooling system in real time, and preprocessed to obtain a power consumption dataset and a chip temperature dataset, and then a power consumption prediction neural network is trained to maintain the optimized control result within the chip critical temperature to maximize the energy saving of the data center; or a data-driven model predictive control method can be used to apply a data-driven control strategy based on behavior theory to a data center rack air cooling control system, improve the control algorithm according to the characteristics of the rack system, and then directly realize the temperature control of the rack type data center.

[0004] However, in the related art, these models usually have a large number of parameters, which not only leads to a huge consumption of computing resources and a prolongation of training time, but also increases the cost of model deployment and operation. In addition, the nonlinear characteristics of the model are mainly expressed through activation functions, making the internal mechanism of the model become black-boxed, which is difficult to intuitively explain and understand, and lacks transparency and credibility in practical application, and needs to be improved. SUMMARY

[0005] The present application provides a method and device for predicting return water temperature of a data center cooling system to solve the problems in the related art that the model has a large number of parameters, the nonlinear characteristics are expressed through complex activation functions, leading to black-boxing and difficulty in intuitively expressing cross-variable relationships, which limits the effectiveness and interpretability of the model in practical application, and affects the understanding and trust of users on the model prediction results.

[0006] The first aspect embodiment of the present application provides a method for predicting return water temperature of a data center cooling system, comprising the following steps: collecting initial operation data of a cooling system in a data center; pre-processing the initial operation data to obtain final operation data that meets a preset condition; inputting the final operation data into a pre-trained time series prediction model to obtain a prediction result of the return water temperature of the cooling system, wherein the time series prediction model has a Kronecker product layer, a kernel matrix layer and a multilayer perception layer.

[0007] Optionally, in an embodiment of the present application, before the operation data is input into the pre-trained time series prediction model, further comprising: constructing a Kronecker product layer of the time series prediction model by using a Kronecker product and a rotation mask, so as to obtain an input matrix of the final operation data by using the Kronecker product layer; constructing a kernel matrix layer of the time series prediction model by combining the input matrix and a kernel function, so as to obtain an output matrix corresponding to the input matrix according to the kernel matrix layer; constructing a multi-layer perception machine layer of the time series prediction model by using the output matrix and a linear multi-layer perception machine, so as to construct the pre-trained time series prediction model.

[0008] Optionally, in an embodiment of the present application, the constructing the pre-trained time series prediction model comprises: obtaining a training data set and a test data set according to the final operation data; and generating the pre-trained time series prediction model based on the training data set, the test data set, a mean squared error (MSE) and a mean absolute error (MAE).

[0009] Optionally, in an embodiment of the present application, the constructing the Kronecker product layer of the time series prediction model by using the Kronecker product and the rotation mask, so as to obtain the input matrix of the final operation data, comprises: obtaining an initial input matrix corresponding to the final operation data based on the final operation data; calculating a Kronecker product matrix of the final operation data by using the initial input matrix and the Kronecker product; and obtaining the input matrix based on the Kronecker product matrix and the rotation mask.

[0010] Optionally, in an embodiment of the present application, a mathematical expression of the multi-layer perception machine layer can be, but is not limited to, as follows:

[0011] y(k) = b + k1x y (1) (k) + … + k n x y (n) (k),

[0012] wherein, y n (k) is an n-order output result of the model, b is a bias term of the multi-layer perception machine, k i is a weight corresponding to the order result.

[0013] The second aspect embodiment of the present application provides a data center cooling system return water temperature prediction device, comprising: an acquisition module configured to acquire initial operation data of a cooling system in a data center; a preprocessing module configured to preprocess the initial operation data to obtain final operation data satisfying a preset condition; and a prediction module configured to input the final operation data into a pre-trained time series prediction model to obtain a prediction result of the cooling system return water temperature, wherein the time series prediction model has a Kronecker product layer, a kernel matrix layer, and a multilayer perceptron layer.

[0014] Optionally, in an embodiment of the present application, the device further comprises: a first construction module configured to construct the Kronecker product layer of the time series prediction model by using a Kronecker product and a rotation mask before inputting the operation data into the pre-trained time series prediction model, so as to obtain an input matrix of the final operation data by using the Kronecker product layer; a second construction module configured to construct the kernel matrix layer of the time series prediction model in combination with the input matrix and a kernel function, so as to obtain an output matrix corresponding to the input matrix according to the kernel matrix layer; and a third construction module configured to construct the multilayer perceptron layer of the time series prediction model by using the output matrix and a linear multilayer perceptron, so as to construct the pre-trained time series prediction model.

[0015] Optionally, in an embodiment of the present application, the third construction module comprises: an acquisition unit configured to acquire a training data set and a test data set according to the final operation data; and a first generation unit configured to generate the pre-trained time series prediction model based on the training data set, the test data set, MSE, and MAE.

[0016] Optionally, in an embodiment of the present application, the first construction module comprises: a second generation unit configured to obtain an initial input matrix corresponding to the final operation data based on the final operation data; a calculation unit configured to calculate a Kronecker product matrix of the final operation data by using the initial input matrix and the Kronecker product; and a third generation unit configured to obtain the input matrix based on the Kronecker product matrix and the rotation mask.

[0017] Optionally, in an embodiment of the present application, a mathematical expression of the multilayer perceptron layer can be, but is not limited to, as follows:

[0018] y(k) = b + k1x y (1) (k) + … + k n x y (n) (k),

[0019] wherein y n (k) is an n-order output result of the model, b is a bias term of the multilayer perceptron, ki to correspond to the weight of the order result.

[0020] The third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for predicting the return water temperature of the cooling system of the data center according to the above-mentioned embodiments.

[0021] The fourth aspect of the present application provides a computer readable storage medium, which stores a computer program, wherein the program is executed by a processor to implement the method for predicting the return water temperature of the cooling system of the data center according to the above-mentioned embodiments.

[0022] The fifth aspect of the present application provides a computer program product, comprising a computer program, wherein the program is executed to implement the method for predicting the return water temperature of the cooling system of the data center according to the above-mentioned embodiments.

[0023] The embodiments of the present application can preprocess the initial operation data of the cooling system in the collected data center, and then obtain the final operation data satisfying certain conditions, and input the final operation data into the pre-trained time series prediction model to obtain the prediction result of the return water temperature of the cooling system, thereby realizing accurate prediction of the return water temperature and the average temperature and humidity of the return air of the cooling system, intuitively expressing the variable relationship, performing detailed variable influence analysis, and providing effective system optimization strategy, so as to improve the overall performance of the cooling system and ensure the safe and efficient operation of the server room. Thus, the problems in the related art, such as the large amount of model parameters, the black-box problem caused by the expression of the nonlinear characteristics through the complex activation function, the difficulty in intuitively expressing the cross-variable relationship, the limitation of the effectiveness and interpretability of the model in actual application, and the influence on the understanding and trust of the user on the prediction result of the model are solved.

[0024] Additional aspects and advantages of the present application will be made apparent by the following description and the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0025] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description, taken in conjunction with the accompanying drawings, in which:

[0026] Figure 1 a flowchart of a method for predicting the return water temperature of the cooling system of the data center according to an embodiment of the present application;

[0027] Figure 2 a structural schematic diagram of a time series prediction model according to an embodiment of the present application;

[0028] Figure 3A block diagram of a prediction of a data center cooling system according to an embodiment of the present application is provided.

[0029] Figure 4 A block diagram of a prediction of an indirect evaporative cooling system according to an embodiment of the present application is provided.

[0030] Figure 5 A block diagram of an overall algorithm of a time series prediction model according to an embodiment of the present application is provided.

[0031] Figure 6 A block diagram of a prediction device of a data center cooling system return water temperature according to an embodiment of the present application is provided.

[0032] Figure 7 A structural diagram of an electronic device according to an embodiment of the present application is provided. DETAILED DESCRIPTION

[0033] Embodiments of the present application are described in detail below with reference to the accompanying drawings, in which the same or similar components have the same or similar designations and functions throughout various figures and / or portions of the drawings. The embodiments described below are examples in which similar or identical components have the same or similar designations and functions throughout the various figures and / or portions of the drawings. The embodiments described below are exemplary and are intended to explain the present application and are not to be understood as limiting the present application.

[0034] A prediction method and device of a data center cooling system return water temperature according to an embodiment of the present application are described below with reference to the accompanying drawings. In view of the problems of a large number of model parameters, black-boxing caused by expressing nonlinear characteristics through a complex activation function, and difficulty in intuitively expressing cross-variable relationships, which limit the effectiveness and interpretability of the model in actual application and affect the understanding and trust of the user on the prediction results of the model, the present application provides a prediction method of a data center cooling system return water temperature. In the method, the initial operation data of the cooling system in the data center can be preprocessed to obtain final operation data that satisfies certain conditions, and the final operation data is input into a pre-trained time series prediction model to obtain a prediction result of the cooling system return water temperature. The method realizes accurate prediction of the cooling system return water temperature and the average temperature and humidity of the return air, intuitively expresses the variable relationships, performs detailed variable influence analysis, and provides effective system optimization strategies, thereby improving the overall performance of the cooling system and ensuring the safe and efficient operation of the server room. Thus, the problems of a large number of model parameters, black-boxing caused by expressing nonlinear characteristics through a complex activation function, and difficulty in intuitively expressing cross-variable relationships, which limit the effectiveness and interpretability of the model in actual application and affect the understanding and trust of the user on the prediction results of the model, are solved.

[0035] Specifically, Figure 1A flow chart of a data center cooling system return water temperature prediction method according to an embodiment of the present application is provided.

[0036] As shown in Figure 1 The data center cooling system return water temperature prediction method includes the following steps:

[0037] In step S101, initial operation data of the cooling system in the data center is collected.

[0038] It can be understood that the initial operation data of the embodiment of the present application can include, but is not limited to, inlet water temperature, return water temperature, IT (Information Technology) device power consumption, water supply pressure difference, return air average temperature, return air average humidity, etc., which can be set by a person skilled in the art according to actual conditions, and the present application does not make specific limitations.

[0039] In addition, it should be noted that the embodiment of the present application can collect the initial operation data of the data center cooling system by using a sampling sensor, or other methods, which are not limited by the present application.

[0040] As a possible implementation, the embodiment of the present application can collect the initial operation data of the cooling system.

[0041] For example, the embodiment of the present application can collect the initial operation data of the cooling center by using a sampling sensor, such as inlet water temperature, return water temperature, IT device power consumption, water supply pressure difference, return air average temperature, return air average humidity, etc., which are not limited by the present application.

[0042] In step S102, the initial operation data is preprocessed to obtain final operation data that meets a preset condition.

[0043] It should be noted that the preprocessing of the embodiment of the present application can include, but is not limited to, data cleaning, data standardization, and processing of missing data, etc., which can be set by a person skilled in the art according to actual conditions, and the present application does not make specific limitations.

[0044] As a possible implementation, the embodiment of the present application can perform data cleaning, data standardization, and processing of missing data on the initial operation data to ensure the consistency and accuracy of the data, and then obtain the final operation data that meets certain conditions. The certain conditions can be set by a person skilled in the art according to actual conditions, and the present application does not make specific limitations.

[0045] Optionally, in an embodiment of the present application, before the running data is input into the pre-trained time series prediction model, further comprising: constructing a Kronecker product layer of the time series prediction model by using the Kronecker product and the rotation mask, to obtain an input matrix of the final running data by using the Kronecker product layer; constructing a kernel matrix layer of the time series prediction model by combining the input matrix and the kernel function, to obtain an output matrix corresponding to the input matrix according to the kernel matrix layer; constructing a multi-layer perception machine layer of the time series prediction model by using the output matrix and the linear multi-layer perception machine, to construct the pre-trained time series prediction model. The mathematical expression of the multi-layer perception machine layer can be, but is not limited to:

[0046] y(k) = b + k1x y (1) (k) + … + k n x y (n) (k),

[0047] wherein y n (k) is an n-order output result of the model, b is a bias term of the multi-layer perception machine, k i is a weight corresponding to the order result.

[0048] In some embodiments, the Kronecker product layer of the embodiment of the present application can include, but is not limited to, a rotation mask. When two identical matrices are subjected to Kronecker product to produce repeated high-order terms, the N-order rotation mask has exactly the same form as the N-order input matrix X N , and can arrange X N after N-order Kronecker product to eliminate the repeated high-order terms.

[0049] It can be understood that the embodiment of the present application can use the Kronecker product layer to perform multi-order Kronecker product operation on the input matrix to form an N-order input matrix to express the high-order nonlinear relationship of the system, and use the rotation mask to perform masking operation on the N-order input matrix output by the Kronecker product layer to eliminate the repeated high-order terms to reduce the parameter quantity.

[0050] In some embodiments, the kernel matrix layer of the embodiment of the present application can include, but is not limited to, N kernel functions, which are applied to the input matrix of the corresponding order to obtain the output matrix of the corresponding order, i.e., the embodiment of the present application can obtain the output matrix of the corresponding order through the N-order input matrix.

[0051] In some embodiments, the multi-layer perception machine layer of the embodiment of the present application can splice the output matrix and output it through a linear multi-layer perception machine, and the prediction result obtained has mathematical representability, which can clearly know how each input variable affects the prediction result, and has transparency and credibility in actual application.

[0052] It can be understood that the multi-layer perceptron layer of the embodiment of the application discards the activation function of the multi-layer perceptron, splices the output matrix of the kernel matrix layer, and inputs the output matrix to the multi-layer perceptron layer for linear combination to obtain a prediction result, and has stronger interpretability. The mathematical expression of the multi-layer perceptron layer can be, but is not limited to, y(k) = b + k1x y (1) (k) + … + k n x y (n) (k).

[0053] Optionally, in an embodiment of the application, a Kronecker product layer of a time series prediction model is constructed by using a Kronecker product and a rotation mask to obtain an input matrix of final running data, including: obtaining an initial input matrix corresponding to the final running data based on the final running data; calculating a Kronecker product matrix of the final running data by using the initial input matrix and the Kronecker product; and obtaining the input matrix based on the Kronecker product matrix and the rotation mask.

[0054] As a possible implementation manner, the embodiment of the application can obtain an initial input matrix based on final running data, calculate a Kronecker product matrix of the final running data by using the initial input matrix and the Kronecker product, and then obtain an input matrix by using a rotation mask.

[0055] For example, the embodiment of the application obtains an initial input matrix based on final running data, performs a multi-order Kronecker product operation on the initial input matrix to form a Kronecker product matrix, to intuitively express the relationship between variables and facilitate user understanding and trust in the prediction result. When two same matrices perform a Kronecker product to generate repeated high-order terms, the rotation mask is used to arrange the high-order repeated terms to be ignored, and then an input matrix is obtained.

[0056] Optionally, in an embodiment of the application, a pre-trained time series prediction model is constructed, including: acquiring a training data set and a test data set according to final running data; and generating the pre-trained time series prediction model based on the training data set, the test data set, MSE, and MAE.

[0057] It can be understood that, after obtaining final running data, the embodiment of the application can divide the final running data into a training data set and a test data set, and perform normalization processing on the training data set and the test data set, so that the input data has a more stable gradient, to reduce the magnitude difference between features and improve the precision of the model.

[0058] For example, in this application embodiment, the final running data can be divided as follows: the first 80% of the final running data is selected as the training dataset and the last 20% is selected as the test dataset, so as to fully capture the dynamic characteristics of the cooling system and evaluate the predictive performance of the model on unseen datasets, providing a reliable basis for model training and validation.

[0059] Furthermore, embodiments of this application can also use MSE and MAE to evaluate the model's prediction accuracy, enabling interpretability analysis, thereby utilizing the variable relationships and influence analysis provided by the model to formulate specific optimization schemes.

[0060] In this embodiment of the application, the formula for calculating MSE can be, but is not limited to, expressed as:

[0061]

[0062] The formula for calculating MAE can be expressed, but is not limited to, as follows:

[0063]

[0064] Among them, y i For the predicted target, These are predicted values.

[0065] In step S103, the final running data is input into a pre-trained time series prediction model to obtain the prediction result of the cooling system return water temperature. The time series prediction model has a Kronecker layer, a kernel matrix layer, and a multilayer perceptron layer.

[0066] It is understood that the time series prediction model in the embodiments of this application may include Kronecker layers, kernel matrix layers, and multilayer perceptron layers, and may also include a dataset loading module, a model configuration module, a toolset module, a testing module, and a parser module, etc. The specific configuration can be made by those skilled in the art according to the actual situation, and this application does not impose any specific limitations.

[0067] In actual implementation, the embodiments of this application can input the final operating data into a pre-trained time series prediction model to obtain the predicted result of the cooling system return water temperature.

[0068] For example, such as Figure 2 As shown, the time series prediction model in this application embodiment includes a three-layer structure: Kronecker layer, kernel matrix layer, and multilayer perceptron layer. It aims to accurately predict the return water temperature and average temperature and humidity of the return air in the cooling system, intuitively express the relationship between variables, conduct detailed variable impact analysis, and provide effective system optimization strategies, thereby improving the overall performance of the cooling system and ensuring the safe and efficient operation of the server room.

[0069] Generally, as Figures 3-5 shown, the embodiments of the present application preprocess the initial operation data collected by the sampling sensor of the cooling system, such as the inlet water temperature, the return water temperature, the power consumption of the IT equipment, the water supply pressure difference, the return air average temperature, the return air average humidity, and the like, to obtain the final operation data satisfying certain conditions, and then input the final operation data into the time series prediction model trained in advance, so as to predict and output important parameters such as the east return water temperature, the west return water temperature, and the return air average temperature and humidity.

[0070] Specifically, after the final operation data of the embodiments of the present application is converted into an initial input matrix through an input layer, a N-order Kronecker product matrix is generated through a multi-order Kronecker product operation. For any Kronecker product matrix X, the sequence number mapping relationship before and after expansion is as follows: X(i,j)=X e (m), where m=(i-1)*N+j. The Kronecker product matrix X after N-order Kronecker product is denoted as X N , where the shape of X N may be but is not limited to represented as (length,…,length), and the dimension is N, and the following relationship is obtained: X N (i,j,…,n)=X e (i)×X e (j)×…×X e (n). Wherein, X N is the result obtained by Kronecker product of N X e matrices, and has a large number of repeated high-order terms.

[0071] Further, in order to solve the problem of repeated high-order terms, the embodiments of the present application create a N-order kernel matrix layer, which has exactly the same form as the N-order Kronecker product matrix X N , and each element in it can be obtained by the following formula, which can be but is not limited to represented as:

[0072]

[0073] In addition, the kernel matrix layer of the embodiments of the present application contains N kernel functions, and the kernel functions are applied to the input matrix of the corresponding order, so as to obtain the output matrix of the corresponding order of the N-order input matrix.

[0074] Further, the output matrix is spliced and further processed in the multi-layer perception layer in the embodiment of the application. Different from the traditional multi-layer perception layer, the multi-layer perception layer in the embodiment of the application does not contain a nonlinear activation function, that is, the output matrix obtained at the previous stage is directly input to the multi-layer perception layer for linear combination, and the linear superposition of the corresponding order is performed to obtain a prediction result. The mathematical expression of the multi-layer perception layer can be, but is not limited to, y(k) = b + k1x y (1) (k) + … + k n x y (n) (k), where the output result y (m) (k) of any order can be, but is not limited to, expressed as y (m) (k) = h m x X N , where in the embodiment of the application, x is redefined through the einsum operation, and the einsum can be described as a…, b…→ ab.

[0075] In addition, it should be noted that the embodiment of the application captures the nonlinear characteristics of the system and the correlation between the covariates through the kernel function, discards the traditional method of capturing nonlinear characteristics through the activation function, and proves to be effective through experiments.

[0076] The embodiment of the application can outperform the existing best model in multiple data sets, greatly reduce the computational requirements required for fitting, and has the following advantages:

[0077] (1) Accurate prediction of cooling system return water temperature: the pre-trained time series prediction model constructed in the embodiment of the application can process a large amount of data, which can include, but is not limited to, inlet water temperature, return water temperature, IT device power consumption, water supply pressure difference, return air average temperature, return air average humidity, etc., and can capture the dynamic characteristics of the cooling system. By effectively utilizing these data, the precision of the model prediction effect can be improved, and reliable prediction results can be provided under various operating conditions.

[0078] (2) Model interpretability: the pre-trained time series prediction model constructed in the embodiment of the application not only can accurately predict the return water temperature of the cooling system, but also can intuitively express the relationship between variables. The model parameters can not only reflect the specific contribution and mutual influence of each variable on the prediction result, but also help users understand the internal mechanism of the model, and can perform detailed impact analysis on the variables, quantify the influence degree of different variables on the return water temperature of the cooling system, and through feature importance analysis and partial dependence plot technology, the roles and functions of each variable in prediction are clear.

[0079] (3) The optimization strategy of the cooling system is proposed and verified: the embodiment of the application can use the variable relationship and influence analysis provided by the constructed pre-trained time series prediction model to develop specific optimization schemes, such as adjusting the supply and return water temperature and pressure, and verify the effectiveness of these optimization strategies through simulation and actual application, improve the efficiency and stability of the cooling system, and ensure that the equipment in the server room operates stably at the best temperature.

[0080] The prediction method of the return water temperature of the data center cooling system proposed by the embodiment of the application will be described in detail in a specific embodiment in combination with Figures 2-5 FIG. 1.

[0081] As shown in Figure 2 FIG. 2, the time series prediction model of the embodiment of the application includes three layers: a Kronecker product layer, a kernel matrix layer, and a multilayer perceptron layer.

[0082] Further, in the embodiment of the application, as shown in Figures 3-5 , the initial operation data of the cooling system collected by the sampling sensor is preprocessed to obtain final operation data that meets certain conditions, and then the final operation data is input into the pre-trained time series prediction model, so as to predict and output important parameters such as the east return water temperature, the west return water temperature, and the average temperature and humidity of the return air.

[0083] Specifically, after the final operation data of the embodiment of the application is converted into an initial input matrix through the input layer, the N-order Kronecker product matrix is generated through the multi-order Kronecker product operation. The Kronecker product matrix X N is the result obtained by Kronecker product of N X e matrices, and has a large number of repeated high-order terms.

[0084] Further, in order to solve the problem of repeated high-order terms, the embodiment of the application creates an N-order kernel matrix layer, applies a kernel function to the input matrix of the corresponding order to obtain the output matrix of the corresponding order.

[0085] Further, the embodiment of the application further processes the output matrix in the multilayer perceptron layer after splicing, and obtains the prediction result through linear superposition of the corresponding order.

[0086] The method for predicting the return water temperature of a data center cooling system proposed in this application preprocesses the initial operating data of the cooling system collected in the data center to obtain final operating data that meets certain conditions. This final operating data is then input into a pre-trained time series prediction model to obtain the predicted return water temperature. This method accurately predicts the return water temperature and the average temperature and humidity of the return air, intuitively expresses the variable relationships, performs detailed variable impact analysis, and provides effective system optimization strategies. This improves the overall performance of the cooling system and ensures the safe and efficient operation of the server room. Therefore, it solves the problems in related technologies, such as the large number of model parameters, the black box effect caused by expressing nonlinear characteristics through complex activation functions, and the difficulty in intuitively expressing cross-variable relationships. These limitations restrict the effectiveness and interpretability of the model in practical applications, affecting users' understanding and trust in the model's prediction results.

[0087] Next, referring to the accompanying drawings, a device for predicting the return water temperature of a data center cooling system according to an embodiment of this application is described.

[0088] Figure 6 This is a block diagram of a device for predicting the return water temperature of a data center cooling system according to an embodiment of this application.

[0089] like Figure 6 As shown, the data center cooling system return water temperature prediction device 10 includes: a data acquisition module 100, a preprocessing module 200, and a prediction module 300.

[0090] The acquisition module 100 is used to acquire the initial operating data of the cooling system in the data center.

[0091] The preprocessing module 200 is used to preprocess the initial running data to obtain the final running data that meets the preset conditions.

[0092] The prediction module 300 is used to input the final running data into a pre-trained time series prediction model to obtain the prediction result of the cooling system return water temperature. The time series prediction model has a Kronecker layer, a kernel matrix layer, and a multilayer perceptron layer.

[0093] Optionally, in one embodiment of this application, it further includes: a first building module, a second building module, and a third building module.

[0094] The first building module is used to construct the Kronecker layer of the time series prediction model using Kronecker layer and rotating mask before inputting the running data into the pre-trained time series prediction model, so as to obtain the input matrix of the final running data using the Kronecker layer.

[0095] The second construction module is configured to construct a kernel matrix layer of the time series prediction model by combining the input matrix and the kernel function, so as to obtain an output matrix corresponding to the input matrix according to the kernel matrix layer.

[0096] The third construction module is configured to construct a multi-layer perception machine layer of the time series prediction model by using the output matrix and the linear multi-layer perception machine, so as to construct the pre-trained time series prediction model.

[0097] Optionally, in an embodiment of the present application, the third construction module comprises an acquisition unit and a first generation unit.

[0098] The acquisition unit is configured to acquire the training data set and the test data set according to the final running data.

[0099] The first generation unit is configured to generate the pre-trained time series prediction model based on the training data set, the test data set, the MSE and the MAE.

[0100] Optionally, in an embodiment of the present application, the first construction module comprises a second generation unit, a calculation unit and a third generation unit.

[0101] The second generation unit is configured to obtain an initial input matrix corresponding to the final running data based on the final running data.

[0102] The calculation unit is configured to calculate a Kronecker product matrix of the final running data by using the initial input matrix and the Kronecker product.

[0103] The third generation unit is configured to obtain the input matrix based on the Kronecker product matrix and the rotation mask.

[0104] Optionally, in an embodiment of the present application, the mathematical expression of the multi-layer perception machine layer can be, but is not limited to, as follows:

[0105] y(k) = b + k1x y (1) (k) + … + k n x y (n) (k),

[0106] wherein y n (k) is an n-order output result of the model, b is a bias term of the multi-layer perception machine, and k i is a weight corresponding to the order result.

[0107] It should be noted that the above description of the embodiment of the method for predicting the return water temperature of the data center cooling system is also applicable to the embodiment of the device for predicting the return water temperature of the data center cooling system, which will not be described here again.

[0108] The data center cooling system return water temperature prediction device provided by the embodiment of the present application can preprocess the collected initial operation data of the cooling system in the data center, and then obtain final operation data meeting certain conditions, and input the final operation data into a pre-trained time sequence prediction model to obtain a prediction result of the cooling system return water temperature, thereby realizing accurate prediction of the cooling system return water temperature and average temperature and humidity of return air, intuitively expressing variable relationships, performing detailed variable influence analysis, and providing effective system optimization strategies, so as to improve the overall performance of the cooling system and ensure the safe and efficient operation of the server room. Thus, the problems in the related art that the model parameter quantity is large, the nonlinear characteristics are expressed through a complex activation function, leading to black-boxing, and the cross-variable relationship is difficult to intuitively express, thereby limiting the effectiveness and interpretability of the model in actual application, and affecting the understanding and trust of the user for the model prediction result are solved.

[0109] Figure 7 A structural schematic diagram of an electronic device according to an embodiment of the present application is provided. The electronic device can include:

[0110] The memory 701, the processor 702, and a computer program stored in the memory 701 and executable on the processor 702.

[0111] The processor 702 implements the data center cooling system return water temperature prediction method provided in the above embodiments when executing the program.

[0112] Further, the electronic device further includes:

[0113] The communication interface 703 is configured to communicate between the memory 701 and the processor 702.

[0114] The memory 701 is configured to store a computer program executable on the processor 702.

[0115] The memory 701 can include a high-speed RAM memory, and can also include a non-volatile memory, for example, at least one disk memory.

[0116] If the memory 701, the processor 702 and the communication interface 703 are implemented independently, the communication interface 703, the memory 701 and the processor 702 can be connected with each other through a bus and complete communication between each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For convenience of representation, Figure 7 Only one thick line is used to represent the bus in the figure, but it does not mean that there is only one bus or only one type of bus.

[0117] Optionally, in a specific implementation, if the memory 701, the processor 702 and the communication interface 703 are integrated on a chip, the memory 701, the processor 702 and the communication interface 703 can complete communication between each other through an internal interface.

[0118] The processor 702 can be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0119] The embodiments of the present application also provide a computer readable storage medium, having stored thereon a computer program, which, when executed by a processor, implements the data center cooling system return water temperature prediction method as above.

[0120] The embodiments of the present application also provide a computer program product, comprising a computer program, which, when executed by a processor, implements the data center cooling system return water temperature prediction method as above.

[0121] In the description of the application, reference to "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" means that a particular feature, structure, material, or characteristic being described is included in at least one embodiment or example of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment or example. Furthermore, the described specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples. In addition, the usage of "N" means at least two, for example, two, three or the like, unless explicitly stated otherwise.

[0122] Furthermore, the terms "first", "second", or the like, are used merely as a designation of certain elements or features, and do not imply or connote relative importance or a specific order of categorization thereof. Accordingly, features described as "first" or "second" can be explicitly or implicitly included in at least one of the features. In the description of the application, the meaning of "N" is at least two, for example, two, three, etc., unless explicitly specified otherwise.

[0123] Any process or method descriptions or blocks in flow charts or otherwise described herein represent embodiments which can be managed as one or more modules, segments, or portions of code which include one or more executable instructions for implementing specific logic functions or steps, and alternate implementations are possible. In some embodiments, the processes and methods described can be executed by one or more apparatuses or devices, either directly or after conversion to another language. Alternate implementations are possible.

[0124] The logic and / or steps represented in the flowcharts and / or described herein, for example, can be considered as a sequence of executable instructions stored in a computer readable medium, which can be executed by an instruction execution system, apparatus or device, such as a computer-based system, a processor-based system, or other system that can fetch the instructions from the instruction execution system, apparatus or device and execute the instructions, or a combination of the above. For the purposes of this specification, a "computer readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus or device. The computer readable medium can be a computer readable storage medium or a computer readable signal medium. The computer readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a propagation medium. The computer readable signal medium can include, but is not limited to, a computer readable medium that facilitates transfer of the program from one place to another. A specific example of a computer readable medium is a non-transitory computer-readable storage medium. A specific example of a computer readable signal medium is a source or destination of the computer readable medium. Another specific example of a computer readable signal medium is a computer readable signal travelling through space. Thus, a computer readable medium can take many forms of hardware to carry out the program for use by or in connection with the instruction execution system, apparatus or device.

[0125] It should be understood that aspects of the application can be implemented in hardware, software, firmware or a combination thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware and in another embodiment, the hardware can be implemented using any or a combination of the following technologies, which are each well known in the art: a discrete logic circuit(s) having logic gates for implementing logic functions upon an application of data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array(s) (PGA), a field programmable gate array (FPGA), etc.

[0126] Those of skill in the art would understand that the steps of the methods carried out above can be carried out wholly or partly by a program instructing relevant hardware, and the program can be stored in a computer readable storage medium, and when executed, includes one or a combination of the steps of the method embodiments.

[0127] In addition, each of the functional units in the various embodiments of the present application can be integrated in one processing module, or each of the units can be physically present separately, or two or more units can be integrated in one module. The integrated module can be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0128] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A method for predicting return water temperature of a data center cooling system, the method comprising: The method comprises the following steps: Collecting initial operation data of a cooling system in a data center; Preprocessing the initial operation data to obtain final operation data meeting preset conditions; Inputting the final operation data into a pre-trained time series prediction model to obtain a prediction result of return water temperature of the cooling system, wherein the time series prediction model has a Kronecker product layer, a kernel matrix layer and a multilayer perceptron layer; Before the operation data is input into the pre-trained time series prediction model, the method further comprises the following steps: Using an N-order Kronecker product and a rotation mask to construct the Kronecker product layer of the time series prediction model to obtain an N-order input matrix of the final operation data by using the Kronecker product layer; Combining the input matrix and N kernel functions to construct the kernel matrix layer of the time series prediction model to obtain an output matrix corresponding to the order of the input matrix according to the kernel matrix layer; Using the output matrix and a linear multilayer perceptron to construct the multilayer perceptron layer of the time series prediction model to construct the pre-trained time series prediction model, wherein the obtained output matrix is directly input into the multilayer perceptron layer for linear combination and linear superposition of the corresponding order to obtain a prediction result.

2. The method of claim 1, wherein, The method of constructing the pre-trained time series prediction model comprises the following steps: Obtaining a training data set and a test data set according to the final operation data; Generating the pre-trained time series prediction model based on the training data set, the test data set, mean square error (MSE) and mean absolute error (MAE).

3. The method of claim 1, wherein, Using a Kronecker product and a rotation mask to construct the Kronecker product layer of the time series prediction model to obtain an input matrix of the final operation data by using the Kronecker product layer comprises the following steps: Obtaining an initial input matrix corresponding to the final operation data based on the final operation data; Calculating a Kronecker product matrix of the final operation data by using the initial input matrix and the Kronecker product; Obtaining the input matrix based on the Kronecker product matrix and the rotation mask.

4. The method of claim 1, wherein, The mathematical expression of the multilayer perceptron layer is: ， wherein, is the model's order output result, is the bias term of the multi-layer perceptron, is the weight corresponding to the order result.

5. A device for predicting the return water temperature of a data center cooling system, characterized in that, The method comprises the following steps: A collection module is configured to collect initial operation data of a cooling system in a data center; A preprocessing module is configured to preprocess the initial operation data to obtain final operation data meeting preset conditions; A prediction module is configured to input the final operation data into a pre-trained time series prediction model to obtain a prediction result of return water temperature of the cooling system, wherein the time series prediction model has a Kronecker product layer, a kernel matrix layer and a multilayer perceptron layer; The method further comprises the following steps: A first construction module is configured to, before the operation data is input into the pre-trained time series prediction model, use an N-order Kronecker product and a rotation mask to construct the Kronecker product layer of the time series prediction model to obtain an N-order input matrix of the final operation data by using the Kronecker product layer. a second constructing module, configured to construct a kernel matrix layer of the time series prediction model in combination with the input matrix and N kernel functions, so as to obtain an output matrix corresponding to an order of the input matrix according to the kernel matrix layer; a third constructing module, configured to construct a multi-layer perception machine layer of the time series prediction model by using the output matrix and the linear multi-layer perception machine, so as to construct the pre-trained time series prediction model, wherein the obtained output matrix is directly input to the multi-layer perception machine layer for linear combination, and linear superposition is performed on the corresponding order to obtain a prediction result.

6. An electronic device, comprising: comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor executing the program to implement the method of predicting the return water temperature of the data center cooling system according to any one of claims 1-4.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that The program is executed by the processor to implement the method of predicting the return water temperature of the data center cooling system according to any one of claims 1-4.

8. A computer program product, characterised in that, comprising a computer program which, when executed, is configured to implement the method of predicting the return water temperature of the data center cooling system according to any one of claims 1-4.

Images