CN113179175A - Real-time bandwidth prediction method and device for power communication network service - Google Patents

Real-time bandwidth prediction method and device for power communication network service Download PDF

Info

Publication number
CN113179175A
CN113179175A CN202110282173.7A CN202110282173A CN113179175A CN 113179175 A CN113179175 A CN 113179175A CN 202110282173 A CN202110282173 A CN 202110282173A CN 113179175 A CN113179175 A CN 113179175A
Authority
CN
China
Prior art keywords
real
bandwidth
time
information
prediction
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.)
Granted
Application number
CN202110282173.7A
Other languages
Chinese (zh)
Other versions
CN113179175B (en
Inventor
陈芳
李子凡
段明雄
赵星宇
孙雨潇
钱升起
张儒依
喻鹏
金卓军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
State Grid Information and Telecommunication Co Ltd
Beijing University of Posts and Telecommunications
Original Assignee
State Grid Information and Telecommunication Co Ltd
Beijing University of Posts and Telecommunications
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by State Grid Information and Telecommunication Co Ltd, Beijing University of Posts and Telecommunications filed Critical State Grid Information and Telecommunication Co Ltd
Priority to CN202110282173.7A priority Critical patent/CN113179175B/en
Publication of CN113179175A publication Critical patent/CN113179175A/en
Application granted granted Critical
Publication of CN113179175B publication Critical patent/CN113179175B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0896—Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities
    • 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
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
    • H04L43/0876—Network utilisation, e.g. volume of load or congestion level
    • H04L43/0888—Throughput

Abstract

The invention provides a real-time bandwidth prediction method and a real-time bandwidth prediction device for electric power communication network services, wherein the method comprises the following steps: performing bandwidth prediction of a target time period based on a trained real-time traffic prediction model to obtain predicted bandwidth information; correcting the predicted bandwidth information based on momentum to obtain corrected predicted bandwidth information; according to the corrected service flow information, bandwidth allocation is carried out by using a target bandwidth allocation model to obtain a target planning path; the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information. By constructing a real-time flow prediction model considering both long-term sample flow information and real-time sample flow information, the model can well take periodicity and real-time requirements into account, and after a prediction result output by the model is obtained, a momentum-based prediction correction algorithm is also provided, so that the prediction result can be effectively corrected, and the prediction precision is further improved.

Description

Real-time bandwidth prediction method and device for power communication network service
Technical Field
The invention relates to the technical field of power communication networks, in particular to a real-time bandwidth prediction method and a real-time bandwidth prediction device for power communication network services.
Background
With the rapid development of smart power grids and the continuous emergence of new power services, the reliable, efficient and stable operation of power communication networks is an important guarantee for the safety and stability of power systems.
The electric power communication network is a key ring for the safe operation of the power grid and the realization of the intellectualization of the power grid, and has been deeply advanced to each stage of power production, allocation, operation and management. Therefore, the establishment of a safe and reliable power communication network is a key link for the construction of the intelligent power grid. The method can accurately predict the traffic information of the service in real time, has great significance on bandwidth allocation, can avoid the situation that the bandwidth allocation is too small to meet the service bandwidth requirement and can also avoid resource waste caused by too much bandwidth allocation, so that the real-time and accurate service traffic prediction and bandwidth allocation are important guarantee for the safe and stable operation of the power communication network, are vital to the normal and stable operation of a data network for the real-time and accuracy of the service traffic prediction, and are also key technologies for improving the operation efficiency of the power network. At present, with the appearance of a series of new grid services, such as video conference services, scheduling automation services, and the like, the fluctuation of service flow is more complicated, and the existing service flow prediction model cannot be well fitted to the existing network flow.
Therefore, how to better realize real-time bandwidth prediction of power communication network services has become an urgent problem to be solved in the industry.
Disclosure of Invention
The invention provides a real-time bandwidth prediction method and a real-time bandwidth prediction device for electric power communication network services, which are used for solving the problem that the real-time bandwidth prediction of the electric power communication network services cannot be well realized in the prior art.
The invention provides a real-time bandwidth prediction method for electric power communication network service, which comprises the following steps:
performing bandwidth prediction of a target time period based on a trained real-time traffic prediction model to obtain predicted bandwidth information;
correcting the predicted bandwidth information based on momentum to obtain corrected predicted bandwidth information;
the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information.
Based on the real-time bandwidth prediction method of the power communication network service provided by the invention, after the step of obtaining the corrected predicted bandwidth information, the method further comprises the following steps:
inputting the corrected predicted bandwidth information into a target bandwidth allocation model to obtain a target planning path;
wherein the target bandwidth allocation model is constructed based on a DDQN algorithm.
Based on the real-time bandwidth prediction method of the power communication network service provided by the invention, the trained real-time traffic prediction model comprises the following steps: a plurality of long-term traffic information modules and a plurality of real-time traffic information modules;
the long-term flow information module carries out modeling on a long period for the prediction of the service flow information according to the long-term sample flow information;
the real-time flow information module carries out modeling on a short period according to the real-time sample flow information and the prediction of the service flow information;
and the training of the real-time flow prediction model is realized through the long-term flow information module and the real-time flow information module together, so that the trained real-time flow prediction model is obtained.
Based on the real-time bandwidth prediction method of the power communication network service provided by the invention, the step of correcting the predicted bandwidth information based on momentum to obtain the corrected predicted bandwidth information specifically comprises the following steps:
if the predicted bandwidth information at the current moment is larger than the predicted bandwidth information at the previous moment, adding a rising correction value to the predicted bandwidth information at the current moment to obtain corrected predicted bandwidth information;
or the predicted bandwidth information at the current moment is less than or equal to the predicted bandwidth information at the previous moment, adding a descending correction value to the predicted bandwidth information at the current moment to obtain corrected predicted bandwidth information.
Based on the real-time bandwidth prediction method of the power communication network service provided by the invention, the state space of the target bandwidth allocation model is as follows:
St＝(At,Bt,Ct)；
wherein A ist＝{A1,t,…,A|V|,tExpressing the number of services on each node at the time t in the network; b ist＝{B1,t,…,B|W|,tExpressing the number of services borne on each link at the moment t of the network; ctTo indicate location information of the current service.
Based on the real-time bandwidth prediction method of the power communication network service provided by the invention, the reward function of the target bandwidth allocation model is as follows:
Figure BDA0002979011090000031
wherein r is2Is a reward when the bandwidth allocated to the service is less than the predicted value of the bandwidth after the action is selected; r is2Is a reward that is fed back when the selected action a is an invalid action; r is3Is to a serviceAfter the paths are allocated, the network can meet the reward when the real-time bandwidth of the service is required.
The invention also provides a real-time bandwidth prediction device for the electric power communication network service, which comprises the following steps:
the prediction module is used for predicting the bandwidth of a target time period based on the trained real-time traffic prediction model to obtain predicted bandwidth information;
the correction module is used for carrying out momentum-based correction on the predicted bandwidth information to obtain corrected predicted bandwidth information;
the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information.
The device further comprises: a distribution module;
the distribution module is used for inputting the corrected predicted bandwidth information into a target bandwidth distribution model to obtain a target planning path;
wherein the target bandwidth allocation model is constructed based on a DDQN algorithm.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the steps of the real-time bandwidth prediction method of the power communication network service.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method for real-time bandwidth prediction of power communication network traffic as described in any one of the above.
According to the real-time bandwidth prediction method and device for the power communication network service, provided by the invention, the real-time flow prediction model which simultaneously considers the long-term sample flow information and the real-time sample flow information is constructed, so that the model can well take periodicity and real-time requirements into consideration, the prediction result is more accurate, and after the prediction result output by the model is obtained, a momentum-based prediction correction algorithm is further provided in the embodiment of the invention, the prediction result can be effectively corrected, and the prediction accuracy is further improved.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a schematic flowchart of a method for predicting a real-time bandwidth of a power communication network service according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a real-time traffic prediction model of an AUGRU according to an embodiment of the present invention;
fig. 3 is a network topology diagram of a simulation network a provided in the embodiment of the present invention;
fig. 4 is a delay diagram of an s1 service according to an embodiment of the present invention;
fig. 5 is a delay diagram of an s2 service according to an embodiment of the present invention;
fig. 6 is a delay diagram of an s3 service according to an embodiment of the present invention;
fig. 7 is a delay diagram of an s4 service according to an embodiment of the present invention;
fig. 8 is a schematic diagram of a real-time bandwidth prediction apparatus for a power communication network service of the power communication network service provided by the present invention;
fig. 9 is a schematic physical structure diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic flow chart of a method for predicting a real-time bandwidth of a power communication network service according to an embodiment of the present invention, as shown in fig. 1, including:
step S1, bandwidth prediction of a target time period is carried out based on a trained real-time traffic prediction model, and predicted bandwidth information is obtained;
specifically, the target time period described in the present invention refers to a specific time period that needs to be predicted, and the time period can be set according to personal needs.
The trained real-time traffic prediction model described in the invention is a prediction model based on the AUGRU, which can well process and predict sequence data, and for the sequence data with overlong sequence length, the accurate prediction can be carried out through an attention mechanism in the AUGRU.
Step S2, correcting the predicted bandwidth information based on momentum to obtain corrected predicted bandwidth information;
the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information.
Specifically, a large number of experiments show that the flow data obtained by predicting the real-time flow prediction model based on the AUGRU has some differences from the true value in terms of numerical values, but the overall change rule is consistent, so that the embodiment of the invention provides a correction method based on momentum by referring to a momentum algorithm, and the accuracy of prediction can be further improved.
After the embodiment of the invention constructs the real-time flow prediction model which simultaneously considers the long-term sample flow information and the real-time sample flow information, the bandwidth prediction can be accurately carried out, and the prediction result output by the model is simultaneously obtained, the embodiment of the invention also provides a prediction correction algorithm based on momentum, which can effectively correct the prediction result and further improve the prediction precision.
Based on any of the above embodiments, after the step of obtaining the corrected predicted bandwidth information, the method further includes:
inputting the corrected predicted bandwidth information into a target bandwidth allocation model to obtain a target planning path;
wherein the target bandwidth allocation model is constructed based on a DDQN algorithm.
Specifically, after the predicted bandwidth information is obtained, how to allocate a reasonable bandwidth and route to the service aiming at the predicted value is still considered, so that the service can be safely, reliably and efficiently transmitted to the destination address in the power communication network.
Therefore, in the embodiment of the invention, a target bandwidth allocation model is constructed based on the DDQN algorithm, and the optimal strategy selection applied to the specific state space and the action space of the target bandwidth allocation model can be found through the thought of the Q learning algorithm, so that the target planning path is obtained.
According to the embodiment of the invention, through the target bandwidth allocation model, after the predicted bandwidth is obtained, the bandwidth allocation is further carried out, so that the problem of practical application is solved more favorably.
Based on any of the above embodiments, the trained real-time traffic prediction model includes: a plurality of long-term traffic information modules and a plurality of real-time traffic information modules;
the long-term flow information module carries out modeling on a long period for the prediction of the service flow information according to the long-term sample flow information;
the real-time flow information module carries out modeling on a short period according to the real-time sample flow information and the prediction of the service flow information;
and the training of the real-time flow prediction model is realized through the long-term flow information module and the real-time flow information module together, so that the trained real-time flow prediction model is obtained.
Specifically, the embodiment of the invention provides a prediction model based on the AUGRU, a real-time sequence is added during modeling, and an attention mechanism is used for replacing an update door mechanism in the GRU to model sequence information. Because the periodicity of the service in the power communication network is generally long, usually in units of days, the AUGRU is a good choice to predict the size of the service in real time. .
The AUGRU directly controls the updating of the hidden state by using the attention coefficient, and the updating formula of the hidden state of the AUGRU is as follows:
Figure BDA0002979011090000071
Figure BDA0002979011090000072
fig. 2 is a schematic structural diagram of a real-time traffic prediction model of an augur according to an embodiment of the present invention, and as shown in fig. 2, the model is structurally divided into two parts, namely a long-term traffic information module and a real-time traffic information module. The long-term traffic information module models traffic information to be predicted over a long period, such as traffic information for the past week. The real-time traffic information module models a real-time module of traffic, for example, the last 30 minutes. The two modules enable the model to learn the flow change rule of the service in a long period, and can adjust the long period result to a certain extent according to the sudden change of the flow of the service in a short period, so that the model takes the prediction real-time property and periodicity into account. Next, the manner of calculation for the attition score in the model of the embodiment of the present invention is as follows:
attention score＝sigmoid(Ht*ht)
Htis the output of the last unit, h, after the long-term traffic information is processedtThe traffic information is input to the real-time traffic information module, i.e., traffic information of the service at time t. It should be noted that, in the long-term sequence module according to the embodiment of the present invention, the time interval of the input sequence sampling is 15 minutes, and the time interval of the real-time traffic sequence sampling is 2 minutes, so that the model can more accurately grasp the real-time traffic information change of the sequence. In addition, the output of the last unit of the real-time traffic information module is the predicted real-time traffic of the service.
According to the embodiment of the invention, the bandwidth prediction can be accurately carried out by simultaneously considering the real-time traffic prediction model of the long-term sample traffic information and the real-time sample traffic information.
Based on any of the above embodiments, the step of performing momentum-based correction on the predicted bandwidth information to obtain corrected predicted bandwidth information specifically includes:
if the predicted bandwidth information at the current moment is larger than the predicted bandwidth information at the previous moment, adding a rising correction value to the predicted bandwidth information at the current moment to obtain corrected predicted bandwidth information;
or the predicted bandwidth information at the current moment is less than or equal to the predicted bandwidth information at the previous moment, adding a descending correction value to the predicted bandwidth information at the current moment to obtain corrected predicted bandwidth information.
Specifically, through a large number of experiments, the embodiment of the invention discovers that the flow data obtained by predicting the real-time flow prediction model based on the AUGRU has some differences from the true value in the value, but the overall change rule is consistent. Based on this point, the embodiment of the present invention proposes a correction method based on momentum by referring to a momentum algorithm, and the specific formula is as follows:
aup＝μupaup+(1-μup)θ
adown＝μdownadown+(1-μdown)θ
wherein, aupIndicating a rising correction value, adownIndicating a falling correction value, muupThe retention in the ascending direction is expressed as a resistance value, μdownIndicating the degree of retention in the descending direction.
Next, how to adjust the prediction result based on the AUGRU by using the prediction correction mechanism is described as follows:
Figure BDA0002979011090000081
after the LSTM model is used for predicting, the flow value at the current moment in the prediction result is compared with the flow value at the previous moment, and if the flow value at the current moment is predictedIf the value is larger than the value of the previous moment, adding a to the flow value of the current momentupMake a correction while applying aupUpdating the value; similarly, if the value at the current moment is smaller than the value at the previous moment, a is added to the flow value at the current momentdownMake a correction while applying adownThe value is updated.
When the bandwidth of the service of the power communication network is predicted, the flow of a certain service is increased suddenly at a certain time, so that the predicted bandwidth and the actual bandwidth deviate, a set of reasonable bandwidth compensation mechanism is established, the mechanism can be adjusted to a certain extent when the bandwidth is allocated to the service according to the predicted and actual flow deviation, the problem of a large amount of packet loss when the service flow is increased due to insufficient allocated bandwidth is avoided, and the real-time dynamic adjustment of the service bandwidth by the power communication system is realized.
Based on any of the above embodiments, after the trained real-time traffic prediction model is established, how to allocate reasonable bandwidth and route to the service aiming at the predicted value is also considered, so that the service can be safely, reliably and efficiently transmitted to the destination address in the power communication network. Aiming at the problem, the embodiment of the invention provides a service bandwidth real-time distribution method based on DDQN.
The DDQN is widely applied as a combination of deep learning and reinforcement learning, and meanwhile, the DDQN algorithm solves the problem that the Q value of the traditional DQN algorithm is excessively high in estimation. DDQN is an extension of Q learning algorithm, which is essentially a table lookup method, and is not applicable when the state space S is very large. The DDQN represents the action state cost function by means of a neural network, enabling the DDQN to be applied to state spaces or more complex scenarios of action spaces.
Like DQN, DDQN also requires a well-defined state space and reward functions in accordance with the context of embodiments of the present invention. First, the embodiments of the present invention need to define a good state space. Let At＝{A1,t，…,A|V|,tDenotes the traffic volume at each node at time t in the network, Bt＝{B1,t,…，B|W|,tIndicates the service carried by each link at the time t of the networkThe number of the cells. The state of the network can be used (A)t,Bt) And (4) defining. In addition, the state of the network will change after a hop node is selected by the service, so that for a certain service, when the service is in a different node, the next hop selectable path is different, and therefore the embodiment of the invention uses CtTo indicate the location information of the current service, which can be represented by a one-hot code, the length of the vector being the number of nodes. If the current service is at the ith node, the ith position in the vector is 1, and the rest are 0. The state space of the DDQN can be represented by S up to this pointt＝(At，Bt,Ct)。
The purpose of the reward function is to evaluate the value after selecting an action in a certain state. Because the embodiment of the invention aims to ensure the bandwidth of the service and simultaneously considers the service delay, the design of the reward function not only considers the rationality of service bandwidth allocation but also ensures that the service can reach the destination address within the acceptable delay. Therefore, the embodiment of the invention takes the bandwidth allocation requirement of whether the service is guaranteed or not as a reward function, and also considers the priority of the service, namely under the limited bandwidth, firstly allocating the bandwidth resource for the service with high priority, and secondly considering the service with low priority.
Meanwhile, since there is a need for the index in the actual power production process, a constraint condition for each index needs to be set. For this purpose, a reward function for reinforcement learning is constructed:
Figure BDA0002979011090000101
wherein r is2Is a reward when the bandwidth allocated to the service after the selection action is less than the predicted value of bandwidth, r2Is a reward, typically a large negative number, that is fed back when the selected action a is an invalid action to avoid the model continuing to select such action. r is3After the path is distributed to the service, the network can meet the real-time bandwidth requirement of the service, and the main factor considered at the moment is the average service delay, so that the reward value at the moment isAverage traffic delay.
In addition, a cost function is defined, and the purpose of reinforcement learning is to find an optimal decision in order to maximize the value of the selection action, and in the DDQN, max is found by using MainNet firstlya′Q(s′，a′；θi) Action (theta)iIs a parameter of MainNet) and then go to TargetNet to find the Q value of this Action to form the Target Q value, which is not necessarily the largest in TargetNet, so that a suboptimal Action that leads to a high valuation can be avoided. The Loss funcition to learn finally is:
L(θ)＝E[(TargetQ-Q(s,a；θi))2]
Figure BDA0002979011090000102
in the above equation, L (θ) represents a loss function in DDQN, and γ is a discount coefficient. Q (s, a; theta)i) Is a function of the action state cost.
When the embodiment of the invention distributes the service paths, the service route planning algorithm based on the deep reinforcement learning algorithm is adopted, and when the DDQN algorithm is adopted for service route planning, the action space and the state space of the DDQN algorithm and the reward function of the DDQN algorithm are defined based on the actual service environment, so that the optimal planning path can be found by converging the DDQN algorithm by reasonably setting the reward function.
In another embodiment, in order to verify the effectiveness of the method proposed by the present patent, a simulation scenario diagram is shown as follows, and the power communication system has 17 nodes and 15 communication links. There are a total of 6 services on the communication network, which are { [0,10], [0,15], [1,15], [2,12], [3,7], [4,14] }. The experimental setup carries a total of 4 traffic items above, s1, s2, s3 and s4, with high to low priority. Fig. 3 is a topological diagram of a simulated network a network according to an embodiment of the present invention, and as shown in fig. 3, a Naive-Bandwidth Allocation Algorithm (N-BA), a High-order Moving Average prediction-Bandwidth Allocation Algorithm (HMAM-DBA), an augur-based real-time traffic prediction Model, and a DDQN-based real-time Bandwidth Allocation method (augur-DDQN) are respectively adopted in the embodiment of the present invention.
Fig. 4 is a delay diagram of the s1 service provided by the embodiment of the present invention, as shown in fig. 4, for the s1 high priority service, since the N-BA algorithm adopts a fixed bandwidth allocation mode, for the high priority service, it always can preferentially meet the bandwidth requirement. Compared with the N-BA algorithm, the HMAM-DBA algorithm can increase the delay of the high-priority service by about 25 percent. The AUGRU-DDQN algorithm provided by the invention adopts the AUGRU algorithm to predict the bandwidth in real time, and then after prediction correction, the prediction accuracy is improved to a certain extent compared with the prediction accuracy of the HMAM-DBA algorithm, and when bandwidth allocation is carried out, the optimal bandwidth allocation formula is searched by a DDQN real-time bandwidth allocation model, so that the time delay is greatly shortened compared with the HMAM-DBA algorithm, and the overall performance is similar to that of the N-BA algorithm.
Fig. 5 is a delay diagram of the s2 service provided by the embodiment of the present invention, and as shown in fig. 5, compared with the N-BA algorithm, the HMAM-DBA algorithm and the AUGRU-DDQN algorithm are significantly improved in delay performance. Compared with an N-BA algorithm, the delay performance of the HMAM-DBA algorithm is improved by 33% -40% approximately under the condition of higher network load; the AUGRU-DDQN algorithm is improved by about 37% -44%. This is due to the fact that the N-BA algorithm allocates too much bandwidth to high priority traffic when the network load is high.
Fig. 6 is a delay diagram of s3 service according to an embodiment of the present invention, and as shown in fig. 6, for s3 service, compared with the N-BA algorithm, the HMAM-DBA algorithm has a delay decreased by about 14%, and the LSTM-DBA algorithm has a delay decreased by about 17%. The reasons for the HMAM-DBA algorithm and the AUGRU-DDQN algorithm to obtain the improvement of the delay performance on the s3 service are mainly two points: firstly, the lowest allocated bandwidth is guaranteed through real-time bandwidth prediction and setting as s3 service through a threshold; secondly, after the first bandwidth allocation is finished, the residual bandwidth is effectively utilized through secondary allocation. Thus, in general, the AUGRU-DDQN algorithm proposed herein performs well at both low and high load on the communication network.
Fig. 7 is a time delay diagram of s4 service provided by the embodiment of the present invention, and as shown in fig. 7, for s4 service with low priority, the performance of the N-DBA algorithm in reducing the time delay is significantly better than that of the HMAM-BA algorithm and the AUGRU-DDQN algorithm. Especially when the communication network is under high load, the delay is reduced by about 65% compared to the latter two. This is because the fixed bandwidth allocation of the N-DBA algorithm always reserves a certain bandwidth for the s4 bandwidth, which also results in the overall performance of the N-DBA algorithm on the s1-s3 service being inferior to the latter two. Although the HMAM-BA algorithm and the AUGRU-DDQN algorithm have high latency on s4 traffic, the results are in an acceptable range because the latency requirement for s4 traffic is not high. The AUGRU-DDQN algorithm performs better than the HMAM-BA algorithm in high load situations because of the greater accuracy of the AUGRU prediction and the benefit of the thresholds set for s1-s 3.
Fig. 8 is a schematic diagram of a real-time bandwidth prediction apparatus for power communication network service of the power communication network service provided in the present invention, as shown in fig. 8, including: the prediction module 810 is configured to perform bandwidth prediction for a target time period based on the trained real-time traffic prediction model to obtain predicted bandwidth information;
the correction module 820 is configured to perform momentum-based correction on the predicted bandwidth information to obtain corrected predicted bandwidth information;
the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information.
The device further comprises: a distribution module;
the distribution module is used for inputting the corrected predicted bandwidth information into a target bandwidth distribution model to obtain a target planning path;
wherein the target bandwidth allocation model is constructed based on a DDQN algorithm.
According to the embodiment of the invention, a real-time flow prediction model which simultaneously considers long-term sample flow information and real-time sample flow information is constructed, so that the model can well take periodicity and real-time requirements into account, the prediction result is more accurate, and after the prediction result output by the model is obtained, a momentum-based prediction correction algorithm is also provided in the embodiment of the invention, the prediction result can be effectively corrected, and the prediction precision is further improved.
Fig. 9 is a schematic physical structure diagram of an electronic device provided in the present invention, and as shown in fig. 9, the electronic device may include: a processor (processor)910, a communication Interface (Communications Interface)920, a memory (memory)990, and a communication bus 940, wherein the processor 910, the communication Interface 920, and the memory 990 communicate with each other via the communication bus 940. Processor 910 may invoke logic instructions in memory 990 to perform a method of real-time bandwidth prediction for power communications network traffic, the method comprising: performing bandwidth prediction of a target time period based on a trained real-time traffic prediction model to obtain predicted bandwidth information; correcting the predicted bandwidth information based on momentum to obtain corrected predicted bandwidth information; the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information.
Furthermore, the logic instructions in the memory 990 may be implemented in software functional units, and may be stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, enable the computer to perform a method for real-time bandwidth prediction of power communication network traffic provided by the above methods, the method comprising: performing bandwidth prediction of a target time period based on a trained real-time traffic prediction model to obtain predicted bandwidth information; correcting the predicted bandwidth information based on momentum to obtain corrected predicted bandwidth information; the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information.
In yet another aspect, the present invention also provides a non-transitory computer readable storage medium, on which a computer program is stored, the computer program being implemented by a processor to perform the real-time bandwidth prediction method for power communication network service provided by the above embodiments, the method including: performing bandwidth prediction of a target time period based on a trained real-time traffic prediction model to obtain predicted bandwidth information; correcting the predicted bandwidth information based on momentum to obtain corrected predicted bandwidth information; the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (10)

1. A real-time bandwidth prediction method for electric power communication network service is characterized by comprising the following steps:
performing bandwidth prediction of a target time period based on a trained real-time traffic prediction model to obtain predicted bandwidth information;
correcting the predicted bandwidth information based on momentum to obtain corrected predicted bandwidth information;
the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information.
2. The method of claim 1, wherein after the step of obtaining the modified predicted bandwidth information, the method further comprises:
inputting the corrected predicted bandwidth information into a target bandwidth allocation model to obtain a target planning path;
wherein the target bandwidth allocation model is constructed based on a DDQN algorithm.
3. The method according to claim 1, wherein the trained real-time traffic prediction model comprises: a plurality of long-term traffic information modules and a plurality of real-time traffic information modules;
the long-term flow information module carries out modeling on a long period for the prediction of the service flow information according to the long-term sample flow information;
the real-time flow information module carries out modeling on a short period according to the real-time sample flow information and the prediction of the service flow information;
and the training of the real-time flow prediction model is realized through the long-term flow information module and the real-time flow information module together, so that the trained real-time flow prediction model is obtained.
4. The method for predicting the real-time bandwidth of the power communication network service according to claim 1, wherein the step of performing momentum-based correction on the predicted bandwidth information to obtain the corrected predicted bandwidth information specifically comprises:
if the predicted bandwidth information at the current moment is larger than the predicted bandwidth information at the previous moment, adding a rising correction value to the predicted bandwidth information at the current moment to obtain corrected predicted bandwidth information;
or the predicted bandwidth information at the current moment is less than or equal to the predicted bandwidth information at the previous moment, adding a descending correction value to the predicted bandwidth information at the current moment to obtain corrected predicted bandwidth information.
5. The method according to claim 2, wherein the state space of the target bandwidth allocation model is:
St＝(At,Bt,Ct)；
wherein A ist＝{A1,t,…,A|V|,tExpressing the number of services on each node at the time t in the network; b ist＝{B1,t,…,B|W|,tExpressing the number of services borne on each link at the moment t of the network;Ctto indicate location information of the current service.
6. The method for predicting bandwidth of power communication network service according to claim 5, wherein the reward function of the target bandwidth allocation model is:
Figure FDA0002979011080000021
wherein r is2Is a reward when the bandwidth allocated to the service is less than the predicted value of the bandwidth after the action is selected; r is2Is a reward that is fed back when the selected action a is an invalid action; r is3After the path is distributed to the service, the network can meet the real-time bandwidth requirement of the service.
7. An apparatus for predicting real-time bandwidth of power communication network traffic, comprising:
the prediction module is used for predicting the bandwidth of a target time period based on the trained real-time traffic prediction model to obtain predicted bandwidth information;
the correction module is used for carrying out momentum-based correction on the predicted bandwidth information to obtain corrected predicted bandwidth information;
the trained real-time flow prediction model is obtained by training together based on long-term sample flow information and real-time sample flow information.
8. The apparatus for real-time bandwidth prediction of electric power communication network traffic according to claim 7, further comprising: a distribution module;
the distribution module is used for inputting the corrected predicted bandwidth information into a target bandwidth distribution model to obtain a target planning path;
wherein the target bandwidth allocation model is constructed based on a DDQN algorithm.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program performs the steps of the method for real-time bandwidth prediction of power communication network traffic according to any of claims 1 to 6.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the steps of the method for real-time bandwidth prediction of power communication network traffic according to any of claims 1 to 6.
CN202110282173.7A 2021-03-16 2021-03-16 Real-time bandwidth prediction method and device for power communication network service Active CN113179175B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110282173.7A CN113179175B (en) 2021-03-16 2021-03-16 Real-time bandwidth prediction method and device for power communication network service

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110282173.7A CN113179175B (en) 2021-03-16 2021-03-16 Real-time bandwidth prediction method and device for power communication network service

Publications (2)

Publication Number Publication Date
CN113179175A true CN113179175A (en) 2021-07-27
CN113179175B CN113179175B (en) 2022-12-20

Family

ID=76922083

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110282173.7A Active CN113179175B (en) 2021-03-16 2021-03-16 Real-time bandwidth prediction method and device for power communication network service

Country Status (1)

Country Link
CN (1) CN113179175B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113904948A (en) * 2021-11-12 2022-01-07 福州大学 5G network bandwidth prediction system and method based on cross-layer multi-dimensional parameters
CN114465915A (en) * 2021-12-22 2022-05-10 天翼云科技有限公司 CDN bandwidth prediction method, device and system and electronic equipment
CN114584494A (en) * 2022-04-22 2022-06-03 缀初网络技术(上海)有限公司 Method for measuring actual available bandwidth in edge cloud network

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102938742A (en) * 2012-11-08 2013-02-20 广东电网公司电力调度控制中心 Communication bandwidth forecasting method and device based on power business requirements
CN104079447A (en) * 2013-12-29 2014-10-01 国家电网公司 Bandwidth prediction method based on adaptive filtering
CN104463351A (en) * 2014-11-15 2015-03-25 国家电网公司 Communication bandwidth prediction method and device based on power business requirements
CN105071992A (en) * 2015-08-26 2015-11-18 国家电网公司 Method for predicting power distribution/utilization service communication bandwidth of transformer substation
US20180137412A1 (en) * 2016-11-16 2018-05-17 Cisco Technology, Inc. Network traffic prediction using long short term memory neural networks
CN109510715A (en) * 2017-09-14 2019-03-22 中国电信股份有限公司 Bandwidth allocation methods, device, data center and storage medium
CN110417650A (en) * 2019-08-02 2019-11-05 广东中兴新支点技术有限公司 Multilink data distribution method, device and electronic equipment
CN111371626A (en) * 2020-03-20 2020-07-03 南京大学 Bandwidth prediction method based on neural network
CN111464323A (en) * 2019-01-18 2020-07-28 北京沃东天骏信息技术有限公司 Node bandwidth scheduling method and scheduling device

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102938742A (en) * 2012-11-08 2013-02-20 广东电网公司电力调度控制中心 Communication bandwidth forecasting method and device based on power business requirements
CN104079447A (en) * 2013-12-29 2014-10-01 国家电网公司 Bandwidth prediction method based on adaptive filtering
CN104463351A (en) * 2014-11-15 2015-03-25 国家电网公司 Communication bandwidth prediction method and device based on power business requirements
CN105071992A (en) * 2015-08-26 2015-11-18 国家电网公司 Method for predicting power distribution/utilization service communication bandwidth of transformer substation
US20180137412A1 (en) * 2016-11-16 2018-05-17 Cisco Technology, Inc. Network traffic prediction using long short term memory neural networks
CN109510715A (en) * 2017-09-14 2019-03-22 中国电信股份有限公司 Bandwidth allocation methods, device, data center and storage medium
CN111464323A (en) * 2019-01-18 2020-07-28 北京沃东天骏信息技术有限公司 Node bandwidth scheduling method and scheduling device
CN110417650A (en) * 2019-08-02 2019-11-05 广东中兴新支点技术有限公司 Multilink data distribution method, device and electronic equipment
CN111371626A (en) * 2020-03-20 2020-07-03 南京大学 Bandwidth prediction method based on neural network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
BIAO ZHANG等: "Neural Machine Translation With GRU-Gated Attention Model", 《IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS》 *
梁根等: "多业务流量预测动态带宽分配方法", 《计算机工程与应用》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113904948A (en) * 2021-11-12 2022-01-07 福州大学 5G network bandwidth prediction system and method based on cross-layer multi-dimensional parameters
CN113904948B (en) * 2021-11-12 2023-11-03 福州大学 5G network bandwidth prediction system and method based on cross-layer multidimensional parameters
CN114465915A (en) * 2021-12-22 2022-05-10 天翼云科技有限公司 CDN bandwidth prediction method, device and system and electronic equipment
CN114584494A (en) * 2022-04-22 2022-06-03 缀初网络技术(上海)有限公司 Method for measuring actual available bandwidth in edge cloud network

Also Published As

Publication number Publication date
CN113179175B (en) 2022-12-20

Similar Documents

Publication Publication Date Title
CN113179175B (en) Real-time bandwidth prediction method and device for power communication network service
CN111953758A (en) Method and device for computing unloading and task migration of edge network
CN111211987B (en) Method and system for dynamically adjusting flow in network, electronic equipment and storage medium
CN110247795B (en) Intent-based cloud network resource service chain arranging method and system
CN107948083A (en) A kind of SDN data centers jamming control method based on enhancing study
CN113098714A (en) Low-delay network slicing method based on deep reinforcement learning
WO2023124947A1 (en) Task processing method and apparatus, and related device
CN113395723B (en) 5G NR downlink scheduling delay optimization system based on reinforcement learning
CN115374853A (en) Asynchronous federal learning method and system based on T-Step polymerization algorithm
CN111740925B (en) Deep reinforcement learning-based flow scheduling method
CN114697229A (en) Construction method and application of distributed routing planning model
CN110198280A (en) A kind of SDN link allocation method based on BP neural network
CN115907038A (en) Multivariate control decision-making method based on federated split learning framework
CN114090108B (en) Method and device for executing computing task, electronic equipment and storage medium
CN111585915A (en) Long and short flow balanced transmission method and system, storage medium and cloud server
CN114501667A (en) Multi-channel access modeling and distributed implementation method considering service priority
CN112231117B (en) Cloud robot service selection method and system based on dynamic vector hybrid genetic algorithm
CN110392377B (en) 5G ultra-dense networking resource allocation method and device
Che et al. Deep reinforcement learning in M2M communication for resource scheduling
CN115314399B (en) Data center flow scheduling method based on inverse reinforcement learning
CN115150335B (en) Optimal flow segmentation method and system based on deep reinforcement learning
CN111340192A (en) Network path allocation model training method, path allocation method and device
CN115499365A (en) Route optimization method, device, equipment and medium
CN112600827B (en) Virtual service migration method and system based on incremental minimum cost maximum flow
Kong et al. Energy saving strategy for task migration based on genetic algorithm

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