Data center green operation evaluation method and device, medium and equipment

Abstract

The invention discloses a data center green operation evaluation method and device, a medium and equipment. The method comprises the following steps: acquiring data center green operation feature data of a data center; the data center green operation feature data is input into a pre-trained data center green operation evaluation model, an evaluation score of the data center is output, and the construction process of the data center green operation evaluation model is as follows: historical data center features are acquired, and a data center green operation index system is constructed; adopting an analytic hierarchy process and an entropy weight method to construct a subjective and objective combined weighted learning sample according to the index system of each data center; and performing pre-training and fine tuning on the deep belief network according to the learning sample, and generating a data center green operation evaluation model.

Description

Technical Field

[0001] The present invention relates to the technical field of data center green operation evaluation, and more specifically, to a data center green operation evaluation method, device, medium and equipment. Background Art

[0002] In order to promote the sustainable development of data centers, it is crucial to develop a scientific and reasonable evaluation system. However, the current evaluation system still has many shortcomings, making it difficult to fully and accurately reflect the objective situation of the data center's operating status, and limiting the scientific planning of the data center's green development path. Therefore, optimizing and improving the evaluation system of data centers is an urgent issue to be solved.

[0003] Most of the existing evaluation systems focus on a single or a few indicators, lacking a comprehensive consideration of the comprehensive performance of data centers. Traditional systems often focus on the evaluation of hard indicators such as energy consumption and computing power of data centers, including power utilization efficiency (PUE) and IT equipment load utilization rate. The performance of data centers in terms of environment, operation, technology and innovation is ignored. This one-sided evaluation method makes it impossible for data centers to take into account the issue of sustainable transformation, resulting in managers being unable to fully consider the long-term development needs of data centers, and then ignoring key factors such as environmental protection and technological innovation, affecting the overall competitiveness and sustainable development of data centers. In addition, the determination of indicator weights remains a major problem in the design of evaluation systems. In traditional methods, the determination of indicator weights often relies on expert experience or simple mathematical and statistical methods, lacks systematicity and objectivity, and is easily interfered by subjective factors, which in turn affects the fairness and scientificity of the evaluation results. In the evaluation of data centers as a multi-dimensional complex system, how to objectively and reasonably determine the weights of each indicator and measure the important criteria for the effectiveness of the evaluation system has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a data center green operation evaluation method, device, medium and equipment.

[0005] According to one aspect of the present invention, a data center green operation evaluation method is provided, comprising:

[0006] Obtain data center green operation characteristic data of the data center;

[0007] The data center green operation feature data is input into the pre-trained data center green operation evaluation model, and the evaluation score of the data center is output, where

[0008] The construction process of the data center green operation evaluation model is as follows:

[0009] Obtain historical data center characteristics and build an indicator system for green operation of data centers;

[0010] The analytic hierarchy process and entropy weight method are used to construct learning samples that combine subjective and objective factors and empower them according to the indicator system of each data center.

[0011] The deep belief network is pre-trained and fine-tuned based on the learning samples to generate a data center green operation evaluation model.

[0012] Optionally, the analytic hierarchy process and entropy weight method are used to construct a learning sample that combines subjective and objective weights according to the indicator system of each data center, including:

[0013] The expert opinions of the hierarchical analysis method are used to weight the original evaluation matrix and obtain the first weight of each evaluation indicator in the indicator system;

[0014] Construct an original evaluation matrix based on multiple data centers and the indicator system of each data center;

[0015] The entropy weight method is used to weight the original evaluation matrix to obtain the second weight of each evaluation indicator in the indicator system;

[0016] A linear method is used to linearly combine the first weight and the second weight to obtain the third weight assigned to each evaluation indicator combination in the indicator system;

[0017] The third weight is used to score each evaluation indicator of the indicator system of each data center to obtain learning samples.

[0018] Optionally, the expert opinions of the hierarchical analysis method are used to weight the original evaluation matrix to obtain the first weight of each evaluation indicator in the indicator system, including:

[0019] Construct a judgment matrix of evaluation indicators of the data center's indicator system based on the expert opinions of the hierarchical analysis method;

[0020] The square root method is used to calculate the initial first weight of each evaluation index according to the judgment matrix;

[0021] Normalizing the initial first weight to obtain a normalized initial first weight;

[0022] Calculate the maximum eigenvalue according to the normalized initial first weight, and calculate the consistency ratio according to the maximum eigenvalue;

[0023] When the consistency ratio is less than a preset threshold, the normalized initial weight in this case is determined as the first weight.

[0024] Optionally, the entropy weight method is used to weight the original evaluation matrix to obtain the second weight of each evaluation indicator in the indicator system, including:

[0025] Normalizing the original evaluation matrix to obtain a normalized original evaluation matrix;

[0026] The variation size, entropy value and information entropy redundancy of each evaluation index are calculated according to the normalized original evaluation matrix;

[0027] According to the variation size, entropy value and information entropy redundancy of each evaluation indicator, the second weight of each evaluation indicator is calculated.

[0028] Optionally, the deep belief network is pre-trained and fine-tuned according to the learning samples to generate a data center green operation evaluation model, including:

[0029] Input the learning samples into the deep belief network to train the restricted Boltzmann machine layer by layer by maximizing the log-likelihood function to generate the initial data center green operation evaluation model;

[0030] The initial data center green operation evaluation model is fine-tuned through the back-propagation algorithm to generate a data center green operation evaluation model.

[0031] According to another aspect of the present invention, there is provided a data center green operation evaluation device, comprising:

[0032] An acquisition module, used to acquire data center green operation characteristic data of the data center;

[0033] The output module is used to input the data center green operation feature data into the pre-trained data center green operation evaluation model and output the evaluation score of the data center, where

[0034] The construction process of the data center green operation evaluation model in the output module is as follows:

[0035] The acquisition submodule is used to obtain historical data center characteristics and build an indicator system for green operation of data centers;

[0036] Construct submodules to construct learning samples that combine subjective and objective factors and empower them according to the indicator system of each data center by using the hierarchical analysis method and entropy weight method;

[0037] The generation submodule is used to pre-train and fine-tune the deep belief network based on the learning samples to generate a data center green operation evaluation model.

[0038] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the storage medium stores a computer program, and the computer program is used to execute the method described in any one of the above aspects of the present invention.

[0039] According to another aspect of the present invention, an electronic device is provided, comprising: a processor; a memory for storing instructions executable by the processor; the processor is configured to read the executable instructions from the memory and execute the instructions to implement the method described in any one of the above aspects of the present invention.

[0040] The present invention uses the analytic hierarchy process and entropy weight method to comprehensively weight the green operation evaluation indicators of the data center to construct a learning sample, and on this basis, a deep belief network is used to achieve a scientific evaluation of the green operation of the data center, as detailed as follows:

[0041] 1. Enhanced the objectivity and accuracy of the evaluation results: The weighting method combining the analytic hierarchy process and the entropy weight method can effectively integrate subjective experience and objective data, ensuring that the weight distribution of the evaluation indicators is more scientific and reasonable. The analytic hierarchy process quantifies subjective experience into weights through expert scoring, while the entropy weight method objectively calculates weights through the discreteness of indicator data. The combination of the two takes into account the professional knowledge of experts in weight distribution and makes full use of the information of the data itself, enhancing the objectivity and accuracy of the evaluation results.

[0042] 2. Capturing complex features and improving the accuracy and reliability of evaluation results: As an efficient deep learning model, the Deep Belief Network (DBN) can accurately model complex systems in a big data environment. Through the pre-training and fine-tuning process, DBN can extract high-order features of data layer by layer, capture subtle changes in the operation of the data center, and improve the accuracy and reliability of the evaluation. In the pre-training stage, unsupervised learning is used to effectively initialize the network weights, avoiding the common gradient vanishing problem of deep networks, while the fine-tuning stage further optimizes the model through supervised learning to improve classification performance. This fully reflects the green operation of the data center. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0044] Figure 1 It is a flow chart of a data center green operation evaluation method provided by an exemplary embodiment of the present invention;

[0045] Figure 2 is another flow chart of a data center green operation evaluation method provided by an exemplary embodiment of the present invention;

[0046] Figure 3 is a structural schematic diagram of a data center green operation evaluation device provided by an exemplary embodiment of the present invention;

[0047] Figure 4 This is a structure of an electronic device provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0048] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0049] It should be noted that the relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0050] Those skilled in the art can understand that the terms "first" and "second" in the embodiments of the present invention are only used to distinguish different steps, devices or modules, etc., and neither represent any specific technical meaning nor indicate the necessary logical order between them.

[0051] It should also be understood that, in the embodiments of the present invention, “plurality” may refer to two or more than two, and “at least one” may refer to one, two or more than two.

[0052] It should also be understood that any component, data or structure mentioned in the embodiments of the present invention can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.

[0053] In addition, the term "and / or" in the present invention is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects before and after are in an "or" relationship.

[0054] It should also be understood that the description of the various embodiments of the present invention focuses on the differences between the various embodiments, and the same or similar aspects thereof can be referenced to each other, and for the sake of brevity, they will not be described one by one.

[0055] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0056] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0057] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0058] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0059] Embodiments of the present invention can be applied to electronic devices such as terminal devices, computer systems, servers, etc., which can operate with many other general or special computing system environments or configurations. Examples of well-known terminal devices, computing systems, environments and / or configurations suitable for use with electronic devices such as terminal devices, computer systems, servers, etc. include, but are not limited to: personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, small computer systems, large computer systems, and distributed cloud computing technology environments including any of the above systems, etc.

[0060] Electronic devices such as terminal devices, computer systems, servers, etc. can be described in the general context of computer system executable instructions (such as program modules) executed by computer systems. Generally, program modules can include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media including storage devices.

[0061] Exemplary Methods

[0062] Figure 1 1 is a flow chart of a data center green operation evaluation method provided by an exemplary embodiment of the present invention. This embodiment can be applied to electronic devices, such as Figure 1 As shown, the data center green operation evaluation method 100 includes the following steps:

[0063] Step 101, obtaining data center green operation characteristic data of the data center;

[0064] Step 102: input the data center green operation feature data into a pre-trained data center green operation evaluation model, and output the evaluation score of the data center, where

[0065] The construction process of the data center green operation evaluation model is as follows:

[0066] Obtain historical data center characteristics and build an indicator system for green operation of data centers;

[0067] The analytic hierarchy process and entropy weight method are used to construct learning samples that combine subjective and objective factors and empower them according to the indicator system of each data center.

[0068] The deep belief network is pre-trained and fine-tuned based on the learning samples to generate a data center green operation evaluation model.

[0069] Specifically, in response to the problems existing in the background technology, the present invention proposes a more complete index system, combines the fusion method of subjective weight and objective weight, and uses the deep belief network (DBN) to optimize and evaluate the weight. This index system not only covers the traditional hard indicators of data centers such as energy efficiency and computing power, but also introduces multi-dimensional soft indicators such as environmental green management, operational green management, improvement and innovation, thus forming a comprehensive and multi-level evaluation framework. In the design process, we pay attention to the correlation between various indicators to avoid the problem of repeated measurement or neglect of important factors in the indicators. The subjective weight part of the indicator is mainly determined by the hierarchical analysis method of the expert scoring method, reflecting the expert's experience and industry knowledge of the importance of the indicator. On this basis, in order to avoid the deviation of the evaluation results that may be caused by relying solely on subjective weights, the entropy weight method, an objective weight determination method based on data, is introduced. The importance weight of each indicator is extracted through the uncertainty of entropy. The subjective and objective fusion method can effectively integrate expert experience and data-driven objective analysis, thereby improving the accuracy and fairness of the evaluation system. In addition, we further introduced the Deep Belief Network (DBN), which uses deep learning to continuously iterate and learn a large amount of data, gradually approaching the optimal weight distribution. This process not only reduces the impact of human factors on weight distribution, but also dynamically adjusts weight distribution according to changes in the data center operating environment, thereby ensuring the flexibility, adaptability and continued effectiveness of the evaluation system.

[0070] The process of AHP is as follows:

[0071] (1) Establish a hierarchical model: divide the decision-making goals, factors to be considered (decision-making criteria) and decision-making objects into the highest level, middle level and lowest level according to their mutual relationships, and draw a hierarchical diagram. The highest level represents the purpose of solving the problem, that is, the goal to be achieved by the hierarchical analysis method. The middle level represents the factors to be considered and the decision-making criteria. The lowest level represents the measures or plans to solve the problem.

[0072] (2) Constructing the judgment matrix: After constructing the indicator hierarchy, it is necessary to construct a judgment matrix for the indicator artificially based on the evaluation scale table. The form of the judgment matrix is ​​generally constructed as follows:

[0073]

[0074] According to the consistent matrix method proposed by Saaty et al., the factors in the middle layer are compared with each other and rated according to their importance. The basic classification includes five items, namely, equal strong, weak strong, strong, very strong, and absolute, and the nominal scales are assigned with measurement values ​​of 1, 3, 5, 7, and 9. Four other scales are set between the five basic scales and are assigned with measurement values ​​of 2, 4, 6, and 8, for a total of nine scales. The meaning of each scale is shown in Table 1.

[0075] Table 1 Ratio scale table

[0076]

[0077] (3) Hierarchical single sorting and consistency test: It is necessary to test the compatibility of the judgment matrix A, that is, to check whether there is inconsistency when comparing different indicators one by one. First, based on the judgment matrix, the square root method is used to calculate W i , W i The elements of are the ranking weights of the relative importance of the factors at the same level to the factors at the previous level. This process is called hierarchical single ranking. Then according to W i The formula for calculating the maximum characteristic root λmax is as follows:

[0078]

[0079] The consistency index CI is calculated by λmax:

[0080]

[0081] According to the CI value, CR is obtained:

[0082]

[0083] Table 2 Average random consistency index RI standard value

[0084]

[0085] Calculate CR according to the formula. If CR < 0.1, the comparison matrix has passed the consistency test. Otherwise, the judgment matrix needs to be readjusted until CR < 0.1 is satisfied.

[0086] In summary, the analytic hierarchy process is a multi-criteria decision analysis method widely used in complex decision-making problems. Its main purpose is to help decision makers decompose problems and conduct multi-level quantitative and qualitative analysis through systematic and structured methods when facing complex and multi-level decision-making problems. AHP can decompose complex problems into multiple levels, so that the factors at each level can be evaluated independently, and the relative importance of each factor can be clarified through paired comparison methods, thereby improving the scientificity and objectivity of decision-making. It not only improves decision-making efficiency, but also enhances the transparency and traceability of decisions, providing strong support for optimizing resource allocation and achieving strategic goals.

[0087] Since the subjective weighting method mainly controls the weights manually and is greatly affected by subjective factors, the objective weighting method is used to adjust the weights to make the weights more scientific. The entropy weight method is a more scientific method among the objective weighting methods. The uncertainty of entropy is used to characterize the amount of information. The larger the entropy value, the more unstable the information is, and the more unstable the impact on the results is. It is believed that the greater the impact, the greater the corresponding weight. On the contrary, when the entropy value is small, it means that the amount of information is small, the degree of uncertainty is smaller, the impact on the stability of the evaluation results is smaller, and the corresponding weight is smaller. [2] .

[0088] The basic steps of the entropy weight method are:

[0089] (1) Since the entropy weight method relies on data to process indicators, the original data must first be normalized. There are many methods for data normalization, the most commonly used are min-max normalization and Z-score normalization. Assume that the original data set is defined as X = (x1, x2, ... x n ), the processing result is Y=(y1,y2,…x n ).

[0090] Min-max normalization is also called discrete normalization in data processing. The transformation relationship is shown in the formula.

[0091]

[0092] Among them, 1<=j<=n. max{x j} is X=(x1,x2,…x n ), min{x j} is X=(x1,x2,…x n ). And The advantage of this method is that the information in the original data is not changed, that is, the relationship between the data is preserved, which achieves the effect of unifying the dimensions without losing much information. The disadvantage is that the maximum and minimum values ​​are not fixed. If new data is added, the maximum and minimum values ​​need to be updated. If a value in the sample set is too large, the normalized value is close to 0, and the difference between the sample values ​​is small.

[0093] Z-score normalization is also one of the classic data processing methods, also known as standard deviation normalization. The results of data normalization are not all mapped to the interval [0, 1], which is different from the min-max normalization method. The normalization formula is shown below.

[0094]

[0095] in, is the sample mean, and s is the sample standard deviation.

[0096] (2) Secondly, in order to avoid the impact of dimension, the indicators should be standardized. According to the meaning of the indicators, the indicators can be divided into positive indicators (the larger the value, the better) and negative indicators (the smaller the value, the better). Assume that n data centers {X1, X2, …, X n}, there are m indicators, among which X i ={x1,x2,…,x m}, and were standardized by the following methods respectively.

[0097] First, you need to select the minimum and maximum values ​​of each indicator, subtract 0.0001 from the minimum value, and add 0.0001 to the maximum value, in order to be compatible with the case where an entire column has the same value.

[0098] X min =min(X 1j ,X 2j ,…,X nj )-0.0001

[0099] X max =max(X 1j ,X 2j ,…,X nj )+0.0001

[0100] For positive indicators,

[0101]

[0102] For negative indicators,

[0103]

[0104] (3) Then it is necessary to calculate the proportion of the j-th indicator in the i-th data center, that is, to calculate the variation size of the indicator.

[0105]

[0106] (4) Calculate the entropy value of the j-th indicator.

[0107]

[0108] where e j ≥0. If p ij = 0, define e j =0.

[0109] (5) Calculate the information entropy redundancy (difference).

[0110] d j =1-e j ,j=1,…,m

[0111] (6) Calculate the weight of each indicator.

[0112]

[0113] (7) Calculate the comprehensive score of each data center.

[0114]

[0115] The entropy weight method determines the weight of the indicator according to the degree of variation of the indicator values ​​of each indicator. This is an objective weighting method that avoids the deviation caused by human factors. Compared with those subjective value assignment methods, it has higher accuracy and stronger objectivity, and can better explain the results. However, the entropy weight method itself has certain defects. It ignores the importance of the indicator itself. Sometimes the determined indicator weights are far from the expected results. At the same time, the entropy weight method cannot reduce the dimension of the evaluation indicator. That is, the entropy weight method conforms to mathematical laws and has strict mathematical significance, but it often ignores the subjective intentions of decision makers. And if the change in the indicator value is very small or suddenly increases or decreases, the entropy weight method has limitations. It should be noted that the number of indicators in the entropy weight method is generally better than the number of objects. It can be used to determine the weight of indicators in any evaluation problem, and can be used to eliminate indicators in the indicator system that do not contribute much to the evaluation results.

[0116] Deep Belief Networks (DBNs) [3] Proposed by Geoffrey Hinton in 2006, it is a classic deep generative model whose network structure includes:

[0117] (1) Restricted Boltzmann Machines (RBM) layer

[0118] Each layer of RBM consists of a visible layer and a hidden layer. The visible layer corresponds to the input data, and the hidden layer extracts features. RBM is an undirected graph model. The connection weights between nodes are symmetric, and there is no connection between nodes in the same layer. Through RBM training, the hidden layer gradually learns and extracts high-level features of the input data.

[0119] Weight matrix W = (w ij ) specifies the visible layer unit v i and hidden unit h j The weight of the edge between them, in addition, for each visible layer unit v i With bias a i , for each hidden unit h j With bias b j , the energy function of RBM is defined as:

[0120] E(v,h)=-∑ i a i v i -∑ j b j h j -∑ i ∑ j v i w ij h j

[0121] The joint probability distribution between the hidden and visible layers is given by the energy function:

[0122]

[0123] Among them, Z is the partition function, which is defined as e under all possible values ​​of the node -E(v,h) The sum of (a normalizing constant that makes the probability distribution sum to 1). Similarly, the marginal distribution of visible layer values ​​can be obtained by summing over all hidden layer configurations:

[0124]

[0125] For m visible layer nodes and n hidden layer nodes, the conditional probability of the visible layer configuration v for the hidden layer configuration h is as follows:

[0126]

[0127] Similarly, the conditional probability of h with respect to v is

[0128]

[0129] Among them, the activation probability of a single node is

[0130]

[0131] Among them, σ represents the Sigmoid function.

[0132] (2) Stacked RBM

[0133] DBN is constructed by stacking multiple RBMs. The hidden layer output of the first RBM is used as the input of the second RBM, and so on. The layer-by-layer greedy training strategy enables each layer to capture different levels of features of the input data.

[0134] (3) Fully connected layer

[0135] After stacking RBMs, DBNs usually add one or more fully connected layers for supervised learning. The fully connected layers further fine-tune the network parameters through the backpropagation algorithm, making the model perform better in classification or regression tasks.

[0136] The training process of DBN can be divided into two stages: pre-training and fine-tuning.

[0137] (1) Pre-training (unsupervised learning):

[0138] Each RBM is trained layer by layer, and the feature representation of the data is learned by maximizing the probability distribution of the input data on the RBM. During the pre-training process, the Contrastive Divergence algorithm is used to update the parameters.

[0139] (2) Fine-tuning (supervised learning):

[0140] Based on the pre-training, the entire network is fine-tuned using the labeled data through the back-propagation algorithm. The fine-tuning stage adjusts the parameters of all layers to improve the accuracy of the model on a specific task.

[0141] The specific contents of a data center green operation evaluation method proposed in the present invention are as follows: Figure 2 As shown:

[0142] 1. Construction of green operation indicator system for data centers

[0143] The green operation evaluation index of the data center is the basis of comprehensive evaluation, and needs to be comprehensively considered from multiple angles and levels to ensure that the selected index can scientifically and reasonably reflect the green operation status of the data center. Therefore, the present invention follows the principles of scientificity and operability when selecting the index, as shown in Table 3.

[0144] Table 3 Principles for selecting evaluation indicators

[0145]

[0146]

[0147] According to national standards, local standards, group standards, industry standards and expert opinions, as well as the above-mentioned indicator selection principles, this paper screened 33 data center green operation evaluation indicators. The data center green evaluation indicator system consists of 10 aspects and 33 indicators, including efficient use of energy, efficient use of water resources, efficient use of computing resources, application of energy-saving technologies, green design and green procurement, green management of equipment, green management of the environment, green management of operations, scientific layout and intensive construction, and improvement and innovation, as shown in Table 4.

[0148] Table 4 Data center green operation evaluation index system

[0149]

[0150]

[0151] 2. Construction of data center green operation evaluation method:

[0152] 2.1 Constructing learning samples based on subjective and objective methods combined with empowerment:

[0153] The present invention constructs a learning sample that combines subjective and objective weights based on the analytic hierarchy process and the entropy weight method. First, a judgment matrix is ​​constructed based on the expert opinions of the analytic hierarchy process, and then the weights are calculated according to the square root method, and the maximum characteristic root and the consistency ratio are further calculated. If the consistency ratio is less than 0.1, it proves that the weight is reasonable. On this basis, the original evaluation matrix is ​​preprocessed, the variation size, entropy value and information entropy redundancy of each indicator are calculated, and finally the weight is calculated based on the above information. Finally, we use a linear method to combine the weights of the analytic hierarchy process and the entropy weight method to obtain weights based on the combination of subjective and objective methods. Specifically, Algorithm 1 shows the weighting mechanism of the analytic hierarchy process and the entropy weight method, where X=(x ij ) represents the original evaluation matrix, W[j] AHP represents the weight of the AHP method, W[j] entropy In the present invention, we score the green operation evaluation index of the data center according to the weight of the combined weighting, and obtain a learning sample based on the subjective and objective method combined with the weighting. This method has higher scientificity and reliability.

[0154]

[0155]

[0156] 2.2 Data Center Green Operation Evaluation Method Based on Deep Belief Network

[0157] The data center green operation evaluation method based on deep belief network mainly includes pre-training and fine-tuning of deep belief network. Pre-training of deep belief network involves training restricted Boltzmann machines (RBMs) layer by layer. In this process, each layer of RBM learns features from the training data and passes these features to the next layer. This layer-by-layer training method not only helps to capture the complex structure of the data, but also effectively avoids the gradient vanishing problem. Through pre-training, DBN can obtain a good initial parameter setting, laying the foundation for subsequent fine-tuning. Next, the fine-tuning step globally optimizes the entire network. Fine-tuning uses the back-propagation algorithm to further adjust the weights of the network by minimizing the prediction error. This process enables the network to more accurately reflect the actual operation of the data center when processing specific green operation evaluation tasks. The detailed algorithm introduction is shown below.

[0158] First, the green operation indicators of existing data centers and their corresponding scores are collected and sorted, and these data will be input into the deep belief network as a training set. The initial stage of the deep belief network includes pre-training, which is performed by training the restricted Boltzmann machine layer by layer. The training process of each layer of the Boltzmann machine first maximizes its log-likelihood function to capture the probability distribution of the input data. After the pre-training is completed, the entire network has learned how to effectively represent the input data. Next, the fine-tuning process is carried out, and the fine-tuning is comprehensively adjusted on the entire deep belief network through the back-propagation algorithm. The network is supervised and learned using the green operation indicators and scores of existing data centers to minimize the error between the predicted score and the actual score. Through multiple iterative optimizations, the deep belief network gradually improves its understanding and prediction capabilities of the green operation of the data center. The core of this fine-tuning process is to adjust the weights and biases of the network so as to output more accurate evaluation results on given input data. After the network training is completed, the green operation indicators of the data center to be evaluated will be input into the trained deep belief network. The network will process and evaluate the new input data based on the previously learned patterns and rules, and output the green operation score of the data center. This score reflects the data center's performance in green operations and can provide important reference for managers.

[0159] The technical solution of the present invention uses the analytic hierarchy process and entropy weight method to comprehensively weight the green operation evaluation indicators of the data center to construct a learning sample, and on this basis, a deep belief network is used to achieve a scientific evaluation of the green operation of the data center, as detailed as follows:

[0160] 1. Enhanced the objectivity and accuracy of the evaluation results: The weighting method combining the analytic hierarchy process and the entropy weight method can effectively integrate subjective experience and objective data, ensuring that the weight distribution of the evaluation indicators is more scientific and reasonable. The analytic hierarchy process quantifies subjective experience into weights through expert scoring, while the entropy weight method objectively calculates weights through the discreteness of indicator data. The combination of the two takes into account the professional knowledge of experts in weight distribution and makes full use of the information of the data itself, enhancing the objectivity and accuracy of the evaluation results.

[0161] 2. Capturing complex features and improving the accuracy and reliability of evaluation results: As an efficient deep learning model, the deep belief network (DBN) can accurately model complex systems in a big data environment. Through the pre-training and fine-tuning process, DBN can extract high-order features of the data layer by layer, capture subtle changes in the operation of the data center, and improve the accuracy and reliability of the evaluation. The pre-training stage uses unsupervised learning to effectively initialize the network weights, avoiding the common gradient vanishing problem of deep networks, while the fine-tuning stage further optimizes the model through supervised learning to improve classification performance. In this way, the deep belief network performs well in processing multi-dimensional complex data and can fully reflect the green operation of the data center.

[0162] Exemplary Devices

[0163] Figure 3 Schematic diagram of the structure of a data center green operation evaluation device provided by an exemplary embodiment of the present invention. Figure 3 As shown, the device 300 includes:

[0164] An acquisition module 310 is used to acquire data center green operation characteristic data of the data center;

[0165] The output module 320 is used to input the data center green operation feature data into the pre-trained data center green operation evaluation model and output the evaluation score of the data center, where

[0166] The construction process of the data center green operation evaluation model in the output module 320 is as follows:

[0167] The acquisition submodule is used to obtain historical data center characteristics and build an indicator system for green operation of data centers;

[0168] Construct submodules to construct learning samples that combine subjective and objective factors and empower them according to the indicator system of each data center by using the hierarchical analysis method and entropy weight method;

[0169] The generation submodule is used to pre-train and fine-tune the deep belief network based on the learning samples to generate a data center green operation evaluation model.

[0170] Optionally, build submodules, including:

[0171] The first acquisition unit is used to weight the original evaluation matrix using the expert opinions of the hierarchical analysis method to obtain the first weight of each evaluation indicator in the indicator system;

[0172] A construction unit, used to construct an original evaluation matrix according to multiple data centers and an indicator system of each data center;

[0173] The second acquisition unit is used to weight the original evaluation matrix using an entropy weight method to obtain a second weight of each evaluation indicator in the indicator system;

[0174] A third acquisition unit is used to linearly combine the first weight and the second weight using a linear method to obtain a third weight assigned to each evaluation indicator combination in the indicator system;

[0175] The fourth acquisition unit is used to score each evaluation indicator of the indicator system of each data center using the third weight to obtain a learning sample.

[0176] Optionally, the first acquisition unit includes:

[0177] Constructing sub-units for constructing a judgment matrix of evaluation indicators of the indicator system of the data center based on the expert opinions of the hierarchical analysis method;

[0178] A first calculation subunit, used for calculating an initial first weight of each evaluation index according to the judgment matrix by using a square root method;

[0179] A first acquisition subunit, configured to perform normalization processing on the initial first weight to obtain a normalized initial first weight;

[0180] A second calculation subunit is used to calculate the maximum characteristic root according to the normalized initial first weight, and calculate the consistency ratio according to the maximum characteristic root;

[0181] The determination subunit is used to determine the normalized initial weight in this case as the first weight when the consistency ratio is less than a preset threshold.

[0182] Optionally, the second acquisition unit includes:

[0183] The second acquisition subunit is used to normalize the original evaluation matrix to obtain a normalized original evaluation matrix;

[0184] The third calculation subunit is used to calculate the variation size, entropy value and information entropy redundancy of each evaluation indicator according to the normalized original evaluation matrix;

[0185] The fourth determination subunit is used to calculate the second weight of each evaluation indicator according to the variation size, entropy value and information entropy redundancy of each evaluation indicator.

[0186] Optionally, generate submodules, including:

[0187] The first generation unit is used to input the learning samples into the deep belief network to train the restricted Boltzmann machine layer by layer by maximizing the log-likelihood function to generate an initial data center green operation evaluation model;

[0188] The second generating unit is used to fine-tune the initial data center green operation evaluation model through a back propagation algorithm to generate a data center green operation evaluation model.

[0189] Exemplary Electronic Devices

[0190] Figure 4 This is a structure of an electronic device provided by an exemplary embodiment of the present invention. Figure 4 As shown, the electronic device 40 includes one or more processors 41 and a memory 42 .

[0191] The processor 41 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0192] The memory 42 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 41 may run the program instructions to implement the methods of the software programs of the various embodiments of the present invention described above and / or other desired functions. In one example, the electronic device may also include: an input device 43 and an output device 44, which are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0193] In addition, the input device 43 may also include, for example, a keyboard, a mouse, etc.

[0194] The output device 44 can output various information to the outside, and can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto.

[0195] Of course, to simplify, Figure 4 Only some of the components related to the present invention in the electronic device are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, the electronic device may further include any other appropriate components according to specific application conditions.

[0196] Exemplary computer program products and computer-readable storage media

[0197] In addition to the above-mentioned methods and devices, an embodiment of the present invention may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the method according to various embodiments of the present invention described in the above-mentioned "Exemplary Method" section of this specification.

[0198] The computer program product may be written in any combination of one or more programming languages ​​to write program code for performing the operations of the embodiments of the present invention, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0199] In addition, an embodiment of the present invention may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, enable the processor to execute the steps of the method according to various embodiments of the present invention described in the above “Exemplary Method” section of this specification.

[0200] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, system or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0201] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details disclosed above are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0202] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0203] The block diagrams of the devices, systems, equipment, and systems involved in the present invention are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, systems, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The word "such as" used here refers to the phrase "such as but not limited to", and can be used interchangeably with it.

[0204] The method and system of the present invention may be implemented in many ways. For example, the method and system of the present invention may be implemented by software, hardware, firmware or any combination of software, hardware, firmware. The above order of steps for the method is only for illustration, and the steps of the method of the present invention are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present invention may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present invention. Thus, the present invention also covers a recording medium storing a program for executing the method according to the present invention.

[0205] It should also be noted that in the system, device and method of the present invention, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. The above description of the disclosed aspects is provided to enable any technician in the field to make or use the present invention. Various modifications to these aspects are very obvious to those skilled in the art, and the general principles defined here can be applied to other aspects without departing from the scope of the present invention. Therefore, the present invention is not intended to be limited to the aspects shown here, but in accordance with the widest range consistent with the principles and novel features disclosed here.

[0206] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present invention to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A data center green operation evaluation method, characterized in that: include: Obtain data center green operation characteristic data of the data center; The data center green operation feature data is input into a pre-trained data center green operation evaluation model, and the evaluation score of the data center is output, where The construction process of the data center green operation evaluation model is as follows: Obtain historical data center characteristics and build an indicator system for green operation of data centers; Adopting the hierarchical analysis method and the entropy weight method to construct a learning sample with subjective and objective combined empowerment according to the indicator system of each data center; The deep belief network is pre-trained and fine-tuned according to the learning samples to generate the data center green operation evaluation model.

2. The method according to claim 1, characterized in that The analytic hierarchy process and entropy weight method are used to construct a learning sample that combines subjective and objective weighting according to the indicator system of each data center, including: The original evaluation matrix is ​​weighted by expert opinions of the hierarchical analysis method to obtain the first weight of each evaluation indicator in the indicator system; Constructing an original evaluation matrix according to a plurality of data centers and an indicator system of each of the data centers; The original evaluation matrix is ​​weighted by using an entropy weight method to obtain a second weight of each evaluation indicator in the indicator system; Using a linear method to linearly combine the first weight and the second weight to obtain a third weight assigned to each evaluation indicator combination in the indicator system; The third weight is used to score each evaluation indicator of the indicator system of each data center to obtain the learning sample.

3. The method according to claim 2, characterized in that The original evaluation matrix is ​​weighted by expert opinions of the hierarchical analysis method to obtain the first weight of each evaluation indicator in the indicator system, including: Constructing a judgment matrix of evaluation indicators of the indicator system of the data center according to the expert opinions of the hierarchical analysis method; Calculating the initial first weight of each evaluation index according to the judgment matrix using the square root method; Normalizing the initial first weight to obtain a normalized initial first weight; Calculating a maximum eigenvalue according to the normalized initial first weight, and calculating a consistency ratio according to the maximum eigenvalue; When the consistency ratio is less than a preset threshold, the normalized initial weight in this case is determined as the first weight.

4. The method according to claim 2, characterized in that: The entropy weight method is used to weight the original evaluation matrix to obtain the second weight of each evaluation indicator in the indicator system, including: Normalizing the original evaluation matrix to obtain a normalized original evaluation matrix; Calculating the variation size, entropy value and information entropy redundancy of each evaluation indicator according to the normalized original evaluation matrix; The second weight of each evaluation indicator is calculated according to the variation size, the entropy value and the information entropy redundancy of each evaluation indicator.

5. The method according to claim 1, characterized in that Pre-training and fine-tuning the deep belief network according to the learning sample to generate the data center green operation evaluation model, including: Inputting the learning samples into the deep belief network to train the restricted Boltzmann machine layer by layer by maximizing the log-likelihood function to generate an initial data center green operation evaluation model; The initial data center green operation evaluation model is fine-tuned through a back-propagation algorithm to generate the data center green operation evaluation model.

6. A data center green operation evaluation device, characterized in that: include: An acquisition module, used to acquire data center green operation characteristic data of the data center; An output module is used to input the data center green operation feature data into a pre-trained data center green operation evaluation model and output the evaluation score of the data center, wherein The construction process of the data center green operation evaluation model in the output module is as follows: The acquisition submodule is used to obtain historical data center characteristics and build an indicator system for green operation of data centers; A construction submodule is used to construct a learning sample that combines subjective and objective weights according to the indicator system of each data center by using the hierarchical analysis method and the entropy weight method; A generation submodule is used to pre-train and fine-tune the deep belief network according to the learning samples to generate the data center green operation evaluation model.

7. The device according to claim 6, characterized in that Build submodules, including: A first acquisition unit is used to weight the original evaluation matrix using expert opinions of the hierarchical analysis method to obtain a first weight of each evaluation indicator in the indicator system; A construction unit, configured to construct an original evaluation matrix according to a plurality of data centers and an indicator system of each of the data centers; A second acquisition unit is used to weight the original evaluation matrix by using an entropy weight method to obtain a second weight of each evaluation indicator in the indicator system; A third acquisition unit, configured to linearly combine the first weight and the second weight using a linear method to acquire a third weight assigned to each evaluation indicator combination in the indicator system; The fourth acquisition unit is used to score each evaluation indicator of the indicator system of each data center by using the third weight to obtain the learning sample.

8. The device according to claim 7, characterized in that The first acquisition unit includes: A construction subunit is used to construct a judgment matrix of evaluation indicators of the indicator system of the data center according to the expert opinions of the hierarchical analysis method; A first calculation subunit, configured to calculate an initial first weight of each evaluation index according to the judgment matrix by using a square root method; A first acquisition subunit, configured to perform normalization processing on the initial first weight to acquire a normalized initial first weight; A second calculation subunit, configured to calculate a maximum characteristic root according to the normalized initial first weight, and calculate a consistency ratio according to the maximum characteristic root; A determination subunit is used to determine the normalized initial weight in this case as the first weight when the consistency ratio is less than a preset threshold.

9. The device according to claim 7, characterized in that The second acquisition unit includes: The second acquisition subunit is used to normalize the original evaluation matrix to obtain a normalized original evaluation matrix; A third calculation subunit is used to calculate the variation size, entropy value and information entropy redundancy of each evaluation indicator according to the normalized original evaluation matrix; The fourth determination subunit is used to calculate the second weight of each evaluation indicator according to the variation size, the entropy value and the information entropy redundancy of each evaluation indicator.

10. The device according to claim 6, characterized in that Generate submodules, including: A first generating unit is used to input the learning sample into the deep belief network to train the restricted Boltzmann machine layer by layer by maximizing the log-likelihood function to generate an initial data center green operation evaluation model; The second generating unit is used to fine-tune the initial data center green operation evaluation model through a back propagation algorithm to generate the data center green operation evaluation model.

11. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and the computer program is used to execute the method according to any one of claims 1 to 5.

12. An electronic device, characterized in that: The electronic device comprises: processor; a memory for storing instructions executable by the processor; The processor is used to read the executable instructions from the memory and execute the instructions to implement the method described in any one of claims 1 to 5.

Images