Bio-Inspired Algorithms and ApplicationsView this Special Issue
ODL Centralized Control of Power Communication Network Based on Bio-Inspired Algorithms and SDN
An ODL centralized control strategy is designed to study the problem of heavy load in the power communication network, in which the machine learning method is introduced. SDN technology is used to establish SDN cluster control structure, and different algorithms, such as genetic algorithm, are utilized to optimize resource scheduling. The results show that the improved algorithm obtains the shortest link path through 28 iterations. At the same time, AHP is used to switch the network spectrum. Moreover, the application effect of the control strategy is simulated and analyzed, and its effect on network communication application is verified.
1. Related Work
Technologies such as cloud computing and big data continue to develop but, at the same time, begin to penetrate into traditional industries. Among them, the construction and management of power grid began to adopt intelligent new technology, and the construction of smart power grid has become an inevitable trend. However, with the growth of the power communication network business, the network load increases gradually. In addition, the problems of resource waste and low compatibility in the traditional electric power communication system affect the quality of electric power service. Therefore, people propose to introduce SDN into network deployment. Software-defined networking (SDN) is also gaining popularity in power systems . It is a new software-defined network model, which separates control and data forwarding functions, so as to realize centralized operation and maintenance control of power communication network resources. At the same time, its programmable advantage can achieve a wide range of power communication business expansion. Standard southbound interface and virtual network layer can promote the close integration of new power equipment and control network. In a word, it liberates a lot of manpower, avoids many configuration failures, and is easy to deploy uniformly and quickly, so as to make up for the defects of traditional communication network single decentralized control, which has been successfully applied in some data centers and cloud computing networks [2, 3]. Therefore, the centralized management and control architecture of distribution communication network based on SDN is designed, and the realization scheme of various control functions is studied, so as to provide a theoretical basis and demonstration reference for the construction of a new generation of power communication network, in which resources are uniformly controlled, network and data are coscheduled, and business needs are rapidly answered. At present, there are many researches on SDN deployment. For example, in April 2014, Beijing Telecom completed the commercial deployment of SDN in cooperation with Huawei, successfully applied SDN technology to IDC (Internet data center) network, and released a series of new IDC businesses based on SDN [4, 5]. In 2017, the ZENIC SDON innovation scheme of ZTE applied SDN technology to the WDM/OTN network, which can provide BoD, PoD, and OoD services on demand and support rapid deployment of multilevel services including L3/L2/L1/L0.
At the same time, aiming at the optimization problem of the power network, some scholars have proposed a large number of artificial intelligence algorithms. For example, Liu applied the genetic algorithm to network QoS optimization and obtained the scheduling sequence with the shortest task span and average task calculation time, which greatly improved the performance of the network ; Wang et al. put forward the bandwidth and capacity optimization method of elastic optical networks (eon) based on mixed-integer linear programming model . The results show that this method can not only improve the overall network capacity but also minimize the number of allocated time slots and service request blocking rate; Rongheng Lin and others proposed a bat SDN network scheduling method, which transforms the resource reservation problem into a multiknapsack problem ; Liang et al. used ant colony algorithm to schedule network resources. Experiments show that this method can effectively reduce the time of network resource scheduling ; Lalitha Devi et al. and Yi et al. proposed a joint beam and power scheduling scheme and solved the scheduling scheme by Bayesian algorithm [10, 11]; Praveena and Vijayarajan proposed that the energy consumption of mobile cloud network through the neural network is the lowest . The results show that through optimization, the power consumption is reduced by 53.68%; Faroqal Tam and others introduced the deep reinforcement learning method to schedule the resources of a 5 g MAC layer wireless network . The results show that this method has a positive impact on network resource scheduling. Aiming at the resource control problem in the SDN network, this study proposes a centralized control scheme based on ODL, so as to better promote the scheduling and effective utilization of SDN network resources. The innovation of this research is that combined with the ODL centralized control method, two methods are proposed for optimization, and a variety of machine learning algorithms are introduced for optimization, so as to greatly improve the efficiency of ODL centralized control and provide a reference for the multialgorithm application of ODL control.
2. The ODL Centralized Control Framework in This Paper
ODL is generally divided into the controller, network application layer, and north-south interface, and each part has different functions. The southbound protocol layer is a module that provides support for different protocols of the device, such as LISP. The northbound interface layer provides a standard set of Rest interfaces to external applications while supporting the cluster working mode. The ODL is the core of SDN control. Among them, the controller supports link management, bandwidth control, application development, and other functions, which is essentially a platform with a high open degree. The network application layer can be used to obtain network information, so as to achieve effective control of network behavior, which has high flexibility. For the north-south interface, the southbound interface is related to format conversion and data transmission, which supports different types of protocols. And northbound interface adopts RESTful protocol to expand interactive interfaces [14–17]. The specific structure is shown in Figure 1.
In order to realize the virtualization expansion of the network, the open switch software is introduced in this design, which is convenient to control virtual machines and traffic effectively.
3. Research on Bandwidth Resource Allocation Combining Genetic and Ant Colony Algorithm
Combined with the background of the ODL overall control framework in Figure 1, this study mainly studies and analyzes from the perspective of spectrum resource allocation and spectrum switching [18–20].
3.1. Power Distribution Communication Service Spectrum Arrangement Based on Genetic Algorithm
With the continuous development of the power distribution business, various types of services have also begun to appear, showing a general trend of diversified development, which further optimizes the network of Figure 2 and strengthens the scheduling of spectrum resources of SDN networks. There are also certain differences in communication nodes for the management of various business facilities, which put forward different requirements for the communication link spectrum. In this kind of business, communication spectrum needs to be allocated, which is studied deeply. Comparing different algorithms, and after analysis, a genetic algorithm is used in this study. The user ids are 1, 2, 3, ..., N, and the L matrix is utilized to select the user’s free spectrum, and then, all are matched one by one with users. It is known that users 1 and 2 are matched with spectrum B and C, respectively, which are represented as and in turn. According to this matching method, chromosome can be obtained. If the evolution rate needs to be compared, can be obtained in binary form.
Chromosomes that meet the requirements can be obtained by random matching of spectrum resources. Then, the corresponding chromosomes are obtained on the basis of the reasonable grouping of genes. For reasonable settings of algorithm parameters, the population size is set to 50.
The basic process of crossing is as follows: the first is to set the crossing probability Pc = 0.6. If the probability is lower than this value, the crossing operation is performed. The crossover points need to be set arbitrarily at first, and the value range is [1, n], and then, the chromosome segments are crossed. Moreover, the remaining fragments are stored in arrays x  and y . Comparing the fragments after crossing, if there are different elements, replacement is performed. The specific mutation is as follows: the value of any generated integer rand key is 1, 2, and 3, which corresponds to the insertion, inversion, and swap operations, respectively.
When using a genetic algorithm, appropriate parameters need to be set. Here, the minimum value of evolution rate and the corresponding continuous algebra are represented as Genemin-imprion-rane and Genedie, respectively, and the maximum and minimum values of iteration are Genemax and Genemin, respectively. In the iteration process, if the conditions are met, the iteration ends, which includes the continuous Genedie generation of iteration or the number of iterations reaching Genemax. The resolution improvement rate is lower than Genemin-imprion-rane.
3.2. Power Distribution Communication Service Spectrum Selection Based on Ant Colony Algorithm(1)Ant colony algorithm (ACO) is introduced in spectrum allocation of power communication network, which is helpful to improve the reliability and efficiency of spectrum allocation [21–23]. In the study of spectrum allocation, a bipartite graph is adopted, where , U, and E represent the user, the authorized user, and the connection edge, respectively. Furthermore, there is , and edge eij connected represents that i is matched to the authorized user i. represents the trace of this edge. If i and j are not connected, can be obtained.(i)Update pheromone. The specific update formula is as follows:where and represent the maximum and minimum pheromone values of in turn. The former value is infinite, and the latter is set to 60. represents the volatilizing rate of pheromone. refers to the increased value in trace on , and its formula is as follows:where and represent the benefit and adjustment coefficient, respectively.(2)Use authorize user nodes to select policies. The formula for selecting node j is as follows:where and represent transition probability and threshold parameters, respectively, and the latter is set to 0.9. Q stands for a random number, and its value ranges from 0 to 1.where i represents the nodes that ants can access and and is not a node in the tabu table.(3)Set algorithm parameters, where Q = 10, and , respectively.(4)Set the number of ants, namely the number of user nodes n.(5)The ending condition of the ant colony algorithm is as follows: the first iteration number reaches ; b iterates continuous , and the improvement rate of the child optimization solution is always less than .
3.3. Bandwidth Resource Optimization Strategy Based on Ant Colony Algorithm and Genetic Algorithm
A genetic algorithm is mainly used to get the best spectrum allocation combination, which is used as the initial pheromone matrix and applied to the ant colony algorithm. In this way, the problem of insufficient initial pheromone is solved, and the efficiency of bandwidth allocation is improved. The specific algorithm is shown in Figure 3.
The optimization process based on a genetic algorithm (GA) is as follows:(1)Determine the end conditions of GA first(2)Then, set the initial value of the ant colony algorithm trace, which is shown as follows:where represents the pheromone, which is set to 60; represents the obtained pheromone increment.(3)Obtain pheromone increment: sort the results of the solution and get the results according to the fitness where the first 10% is represented as ; starts at 0, and if some solution in passes through , then plus 20.
The maximum point of the function can be obtained by a genetic algorithm, which is helpful to improve the network performance.
The objective function satisfies the principle of optimal network average benefit, and the specific form is as follows:where a and b represent allocation matrix and benefit matrix, respectively; represents whether m frequency band is allocated to n users; represents the distributed network benefit. X represents spectrum allocation, and each x keeps corresponding to a set of matrix a and b. Spectrum and the number of users are set to 64 and 16, respectively.
4. Power Service Switching of SDN Multidomain Network
4.1. Switching Strategy of Multidomain Network
Multidomain network switching can rearrange different services. In practice, multiple types of distribution communication network services are involved, in which service quality must be guaranteed based on appropriate policies. When switching target domains, it is necessary to reallocate network resources. In this case, a resource allocation strategy is designed to allocate resources based on service quality requirements, which is shown as follows:(1)If the business does not put forward high requirements for service quality, it only needs to allocate certain free resources in the target domain.(2)If the service quality requirements are high, the service can be divided into two types of switching, namely low-to-high-capacity domain and high-to-low-capacity domain. For the former, a high-capacity network has more resources, which can improve the service quality. Therefore, the service requirements can be met as long as appropriate resources are configured. For the latter, more emphasis should be given to quality requirements, and appropriate resources should be allocated.
The switching problem of the multidomain network is studied to improve the rationality of resource allocation and meet the basic requirements of service quality. The controller needs to be used in network topology management. The basic control process is as follows:(1)Determine the original domain and target domain reasonably first, and obtain the control command(2)Then, locate the services involved, and obtain information such as the status of subdomains(3)Reallocate business resources in combination with resource usage information(4)After the completion of the allocation, build the switched link, then migrate the switched services, and record the relevant information in the database
4.2. Switching Algorithm between Domains Based on Analytic Hierarchy Process
At present, there are many scholars have studied the switching algorithm between domains, and a variety of algorithms are proposed. Algorithms have different characteristics and can achieve different effects. In this paper, the switching algorithm of a multidomain network is studied, and the analytic hierarchy process (AHP) is introduced. This method analyzes many factors that affect things at multiple levels, and the relative importance of things is determined based on the comparison between the two. According to the process of AHP hierarchy analysis, a multiobjective decision algorithm is constructed, and the adopted decision indicators include packet loss rate, delay, and bandwidth. The basic process of the algorithm is shown in Figure 2 [24–28].
5. Analysis of Simulation Results
5.1. Simulation Results of Bandwidth Resource Optimization
It is necessary to set the parameters of the genetic algorithm reasonably, in which the probability of mutation and crossover is 0.03 and 0.70, respectively; the number of iterations and population size are 40 and 50, respectively. After the parameter setting is completed, the MATLAB tool is used to simulate the designed algorithm, and the data and models refer to Li Ling. The results obtained after several iterations are shown in Figure 4.
It can be seen from Figure 7 that the average network benefit shows an upward trend with the increase of the number of evolution.
In the above process, matrix a actually belongs to a distribution mode, that is, m spectrum fragments can be allocated to n users. The ant colony algorithm can solve two problems: the first is to arrange the spectrum, especially when new fragments are obtained; the second is to arrange the continuous spectrum.
Given that T(i, j) represents the pheromone distribution matrix, and its initial value is the connection relation of a and b. The new spectrum number is K, which is extended to m + k dimensions, and then, the bandwidths are divided into multiple subsets. Here, the new set is expressed as S, and based on the ant colony algorithm, each subset is processed, and thus, the corresponding spectrum connection mode can be obtained. The ant colony algorithm is used for simulation, and the specific parameter settings are shown in the following Table 1.
After setting each parameter, the simulation can be carried out, and the final results are shown in Figure 5.
Based on the above results, it can be seen that the ant colony algorithm can not only obtain the optimal loop route but also has fewer iterations and higher convergence. When the number of iterations reaches 28 times, the minimum value is obtained. Therefore, the application of this method can meet the requirements of node allocation and is suitable for power grid communication resource allocation.
5.2. Simulation Results of Interdomain Handoff
MATLAB tool is used to simulate the designed algorithm to realize the selection of multidomain network. In addition, bandwidth, packet loss rate, delay, and network hop are utilized to evaluate the performance of network.
5.2.1. Power Distribution Video Surveillance Service
In the operation process of a power communication system, there are many kinds of monitoring services, which generally have relatively low requirements for packet loss rate and pay more attention to delay and bandwidth.
In this process, the eig function is mainly used to calculate the eigenvalue and eigenvector 1` of C. The calculated CR is lower than 0.1, and thus, the consistency test is passed. On this basis, the combination of the single factor and weight coefficient of business 1 is further obtained. The comparison matrix is shown in Table 2.
The total hierarchical distribution coefficient is obtained based on the decision coefficient, and the utility comparison diagram of business 1 is .
According to the above information, 0.2823 and 0.7177, respectively, represent the utility values of service 1 in access domains 1 and 2. Therefore, domain 2 is the optimal switching network for this service.
5.2.2. Electricity Information Collection Service
This service has high requirements for bandwidth and delay, and the corresponding comparison matrix is shown in Table 3:
The combinations of single factor and weight coefficient for business 1 are as follows:
The total hierarchical distribution coefficient is obtained based on the decision coefficient, and the utility comparison diagram of business 2 is .
According to the above information, 0.6068 and 0.3932, respectively, represent the utility values of business 2 in the optimal access domain 1 and 2. It can be clearly seen that the optimal switching network for this business is domain 1.
5.2.3. Distributed Generation Access Service
The comparison matrix for this service is shown in Table 4.
The combinations of single factor and vector weight coefficient are further obtained as follows:
The total hierarchical distribution coefficient is obtained based on the decision coefficient, and the utility comparison diagram of business 3 is .
According to the above information, 0.7167 and 0.2833, respectively, represent the utility values of business 3 in the best access domain 1 and 2. It is obvious that domain 1 should be selected for this business.
5.2.4. Power Distribution Automation Service
The comparison matrix for this service is shown in Table 5.
The combinations of single factor and weight coefficient are shown as follows:
The total hierarchical distribution coefficient is obtained based on the decision coefficient, and the utility comparison diagram of business 4 is .
According to the above information, 0.4756 and 0.5244, respectively, represent the utility values of business 4 in the best access domain 1 and 2. It is obvious that domain 2 should be selected for this business.
To sum up, it can be seen that the performance of the SDN network is improved to a certain extent by introducing the spectrum scheduling of the improved genetic algorithm and switching method of AHP. After parameters are optimized, the optimal spectrum resource optimization path of the SDN network can be obtained. Moreover, the switching of AHP shows that domain 2 is the best access domain, so as to better improve the performance of the whole network.
The experimental data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that they have no conflicts of interest regarding this work.
L.-H. Chang, “Application-based online traffic classification with deep learning models on SDN networks,” Advances in Technology Innovation, vol. 5, no. 4, pp. 216–229, 2020.View at: Google Scholar
S. Rui, “A hybrid SDN solution for mobile networks,” Computer Networks, vol. 190, 2021.View at: Google Scholar
Ma Zheng, “Research on the on-demand scheduling algorithm of intelligent routing load based on SDN,” International Journal of Internet Protocol Technology, vol. 14, no. 1, 2021.View at: Google Scholar
Z. Liu, “QoS oriented task scheduling based on genetic algorithm in cloud computing,” Computer Systems Science and Engineering, vol. 30, no. 6, pp. 481–487, 2015.View at: Google Scholar
K. Lalitha Devi and S. Valli, “Multi-objective heuristics algorithm for dynamic resource scheduling in the cloud computing environment,” The Journal of Supercomputing, pp. 1–29, 2021.View at: Google Scholar
A. Praveena and V. Vijayarajan, “ENergy efficient resource scheduling using optimization based neural network in mobile cloud computing,” Wireless Personal Communications, vol. 114, pp. 1–20, 2020.View at: Google Scholar
M. Sharma, “To eliminate the threat of a single point of failure in the SDN by using the multiple controllers,” International Journal of Recent Technology and Engineering, vol. 9, no. 2, pp. 234–241, 2020.View at: Google Scholar
A. Shahid, “Securing genetic algorithm enabled SDN routing for blockchain based internet of things,” IEEE Access, vol. 9, Article ID 139739, 2021.View at: Google Scholar
Z. Stanislav Igorovych, “Genetic algorithm for SDN protection against network attacks[J],” System research and information technologies, vol. 0, no. 2, p. 14, 2016.View at: Google Scholar
L. Yan, “Voltage sag severity analysis based on improved FP-Growth algorithm and AHP algorithm,” Journal of Physics: Conference Series, vol. 1732, no. 1, Article ID 012088, 2021.View at: Google Scholar