A Novel Self-Adaptation Approach for Multi-Domain Communication Considering Heterogenerous Power Service in Data Centers

Abstract

With the rapid development of cloud computing, artificial intelligence, etc., data centers have become a flexible load due to their adjustable capabilities and are able to take on power service. However, the existing communication architecture is monolithic and poorly adapted between services and communications, which poses a challenge for data centers to participate in power service. Firstly, this article constructs a multi-domain communication architecture for data center management systems that includes local and remote communication, and analyzes the key technologies supporting this architecture. Secondly, based on differentiated service types and communication requirements, an evaluation system for the adaptability between service and communication technology was constructed, and a communication mode adaptation method based on fuzzy analytic hierarchy process (FAHP)-improved CRITIC combined weighting model and grey Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) was proposed to achieve the analysis of the adaptability between differentiated service and multi-domain communication technology. According to the analysis of examples, the architecture and adaptation algorithm proposed in this article provide an effective theoretical basis and solution for the selection of multi-domain communication technologies for data center management system service.

1. Introduction

With the large-scale application of artificial intelligence technology, the number of data center racks has surged globally, and data centers have gradually become bridges connecting various industries [1]. However, large data centers are accelerating global carbon emissions. According to statistics, the power consumption of global data centers in 2020 has exceeded the total electricity consumption of some countries [2]. As a new type of power load, the power consumption of data centers is mainly composed of auxiliary equipment such as IT equipment, refrigeration systems, and lighting [3]. Partial loads have strong flexibility in both temporal and spatial scales. Scholars have studied the collaborative flexibility of data center operations at the spatiotemporal scale to fully absorb new energy, reduce carbon emissions, and cope with global temperature rise. The data center management system, as a system for information collection, processing, real-time monitoring, and dynamic aggregation and flexible control of load resources, is of great significance for collecting, analyzing, and providing feedback on the internal composition of the data center. Data centers must meet the needs of reducing carbon emissions and lowering global temperatures. In the trend of complex load resources and diversified load structures in data centers, building a safe, reliable, and practical data center management system is the key to building data center informatization [4], and it is also the foundation for the coordinated operation of electricity and computing power. In 2016, 178 countries around the world approved the Paris Agreement in Paris to address global warming and reduce carbon emissions. As a major energy consumer, data centers will become increasingly important in this agreement. Building a data center management system can facilitate the coordinated operation of computing power and electricity, assist in the precise and efficient management of new energy absorption and carbon emissions, and promote the achievement of the goal of countries signing the Paris Agreement.

Modern information and communication technology, as the key to building and carrying out data center management system services, can help data center management systems achieve precise monitoring and regulation, effectively support the construction of normalized demand side management work in the power grid, and improve the quality of energy services (QoS). However, with the surge of computing power racks, data management systems under low-carbon requirements are facing challenges such as a surge in data volume, diversification and differentiation of communication services, and difficulty in adapting appropriate communication technologies to power communication services.

Reference [5] introduced a method to evaluate communication technologies in power systems, combining the Analytic Hierarchy Process (AHP) with triangular fuzzy numbers (TFN). This method, utilizing a fuzzy judgment matrix and AHP, effectively ranks technologies and identifies Power Line Carrier as the superior option. Reference [6] advanced electric power communication network security by introducing a fuzzy AHP. This work significantly bolsters network reliability and establishes a structured risk assessment framework, offering a notable contribution to power system security management. The AHP method was used in Reference [7] to select the best communication technologies for electrical substations. Optical fiber, NBPLC, and LoRa were identified for urban and suburban areas, while satellite communication was preferred for rural areas. This aids in improving the efficiency and reliability of electrical distribution networks. The F-TOPSIS method was used in Reference [8] to select smart grid communication technologies based on cost, security, and reliability. This approach provides a systematic framework for comparing technologies and guides the deployment of efficient and secure smart grid communication infrastructures. Reference [9] introduced a layered architecture integrating fog and cloud computing to enhance smart city communication networks, focusing on smart energy systems. However, this approach overlooks the AHP method’s subjectivity, potentially leading to judgment errors. Additionally, a single form of communication technology cannot meet the diverse needs of power communication.

In response to the above issues, this article first analyzes the communication requirements for differentiated data center management service, integrates multiple heterogeneous communication methods, and constructs a communication architecture for a data center management system that includes local and remote communication, Furthermore, we analyze the key technologies supporting this architecture. To address the issue of poor adaptability between differentiated services and modern communication technology, a comprehensive and scalable communication index system for data center management systems is established. A subjective and objective combination weighting model based on fuzzy analytic hierarchy process and fusion entropy weight CRITIC method is proposed to weight the differentiated service communication demand index. Then, the grey correlation analysis method is used to improve the approximation ideal solution ranking method to select the optimal local and remote communication technologies for specific power services. The subjective–objective weighting method fully considers both subjective decision-making needs and the objective attributes of the indexes, ensuring accurate and reasonable weight results. The improved TOPSIS method based on grey relational analysis simultaneously considers the proximity and correlation between alternative energy storage solutions and the ideal solution, enhancing the ability to handle uncertain information. Finally, the effectiveness of the proposed method is verified by taking four differentiated power load management services as examples.

2. Communication Architecture

2.1. Multi-Domain Communication Architecture of Data Center Management Systems Including Local and Remote Communication

The communication system is a core component of the data center management system, and advanced and comprehensive information and communication technology and standardized protocols provide important technical support for the data center management system [10]. Modern communication technology can adapt to the access and scheduling of various power resources such as IT equipment, refrigeration systems, lighting, etc. inside data centers [11], achieving intelligent operation of data centers and providing important guarantees for the safe and stable operation, optimized scheduling, and intelligent management of data centers [12].

Traditional research on communication network architecture for data center management systems often focuses on adaptation analysis of a single communication technology, lacking consideration of the demand for heterogeneous communication fusion networking in differentiated and diversified power service under the background of the gradual popularization of modern communication technology [13].

This article constructs a multi-domain communication architecture for a data center management system that includes local and remote communication. Considering the service communication needs of the data center management system, heterogeneous communication fusion networking is achieved from local and remote communication, improving the carrying capacity of power service. Based on cloud edge collaboration technology, network costs are saved, ensuring efficient consumption of new energy, achieving unified management and precise regulation of load resources, and achieving precise load control and other data center load management [14].

Figure 1 is a schematic diagram of the multi-domain communication network architecture of the data center management system. The communication system has a layered architecture, including terminal layer, access layer, network layer, platform layer, and service layer, and provides secure and reliable communication protocols. The terminal layer is mainly composed of the main internal structures of the data center, including IT equipment, refrigeration systems, lighting, and other auxiliary equipment. Intelligent terminals can perceive the real-time status of various load resources, complete data collection and status monitoring of load resources, and upload the data. In addition, the terminal layer also includes edge computing and edge proxy functions.

The access layer is composed of local communication technologies such as RS485, HPLC, WiFi, LoRa, and Bluetooth. It establishes communication connections with terminal equipment using various communication protocols and is equipped with an access security firewall to ensure the security of data uploads from terminals and the issuance of platform instructions. It collaborates with the network layer to upload service data, thus requiring unified access to load terminals at the access layer.

The network layer consists of remote communication technologies such as wireless private network, wireless public network, fiber optic private network, and medium voltage power line carrier communication. Various types of information transmit in the data center management system over a wide area between the platform layer and the access layer, and equip with a network firewall to ensure the security of remote data transmission.

The platform layer serves as the cloud computing center and is composed of platforms. The platform layer is based on artificial intelligence, big data, and other technologies to achieve centralized storage, analysis, and processing of load data, then making intelligent decisions, control, and services through software platforms. It utilizes cloud edge collaboration to achieve efficient resource utilization.

The service layer consists of differentiated data center management system services, including precise load control, video monitoring, and electricity information collection. The service layer is responsible for transmitting specific service requirements to the platform layer for processing the uploaded load information and sending control instructions to the intelligent perception devices at the terminal layer.

2.2. Key Technologies Supporting the Communication Network Architecture of Data Center Management Systems

2.2.1. Heterogeneous Communication Network Fusion

The traditional load management system’s single communication network technology is gradually difficult to carry differentiated and diversified data center power services, while the data center management system uses remote communication and local communication technologies at the network layer and access layer, respectively, to jointly carry power services and provide diversified communication services. In response to the problem of heterogeneity in local and remote communication technologies, the data center management system adopts heterogeneous communication network fusion to achieve organic integration and efficient interconnection between local and remote heterogeneous communication technologies [11], fully utilizing different network characteristics to ensure the service quality of power service communication.

2.2.2. Edge Computing

The large-scale construction of data center racks has caused problems such as high data processing pressure on the cloud platform of the data center management system [15], untimely response to load resources, and poor service quality. As shown in Figure 2, edge computing technology realizes edge intelligent services for the data center management system, enables the differentiated load management service to operate locally, and relieves the pressure of cloud platform data processing by establishing an open platform with core capabilities such as storage and computing in the area close to the edge data source [16].

Compared to traditional load management systems, data center systems based on cloud edge collaboration have the advantages of low latency and low bandwidth usage [13]. This improves the response speed of load resources while saving network deployment costs, achieving full utilization of demand side resources, ensuring efficient consumption of new energy, and improving the overall efficiency of energy utilization [17].

3. Service Requirements Analysis of the Data Center Management System

With the continuous expansion of data and service scale, data centers are currently moving towards large-scale development. The continuous emergence of data center management services has put forward higher requirements for the reliability, security, and other performance [18].

Control service is a key service that monitors, controls, and optimizes data centers in real time through communication networks and control equipment, aiming to ensure the safe and stable operation of data centers and green economic operation. The control service of data center management system mainly includes resource scheduling optimization, security management, energy management, etc. Control service schedules and optimizes data center resources to ensure efficient utilization of resources, including dynamically adjusting resource allocation according to service needs, optimizing server load distribution, and reasonably allocating network bandwidth to improve the overall performance and efficiency of the data center. It is necessary to ensure the high reliability of communication networks and control equipment to prevent data loss caused by communication or control failures, which may lead to operational failures and accidents in the data center. It is necessary to ensure the security of communication networks and control equipment to prevent information leakage or control failure in data centers caused by attacks or tampering on communication networks or control equipment, which may affect the security of user data usage. In the context of a green economy, energy management in data centers has become an important part of control related services. By monitoring and adjusting the load usage of data centers, utilizing the flexible spatiotemporal characteristics of data center loads, we can fully absorb new energy and reduce operating costs. Therefore, the control service of data center management system requires high requirements for real-time management, reliability, and security of communication.

The collection service provides terminal device data for the data center management system, and is the basic service for the data center management system to monitor and regulate controllable load terminals. This type of service mainly includes equipment status detection, electricity information collection, fault detection and diagnosis, etc. With the deepening construction of data centers and the explosive increase in collection terminal equipment, it is necessary to consider the deployment cost of communication technology. The large-scale integration of distributed renewable energy connecting to the grid increases the difficulty of deploying communication equipment, and the large-scale distributed collection equipment also puts higher requirements on the transmission distance and reliability of communication technology [14].

The service conducts centralized regulation and integration of the comprehensive energy, energy storage equipment, and schedulable load of the data center through the communication network and intelligent equipment, and reflects the supply and demand status and operating costs of the power system, encourages the active participation and reasonable dispatching of the data center operators, coordinates the power distribution among the grid companies, virtual power plant aggregators, and data center entities, improves the power supply and demand situation, and fully consumes renewable energy through the transmission of real-time price. The service of the data center management system mainly includes electricity market transactions, clearing transactions, etc. By integrating the access objects of the grid side and the user side into the data center management system, control of the two terminals is achieved, and user transactions throughout the entire chain of the data center management system are completed. Because this type of service involves real-time transmission and feedback of electricity prices and data information, communication network latency requirements are high to ensure fast response and real-time control of the service. To prevent the theft, tampering, or destruction of electricity market transaction data and information, the communication technology carrying this service should have a high degree of security protection capability. In addition, to support power market data transmission and information sharing between different regions and regions, this type of service requires high communication network transmission distance. The adaptability evaluation system of service and communication technology is shown in the Figure 3.

4. Service and Communication Technology Adaptation Method

4.1. Construction of an Adaptability Evaluation System of Service and Communication Technology

To accurately describe the communication service requirements and communication technology performance of the data center management system, we consider selecting six typical factors that affect the adaptability between service and communication technology, and construct an evaluation system for the adaptability between service and communication technology of the data center management system [15], as shown in Figure 2. This system is based on the Analytic Hierarchy Process [16] and is divided into target layer, index layer, and scheme layer. The target layer evaluates the adaptability between a certain load management service and communication technology. The index layer reflects the influencing factors of service requirements and communication technology performance. The solution layer is an alternative local and remote communication technology.

4.2. Determination of Index Weights Based on the FAHP-Improved CRITIC Combined Weighting Model

There are primarily two methods for determining index weights: subjective weighting and objective weighting. Subjective weighting relies on the subjective experience of decision makers or experts to determine the weights of evaluation indexes. However, this method can be influenced by various factors such as the knowledge structure, work experience, and preferences of decision makers, which may not fully reflect the importance of the evaluation indexes. Objective weighting determines the weight of evaluation indexes based on the amount of discriminative information and the interrelationships provided by the indexes. However, objective weighting might overlook the rich experience of experts and scholars and have a strong dependence on samples. The results may not reflect the actual conditions.

Compared to subjective and objective weighting, combined weighting can integrate the subjective and objective weights of various evaluation indexes, reflecting both the objective information of the indexes and the subjective judgment of the evaluators. Therefore, it can accurately reflect the actual weights of each index.

Based on the adaptability evaluation system, this paper comprehensively utilizes the FAHP method and the improved CRITIC method to weight communication indexes, taking into account both subjective and objective factors. The subjective and objective combination weighting model for communication adaptability of the data center management system is shown in the Figure 4.

4.2.1. Subjective Weighting

The AHP method is a decision-making tool. It is widely used in multi-criteria decision-making problems. It relies on expert evaluation when conducting decision analysis, and the main characteristics of this subjective evaluation are fuzziness and uncertainty. FAHP takes into account the characteristics in the process of pairwise comparison of indexes and introduces fuzzy theory based on the traditional AHP method, which greatly reduces judgment errors.

This paper uses the FAHP method based on triangular fuzzy numbers. The steps to weight the indexes of the adaptability evaluation system using the FAHP method are as follows.

Step 1. Establishing a fuzzy judgment matrix.

Experts and relevant practitioners score the selected communication indexes in pairs using a triangular fuzzy scale (as shown in

Table 1. Fuzzy triangular scale.

Scale	Definition	Triangular Fuzzy Numbers
1	equally important	(1,1,1)
3	slightly important	(2,3,4)
5	obviously important	(4,5,6)
7	very important	(6,7,8)
9	extremely important	(9,9,9)
2		(1,2,3)
4	interval value between two adjacent levels	(3,4,5)
6		(5,6,7)
8		(7,8,9)

), thereby establishing a fuzzy judgment matrix ${\tilde{A}}^{\dot{k}}={({\tilde{a}}_{ij}^{\dot{k}})}_{n\times n}$, where ${\tilde{a}}_{ji}^k={({\tilde{a}}_{ij}^k)}^{-1}=(1/u_{ij},1/h_{ij},1/l_{ij})$:

$$\begin{array}{c}{\tilde{A}}^k={({\tilde{a}}_{\tilde{i}\tilde{j}}^{\tilde{k}})}_{n\times n}=\\ \left[\begin{array}{cccc}(1,1)& (u_{12},h_{12},l_{12})& \cdots & (u_{1n},h_{1n},l_{1n})\\ (\frac{1}{u_{21}},\frac{1}{h_{21}},\frac{1}{l_{21}})& (1,1,1)& \cdots & (u_{2n},h_{2n},l_{2n})\\ ⋮& ⋮& \ddots & ⋮\\ (\frac{1}{u_{n1}},\frac{1}{h_{n1}},\frac{1}{l_{n1}})& (\frac{1}{u_{n2}},\frac{1}{h_{n2}},\frac{1}{l_{n2}})& \cdots & (1,1,1)\end{array}\right]\end{array}$$

(1)

where k (k = 1, 2, …, K) represents the expert number and i (i = 1, 2, …, n) represents the index number. The judgment matrices from a total of K decision makers are averaged to update the decision matrix, obtaining $\tilde{A}$:

$${\tilde{a}}_{ij}=\frac{{\sum }_{k=1}^K{\tilde{a}}_{ij}^k}{K}$$

(2)

$$\tilde{A}={\left({\tilde{a}}_{ij}\right)}_{n\times n}={\left(\frac{{\sum }_{k=1}^K{\tilde{a}}_{ij}^{\overline{k}}}{K}\right)}_{n\times n}$$

(3)

Step 2. Consistency check.

For the updated decision matrix, the median of the elements represented by triangular fuzzy numbers is taken to form the median matrix $m=\left(sij\right)n$ × n. Calculate the largest eigenvalue λmax of matrix m, and then determine the consistency index CI using the formula. The random consistency index RI can be obtained from a lookup table. If the ratio of CI to RI is less than 0.1, the consistency check is passed. Otherwise, the fuzzy judgment matrix needs to be reconstructed.

Step 3. Calculation of subjective weights.

For triangular fuzzy numbers $M1=(l1,h1,u1)$ and $M2=(l2,h2,u2)$, the following computational rules are adhered to:

$$\left\{\begin{array}{c}{M}_1\oplus M_2=(l_1+l_2,h_1+h_2,u_1+u_2)\hfill \\ M_1\otimes M_2=(l_1{l}_2,h_1{h}_2,u_1{u}_2)\hfill \\ M_1^{-1}=(l_1^{-1},h_1^{-1},{u_1}^{-1})\hfill \end{array}\right.$$

(4)

Based on Formulas (5) and (6), the geometric mean of the fuzzy comparison values of each index, denoted as ${\tilde{r}}_i$, and the fuzzy weights of each index, denoted as ${\tilde{\nu }}_i$, are calculated sequentially:

$${\tilde{r}}_i={\left(\prod _{j=1}^n{\tilde{a}}_{ij}\right)}^{1/n}$$

(5)

$$\begin{array}{cc}\hfill {\tilde{\mathcal{W}}}_i& ={\tilde{r}}_i\otimes {\left({\tilde{r}}_1\oplus {\tilde{r}}_2\oplus \cdots \oplus {\tilde{r}}_n\right)}^{-1}\hfill \\ \hfill & =\left(l{w}_i,h{w}_i,u{w}_i\right)\hfill \end{array}$$

(6)

Given that the fuzzy weights are represented by triangular fuzzy numbers, they require defuzzification by a specified formula:

$$M_i=\frac{l{w}_i+h{w}_i+u{w}_i}{3}$$

(7)

The final step involves normalizing these weights to derive the subjective weights of each index, which can be expressed as:

$$N_i=\frac{M_i}{{\sum }_{i=1}^n{M}_i},i=1,2,\dots ,n$$

(8)

4.2.2. Objective Weighting

Objective weighting methods determine the objective weight of selected indexes by exploring objective factors such as volatility, amount of information, and conflict among all alternative solutions. The CRITIC method adopted in this paper is an objective weighting method that considers the variability and conflict of index data. Variability reflects the differences of an evaluation index across various alternatives. The greater the variability, the greater the differentiation capability of the index data within the adaptability evaluation system, and thus, its objective importance is higher. Conflict reflects the correlation between an evaluation index and other indexes. Lower conflict means a higher correlation with other indexes. When an index has a similar influence on the adaptability evaluation system as other indexes, its objective importance is considered lower. Additionally, variability represents the objective attributes of the index data itself, while conflict represents the objective attributes between index data. The specific calculation steps of the CRITIC method are as follows.

Step 1. Construction of evaluation matrices and standardization.

Construct the original evaluation matrix X from the m alternatives at the solution layer and the n indexes at the index layer of the adaptability evaluation system:

$$X=\left[\begin{array}{cccc}{x}_{11}& x_{12}& \cdots & x_{1n}\\ x_{21}& x_{22}& \cdots & x_{2n}\\ ⋮& ⋮& \ddots & ⋮\\ x_{m1}& x_{m2}& \cdots & x_{mn}\end{array}\right]$$

(9)

where $x_{ij}$ represents the performance data of the j-th index of the i-th alternative, where $i=1,2,\dots ,m$ and $j=1,2,\dots ,n$.

According to the influence of each index on the adaptation results, the indexes can be categorized into benefit-type and cost-type indexes. For benefit-type indexes, the values of the indexes are directly proportional to their adaptation results, while cost-type indexes are inversely proportional. The purpose of data standardization is to make different kinds of indexes have the same impact on the results. For benefit-type indexes, carry out the following steps:

$$x_{ij}^{\prime }=\frac{x_{ij}-min\left(x_j\right)}{max\left(x_j\right)-min\left(x_j\right)}$$

(10)

For cost-type indexes, carry out the following steps:

$$x_{ij}^{\prime }={\displaystyle \frac{max\left(x_j\right)-x_{ij}}{max\left(x_j\right)-min\left(x_j\right)}}$$

(11)

Finally, we obtain the standard evaluation matrix $X^{\prime }$:

$${\mathit{X}}^{\prime }=\left[\begin{array}{cccc}{x}_{11}^{\prime }& x_{12}^{\prime }& \dots & x_{1n}^{\prime }\\ x_{21}^{\prime }& x_{22}^{\prime }& \dots & x_{2n}^{\prime }\\ ⋮& ⋮& \ddots & ⋮\\ x_{m1}^{\prime }& x_{m2}^{\prime }& \dots & x_{mn}^{\prime }\end{array}\right]$$

(12)

Step 2. Calculation of objective weights.

The CRITIC method considers volatility and conflict, represented by the standard deviation and correlation coefficient, respectively. The greater the standard deviation of the index data, the higher its volatility, and thus, the higher its weight. The smaller the correlation coefficient between indexes, the greater the conflict and, consequently, the greater the weight. According to the standard evaluation matrix $X^{\prime }$, the calculation formula for the standard deviation of each index data and the correlation coefficient between indexes is as follows:

$${\sigma }_j=\sqrt{\frac{{\sum }_{i=1}^m{(x_{ij}^{\prime }-{\overline{x}}_j)}^2}{m}}$$

(13)

$$r_{ij}=\mathrm{cov}(X_i^{\prime },X_j^{\prime })/({\sigma }_i,{\sigma }_j)$$

(14)

where ${\sigma }_j$ represents the standard deviation of the j-th index data, ${\overline{x}}_j$ represents the mean of the j-th index data, $r_{\dot{y}}$ represents the correlation coefficient between the i-th and j-th evaluation indexes, and $X_i^{\prime }$ and $X_j^{\prime }$ are the i-th row and j-th row, respectively, of the standard evaluation matrix $X^{\prime }$.

The information-carrying capacity of volatility and conflict for each index is defined as $C_j$, as shown in the Formula (15). The larger $Cj$ is, the larger the combined information carrying capacity of volatility and conflict for the j-th index is, and the higher the relative importance is. Finally, the CRITIC objective weight of the j-th index is calculated according to Formula (16):

$$C_j={\sigma }_j\sum _{i=1}^m(1-r_{ij})$$

(15)

$$w_j^c=\frac{C_j}{{\sum }_{j=1}^n{C}_j}=\frac{{\sigma }_j{\sum }_{i=1}^m(1-r_j)}{{\sum }_{j=1}^n\left[{\sigma }_j{\sum }_{i=1}^m(1-r_{ij})\right]}$$

(16)

As an objective weighting method, CRITIC considers the volatility of the index itself and the conflict between indexes; however, it does not take into account the amount of information, which can also be an objective factor affecting weights. The Entropy Weight Method (EWM) addresses this by applying the principle of entropy in physics to calculate weights based on the amount of information carried by the index data. Entropy measures the uncertainty of information. The greater the amount of information in the data, the smaller the uncertainty and the smaller the entropy value. EWM is introduced to improve the CRITIC method so that the improved CRITIC method comprehensively considers objective factors such as the volatility, conflict, and amount of information of the index. The improvement process is as follows.

To calculate the entropy value, the first step involves normalizing the standard evaluation matrix $X^{\prime }$ according to the formula below:

$$p_{ij}=x_{ij}^{\prime }/\sum _{i=1}^m{x}_{ij}^{\prime }$$

(17)

The entropy value of each evaluation index is calculated according to the following formula, where $k=1/lnm$ and m represents the number of alternatives:

$$e_j=-k\sum _{i=1}^m\left(p_{ij}\ast ln{p}_{ij}\right)$$

(18)

Entropy values are incorporated as a reference factor into the weight calculation formula of the CRITIC weighting method, where $(1-e_j)$ represents the amount of information of the j-th index. The improved objective weighting formula is as follows:

$$w_j=\frac{(1-e_j+{\sigma }_j){\sum }_{i=1}^m(1-r_j)}{{\sum }_{j=1}^n\left[(1-e_j+{\sigma }_j){\sum }_{i=1}^m(1-r_j)\right]}$$

(19)

4.2.3. Calculation of Combination Weight

The FAHP weighting method has inherent subjectivity and ignores objective information. The improved CRITIC weighting method is sample-dependent. A method that combines subjective and objective weighting can comprehensively consider both subjective and objective factors, reducing the limitations of both. The combined weighting formula is as follows:

$$W_j=\frac{N_j\ast w_j}{{\sum }_{j=1}^n(N_j\ast w_j)}$$

(20)

The vector of combined weights for the n indexes is $W=(W_1,W_2,\dots ,W_n).$

4.3. Data Preprocessing

The numerical changes of large-span indexes such as bandwidth have a greater influence on the adaptability evaluation results. Low discrimination indexes such as bit error rate have a relatively small influence on the adaptability evaluation results due to numerical changes. Therefore, to meet the requirements of adaptability evaluation, the large-span indexes and low-discrimination indexes are mapped to a reasonable interval. The mapping formula is as follows:

$$f\left(x\right)=\left|\begin{array}{c}lgx\end{array}\right|$$

(21)

where x represents the original value of the index with a large span or low discriminability and $f\left(x\right)$ is the value after reasonable mapping processing. The preprocessed data are maintained within a relatively reasonable range, preparing for subsequent adaptability evaluation.

4.4. Sorting and Decision Making Based on the GTOPSIS Method

The fundamental principle of the TOPSIS method involves sorting alternatives by calculating their Euclidean distances from both the positive ideal solution and the negative ideal solution, respectively. The positive ideal solution is identified by maximizing benefit-type indexes and minimizing cost-type indexes, whereas the negative ideal solution is determined by maximizing cost-type indexes and minimizing benefit-type indexes. This method is widely used in multi-criteria decision making, with strong operability and objective and comprehensive decision results. However, TOPSIS is sensitive to outliers, which may affect the accuracy of the results.

GRA is a method of ranking by calculating the grey relational degree between sub-sequences and the reference sequence. Due to its capacity to handle uncertain information, GRA is relatively robust to outliers, making it less susceptible to the influence of extreme values. Nevertheless, the selection of the reference sequence in GRA involves a high degree of subjectivity.

Consequently, this paper proposes the GTOPSIS method, which combines GRA with the TOPSIS method. GTOPSIS seeks a hybrid distance based on the principle of minimum information entropy, which is then used to calculate a relative hybrid closeness for ranking and decision making. The decision results derived from the improved ranking method are more stable and reliable. The specific steps of the decision-making method based on GTOPSIS are as follows.

4.4.1. TOPSIS

Step 1. Utilize the original evaluation matrix X as the original decision matrix for the TOPSIS method, and normalize it:

$$z_{ij}=x_{ij}/\sqrt{{\displaystyle \sum _{i=1}^m}{x_{ij}}^2}$$

(22)

where $z_{ij}$ represents the elements of the normalized decision matrix, with $i=1,2,\dots ,m$ and $j=1,2,\dots ,n$.

The normalized decision matrix is weighted using the combined weights calculated previously, resulting in the weighted decision matrix $V={\left(v_{ij}\right)}_{m\times n}:$

$$v_{ij}=W_j{z}_{ij}$$

(23)

Step 2. Calculate the positive ideal solution $A^{+}$ and negative ideal solution $A^{-}$ according to the following formula:

$$A^{+}=\left(v_1^{+},v_2^{+},\cdots ,v_n^{+}\right)$$

(24)

$$A^{-}=\left(v_1^{-},v_2^{-},\cdots ,v_n^{-}\right)$$

(25)

where, for benefit-type indexes, ${v_j}^{+}=max\left\{v_{1j},v_{2j},\cdots ,v_{mj}\right\},v_j^{-}=min\{v_{1j},v_{2j},\cdots ,v_{mj}\}$; for cost-type indexes, ${v_j}^{+}=min\{v_{1j},v_{2j},\cdots ,v_{mj}\}{v}_j^{-}=max\left\{v_{1j},v_{2j},\cdots ,v_{mj}\right\}$.

Step 3. Calculate the Euclidean distance $D_i^{+}$ from the alternative solutions to the positive ideal solution and the Euclidean distance $D_i^{-}$ from the alternative solutions to the negative ideal solution, which can be represented as:

$${D_i}^{+}=\sqrt{{\displaystyle \sum _{j=1}^m}{\left({\nu }_{ij}-{\nu }_j^{+}\right)}^2}$$

(26)

$$D_i^{-}=\sqrt{{\displaystyle \sum _{j=1}^m}{(v_{ij}-v_j^{-})}^2}$$

(27)

4.4.2. GRA

The positive ideal solution $A^{+}$ and the negative ideal solution $A^{-}$, obtained from the TOPSIS method, serve as the reference sequences for the GRA method. The grey relational degree $G_i^{\mp }$ between each alternative solution and the positive and negative ideal solutions is calculated using the following formula, where $\rho$, the resolution factor, is typically set to 0.5:

$$G_i^{\mp }=\frac{{\sum }_{j=1}^n{r}_{ij}^{\mp }}{n}$$

(28)

$$r_{ij}^{\mp }=\frac{\underset{1\le i\le m1\le j\le n}{min}\left|v_j^{\mp }-v_{ij}\right|+\rho \underset{1\le i\le m}{max}\underset{1\le j\le n}{max}\left|v_j^{\mp }-v_{ij}\right|}{\left|v_j^{\mp }-v_{ij}\right|+\rho \underset{1\le i\le m}{max}\underset{1\le j\le n}{max}\left|v_j^{\mp }-v_{ij}\right|}$$

(29)

where $v_{ij}$ is an element of the weighted decision matrix V and $v_{ij}^{\mp }$ is an element of the positive and negative ideal solutions.

4.4.3. Hybrid Sorting Method

Step 1. Calculate the hybrid distance $H_i^{\mp }$. The calculation of the hybrid distance employs an optimization model based on the principle of minimum information entropy, with the optimization objective function as follows:

$$\begin{array}{cc}\hfill min{Z}_i^{\mp }& =\sum _{i=1}^m\alpha ·H_i^{\mp }·ln\left(\frac{H_i^{\mp }}{nrm\left(D_i^{\mp }\right)}\right)\hfill \\ & +\sum _{i=1}^m(1-\alpha )·H_i^{\mp }·ln\left(\frac{H_i^{\mp }}{nrm\left(G_i^{\mp }\right)}\right)\hfill \end{array}$$

(30)

Subject to:

$${\sum }_{i=1}^m{H}_i^{\mp }=1,0\le H_i^{\mp }\le 1,\alpha =0.5_{\leftarrow }$$

By applying the Lagrange multiplier method, it can be obtained that:

$$H_i^{\mp }=\frac{{(G_i^{\mp }·D_i^{\mp })}^{1/2}}{{\sum }_{i=1}^n{\left(G_i^{\mp }·D_i^{\mp }\right)}^{1/2}}$$

(31)

Step 2. Calculate the relative hybrid closeness. According to the formula, calculate the relative hybrid closeness $Scor{e}_i$ of the i-th alternative solution to the ideal solution. The alternative solutions can then be ranked from highest to lowest based on this closeness, with a higher relative hybrid closeness indicating better adaptability:

$$Scor{e}_i=\frac{H_i^{-}}{H_i^{+}+H_i^{-}}$$

(32)

5. Example Analysis

To reflect service differentiation, this article selects four service scenarios with significantly different communication needs for communication technology adaptation, namely precise load control, electricity information collection, load management terminal monitoring, and electricity market trading. Based on the data center management system service and communication technology adaptation evaluation system constructed in this article, MatlabR2023b software (version number: 23.2) is used to implement adaptation algorithms, and remote and local communication technology adaptation decisions are made for four differentiated services. Finally, the communication technology adaptation results of the proposed services are analyzed and summarized.

5.1. Communication Technology Performance Index Data

Based on the adaptability evaluation system proposed in this paper, indexes such as latency, bandwidth, bit error rate, security, cost, and transmission distance are selected for adaptability analysis. This paper obtains remote and local communication technology performance index data based on reference [19,20,21], as shown in

Table 2. Remote and local communication technology performance index data.

	C1	C2	C3	C4	C5	C6
D1	1 Gbps	2 ms	${10}^{-10}$	100	100	15 km
D2	200 Mbps	60 ms	${10}^{-6}$	60	40	20 km
D3	100 Mbps	30 ms	${10}^{-7}$	100	60	15 km
D4	500 kbps	100 ms	${10}^{-5}$	80	60	10 km
D5	1 Mbps	5 ms	${10}^{-9}$	100	100	1200 m
D6	2 Mbps	10 ms	${10}^{-6}$	100	60	1600 m
D7	200 Mbps	15 ms	${10}^{-6}$	80	80	300 m
D8	50 kpbs	100 ms	${10}^{-3}$	100	40	2000 m
D9	20 Mbps	20 ms	${10}^{-6}$	80	60	100 m

. Notably, a hundred-point scale is used to quantify qualitative indexes such as security and cost.

5.2. Calculate the Weight of Differentiated Service Communication Requirement Indexes

Based on the FAHP method, referring to Table 1 and Formulas (1)–(8), experts scored to obtain the fuzzy judgment matrix. After checking for consistency, the subjective weights of communication indexes for the four services with differentiated communication requirements are calculated, with the results shown in

Table 3. Subjective weights of four communication demand indexes for power service.

	C1	C2	C3	C4	C5	C6
B1	0.0562	0.2857	0.2469	0.2793	0.0583	0.0736
B2	0.0653	0.1208	0.2969	0.1022	0.2205	0.1943
B3	0.2106	0.0885	0.1509	0.2763	0.1956	0.0781
B4	0.0643	0.2095	0.0956	0.2941	0.0972	0.2393

.

Based on the improved CRITIC method, utilizing Table 2 and Formulas (9)–(19), the original evaluation matrix is normalized. Subsequently, the standard deviation, correlation coefficient, and entropy value are calculated as the basis for assessing the conflict, volatility, and amount of information of the indexes, respectively. Finally, considering all three factors comprehensively, objective weights are assigned to each index. The objective weight vector is $\mathit{\epsilon }$ = (0.2356, 0.1308, 0.1336, 0.1421, 0.1855, 0.1724). According to Formula (20), the objective weights are combined with the subjective weights of the communication indexes for the four services to obtain the combined weights, as shown in

Table 4. Combined weights of the communication indexes for the four services.

	C1	C2	C3	C4	C5	C6
B1	0.0902	0.2546	0.2247	0.2704	0.0737	0.0864
B2	0.0963	0.0989	0.2483	0.0909	0.256	0.2096
B3	0.2912	0.0679	0.1184	0.2305	0.213	0.079
B4	0.0968	0.1752	0.0817	0.2672	0.1153	0.2638

.

According to Table 4, the top three indexes for combined weight ranking in the precision load control service are security, latency, and bit error rate. For the electricity information collection service, the top three indexes for combined weight ranking are bit error rate, cost, and transmission distance. In the load management terminal monitoring service, the top three indexes for combined weight ranking are bandwidth, cost, and security. For the power market transaction service, the top three indexes for combined weight ranking are security, transmission distance, and latency.

The precise load control service requires that communication content is protected from unauthorized access, tampering, or theft, ensuring that load control commands are transmitted swiftly and accurately. Consequently, the security, latency, and bit error rate of communication technologies are assigned higher priorities. As a basic service, the electricity information collection service necessitates the accuracy of collected information. The deployment of large-scale collection equipment increases the significance of bit error rate, transmission distance, and cost in the communication technology supporting this service. The load management terminal monitoring service involves transmitting large volumes of sensitive video and image information, imposing higher requirements on communication bandwidth and security. The deployment of high-performance communication infrastructure further highlights the significance of communication costs for this service. The power market transaction service encompasses participants distributed across various geographical locations, necessitating the transmission of substantial volumes of confidential transaction data. Additionally, it must respond to market changes in real time. To guarantee the efficient operation of the electricity market, communication technology must exhibit high security, extensive transmission distances, and minimal latency.

In summary, the weighted values derived from a combination of subjective and objective weighting are consistent with the communication requirements of differentiated services, validating the effectiveness of the proposed combined weighting method.

5.3. Differentiated Service and Communication Technology Adaptation

Based on the GTOPSIS method, an adaptability analysis of remote and local communication technologies for differentiated services is conducted. Taking the precise load control service as an example, according to Table 2, the initial decision matrix is formed using the performance index data of remote communication technologies, which have been preprocessed for bandwidth and bit error rate. The weighted normalized decision matrix is obtained using Formulas (22) and (23) and the combined weights for precise load control. According to Formulas (24)–(31), the hybrid distance of each remote communication technology to the positive and negative ideal solutions is calculated. A higher resulting relative hybrid closeness indicates better adaptability between the service and remote communication technology.

During the adaptation of local communication technologies, the original decision matrix is formed from the performance index data of local communication technologies, which has been preprocessed. By executing the same steps above, the local communication adaptation results of the precise load control service can be determined.

Figure 5 respectively illustrates the relative hybrid closeness results of adapting the four services to remote and local communication technologies, where B1 represents precise load control service, B2 represents electricity information collection service, B3 represents load management terminal monitoring service and B4 represents power market transaction service.

5.4. Adaptability Result Analysis

(1) For the precise load control service, the most suitable remote communication technology is fiber optic, which has good real-time performance, high reliability, and security, providing relatively safer and more reliable communication services for the precise load control service. The second choice is wireless private networks. Wireless public networks, due to their security, bit error rate, and latency, perform the worst among the four types of remote communication technologies, making them the least suitable for the precision load control service.

The optimal local communication technology is RS485, known for its strong anti-interference capability, and relatively high real-time performance, reliability, and security, making it meet the requirements of the precise load control service. The second choice is High-speed Power Line Carrier (HPLC). LoRa has the lowest adaptability to this service due to its poor performance in terms of latency, bit error rate, and security.

(2) For electricity information collection service, the most suitable remote communication technology is wireless public network, which is in line with the current situation where most collection services use wireless public network communication methods. Wireless public network has low cost, convenient deployment, fast transmission rate, and low bit error rate, which meets the communication needs of electricity information collection service. The second most suitable remote communication technology is wireless private network, which has higher security but also higher cost compared to wireless private network. With the deepening of data center management system construction, higher requirements are put forward for data transmission security. Wireless private networks can also provide a more secure communication channel for electricity information collection service. The remote communication technology with the lowest adaptability is fiber optic communication, which is not suitable for large-scale deployment due to its high construction and renovation costs.

The optimal local communication technology is HPLC, which, compared to narrowband power line communication, offers a wider transmission bandwidth and higher speed. It is less susceptible to interference and has a lower bit error rate. It can transmit over longer distances without the need for additional wiring, significantly reducing construction and maintenance costs. The second-best local communication technology is RS485. The least suitable is WiFi.

(3) For the load management terminal monitoring service, which requires the uploading of collected data and video information, there is a significant requirement for communication bandwidth. The optimal remote communication technology for this service is wireless private networks, which offer relatively superior performance in terms of bandwidth, security, and cost, ensuring the safe and rapid upload of video information and power data. Their capability to cover large transmission distances meets the wide-area coverage needs of the service. The second-best communication technology is wireless public networks. Their high bandwidth and low cost can also provide a relatively convenient and high-speed channel for the load management terminal monitoring service.

The optimal local communication technology for this service is WiFi, due to its large bandwidth, good real-time performance, high reliability, higher security, and low cost, meeting the performance requirements of the load management terminal monitoring service. The second-best local communication technology is Bluetooth. Although LoRa demonstrates good performance in terms of security and cost, its insufficient bandwidth makes it difficult to support this service.

(4) For the power market transaction service, the optimal remote communication technology is fiber optic, followed by wireless private networks. Both technologies exhibit superior performance in terms of security, latency, and transmission distance. In contrast, wireless public networks and MVPLC perform poorly in security and latency, making them less suitable for this service.

The optimal local communication technology for this service is HPLC, followed by RS485. Both provide communication channels with high reliability, low latency, long-distance transmission, and high security, which can meet the communication requirements of the power market transaction service. However, local communication technologies such as LoRa, Bluetooth, and WiFi lack comprehensive performance in terms of latency, security, and transmission distance, resulting in lower adaptability to this service.

This section conducts a study on the adaptability of communication technology through four differentiated services, and obtains that the order of communication technology adaptation fully accords with the current situation of data center management communication networking, confirming the feasibility of the communication technology adaptation algorithm in this article.

6. Conclusions

This article proposes a multi-domain communication architecture for data center management systems, which includes local and remote communication, to address the issues of single communication methods and poor adaptability between service and communication technologies. On the basis of the suitability evaluation system for the data center management system service and communication technology, the subjective and objective combination weighting method proposed in this article is used to weight the evaluation system indexes. Finally, based on the GRA-TOPSIS method, an adaptability analysis was conducted between differentiated services and local and remote communication technologies. The experimental results indicate that the architecture and adaptation algorithm proposed in this article provide a theoretical basis and solution for the selection of multi-domain communication technologies for a data center management system service. As data centers play an increasingly important role in energy management and other critical infrastructure, the security and privacy protection of their communication networks become particularly crucial. Future research can explore ways to enhance the security of existing architectures, including strategies to combat cyberattacks and data breaches.