CN111711538B - Power network planning method and system based on machine learning classification algorithm - Google Patents
Power network planning method and system based on machine learning classification algorithm Download PDFInfo
- Publication number
- CN111711538B CN111711538B CN202010514935.7A CN202010514935A CN111711538B CN 111711538 B CN111711538 B CN 111711538B CN 202010514935 A CN202010514935 A CN 202010514935A CN 111711538 B CN111711538 B CN 111711538B
- Authority
- CN
- China
- Prior art keywords
- service
- machine learning
- classifiers
- algorithm
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000007635 classification algorithm Methods 0.000 title claims abstract description 24
- 238000010801 machine learning Methods 0.000 title claims abstract description 24
- 238000013439 planning Methods 0.000 title description 9
- 238000012545 processing Methods 0.000 claims abstract description 25
- 238000005457 optimization Methods 0.000 claims abstract description 23
- 238000013468 resource allocation Methods 0.000 claims abstract description 13
- 238000001228 spectrum Methods 0.000 claims abstract description 7
- 238000013480 data collection Methods 0.000 claims abstract description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 26
- 239000013598 vector Substances 0.000 claims description 20
- 238000004891 communication Methods 0.000 claims description 18
- 230000005540 biological transmission Effects 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 11
- 238000009826 distribution Methods 0.000 claims description 10
- 230000009977 dual effect Effects 0.000 claims description 8
- 230000035772 mutation Effects 0.000 claims description 6
- 238000011144 upstream manufacturing Methods 0.000 claims description 5
- 230000003595 spectral effect Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 9
- 238000004590 computer program Methods 0.000 description 8
- 238000011161 development Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 4
- 238000012706 support-vector machine Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 238000007726 management method Methods 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 210000000349 chromosome Anatomy 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 238000012384 transportation and delivery Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0803—Configuration setting
- H04L41/0823—Configuration setting characterised by the purposes of a change of settings, e.g. optimising configuration for enhancing reliability
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/24—Traffic characterised by specific attributes, e.g. priority or QoS
- H04L47/2441—Traffic characterised by specific attributes, e.g. priority or QoS relying on flow classification, e.g. using integrated services [IntServ]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention provides a network optimization method and a system based on a machine learning classification algorithm, which comprises the following steps: step1, a service classifier in a data link layer classifies the services collected by a data collection layer and attaches corresponding class labels; and 2, the service classifier transmits the classification result to a master controller positioned on the data processing layer, the master controller calculates an optimal scheduling strategy according to the classification result, performs carrier scheduling and base station resource allocation on the service nodes according to the optimal scheduling strategy, and controls the transmitting power of the service nodes to realize network optimization. Compared with the traditional method for distributing resources according to the service priority, the method takes the influence between the resource and other services into consideration, thereby improving the utilization rate of the frequency spectrum resources and realizing higher network utility.
Description
Technical Field
The invention relates to the field of power grid service classification, in particular to a power network planning method and system based on a machine learning classification algorithm.
Background
With the continuous development of wireless communication technology, the 5G era of high speed, low power consumption and low time delay is coming, and the communication of the human society is gradually becoming smooth at present. The 5G has the characteristics of ultrahigh bandwidth, ultralow time delay and capability of supporting mass access, so that the wireless network access control system has infinite possibility in the development of the vertical industry and provides a foundation for the idea of interconnection of everything.
The power network is an indispensable infrastructure of modern society and is also an energy support for national development, and the supply of electric energy cannot be separated no matter in industrial manufacturing, social production or daily life of people. Efficient development and management of smart grids has been an important subject of research in academia and industry. With the rapid development of social economy and the continuous increase of power consumption demand in China, the coverage area and the application range of a power grid are gradually enlarged, a large-scale smart power grid coverage system is gradually formed, and in the process, the communication technology plays an extremely important role in influencing, and is the most basic link for ensuring the maximization of the operation effect of the smart power grid.
The intelligent power grid transmits bidirectional power and information flow, and the communication network is a neural network of the intelligent power grid, can transmit the operation state of the power grid and equipment data to a central station, and can perform real-time monitoring and control. With the accelerated development of the field of power distribution and the widespread of the Internet of Things (IoT), a large number of sensors, actuators and other terminals are present in the power grid, so that the types of power grid services are various and difficult to predict. In order to enable the intelligent operation of the power grid to have universal sensing and processing capabilities, the terminals urgently need universal communication coverage and provide a multi-Service transmission link capable of guaranteeing Quality of Service (QoS). Therefore, it is necessary to establish a smart grid which mainly covers the coverage area of the power communication technology and has a perfect and scientific communication mechanism to realize smooth and stable operation among the power devices.
In reality, with the explosive growth of power grid services, in the network planning process of an intelligent power grid, the identification and classification of the services are crucial to the network performance, and network resources capable of guaranteeing the service QoS requirements are distributed according to the classification results, so that the maximum network utility is achieved under the condition that the network facility cost is limited. In addition, since the type and amount of information exchanged over the communication infrastructure grows exponentially, some enhanced information is only available within a predefined time window, which will not be available if the communication delay exceeds a predefined time window and, in the worst case, may even lead to network damage. Conventional communication techniques provide only "best effort" services and do not guarantee timeliness or real-time delivery.
In the existing related research, the classification of the services in the network is mostly performed according to the multi-layer protocol labels or port numbers, and then on the basis of the results of service perception and classification, the service-oriented perception control protocol uses methods such as classification shaping and queue management to ensure the service quality of key services. On one hand, due to the use of dynamic port numbers and processing of some encrypted data packets, the common classification methods are difficult to accurately identify; on the other hand, the QoS requirements of different services are very different, and the traditional resource allocation scheme is not favorable for improving the network utility.
Disclosure of Invention
The invention aims to classify according to the statistical characteristics of data streams in network services and design a resource allocation scheme of a Radio Access Network (RAN) of a power communication network on the basis of classification so as to achieve the aim of optimizing the power communication network. The flow characteristics are used for identifying not only the type of the exited service, but also a high identification degree for the encrypted or unknown service of an application layer.
The invention provides a network optimization method based on a machine learning classification algorithm.A data transmission structure between a power grid and an Internet of things comprises a data acquisition layer, a data link layer and a data processing layer from downstream to upstream; the network optimization method comprises the following steps:
step1, a service classifier in a data link layer classifies the services collected by a data collection layer and attaches corresponding class labels;
and 2, the service classifier transmits the classification result to a master controller positioned on the data processing layer, the master controller calculates an optimal scheduling strategy according to the classification result, performs carrier scheduling and base station resource allocation on the service nodes according to the optimal scheduling strategy, and controls the transmitting power of the service nodes to realize network optimization.
Further, in step1, when classifying the services, the data is classified into three categories including a control category, an information collection category and a mobile application category; each large class is divided into a plurality of small classes.
Further, the traffic classifier is obtained by the following steps:
s101, capturing a data packet in a power communication network to obtain a flow set;
s102, extracting statistical characteristics of the data packets, wherein the statistical characteristics comprise the sizes of the data packets, the sizes of the previous data packet and the next data packet and the time interval between the previous data packet and the previous data packet; and normalizing the statistical characteristics to obtain a characteristic vector xiThe dimension of the feature vector is equal to the feature number of the sample;
s103, establishing a sample training set D { (x)1,y1),(x2,y2),...,(xm,ym) }; wherein x is a feature vector after the statistical feature normalization of the power grid service flow, and y represents a specific service category; establishing k (k-1)/2 two classifiers, and training the two classifiers by adopting the established sample training set, wherein k is the number of the service types; the trained k (k-1)/2 secondary classifiers form a traffic classifier;
in step1, the service x to be classified is respectively put into a service classifier formed by trained k (k-1)/2 classifiers, and the two classifiers in the service classifier vote on the type of the service, so that the service type with the largest number of votes is obtained.
Further, in step S104, when a second classifier is established, the following steps are included;
s1041, introducing a relaxation variable into a definition formula of the classifier;
s1042, making a Lagrange equation for the two classifiers, and solving a partial derivative to obtain a dual problem;
and S1043, solving parameters omega and b of the two classifiers by adopting an SMO algorithm to the dual problem, and completing the establishment of the two classifiers.
Further, in step2, when the master controller performs carrier scheduling and base station resource allocation for the service node, the scheduling policy is expressed as binary variable,
wherein, Z is a set of integers,indicating that the mth service occupies the qth sub-carrier and is connected to the mth servicel base stations, otherwiseVector P ═ P for transmission power of each convergence plane service forwarding node1,p2,...,pM]And (4) showing.
The transmission rate of the transmission channel of the mth service is:
wherein sigma2Is the size of white gaussian noise, and is,representing the loss experienced by the service m in the process of transmitting to the base station l, and complying with a Hata model, wherein U is the sub-carrier bandwidth;
the throughput of the control type service, the information acquisition type service and the mobile application type service is the sum of the transmission rates of the corresponding type services, namely
Wherein y isiIndicating the traffic type of the ith traffic.
Considering that different types of traffic have different throughput requirements, three weights can be introduced into the system throughput and γ ═ γ can be satisfiedco,γin,γmoGamma is more than or equal to 0, and the system throughput is the weighted sum of three types of throughput:
obtaining a utility function:
further, the method also comprises the following steps of constructing a scheduling policy constraint condition according to the QoS requirements of different power services: at least comprises the following steps: a utility function; the throughput constraint of the information acquisition service and the mobile application service; controlling the time delay constraint of the service; the base station backhaul capacity constraint.
Further, all constraint conditions of the scheduling strategy are integrated, and a heuristic algorithm and an SQP algorithm are adopted to solve the problemFirstly, decomposing the integer into an inner layer and an outer layer, and searching the optimal integer variable by the outer layer by adopting a differential evolution algorithmDetermining a subcarrier distribution mode pi; and the inner layer uses an SQP algorithm to solve the optimal power allocation scheme P under the premise that the current subcarrier allocation is known.
Further, when solving, the method comprises the following steps:
s301, randomly initializing N scheduling strategies IIiAs an initial population;
s302, calculating each determined scheduling strategy IIiUnder the premise, the power distribution scheme P is solved by utilizing SQP algorithm(i)=SQP(Πi) And the spectral efficiency eff obtained by the solution(i)As the fitness value, the higher the fitness value is, the better the policy is. By Π*Representing an optimal strategy;
and S303, updating iteration is needed for the initial population, and mutation and cross operation are carried out on the initial population.
Further, after step S303, the following steps may be further included: s304, judging whether the crossed and mutated new strategy replaces the old strategy or not so as toThe frequency spectrum efficiency is used as a fitness value, if the fitness value is high, the replacement is carried out, otherwise, the old strategy is kept, the population is replaced, and meanwhile, the optimal strategy II is updated*。
The network optimization system based on the machine learning classification algorithm comprises: a data acquisition layer, a data link layer and a data processing layer which are arranged between the power grid and the Internet of things from downstream to upstream, wherein,
the data acquisition layer is used for acquiring services;
the data link layer is used for classifying the services acquired by the data acquisition layer, attaching corresponding class labels and uploading classification results to the data processing layer;
and the data processing layer is used for calculating an optimal scheduling strategy according to the classification result, performing carrier scheduling and base station resource allocation on the service node according to the optimal scheduling strategy, and controlling the transmitting power of the service node.
Compared with the traditional method for distributing resources according to the service priority, the power network planning method and the system based on the machine learning classification algorithm consider the influence between the power network planning method and other services, thereby improving the utilization rate of frequency spectrum resources and realizing higher network utility. The advantages are obvious under the condition that the number of carriers is small, namely, the frequency spectrum resources are not quite abundant; the method comprises the steps of classifying according to statistical characteristics of data streams in network services, and designing a resource allocation scheme of a Radio Access Network (RAN) of the power communication network on the basis of classification so as to achieve the purpose of optimizing the power communication network. The flow characteristics are used for identifying not only the type of the exited service, but also a high identification degree for the encrypted or unknown service of an application layer.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:
FIG. 1 is an overall framework diagram of a power network planning method based on a machine learning classification algorithm according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of a power network planning method based on a machine learning classification algorithm according to an embodiment of the present invention;
FIG. 3 is a flow chart of a method for solving a mixed integer nonlinear programming problem of a differential evolution algorithm according to another embodiment of the present invention;
in the figure: 1-a data acquisition layer; 2-data link layer; 3-data processing layer
Detailed Description
The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
The following detailed description is exemplary in nature and is intended to provide further details of the invention. Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention.
As shown in fig. 1, the invention provides a power network planning method based on a machine learning classification algorithm, a data transmission structure between a power grid and an internet of things includes a data acquisition layer 1, a data link layer 2 and a data processing layer 3 from downstream to upstream, a service classifier is established in the data link layer 2, and it is to be noted that the data acquisition layer includes various aspects such as building electricity, high-voltage power distribution, residential electricity, mobile electricity and the like; and the data link layer correspondingly sends the data link layer to a data processing layer according to the service type, and the data processing layer comprises a WIFI access point, a GSM base station, an ID-CDMA base station, a space satellite type access point and the like.
The method specifically comprises the following steps:
step1, classifying the services collected by the data collection layer by a service classifier according to QoS requirements by using a classification algorithm in machine learning, and attaching corresponding class labels.
And 2, the service classifier transmits the classification result to a master controller positioned on the data processing layer, the master controller calculates an optimal scheduling strategy according to the classification result, performs carrier scheduling and base station resource allocation on the service nodes according to the optimal scheduling strategy, and controls the transmitting power of the service nodes to realize network optimization.
When the step1 is implemented, referring to the architecture of fig. 1, all the electric power related services are converged to a 'data link layer' from the intelligent terminal of the data acquisition layer 1, and then transmitted to the base station from various nodes. Therefore, based on the architecture of fig. 1, the present invention establishes a service classifier at the data link layer, classifies the services converged to the plane by a machine learning classification algorithm, and attaches a corresponding class label. In the future of the 5G-enabled smart grid, the power services may be roughly classified into a control class, an information collection class and a mobile application class, however, for the present invention, in order to increase the efficiency of limited resources, a refined service class is dedicated, for example, for services with large differences in individual characteristics in the control class, such as distribution automation, accurate power load control, etc., the service may be finely classified into a plurality of sub-classes, and the specific classification may refer to table 1 in the exemplary embodiment.
TABLE 1 typical grid services
In the embodiment of the application, five typical grid services in table 1 are selected for the description of the implementation method, and in an actual application scene, the types of the grid services can be far larger than the five typical grid services, but always belong to three major categories, namely a control category, an information acquisition category and a mobile application category. In addition, in the embodiment of the present application, a Support Vector Machine (SVM) in a Machine learning classification algorithm is taken as an example to classify the grid service.
In implementing step1, the conventional SVM algorithm is only applicable to two classes, and for k ═ 5 traffic types, k (k-1)/2 classifiers need to be established.
The multi-class classifier building process is described below:
s101, capturing a data packet in a power communication network by adopting Wireshark software to obtain a flow set;
s102, extracting statistical characteristics of the data packets, wherein the statistical characteristics comprise the sizes of the data packets, the sizes of the previous data packet and the next data packet and the time interval between the previous data packet, and then normalizing the statistical characteristics to obtain a characteristic vector xiIts dimension is equal to the number of features of the sample.
S103, establishing a sample training set D { (x)1,y1),(x2,y2),...,(xm,ym)}. The training set of each two classifiers is different, and yiE { -1,1 }. And each sample training set is used for training a corresponding classifier, wherein x is a feature vector obtained by normalizing the statistical features of the power grid service flow, and y represents a specific service class.
S104, respectively putting the service x to be classified into k (k-1)/2 trained two classifiers, and voting and selecting by each two classifiers to obtain the service type with the largest number of votes; where k represents the number of services.
The specific process of establishing the two classifiers is described below, and the mathematical expression form is as follows:
s1041, introducing a relaxation variable into a definition formula of the classifier;
vector ω represents the normal vector of the hyperplane used for the partition, C>0 and is a constant, ξiIs a relaxation variable and satisfies xii≧ 0, training set sample denoted D { (x)1,y1),(x2,y2),...,(xm,ym)}。ξiThe introduction of (C) means that the support vector machine is allowed to go wrong on some samples, when C → ∞ then this means that no sample errors are allowed. In addition, the reciprocal of | | ω | | | represents the "interval" between different samples, and in order to ensure that the classification effect is better, the larger the "interval" is, the better.
S1042, it can be seen that formula 1 is a convex quadratic optimization problem, and in order to solve the problem more efficiently, the dual problem is solved by using the lagrangian multiplier method. The Lagrange equation is
Wherein the Lagrangian alphai≥0,μi≥0。
Let L (ω, b, α, ξ, μ) be 0 for ω, b, ξ
Substituting into equation 2 to get the dual problem
In addition, because the original problem has inequality constraint, the dual condition still needs to satisfy the KKT condition.
Wherein f (x) ωTx+b。
S1043, solving the dual problem by adopting an SMO (sequential minor optimization) algorithm, wherein the obtained optimal solution shows that the classification effect is best, and finishing the establishment of the two classifiers. The SMO algorithm is briefly described below.
Step1. random initialization vector α ═ α (α)1,α2,...,αm) And satisfies the constraint condition (8).
Step2. select the variable that violates the KKT condition to the greatest extent(general preference is satisfied withThe variable(s).
and Step5, returning to Step2 before the target function is not converged, otherwise, terminating the program.
Since the training samples are not necessarily linearly separable, a Gaussian kernel function k (x) is usedi,xj) Instead of the formerThe operation, namely mapping the original problem to a higher-dimensional, linearly separable feature space.
After the accurate service classification process is realized, the following steps are carried out.
Step2, the service classifier transmits the classification result to a master controller positioned on a data processing layer, the master controller calculates an optimal scheduling strategy according to the classification result, performs carrier scheduling and base station resource allocation on the service nodes according to the optimal scheduling strategy, and controls the transmitting power of the service nodes to realize network optimization;
aiming at the classification result, the invention designs a method suitable for uplinkAnd the RAN side resource allocation scheme of the link faces different service QoS requirements. In the embodiment of the present application, there are L base stations, and the backhaul capacity of each base station is denoted as RlL base stations share a section of frequency band resource W, and are divided into Q subcarriers with the bandwidth of U by utilizing the OFDM technology.
At a certain moment, M power grid services of unknown types arrive at a convergence plane, after class labels of the table 1 are attached to the class labels through a classifier, through signaling interaction, a master controller of a data processing plane distributes radio resource on an RAN side, and finally the radio resource is sent to each base station through a forwarding node. The classified traffic is represented as Y ═ Y i1, 2.., M }, wherein yi∈{1,2,3,4,5}。
For different types of power grid services, the master controller needs to design a scheduling strategy, that is, a subcarrier is allocated to each service, and a base station to which the service is connected is determined, so that the maximum system utility is realized on the premise that physical resources and facility cost are not changed. The invention adopts binary variable to express the scheduling strategy asWherein Z is a set of integers,indicating that the mth service occupies the qth sub-carrier and is connected to the lth base station, otherwiseVector P ═ P for transmission power of each convergence plane service forwarding node1,p2,...,pM]And (4) showing.
The highest transmission rate r of the channel carrying the service mmCan be obtained according to the shannon formula. Wherein sigma2Is the size of white gaussian noise, and is,which represents the loss experienced by the service m during transmission to the base station l, in the embodiment of the present application, a Hata model is used for description.
The throughput of the control type service, the information acquisition type service and the mobile application type service is the sum of the transmission rates of the corresponding type services, namely
Considering that different types of traffic have different throughput requirements, three weights can be introduced into the system throughput and γ ═ γ can be satisfiedco,γin,γmoGamma is more than or equal to 0, and the system throughput is the weighted sum of three types of throughput:
a utility function is obtained.
The improvement of network utility is based on the premise of guaranteeing service quality of service, and for service flows of information collection type and mobile application type, the guarantee of throughput is crucial to user experience, so a lower limit value, C, is set for total throughput3,C4,C5Indicating a throughput value that can satisfy the traffic QoS.
The control type service is used as a link of power grid control, and is directly related to power grid safety, and the service has extremely high requirements on communication transmission delay and channel reliability. Therefore, the distribution automation service and the accurate power load control service need to strictly meet the time delay requirement, and the embodiment of the application refers to the concept of the average arrival rate lambda and is applied to time delay calculation of each service flow. Wherein v isiIndicates the packet length, λ, of service iiAnd the average reaching rate of the service i is represented, namely an upper limit value is set for the average time delay of the service i.
The performance of the base station, which is an important node in information transmission, affects the system utility. The base stations have limited backhaul capacity so that each base station must not receive traffic more than its backhaul capacity. Wherein R islIndicating the backhaul capacity of base station i.
For simplicity, it is assumed that each service can occupy only one subcarrier when transmitting with a base station and is connected with at most one base station.
In order to improve the spectrum utilization rate, in the embodiment of the present application, it is assumed that each subcarrier may be occupied by one or more services, and therefore there may be co-channel interference, so that an upper limit P is set for the sum of the transmit powers of forwarding nodes occupying the same subcarrierT。
piMore than or equal to 0, i ═ 1,2,.., M (formula 22)
In summary, the policy scheduling optimization problem can be summarized as follows.
C5:pi≥0,i=1,2,...,M
Due to the constraints of C2 and C7, it is clear that the problem is a non-convex non-linear optimization problem. In order to facilitate problem solving, the embodiment of the application provides a solving idea based on a heuristic algorithm. Taking into account integer variablesThe value set is {0,1}, so that integer variables are fixed, the MINLP problem is converted into the NLP subproblem, and then the SQP solver is used for solving the NLP subproblem. The solving method is divided into an inner layer and an outer layer, and the outer layer adopts a differential evolution algorithm to search the optimal integer variableDetermining a subcarrier distribution mode pi; and the inner layer uses an SQP algorithm to solve the optimal power allocation scheme P under the premise that the current subcarrier allocation is known.
The differential evolution algorithm is a self-organizing minimization method, and a user only needs few inputs. The direction of optimizing search is guided by group intelligence generated by mutual cooperation and competition among individuals in the group. The basic idea of the algorithm is as follows: starting from a randomly generated initial population, new individuals are generated by summing the vector difference of any two individuals in the population with the third individual, then the new individuals are compared with the corresponding individuals in the current population, if the fitness of the new individuals is better than that of the current individuals, the new individuals are used for replacing the old individuals in the next generation, otherwise, the old individuals are still stored. Through continuous evolution, excellent individuals are reserved, inferior individuals are eliminated, and search is guided to approach to the optimal solution. Compared with the traditional evolutionary algorithm, the traditional method is that a predetermined probability distribution function is used for determining vector disturbance, a self-organization program of the differential evolutionary algorithm utilizes two randomly selected different vectors in a population to interfere one existing vector, and each vector in the population interferes, so that the following embodiment is adopted, and the introduction of the iterative process specifically comprises the following steps:
s301, randomly initializing N scheduling strategies ΠiAs the starting population.
S302, calculating each determined scheduling strategy IIiUnder the premise, the power distribution scheme P is solved by utilizing SQP algorithm(i)=SQP(Πi) And the spectral efficiency eff obtained by the solution(i)As the fitness value, the higher the fitness value is, the better the policy is. By Π*Representing the optimal strategy.
And S303, updating and iterating the initial population, namely performing mutation and cross operation on the initial population. I.e. randomly selecting three chromosomes a in the population1,a2,a3(a1≠a2≠a3) If there is a variant chromosomeV is a mutation operator. Since the conventional differential evolution algorithm processes continuous variables, the method has the advantages of simple operation, low cost and high efficiencyIs a variable of 0 to 1, so that the present invention generates the variation vector in the form of a power function
The specific algorithm flow of the variation and intersection is as follows:
s304, selecting operation, namely judging whether the crossed and mutated new strategy replaces the old strategySlightly, the frequency spectrum efficiency is used as a fitness value, if the fitness value is high, the replacement is carried out, otherwise, the old strategy is kept, the population is replaced, and the optimal strategy pi is updated at the same time of replacing the population*。
And finishing the first iteration, and repeating the processes of mutation, intersection and selection before the termination condition is not met, so as to finally obtain the optimal solution. A general termination condition is set to reach the number of iterations, or the fitness value reaches some desired value.
The invention also provides a network optimization system based on the machine learning classification algorithm, which comprises the following components: a processor and a memory coupled to the processor, the memory storing a computer program which, when executed by the processor, performs the method steps of the above-described method for network optimization based on a machine learning classification algorithm.
The invention also provides another network optimization system based on the machine learning classification algorithm, which comprises the following steps: the system comprises a data acquisition layer, a data link layer and a data processing layer, wherein the data acquisition layer, the data link layer and the data processing layer are positioned between a power grid and the Internet of things from downstream to upstream; the data link layer is used for classifying the services acquired by the data acquisition layer, attaching corresponding class labels and uploading classification results to the data processing layer; and the data processing layer is used for calculating an optimal scheduling strategy according to the classification result, performing carrier scheduling and base station resource allocation on the service node according to the optimal scheduling strategy, and controlling the transmitting power of the service node.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims.
Claims (8)
1. A network optimization method based on a machine learning classification algorithm is characterized in that a data transmission structure between a power grid and an Internet of things comprises a data acquisition layer, a data link layer and a data processing layer from downstream to upstream; the network optimization method comprises the following steps:
step1, a service classifier in a data link layer classifies the services collected by a data collection layer and attaches corresponding class labels;
step2, the service classifier transmits the classification result to a master controller positioned on a data processing layer, the master controller calculates an optimal scheduling strategy according to the classification result, performs carrier scheduling and base station resource allocation on the service node according to the optimal scheduling strategy, and controls the transmitting power of the service node;
the traffic classifier is obtained by the following steps:
s101, capturing a data packet in a power communication network to obtain a flow set;
s102, extracting statistical characteristics of the data packets, wherein the statistical characteristics comprise the sizes of the data packets, the sizes of the previous data packet and the next data packet and the time interval between the previous data packet and the previous data packet; and normalizing the statistical characteristics to obtain a characteristic vector xiThe dimension of the feature vector is equal to the feature number of the sample;
s103, establishing a sample training set D { (x)1,y1),(x2,y2),...,(xm,ym) }; wherein x is a feature vector after the statistical feature normalization of the power grid service flow, and y represents a specific service category; establishing k (k-1)/2 two classifiers, and training the two classifiers by adopting the established sample training set, wherein k is the number of the service types; the trained k (k-1)/2 secondary classifiers form a traffic classifier;
in step1, the service x to be classified is respectively put into a service classifier formed by trained k (k-1)/2 classifiers, and the two classifiers in the service classifier vote on the type of the service, so that the service type with the largest number of votes is obtained.
2. The method for optimizing network based on machine learning classification algorithm according to claim 1, wherein in step1, the services are classified into three categories: control class, information collection class, mobile application class.
3. The network optimization method based on the machine learning classification algorithm according to claim 1, wherein when the two classifiers are established, the method comprises the following steps;
s1041, introducing a relaxation variable into a definition formula of the classifier;
s1042, making a Lagrange equation for the two classifiers, and solving a partial derivative to obtain a dual problem;
and S1043, solving parameters of the two classifiers by adopting an SMO algorithm to the dual problem, and completing the establishment of the two classifiers.
4. The network optimization method based on machine learning classification algorithm as claimed in claim 1, wherein in step2, when the master controller performs carrier scheduling and base station resource allocation for the service nodes, binary variables are used to express the scheduling policy as,
5. The machine learning classification algorithm-based network optimization method according to claim 4, further comprising constructing scheduling policy constraints according to QoS requirements of different power services, including at least: a utility function; the throughput constraint of the information acquisition service and the mobile application service; controlling the time delay constraint of the service; the base station backhaul capacity constraint.
6. The network optimization method based on the machine learning classification algorithm according to claim 4, characterized in that all constraint conditions of the scheduling strategy are integrated and solved by adopting a heuristic algorithm and an SQP algorithm; firstly, decomposing the integer into an inner layer and an outer layer, and searching the optimal integer variable by the outer layer by adopting a differential evolution algorithmDetermining a subcarrier distribution mode pi; and the inner layer uses an SQP algorithm to solve the optimal power allocation scheme under the premise that the current subcarrier allocation is known.
7. The network optimization method based on the machine learning classification algorithm according to claim 6, characterized by comprising the following steps when solving:
s301, randomly initializing N scheduling strategies IIiAs an initial population;
s302, calculating each determined scheduling strategy IIiUnder the premise, the power distribution scheme P is solved by utilizing SQP algorithm(i)=SQP(Πi) And the spectral efficiency eff obtained by the solution(i)As a fitness value; by Π*Representing an optimal strategy;
s303, the initial population needs to be updated and iterated, and mutation and cross operation are carried out on the initial population;
s304, judging whether the new strategy after crossing and mutation replaces the old strategy or not, taking the frequency spectrum efficiency as a fitness value, if the fitness value is high, replacing, otherwise, keeping the old strategy, and updating the optimal strategy II while replacing the population*。
8. The method for network optimization based on machine learning classification algorithm according to claim 7, wherein steps S303 and S304 are repeatedly executed until the termination condition is satisfied: the set maximum number of iterations is reached or the fitness value reaches the set desired value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010514935.7A CN111711538B (en) | 2020-06-08 | 2020-06-08 | Power network planning method and system based on machine learning classification algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010514935.7A CN111711538B (en) | 2020-06-08 | 2020-06-08 | Power network planning method and system based on machine learning classification algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111711538A CN111711538A (en) | 2020-09-25 |
CN111711538B true CN111711538B (en) | 2021-11-23 |
Family
ID=72539279
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010514935.7A Active CN111711538B (en) | 2020-06-08 | 2020-06-08 | Power network planning method and system based on machine learning classification algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111711538B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112367708B (en) * | 2020-10-30 | 2023-05-26 | 新华三技术有限公司 | Network resource allocation method and device |
CN116128256B (en) * | 2023-04-12 | 2023-06-20 | 国网吉林省电力有限公司信息通信公司 | Power business data processing method |
CN116582450B (en) * | 2023-07-13 | 2023-12-01 | 北京智芯微电子科技有限公司 | Service access sensing method and system for low-voltage distribution network |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106385387A (en) * | 2016-09-27 | 2017-02-08 | 中国科学院空间应用工程与技术中心 | Resource scheduling method of information network links, system and application |
CN110290077A (en) * | 2019-06-18 | 2019-09-27 | 西安交通大学 | A kind of industrial SDN resource allocation methods based on real time business configuration |
CN110381541A (en) * | 2019-05-28 | 2019-10-25 | 中国电力科学研究院有限公司 | A kind of smart grid slice distribution method and device based on intensified learning |
CN111246586A (en) * | 2020-01-20 | 2020-06-05 | 中国电力科学研究院有限公司 | Method and system for distributing smart grid resources based on genetic algorithm |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9986580B2 (en) * | 2013-10-16 | 2018-05-29 | Empire Technology Development Llc | Dynamic frequency and power resource allocation with granular policy management |
CN109462521B (en) * | 2018-11-26 | 2020-11-20 | 华北电力大学 | Network flow abnormity detection method suitable for source network load interaction industrial control system |
-
2020
- 2020-06-08 CN CN202010514935.7A patent/CN111711538B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106385387A (en) * | 2016-09-27 | 2017-02-08 | 中国科学院空间应用工程与技术中心 | Resource scheduling method of information network links, system and application |
CN110381541A (en) * | 2019-05-28 | 2019-10-25 | 中国电力科学研究院有限公司 | A kind of smart grid slice distribution method and device based on intensified learning |
CN110290077A (en) * | 2019-06-18 | 2019-09-27 | 西安交通大学 | A kind of industrial SDN resource allocation methods based on real time business configuration |
CN111246586A (en) * | 2020-01-20 | 2020-06-05 | 中国电力科学研究院有限公司 | Method and system for distributing smart grid resources based on genetic algorithm |
Also Published As
Publication number | Publication date |
---|---|
CN111711538A (en) | 2020-09-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111711538B (en) | Power network planning method and system based on machine learning classification algorithm | |
Ye et al. | Dynamic radio resource slicing for a two-tier heterogeneous wireless network | |
Peng et al. | Deep reinforcement learning based resource management for multi-access edge computing in vehicular networks | |
Chen et al. | Multi-tenant cross-slice resource orchestration: A deep reinforcement learning approach | |
Yu et al. | An intelligent game-based offloading scheme for maximizing benefits of IoT-edge-cloud ecosystems | |
CN111372314A (en) | Task unloading method and task unloading device based on mobile edge computing scene | |
US11496230B2 (en) | Systems and methods for mapping resource blocks to network slices | |
CN111277437A (en) | Network slice resource allocation method for smart power grid | |
CN105704824B (en) | A kind of method of the multi dimensional resource distribution of wireless network | |
US9369925B2 (en) | System and methods to achieve optimum efficiency-jain fairness in wireless systems | |
Hassan et al. | Interference management in cellular-connected Internet of Drones networks with drone-pairing and uplink rate-splitting multiple access | |
Ye et al. | Pricing-based resource allocation in virtualized cloud radio access networks | |
Hao et al. | Interference-aware resource optimization for device-to-device communications in 5G networks | |
Chien et al. | Resource management in 5g mobile networks: Survey and challenges | |
NaderiAlizadeh et al. | State-augmented learnable algorithms for resource management in wireless networks | |
Lotfi et al. | Attention-based open RAN slice management using deep reinforcement learning | |
Kaur et al. | Intelligent spectrum management based on reinforcement learning schemes in cooperative cognitive radio networks | |
CN110691383A (en) | Resource allocation method and device | |
CN113316156B (en) | Intelligent coexistence method on unlicensed frequency band | |
Luan et al. | Distributed network topology formation and resource allocation for clustered Machine-to-Machine communication networks | |
CN113743800A (en) | Power distribution network user scheduling method based on user priority | |
Li et al. | Service Priority Based WiFi6 Optimized Handover Algorithm | |
Jia et al. | Dispatching and Control Information Freshness-Aware Federated Learning for Simplified Power IoT | |
Bian et al. | Rate Adaptation Algorithm with LSTM in IEEE 802.11 ac | |
Alenezi et al. | Federated Reinforcement Learning with Knowledge Transfer for Network Selection in Hybrid WiFi-VLC Networks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |