General inductance computing resource deployment method based on multi-population differential evolution algorithm

Abstract

The invention discloses a universal computing resource deployment method based on a multi-population differential evolution algorithm, and belongs to the technical field of universal computing resource deployment. Comprising the following steps: S101, analyzing power business scene requirements and general inductance network resource performance indexes, and establishing an evaluation index system; s102, combining a BWM method with an entropy weight method, and performing adaptive analysis on the power business demands and the general inductance computing resources to obtain demand weights of different power businesses on the general inductance computing resources; s103, according to an adaptive analysis result of the power business to the general inductance computing resources, establishing a general inductance computing resource deployment multi-dimensional optimization problem model; and S104, adopting a multi-population differential evolution algorithm to calculate a universal computing resource deployment multi-dimensional optimization problem model. Compared with other resource deployment optimization schemes, the method of the invention adopts a scheme of combining power business suitability analysis and a general inductance calculation resource deployment optimization problem, and is more in line with the actual demand of power business, and the proposed optimization problem is more reasonable.

Description

Technical Field

[0001] The present invention relates to the technical field of communication-sensing-computing resource deployment, and particularly to a communication-sensing-computing resource deployment method based on a multi-population differential evolution algorithm. Background Art

[0002] Under the development and integration trend of power industry technology and Internet of Things technology, the construction of a new power system has put forward new requirements for the comprehensive performance of communication-sensing-computing networks. There are different power services in different network levels such as substations, units, and grids. Aiming at the service requirements in these different network level scenarios, coordinating service quality and resource costs, and providing appropriate resource deployment solutions for communication, sensing, and computing to meet the requirements of real-time sensing, high-speed transmission, and accurate calculation of data in different service scenarios and achieving the coordination of service quality and cost is crucial.

[0003] Currently, most research on resource deployment focuses on edge computing technology, lacking research on the deployment of communication and sensing resources; some research only considers the optimization problem of a single attribute, with poor optimization effects, such as user experience or resource deployment cost, and at the same time lacks a resource deployment method that combines the actual needs of power services. Therefore, a communication-sensing-computing resource deployment method based on a multi-population differential evolution algorithm is needed to better meet the actual needs of power services. Summary of the Invention

[0004] The object of the present invention is to propose a communication-sensing-computing resource deployment method based on a multi-population differential evolution algorithm, including the following steps:

[0005] S101: Analyze the requirements of power service scenarios and the performance indicators of communication-sensing-computing network resources, and establish an evaluation index system;

[0006] S102: Combine the BWM method and the entropy weight method to perform an adaptation analysis on power service requirements and communication-sensing-computing resources, and obtain the demand weights of different power services for communication-sensing-computing resources;

[0007] S103: According to the adaptation analysis results of power services for communication-sensing-computing resources, establish a multi-dimensional optimization problem model for communication-sensing-computing resource deployment;

[0008] S104: Use the multi-population differential evolution algorithm to calculate the multi-dimensional optimization problem model of communication-sensing-computing resource deployment.

[0009] Further, the resource domain indicators of the evaluation index system in S101 include communication resources, sensing resources, and computing resources.

[0010] Further, S102 specifically includes:

[0011] According to the three indicators X = (X1, X2, X3) of the S101 resource domain, where X1, X2, and X3 are communication resources, sensing resources, and computing resources respectively, the expert selects the optimal indicator X with the strongest importance in the power business B and the worst indicator X with the weakest importance W ;

[0012] Use a 1-9 scoring system to score the remaining indicators except the optimal indicator X B and the worst indicator X W . Compare the remaining indicators with the optimal indicator pairwise to obtain the optimal comparison vector A B = (a B,1 , a B,2 , a B,3 ), compare the remaining indicators with the worst indicator pairwise to obtain the worst comparison vector A W = (a 1,W , a 2,W , a 3,W );

[0013] Solve for the optimal subjective weight values of each indicator using the following constrained optimization model:

[0014]

[0015] Among them: is the weight of the indicator X j ; a B,j is the importance value of X B to X j ; a j,W is the importance value of X j to X W ;

[0016] Select communication-sensing-computing resource devices in different price ranges to construct the performance-cost matrix of the indicators, and use the following normalization method to normalize the values of the communication-sensing-computing resources:

[0017]

[0018] Among them, r ij is the target variable, that is, the cost of the communication-sensing-computing resources; then calculate the proportion of the communication-sensing-computing resource indicators in the total resource indicators in different price ranges:

[0019]

[0020] Calculate the entropy value of different indicators j:

[0021]

[0022] Among them, j = 1, 2, 3 respectively represent communication, sensing, and computing resources, where n is the number of different price ranges investigated; calculate the entropy weight ω of different metrics j j :

[0023]

[0024] Among them, m is the number of metrics.

[0025] Furthermore, S103 includes:

[0026] Obtain that the total amount of power services existing in the substation area is ρ total , which includes the service volumes of video, collection, and control types being ω A ρ total , ω B ρ total , ω C ρ total Among them, ω A +ω B +ω C = 1. Considering comprehensively the processing time of the services and the cost incurred in deploying communication, sensing, and computing resources, the following expression for the effectiveness of the communication, sensing, and computing resource deployment plan is obtained:

[0027]

[0028] Among them, M = 1, 2, 3 respectively represent power services such as video, collection, and control types, and N = 1, 2, 3 respectively represent communication, sensing, and computing resources, is the result of the adaptability analysis weight of the nth resource for the kth service, and cost n,k is the cost of resource deployment. The specific formula is as follows:

[0029] cost n,k = α n,k ·cost n

[0030] Among them, α n,k is the number of unit communication, sensing, and computing resources deployed in the substation area, and cost n is the cost required for unit communication, sensing, and computing resources. time n,k is the processing time of the service. The specific formula is as follows:

[0031]

[0032] Among them, β i is the initial communication, sensing, and computing resources, and β n is the service processing capacity that unit communication, sensing, and computing resources can provide;

[0033] Obtain the multi-dimensional optimization problem model of the integrated sensing and computing resource deployment from the effectiveness expression of the integrated sensing and computing resource deployment solution:

[0034]

[0035]

[0036] where cost max is the maximum affordable cost limit, and (α n ) max is the maximum allowable integrated sensing and computing resource limit.

[0037] Furthermore, the multi-population differential evolution algorithm in S104 is as follows:

[0038] 1) Initialization:

[0039] Set the initial population and the population size I, and each population contains x individuals where t represents the population at the t-th iteration, and the initial iteration number t = 0; i represents the population number; is the j-th gene of the x-th individual in the i-th population, representing the integrated sensing and computing resources deployed for the k-th service, as shown specifically below

[0040]

[0041] where N is the integrated sensing and computing resources, and M is the type of service; each individual in the population represents an integrated sensing and computing deployment solution, and integrated sensing and computing resources are separately arranged for each different power service existing in the substation area. Each individual has N*M genes;

[0042] 2) Set the fitness function:

[0043] Analyze the quality of all individuals in the population, eliminate the individual with the worst fitness, and set the fitness function as The optimization goal is to minimize E. Therefore, the smaller the fitness function of an individual, the better it is. The individual with the largest fitness function should be eliminated, that is, the worst integrated sensing and computing deployment solution;

[0044] 3) Differential mutation operation:

[0045] Randomly select two individuals in population i, perform vector difference operation, and then perform vector sum operation with the third individual to finally obtain the mutation intermediate variable The specific process is as follows:

[0046]

[0047] where x1, x2, x3 ∈ {1, 2, …, N*M}, and are three different individuals in population i; F is the mutation scaling factor, which is used to regulate the weight of the difference vector.

[0048] 4) Crossover operation:

[0049] The individual and the mutant intermediate vector replace the genes at the same positions with a certain probability to generate a new trial individual The genes of the trial individual come from the individual and the mutant intermediate vector Specifically, it is selected according to the crossover probability factor δ; to ensure that the individual can evolve in the next generation, it is required that the trial individual has at least one gene from the mutant intermediate vector For this purpose, a random integer variable jrand ∈ {1, 2,..., N*M} is generated; if j = jrand, the j-th gene of the trial individual will adopt the j-th gene of the mutant intermediate vector The specific method is as follows:

[0050]

[0051]

[0052]

[0053]

[0054] 5) Selection operation:

[0053] After mutation and crossover, the trial individuals The trial individual and the individual will be selected through survival of the fittest; the fitness of different individuals is evaluated through the fitness function. In the minimization optimization problem, the individual with a smaller fitness function value is selected as the next generation; the specific method is as follows:

[0054]

[0055] 6) Immigration operation:

[0056] Population interaction is connected through the immigration operator, and the immigration operator transfers the optimal individuals of each population Regularly introduce into other populations to replace the worst individuals in other populations Realize information exchange between populations; the specific method is as follows:

[0057]

[0058]

[0059] Where I is the number of populations; the immigration operation largely avoids the local optimization problem, making the final solution of the communication-sensing-computation deployment plan obtained by terminating the loop closest to the optimal deployment plan to the greatest extent;

[0060] 7) Artificial selection operation:

[0061] After each generation of evolution, select the optimal individual of the population through the artificial selection operator and put it into the elite population for preservation; during the population evolution process, the elite population does not participate in the evolution; the elite population stores the optimal individuals in each population from the initial moment to the iteration number t, that is, the optimal communication-sensing-computation resource deployment plan. Only when a better individual appears in each population will the elite population change;

[0062] 8) Termination operation:

[0063] In multiple rounds of evolution operations, the elite population constantly changes. If there are T max rounds of evolution during which the elite population remains unchanged, then terminate the algorithm and output the optimal individual in the current elite population as the optimal solution to this optimization problem, that is, the optimal communication-sensing-computation resource deployment plan.

[0064] The beneficial effects of the present invention are as follows:

[0065] 1. The method of the present invention adopts a scheme that combines power service adaptability analysis with the optimization problem of communication-sensing-computation resource deployment. Compared with other resource deployment optimization schemes, it is more in line with the actual needs of power services, and the proposed optimization problem is more reasonable.

[0066] 2. In terms of adaptability analysis, the method of combining weights of the BWM method and the improved entropy weight method is adopted, which can highlight the variability of each index in terms of cost, and at the same time has wide applicability for promotion to other objectives.

[0067] 3. In solving the optimization problem, by integrating the actual power service and the adaptation weight results of communication-sensing-computation resources, the multi-population differential evolution algorithm is optimized, solving the problem of poor optimization effect caused by a single attribute, making sensing more real-time, transmission faster, and calculation more accurate. Brief Description of the Drawings

[0068] Figure 1 It is a flowchart of the communication-sensing-computation resource deployment method based on the multi-population differential evolution algorithm of the present invention;

[0069] Figure 2 It is a multi - dimensional index system diagram for evaluating the differentiated power service and the requirements of communication, sensing, and computing resources;

[0070] Figure 3 It is a flow chart of the optimized multi - population differential evolution algorithm. Specific implementation manner

[0071] The present invention proposes a method for deploying communication, sensing, and computing resources based on a multi - population differential evolution algorithm. The following further explains the present invention with reference to the accompanying drawings and specific embodiments.

[0072] Figure 1 It is a flow chart of the method for deploying communication, sensing, and computing resources based on the multi - population differential evolution algorithm of the present invention, which is specifically as follows:

[0073] S101: Conduct a differentiated analysis of communication performance requirements for power services such as video, acquisition, and control. Different resources each have corresponding indicators. For example, communication resource indicators: communication bandwidth, delay; sensing resource indicators: number of sensing connections, sensing coverage radius; computing resource indicators: computing storage space, computing speed. However, some indicators may cover multiple resources. For example, bandwidth can be an indicator of both communication resources and sensing resources. Therefore, how to independently and uniformly describe the size of each resource is particularly crucial. Because no matter what kind of device, energy is required as the driving force for its operation, and generally, the greater the energy consumed, the greater its efficiency. So here we consider the resource occupancy as an indicator to uniformly describe the size of each resource.

[0074] In order to measure the relationship between communication, sensing, computing domains and cost indicators, conduct research on relevant devices on the market, select typical communication, sensing, and computing devices, and investigate their various performance index parameters and costs. Determine the evaluation system for the adaptability assessment of communication, sensing, and computing resources and differentiated power services, and determine various evaluation indicators in the differentiated power service scenario. Its system is as Figure 2 , starting from two aspects: the resource domain and the cost domain, and further refining each indicator. Among them, the resource domain indicators are: communication resources, sensing resources, and computing resources.

[0075] S102: Based on the multi - dimensional index evaluation system, use a method combining the BWM method and the entropy weight method to conduct an adaptability analysis of power service requirements and communication, sensing, and computing resources, and obtain the demand weights of different power services for communication, sensing, and computing resources. The specific steps are as follows:

[0076] First, for the three indicators in the resource domain of the multi - dimensional index evaluation system proposed in step S101: X=(X1, X2, X3), the expert selects the optimal indicator X with the strongest importance in the power service Band the worst indicator X with the weakest importance W . Use a 1-9 scoring system to score the remaining indicators (indicators other than the optimal and worst indicators) respectively. Among them, the remaining indicators are compared pairwise with the optimal indicator to obtain the optimal comparison vector A B =(a B,1 , a B,2 , a B,3 ). Then, the optimal BWM weight values of each indicator are solved by the following constrained optimization model:

[0077]

[0078] Among them: is the weight of indicator X j ; a B,j is the importance value of X B to X j ; a j,W is the importance value of X j to X W . The final BWM weight calculation results are shown in Table 1:

[0079] Table 1 Demand resource weights

[0080]

[0081] Among them, video services need to perform certain compression processing on data, so they have relatively large demands for computing resources and sensing resources, and relatively small preference demands for communication resources; acquisition services need to sense a large number of sensors to collect data and centrally process the collected data at the same time, so they have the largest preference demand for sensing resources, followed by the demand for computing resources, and the smallest demand for communication resources; control services are very sensitive to latency and need to ensure high-speed and reliable data transmission, so they have very large demands for communication resources, but almost no demands for sensing resources and computing resources.

[0082] Then, the objective weight is calculated using the entropy weight method based on the target: First, conduct a full investigation of communication, sensing, and computing resource devices on the market, select communication, sensing, and computing resource devices in different price ranges to construct the performance-cost matrix of the indicators. To simplify the problem and calculation, the performance of different resources is represented by a unified power in the example. Then, the following normalization method is used to normalize the values of communication, sensing, and computing resources:

[0083]

[0084] Among them, r ijis the target variable, which is the cost of the communication-sensing-computing resources here. Then, calculate the proportion of the communication-sensing-computing resource indicators in the total resource indicators within different price ranges, as follows:

[0085]

[0086] Calculate the entropy value of different indicators j (j = 1, 2, 3 represent communication, sensing, and computing resources respectively, where n is the number of different price ranges investigated and can be changed according to the actual situation), as follows:

[0087]

[0088] Calculate the entropy weight ω of different indicators j j (where m is the number of indicators), as follows:

[0089]

[0090] The final calculation results are shown in Table 2:

[0091] Table 2 Objective weights

[0092] Normalized value Normalized value Normalized value Normalized value Entropy weight value Communication resources 0 0.5321 1 0.4398 0.3412 Sensing resources 1 0.7523 0.5649 0 0.3030 Computing resources 0.3456 1 0 0.8322 0.3559

[0093] Among them, the computing resources have the largest weight because the computing resources on the market, such as CPUs and GPUs, have relatively high prices, and the price changes of devices with different brands and computing powers have a large span, so their entropy values are relatively high, and thus they have a large weight.

[0094] S103: According to the above analysis results of the adaptation of power services and communication-sensing-computing resources, model the resource deployment problem. Here, in order to simplify the problem, we only consider modeling the deployment of communication-sensing-computing resources for a single substation area. The method of extending to multiple substation areas is similar. First, the total amount of power services existing in this substation area is ρ total , which includes the service volumes of video, collection, and control types, namely ω A ρ total , ω B ρ total , ω C ρ total , where ω A +ω B +ω C = 1, and the specific values depend on the actual situation of the substation area and are randomly taken here. Considering the processing time of the services and the cost of deploying communication-sensing-computing resources, the following expression for the effectiveness of the communication-sensing-computing resource deployment plan can be obtained:

[0095]

[0096] where k = 1, 2, 3 represent power services such as video, acquisition, and control, respectively, and n = 1, 2, 3 represent communication, sensing, and computing resources, respectively. is the adaptability analysis weight result of the nth resource for the kth service, cost n,k is the cost of resource deployment, and the specific formula is as follows:

[0097] cost n,k = α n,k ·cost n

[0098] where α n,k is the number of unit communication, sensing, and computing resources deployed in the substation area, and cost n is the cost required for unit communication, sensing, and computing resources. time n,k is the processing time of the service, and the specific formula is as follows:

[0099]

[0100] where β i is the initial communication, sensing, and computing resources, and β n is the service processing capacity that unit communication, sensing, and computing resources can provide. The more communication, sensing, and computing resources are deployed, the faster the service processing speed, but the cost will also increase accordingly. The service processing speed and cost restrict each other, and the following communication, sensing, and computing resource optimization problem can be obtained:

[0101]

[0102] where cost max is the maximum acceptable cost limit, and (α n ) max is the maximum allowable communication, sensing, and computing resource limit.

[0103] S104: Use the multiple population differential evolution algorithm (MPDEA) to solve the optimization problem proposed in step S103. The algorithm flow is as Figure 3 shown. Compared with the standard genetic algorithm, MPDEA has the characteristics of good stability and fast convergence speed, and it is more difficult to fall into the local optimal trap. The specific steps of MPDEA are as follows:

[0104] 1) Initialize the algorithm:

[0105] Set the initial population and the population size I, and each population contains x individuals where t represents the population at the tth iteration, and the initial iteration number t = 0; i represents the population number. It is the j-th gene of the individual numbered x in the i-th population, representing the integrated sensing, communication, and computing resources for the k-th service deployment, as specifically shown below.

[0106]

[0107] Among them, N is the integrated sensing, communication, and computing resources, and M is the service types. Each individual in the population represents an integrated sensing, communication, and computing deployment plan. Here, integrated sensing, communication, and computing resources are separately arranged for each different power service existing in the transformer substation area. Therefore, each individual has N*M genes.

[0108] 2) Set the fitness function:

[0109] The fitness function determines the evolution direction of the population, analyzes the quality of all individuals in the population, and eliminates the individual with the worst fitness. According to the optimization problem in step S103, we set the fitness function as The optimization goal is to minimize E. Therefore, the individual with a smaller fitness function is better, and the individual with the largest fitness function (i.e., the worst integrated sensing, communication, and computing deployment plan) should be eliminated.

[0110] 3) Differential mutation operation:

[0111] Randomly select two individuals in population i, perform vector difference operation, and then perform vector sum operation with the third individual to finally obtain the mutation intermediate variable. The specific process is as follows:

[0112]

[0113] where x1, x2, x3 ∈ {1, 2,..., N*M}, and are three different individuals in population i; F is the mutation scaling factor used to regulate the weight of the difference vector. This mutation intermediate variable is used to generate a new integrated sensing, communication, and computing deployment plan, and the specific operation is shown in step 4.

[0114] 4) Crossover operation:

[0115] Replace the genes at the same positions in the individual and the mutation intermediate vector with a certain probability to generate a new trial individual The genes of the trial individual come from the individual and the mutation intermediate vector The specific selection is based on the crossover probability factor δ. To ensure that the individual can evolve in the next generation, it is required that the trial individual has at least one gene from the mutation intermediate vector To this end, a random integer variable jrand∈{1,2,...,N*M} is generated. If j=jrand, then the experimental individual The jth gene The mutated intermediate vector The jth gene The specific method is as follows:

[0116]

[0117] Among them, rand(j)∈[0,1], if the random number rand(j) corresponding to the j-th gene does not exceed the crossover probability factor δ, then the test individual The jth gene Derived from the mutated intermediate vector The jth gene Otherwise, the jth gene Will come from individuals The jth gene

[0118] 5) Select an operation:

[0119] Generate test individuals after mutation and crossover Test individuals and individuals The survival of the fittest will be carried out. The fitness of different individuals is evaluated through the fitness function. In the minimization optimization problem, the individuals with smaller fitness function values are selected as the next generation. The specific method is as follows:

[0120]

[0121] 6) Immigration operations:

[0122] Population interactions are linked through immigration operators. Immigration operators transfer the best (smallest effect) individuals from each population to Regularly introduced into other populations to replace the worst individuals in other populations Realize information exchange between populations. The specific method is as follows:

[0123]

[0124] Where I is the population size. The migration operation can avoid local optimization problems to a large extent, so that the synaesthesia deployment plan obtained by terminating the cycle is close to the optimal deployment plan to the greatest extent.

[0125] 7) Manual selection operation:

[0126] After each generation of evolution, the optimal individual of the population is selected through the artificial selection operator and placed in the elite population for preservation. During the population evolution process, the elite population does not participate in the evolution. The elite population stores the optimal individuals in each population from the initial moment to the iteration number t, that is, the optimal communication-sensing-computation resource deployment scheme. Only when a better individual appears in each population will the elite population change.

[0127] 8) Termination operation:

[0128] In multiple rounds of evolution operations, the elite population keeps changing. If there are T max rounds of evolution during which the elite population remains unchanged, the algorithm is terminated, and the optimal individual in the current elite population is output as the optimal solution to the optimization problem, that is, the optimal communication-sensing-computation resource deployment scheme.

[0129] The final simulation output of the communication-sensing-computation resource deployment scheme results is shown in Table 3:

[0130] Table 3 Final deployment scheme

[0131] Communication resources Sensing resources Computing resources Video category 0.7313 4.5084 4.0832 Collection category 0.4270 3.9271 1.4474 Scheduling category 7.5631 0 0

[0132] The simulation-related parameters are shown in Table 4:

[0133] Table 4 Related simulation parameters

[0134] Total business volume 100 Proportion of video category services 0.5 Proportion of collection category services 0.25 Proportion of scheduling category services 0.25 Maximum economic limit 100 Economic cost of resources 1 Business capacity of resources 1 Default initial existing resources 1

[0135] In summary, using the above simulation parameters and combining the method of the present invention to adopt a scheme that combines power service adaptability analysis with the communication-sensing-computation resource deployment optimization problem, compared with other resource deployment optimization schemes, it is more in line with the actual needs of power services, and the proposed optimization problem is more reasonable.

Claims

1. A method for deploying communication and sensing computing resources based on a multi-population differential evolution algorithm, characterized in that It includes the following steps: S101: Analyze the requirements of power business scenarios and the performance indicators of communication, sensing, and computing network resources, and establish an evaluation index system; S102: Combine the BWM method with the entropy weight method to perform an adaptation analysis of power business requirements and communication, sensing, and computing resources, and obtain the demand weights of different power businesses for communication, sensing, and computing resources; S103: Establish a multi-dimensional optimization problem model for the deployment of communication, sensing, and computing resources according to the adaptation analysis results of power businesses for communication, sensing, and computing resources; S104: Use the multi-population differential evolution algorithm to calculate the multi-dimensional optimization problem model for the deployment of communication, sensing, and computing resources.

2. The method for deploying communication and sensing computing resources based on the multi-population differential evolution algorithm according to claim 1, wherein The resource domain indicators of the evaluation index system in S101 include communication resources, sensing resources, and computing resources.

3. The method for deploying communication and sensing computing resources based on the multi-population differential evolution algorithm according to claim 2, wherein S102 specifically includes: According to the three indicators \(X=(X_1,X_2,X_3)\) of the S101 resource domain, where \(X_1\), \(X_2\), and \(X_3\) are communication resources, sensing resources, and computing resources respectively, the optimal indicator \(X\) with the strongest importance and the worst indicator \(X\) with the weakest importance in the power business are selected by experts. B and the worst indicator \(X\) with the weakest importance W ; Use a 1-9 scoring system to score the remaining indicators except for the optimal indicator X B and the worst indicator X W . Make pairwise comparisons between the remaining indicators and the optimal indicator to obtain the optimal comparison vector A B =(a B,1 , a B,2 , a B,3 ). Make pairwise comparisons between the remaining indicators and the worst indicator to obtain the worst comparison vector A W =(a 1,W , a 2,W , a 3,W ); Solve the optimal subjective weight value of each index by the following constraint optimization solution model: Wherein: is the weight of index X j ; a B,j is the importance value of X B to X j ; a j,W is the importance value of X j to X W ; Select communication, sensing, and computing resource devices in different price ranges to construct a performance-cost matrix of the index, and use the following normalization method to normalize the values of communication, sensing, and computing resources: where r ij is the target variable, i.e., the cost of the joint communication and sensing computing resources; then calculate the proportion of the joint communication and sensing computing resource metrics in the overall resource metrics within different price ranges: Calculate the entropy value of different indexes j: where j = 1, 2, 3 represent communication, sensing, and computing resources respectively, and n is the number of different price ranges investigated; Calculate the entropy weight ω of different indicators j j : where m is the number of indexes.

4. The method for deploying communication sensing computing resources based on the multi-population differential evolution algorithm according to claim 3, characterized in that S103 includes: Obtain the total amount of power services existing in the substation area as ρ total , among which the traffic volumes of video, acquisition, and control services are ω A ρ total , ω B ρ total , ω C ρ total , among which ω A +ω B +ω C =1. Considering comprehensively the processing time of services and the cost of deploying communication, sensing, and computing resources, the following expression for the effectiveness of the communication, sensing, and computing resource deployment plan is obtained: Where M = 1, 2, 3 represent power services such as video, acquisition, and control, respectively, and N = 1, 2, 3 represent communication, sensing, and computing resources, respectively. is the adaptability analysis weight result of the nth resource for the kth service, cost n,k is the cost of resource deployment, and the specific formula is as follows: cost n,k = α n,k · cost n where α n,k is the number of unit communication, sensing, and computing resources deployed in the substation area, and cost n is the cost required for unit communication, sensing, and computing resources, and time n,k is the processing time of the service. The specific formula is as follows: where β i is the initial synesthesia computing resource, and β n is the service processing capability that can be provided by the unit synesthesia computing resource; Obtain the multi-dimensional optimization problem model for the deployment of communication, sensing, and computing resources from the effectiveness expression of the communication, sensing, and computing resource deployment plan; where cost max is the maximum affordable cost limit, (α n ) max is the maximum available communication and sensing resource limit.

5. The method for deploying communication and sensing computing resources based on the multi-population differential evolution algorithm according to claim 4, wherein The multi-population differential evolution algorithm in S104 is specifically as follows: 1) Initialization: Set the initial population and the population size I, where each population contains x individuals where t represents the population at the t-th iteration, and the initial iteration number t = 0; i represents the population number; is the j-th gene of the individual numbered x in the i-th population, representing the integrated sensing and computing resources for the k-th service deployment, as shown below where N is the communication, sensing, and computing resources, and M is the type of business; each individual in the population represents a communication, sensing, and computing deployment plan, and communication, sensing, and computing resources are arranged separately for each different power business existing in the substation area. Each individual has N * M genes; 2) Set the fitness function: Analyze the quality of all individuals in the population, eliminate the individual with the worst fitness, and set the fitness function as The optimization goal is to minimize E. Therefore, the individual with a smaller fitness function is better, and the individual with the largest fitness function, that is, the worst tactile communication computing deployment scheme, should be eliminated. 3) Differential mutation operation: Randomly select two individuals from population i, perform a vector difference operation, and then perform a vector sum operation with a third individual to finally obtain a mutation intermediate variable The specific process is as follows: where \(x_1, x_2, x_3\in\{1, 2, \ldots, N\times M\}\), and are three different individuals in population \(i\); \(F\) is the mutation scaling factor, which is used to regulate the weight of the difference vector; 4) Crossover operation: An individual is replaced with genes at the same positions in the mutant intermediate vector with a certain probability to generate a new trial individual The genes of the trial individual are derived from the individual and the mutant intermediate vector Specifically, the selection is based on the crossover probability factor δ; to ensure the evolution of individuals in the next generation, it is required that the trial individual has at least one gene from the mutant intermediate vector To this end, a random integer variable jrand ∈ {1, 2, …, N*M} is generated; If j = jrand, then the j-th gene of the trial individual will be mutated using the j-th gene of the intermediate mutation vector The specific method is as follows: ​ where rand(j) ∈ [0, 1]. If the random number rand(j) corresponding to the j-th gene does not exceed the crossover probability factor δ, then the j-th gene of the test individual is derived from the j-th gene of the mutation intermediate vector Otherwise, the j-th gene will be the j-th gene from the individual ​ 5) Selection operation: Generate trial individuals after mutation and crossover Trial individuals And individuals Will undergo survival of the fittest; evaluate the fitness of different individuals through the fitness function, and in the minimization optimization problem, select the individuals with smaller fitness function values as the next generation; the specific method is as follows: 6) Immigration operation: The population interaction is connected by the immigration operator, which introduces the optimal individuals of each population into other populations regularly to replace the worst individuals in other populations to achieve information exchange between populations; the specific method is as follows: where I is the number of populations; the immigration operation largely avoids the local optimization problem, so that the communication, sensing, and computing deployment plan obtained by finally terminating the loop is closest to the optimal deployment plan to the greatest extent; 7) Artificial selection operation: After each generation of evolution, select the optimal individual of the population through the artificial selection operator and put it into the elite population for preservation; during the population evolution process, the elite population does not participate in the evolution; the elite population stores the optimal individuals in each population from the initial moment to the iteration number t, that is, the optimal communication, sensing, and computing resource deployment plan. The elite population will only change when a better individual appears in each population; 8) Termination operation: In multiple rounds of evolutionary operations, the elite population constantly changes. If there are T max rounds of evolution during which the elite population remains unchanged, the algorithm is terminated, and the optimal individual in the current elite population is output as the optimal solution to the optimization problem, that is, the optimal solution for the deployment of cross-sensory computing resources.

Images