Flow prediction method and system for power optical communication network, and storage medium

Abstract

The invention discloses a flow prediction method and system for a power optical communication network and a storage medium, and the method comprises the steps: obtaining target flow related data of a target power optical communication network, and obtaining each flow related mode component; wherein the flow related mode components reflect various factors which influence the target flow related data, the related weight of each flow related mode component is obtained, and then the flow key mode component is determined; and sequentially inputting the first traffic key mode components and the second traffic key mode components obtained by division into a target communication traffic prediction model constructed by a long-short-term memory network technology to obtain a component prediction result of each second traffic key mode component, and obtaining a target communication traffic prediction result of each second traffic key mode component based on the related weights and the corresponding component prediction results. And obtaining a target flow prediction result. According to the traffic prediction method and system of the power optical communication network and the storage medium provided by the embodiment of the invention, the accuracy of traffic prediction of the power optical communication network is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a traffic prediction method, system and storage medium for a power optical communication network. Background Art

[0002] The power optical communication network is a dedicated optical fiber communication infrastructure for the power system. Using optical fiber as the transmission medium, it bears the communication requirements of power grid production, control and management services. By predicting the traffic of the power optical communication network and analyzing the traffic change trend, the routing strategy can be dynamically adjusted or a backup channel can be enabled to prevent network paralysis caused by sudden traffic.

[0003] However, since the power optical communication network is easily affected by various factors, the existing technology directly predicts the traffic of the power optical communication network without considering the influence of external factors, resulting in low accuracy of traffic prediction for the power optical communication network.

[0004] Therefore, how to improve the accuracy of traffic prediction for the power optical communication network has become a technical problem that needs to be urgently solved by those skilled in the art. Summary of the Invention

[0005] The present invention provides a traffic prediction method, system and storage medium for a power optical communication network to solve the technical problem that since the power optical communication network is easily affected by various factors, the existing technology directly predicts the traffic of the power optical communication network without considering the influence of external factors, resulting in low accuracy of traffic prediction for the power optical communication network.

[0006] To solve the above technical problem, an embodiment of the present invention provides a traffic prediction method for a power optical communication network.

[0007] Obtain target traffic-related data of a target power optical communication network; wherein, the target traffic-related data includes traffic data and external influence data; perform modal decomposition on each target traffic-related sequence data constructed by the target traffic-related data to obtain each traffic-related modal component of each target traffic-related data; wherein, the traffic-related modal component reflects each factor that affects the target traffic-related data; Based on the importance analysis result of each traffic-related modal component, determine the corresponding correlation weight; based on the correlation weights of each traffic-related modal component of each target traffic-related data, determine the traffic key modal component of each target traffic-related data; Partition the dataset for all the traffic key modal components to obtain the first traffic key modal component and the second traffic key modal component; sequentially input the first traffic key modal component and the second traffic key modal component into the target communication traffic prediction model constructed by long short-term memory network technology to obtain the component prediction results of each of the second traffic key modal components; Based on the relevant weights of each of the second traffic key modal components and the corresponding component prediction results, obtain the target traffic prediction result of the target power optical communication network.

[0008] As one of the preferred solutions, determining the traffic key modal component of each target traffic-related data based on the relevant weights of each traffic-related modal component of each target traffic-related data includes: Perform a descending order sorting on the relevant weights of all the traffic-related modal components of each target traffic-related data to obtain the component weight sequence of each target traffic-related data; Input the relevant weights of each traffic-related modal component in the component weight sequence into the cumulative contribution calculation expression to calculate the cumulative contribution sequence corresponding to the component weight sequence; wherein, the cumulative contribution calculation expression is designed to sequentially accumulate the descending order arrangement results of the relevant weights of all the traffic-related modal components of each target traffic-related data; Taking the first element greater than the preset contribution rate threshold in the cumulative contribution sequence as the segmentation point to obtain the traffic key modal component of each target traffic-related data.

[0009] As one of the preferred solutions, the external influence data at least includes grid load data, optical fiber status data, electromagnetic field data, and temperature and humidity data.

[0010] As one of the preferred solutions, determining the corresponding relevant weight based on the importance analysis result of each traffic-related modal component includes: Input all the traffic-related modal components of each target traffic-related data into the random forest algorithm to calculate the importance scores of each traffic-related modal component; Perform a normalization process on the importance scores of all the traffic-related modal components of each target traffic-related data to obtain the relevant weights of each traffic-related modal component.

[0011] As one of the preferred solutions, the cumulative contribution calculation expression is designed as: wherein, is the i-th element in the cumulative contribution sequence, is the j-th element in the component weight sequence.

[0012] As one of the preferred solutions, the step of sequentially inputting the first traffic key modal component and the second traffic key modal component into the target communication traffic prediction model to obtain the component prediction result of each second traffic key modal component includes: Input the first traffic key modal component obtained from the historical target traffic-related data into the first target communication traffic prediction model for prediction training; Use the second target communication traffic prediction model obtained from the prediction training to perform prediction analysis on the second traffic key modal component obtained from the real-time target traffic-related data, and obtain the component prediction result of each second traffic key modal component.

[0013] As one of the preferred solutions, the method further includes: During the prediction training process, use the sparrow search algorithm to feedback and optimize the parameters of the first target communication traffic prediction model to obtain the target parameter combination of the second target communication traffic prediction model.

[0014] As one of the preferred solutions, the method further includes: Calculate each performance evaluation index of the second target communication traffic prediction model based on the difference characteristics between the target traffic prediction result of the real-time target traffic-related data and the corresponding target monitoring result; Use all the performance evaluation indexes to feedback and optimize the target parameter combination of the second target communication traffic prediction model.

[0015] As one of the preferred solutions, the performance evaluation indexes at least include root mean square error, mean absolute error, mean absolute percentage error, and coefficient of determination.

[0016] Another embodiment of the present invention provides a traffic prediction system for a power optical communication network, including: A data modal decomposition module, configured to obtain target traffic-related data of a target power optical communication network; wherein, the target traffic-related data includes traffic data and external influence data; perform modal decomposition on each target traffic-related sequence data constructed from the target traffic-related data to obtain each traffic-related modal component of each target traffic-related data; wherein, the traffic-related modal component reflects each factor that affects the target traffic-related data; A data modality screening module, configured to determine corresponding correlation weights based on the importance analysis results of each of the traffic-related modality components; and determine the traffic key modality components of each of the target traffic-related data based on the correlation weights of each of the traffic-related modality components of each of the target traffic-related data. A key modality prediction module, configured to partition the dataset of all the traffic key modality components to obtain a first traffic key modality component and a second traffic key modality component; and sequentially input the first traffic key modality component and the second traffic key modality component into a target communication traffic prediction model constructed by long short-term memory network technology to obtain the component prediction results of each of the second traffic key modality components. A prediction result reconstruction module, configured to obtain the target traffic prediction result of the target power optical communication network based on the correlation weights of each of the second traffic key modality components and the corresponding component prediction results.

[0017] As one of the preferred solutions, determining the traffic key modality components of each of the target traffic-related data based on the correlation weights of each of the traffic-related modality components of each of the target traffic-related data includes: Sorting the correlation weights of all the traffic-related modality components of each of the target traffic-related data in descending order to obtain a component weight sequence of each of the target traffic-related data. Inputting the correlation weights of each of the traffic-related modality components in the component weight sequence into a cumulative contribution calculation expression to calculate a cumulative contribution sequence corresponding to the component weight sequence; wherein, the cumulative contribution calculation expression is designed to sequentially accumulate the descending order results of the correlation weights of all the traffic-related modality components of each of the target traffic-related data. Using the first element greater than a preset contribution rate threshold in the cumulative contribution sequence as a segmentation point to obtain the traffic key modality components of each of the target traffic-related data.

[0018] As one of the preferred solutions, the external influence data at least includes grid load data, optical fiber status data, electromagnetic field data, and temperature and humidity data.

[0019] As one of the preferred solutions, determining the corresponding correlation weights based on the importance analysis results of each of the traffic-related modality components includes: Inputting all the traffic-related modality components of each of the target traffic-related data into a random forest algorithm to calculate the importance scores of each of the traffic-related modality components. Normalize the importance scores of all the traffic-related modal components of each of the target traffic-related data to obtain the correlation weights of each of the traffic-related modal components.

[0020] As one of the preferred solutions, the cumulative contribution calculation expression is designed as: Wherein, is the i-th element in the cumulative contribution sequence, is the j-th element in the component weight sequence.

[0021] As one of the preferred solutions, the step of sequentially inputting the first traffic key modal component and the second traffic key modal component into the target communication traffic prediction model to obtain the component prediction result of each of the second traffic key modal components includes: Input the first traffic key modal component obtained from the historical target traffic-related data into the first target communication traffic prediction model for prediction training; Use the second target communication traffic prediction model obtained from the prediction training to perform prediction analysis on the second traffic key modal component obtained from the real-time target traffic-related data to obtain the component prediction result of each of the second traffic key modal components.

[0022] As one of the preferred solutions, the system further includes: During the prediction training process, use the sparrow search algorithm to perform feedback optimization on the parameters of the first target communication traffic prediction model to obtain the target parameter combination of the second target communication traffic prediction model.

[0023] As one of the preferred solutions, the system further includes: Calculate each performance evaluation index of the second target communication traffic prediction model based on the difference characteristics between the target traffic prediction result of the real-time target traffic-related data and the corresponding target monitoring result; Use all the performance evaluation indexes to perform feedback optimization on the target parameter combination of the second target communication traffic prediction model.

[0024] As one of the preferred solutions, the performance evaluation indexes at least include root mean square error, mean absolute error, mean absolute percentage error, and coefficient of determination.

[0025] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the device where the computer-readable storage medium is located executes the computer program, it implements the traffic prediction method of a power optical communication network as described above.

[0026] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following: Obtain the target traffic-related data of the target power optical communication network, and at the same time consider the traffic data and external influence data; perform modal decomposition on each target traffic-related sequence data constructed from the target traffic-related data to obtain each traffic-related modal component of each target traffic-related data. By modal decomposition, the mixed signal is separated into independent modal components, solving the problem that traditional methods ignore external factors. Each modal component corresponds to a potential influencing factor; by screening the key traffic modal components, the interference of noise components can be avoided, the error of subsequent traffic prediction for the target power optical communication network can be reduced, and the accuracy of traffic prediction for the power optical communication network is improved; after dividing the data set, use the long short-term memory network technology for phased prediction. The first key traffic modal component obtained from historical data is used to train and optimize the basic model, and the second key traffic modal component obtained from real-time data is used for real-time prediction. By setting different weights for different component prediction results according to the importance analysis result of the components, the target traffic prediction result of the target power optical communication network is obtained, further improving the accuracy of traffic prediction for the power optical communication network. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of a traffic prediction method for a power optical communication network in one embodiment of the present invention; Figure 2 It is a structural block diagram of a traffic prediction system for a power optical communication network in one embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0029] In the description of the present application, the terms "first", "second", "third", etc. are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined with "first", "second", "third", etc. may explicitly or implicitly include one or more of such features. In the description of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0030] In the description of the present application, it should be noted that unless otherwise clearly specified and defined, the terms "installation", "connection", and "linkage" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for the purpose of illustration, rather than indicating or implying that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0031] In the description of the present application, it should be noted that unless otherwise defined, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0032] An embodiment of the present invention provides a method for predicting the traffic of a power optical communication network. Specifically, please refer to Figure 1 , Figure 1 which shows a flowchart of a method for predicting the traffic of a power optical communication network in one of the embodiments of the present invention.

[0033] The flowchart of a method for predicting the traffic of a power optical communication network in one of the embodiments of the present invention includes the following steps S1 to S4, specifically as follows: Step S1: Obtain the target traffic-related data of the target power optical communication network; wherein, the target traffic-related data includes traffic data and external influence data; perform modal decomposition on each target traffic-related sequence data constructed from the target traffic-related data to obtain each traffic-related modal component of each target traffic-related data; wherein, the traffic-related modal component reflects each factor that affects the target traffic-related data.

[0034] Step S2: Determine the corresponding correlation weights based on the importance analysis results of each traffic-related modal component; determine the traffic key modal component of each target traffic-related data based on the correlation weights of each traffic-related modal component of each target traffic-related data.

[0035] Step S3: Divide the dataset for all traffic key modal components to obtain the first traffic key modal component and the second traffic key modal component; sequentially input the first traffic key modal component and the second traffic key modal component into the target communication traffic prediction model constructed by long short-term memory network technology to obtain the component prediction results of each second traffic key modal component.

[0036] Step S4: Based on the relevant weights of each second traffic key modal component and the corresponding component prediction results, obtain the target traffic prediction result of the target power optical communication network.

[0037] It should be noted that modal decomposition is a method of decomposing complex signals or systems into several inherent modes or characteristic vibration modes, which is widely used in the fields of signal processing, structural dynamics, fluid mechanics, etc. In the power optical communication network, it can be used to analyze the multi-scale characteristics of time series data such as traffic and electromagnetic interference; long short-term memory network technology is a special recurrent neural network that is good at modeling the long-term dependence relationship of time series data and can be used for prediction. Both modal decomposition and long short-term memory network technology are well-known technologies and will not be elaborated in this embodiment.

[0038] The traffic prediction method for a power optical communication network provided by the embodiment of the present invention has the following beneficial effects compared with the prior art: Obtain the target traffic-related data of the target power optical communication network, and consider both traffic data and external influence data at the same time; perform modal decomposition on each target traffic-related sequence data constructed from the target traffic-related data to obtain each traffic-related modal component of each target traffic-related data. By modal decomposition, the mixed signal is separated into independent modal components, solving the problem that traditional methods ignore external factors. Each modal component corresponds to a potential influencing factor; by screening the traffic key modal components, the interference of noise components is avoided, which can reduce the error of subsequent traffic prediction for the target power optical communication network and improve the accuracy of traffic prediction for the power optical communication network; after dividing the dataset, use long short-term memory network technology for phased prediction. The first traffic key modal component obtained from historical data is used to train and optimize the basic model, and the second traffic key modal component obtained from real-time data is used for real-time prediction. By setting different weights for different component prediction results according to the importance analysis results of the components, the target traffic prediction result of the target power optical communication network is obtained, further improving the accuracy of traffic prediction for the power optical communication network.

[0039] In one embodiment, the external influence data in step S1 at least includes grid load data, optical fiber status data, electromagnetic field data, and temperature and humidity data.

[0040] It should be noted that when the power grid load increases, the power communication network needs to transmit more monitoring, protection, and control instructions, resulting in increased traffic; fiber bending or aging will cause the bit error rate to rise, and the retransmission mechanism will increase redundant traffic; strong electromagnetic fields will introduce noise, reduce the signal-to-noise ratio, and then cause the optical communication bit error rate to rise, and the retransmission mechanism increases redundant traffic; high temperature may cause the laser wavelength to shift, increasing the optical power loss, and rising humidity may exacerbate the microbending loss, thereby causing the bit error rate to rise and increasing redundant traffic.

[0041] In one embodiment, in step S2, based on the importance analysis results of each traffic-related modal component, the corresponding relevant weights are determined, including: Input all traffic-related modal components of each target traffic-related data into the random forest algorithm, and calculate the importance scores of each traffic-related modal component; perform normalization processing on the importance scores of all traffic-related modal components of each target traffic-related data to obtain the relevant weights of each traffic-related modal component.

[0042] It should be noted that the random forest algorithm is a machine learning algorithm based on ensemble learning. By constructing multiple decision trees and synthesizing their prediction results, the accuracy and robustness of the model are improved. If a certain feature contributes more to the model prediction, its importance is higher. The random forest algorithm is a well-known technology and will not be elaborated in this embodiment.

[0043] A traffic prediction method for a power optical communication network provided in this embodiment can clearly distinguish the core factors by setting different weights according to the importance scores of each traffic-related modal component, thereby improving the accuracy of subsequent traffic prediction for the power optical communication network.

[0044] In one embodiment, in step S2, based on the relevant weights of each traffic-related modal component of each target traffic-related data, the traffic key modal components of each target traffic-related data are determined, including steps S201 to S203.

[0045] Step S201: Sort the relevant weights of all traffic-related modal components of each target traffic-related data in descending order to obtain the component weight sequence of each target traffic-related data; Step S202: Input the relevant weights of each traffic-related modal component in the component weight sequence into the cumulative contribution calculation expression, and calculate the cumulative contribution sequence corresponding to the component weight sequence; among them, the cumulative contribution calculation expression is designed to sequentially accumulate the descending order results of the relevant weights of all traffic-related modal components of each target traffic-related data; Step S203: Use the first element in the cumulative contribution sequence that is greater than the preset contribution rate threshold as the segmentation point to obtain the traffic key modal components of each target traffic-related data.

[0046] In one embodiment, the cumulative contribution calculation expression in step S202 is designed as: Wherein, is the i-th element in the cumulative contribution sequence, is the j-th element in the component weight sequence.

[0047] A traffic prediction method for a power optical communication network provided in this embodiment, by descendingly sorting the correlation weights of all traffic-related modal components of each target traffic-related data, clearly distinguishing the core traffic-related modal components from the secondary traffic-related modal components, providing an ordered input for subsequent cumulative contribution calculation, adopting a recursive cumulative contribution calculation expression, intuitively showing the overall contribution of the first N traffic-related modal components, and avoiding the subjectivity of setting artificial experience thresholds.

[0048] In one embodiment, in step S3, the first traffic key modal component and the second traffic key modal component are sequentially input into the target communication traffic prediction model to obtain the component prediction results of each second traffic key modal component, including: Input the first traffic key modal component obtained from historical target traffic-related data into the first target communication traffic prediction model for prediction training; Use the second target communication traffic prediction model obtained from the prediction training to perform prediction analysis on the second traffic key modal component obtained from real-time target traffic-related data to obtain the component prediction results of each second traffic key modal component.

[0049] In one embodiment, the sparrow search algorithm is used to perform feedback optimization on the parameters of the first target communication traffic prediction model during the prediction training process to obtain the target parameter combination of the second target communication traffic prediction model.

[0050] A traffic prediction method for a power optical communication network provided in this embodiment, by using the long short-term memory network technology to perform phased prediction after dividing the data set, fully mining the long-term rules from the first traffic key modal component obtained from historical data, training and optimizing the basic model, and using the optimized model to perform real-time prediction on the second traffic key modal component obtained from real-time data, improving the accuracy of traffic prediction for the power optical communication network.

[0051] In one embodiment, various performance evaluation indexes of the second target communication traffic prediction model are calculated based on the difference characteristics between the target traffic prediction result of the real-time target traffic-related data and the corresponding target monitoring result; the target parameter combination of the second target communication traffic prediction model is feedback-optimized with all the performance evaluation indexes.

[0052] In one embodiment, the performance evaluation index includes at least root mean square error, mean absolute error, mean absolute percentage error and determination coefficient.

[0053] Another embodiment of the present invention provides a flow prediction system for an electric power optical communication network. For details, see Figure 2 , Figure 2 The structure block diagram of a flow prediction system of a power optical communication network in one embodiment of the present invention is shown, comprising: The data modal decomposition module 11 is used to obtain target flow-related data of the target electric power optical communication network; wherein the target flow-related data includes flow data and external influence data; modal decomposition is performed on each target flow-related sequence data constructed by the target flow-related data to obtain each flow-related modal component of each target flow-related data; wherein the flow-related modal component reflects each factor that affects the target flow-related data; The data modality screening module 12 is used to determine the corresponding relevant weight based on the importance analysis result of each flow-related modal component; determine the flow key modal component of each target flow-related data based on the relevant weight of each flow-related modal component of each target flow-related data; The key modal prediction module 13 is used to divide the data set of all traffic key modal components to obtain the first traffic key modal component and the second traffic key modal component; the first traffic key modal component and the second traffic key modal component are sequentially input into the target communication traffic prediction model constructed by the long short-term memory network technology to obtain the component prediction result of each second traffic key modal component; The prediction result reconstruction module 14 is used to obtain the target flow prediction result of the target power optical communication network based on the relevant weight of each second flow key modal component and the corresponding component prediction result.

[0054] In one embodiment, the external impact data in the data modal decomposition module 11 at least includes grid load data, optical fiber status data, electromagnetic field data, and temperature and humidity data.

[0055] It should be noted that when the grid load increases, the power communication network needs to transmit more monitoring, protection and control instructions, resulting in an increase in traffic; bending or aging of optical fibers will lead to an increase in bit error rates, and the retransmission mechanism will increase redundant traffic; strong electromagnetic fields will introduce noise and reduce the signal-to-noise ratio, which will lead to an increase in the optical communication bit error rate, and the retransmission mechanism will increase redundant traffic; high temperatures may cause laser wavelength shift and increase optical power loss, and increased humidity may aggravate micro-bending losses, which will lead to an increase in bit error rates and increase redundant traffic.

[0056] In one embodiment, based on the importance analysis results of each traffic-related modal component in the data modal screening module 12, the corresponding correlation weights are determined, including: Input all traffic-related modal components of each target traffic-related data into the random forest algorithm, and calculate the importance scores of each traffic-related modal component; perform normalization processing on the importance scores of all traffic-related modal components of each target traffic-related data to obtain the correlation weights of each traffic-related modal component.

[0057] It should be noted that the random forest algorithm is a machine learning algorithm based on ensemble learning. By constructing multiple decision trees and synthesizing their prediction results, the accuracy and robustness of the model are improved. If a certain feature contributes more to the model prediction, its importance is higher. The random forest algorithm is a well-known technology and will not be elaborated in this embodiment.

[0058] The traffic prediction system of the power optical communication network provided in this embodiment can clearly distinguish the core factors by setting different weights according to the importance scores of each traffic-related modal component, thereby improving the accuracy of subsequent traffic prediction for the power optical communication network.

[0059] In one embodiment, based on the correlation weights of each traffic-related modal component of each target traffic-related data in the data modal screening module 12, the traffic key modal components of each target traffic-related data are determined, including: Sort the correlation weights of all traffic-related modal components of each target traffic-related data in descending order to obtain the component weight sequence of each target traffic-related data; Input the correlation weights of each traffic-related modal component in the component weight sequence into the cumulative contribution calculation expression, and calculate the cumulative contribution sequence corresponding to the component weight sequence; among them, the cumulative contribution calculation expression is designed to accumulate the descending order results of the correlation weights of all traffic-related modal components of each target traffic-related data in turn; Taking the first element in the cumulative contribution sequence that is greater than the preset contribution rate threshold as the segmentation point, the traffic key modal components of each target traffic-related data are obtained.

[0060] In one embodiment, the cumulative contribution calculation expression is designed as: Among them, is the i-th element in the cumulative contribution sequence, is the j-th element in the component weight sequence.

[0061] A traffic prediction system for a power optical communication network provided by this embodiment sorts the correlation weights of all traffic-related modal components of each target traffic-related data in descending order, clearly differentiates the core traffic-related modal components from the secondary traffic-related modal components, provides an ordered input for subsequent cumulative contribution calculation, adopts a recursive cumulative contribution calculation expression, intuitively shows the overall contribution of the first N traffic-related modal components, and avoids the subjectivity of setting artificial experience thresholds.

[0062] In one embodiment, in the key modal prediction module 13, the first traffic key modal component and the second traffic key modal component are sequentially input into the target communication traffic prediction model to obtain the component prediction results of each second traffic key modal component, including: Input the first traffic key modal component obtained from historical target traffic-related data into the first target communication traffic prediction model for prediction training; Use the second target communication traffic prediction model obtained from the prediction training to perform prediction analysis on the second traffic key modal component obtained from real-time target traffic-related data to obtain the component prediction results of each second traffic key modal component.

[0063] In one embodiment, the sparrow search algorithm is used to perform feedback optimization on the parameters of the first target communication traffic prediction model during the prediction training process to obtain the target parameter combination of the second target communication traffic prediction model.

[0064] A traffic prediction system for a power optical communication network provided by this embodiment, after dividing the data set, adopts the long short-term memory network technology for phased prediction. The first traffic key modal component obtained from historical data fully mines the long-term law, trains and optimizes the basic model, and uses the optimized model to perform real-time prediction on the second traffic key modal component obtained from real-time data, improving the accuracy of traffic prediction for the power optical communication network.

[0065] In one embodiment, calculate each performance evaluation index of the second target communication traffic prediction model based on the difference characteristics between the target traffic prediction result of the real-time target traffic-related data and the corresponding target monitoring result; use all performance evaluation indexes to perform feedback optimization on the target parameter combination of the second target communication traffic prediction model.

[0066] In one embodiment, the performance evaluation indexes at least include root mean square error, mean absolute error, mean absolute percentage error, and coefficient of determination.

[0067] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the device where the computer-readable storage medium is located executes the computer program, the steps in the above-mentioned traffic prediction method for a power optical communication network are implemented.

[0068] The above embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the appended claims.

Claims

1. A flow prediction method for an electric power optical communication network, characterized in that: The method comprises: Acquire target flow-related data of a target electric power optical communication network; wherein the target flow-related data includes flow data and external influence data; perform modal decomposition on each target flow-related sequence data constructed by the target flow-related data to obtain each flow-related modal component of each target flow-related data; wherein the flow-related modal component reflects each factor affecting the target flow-related data; Based on the importance analysis result of each of the flow-related modal components, determine the corresponding relevant weight; based on the relevant weight of each of the flow-related modal components of each of the target flow-related data, determine the flow key modal component of each of the target flow-related data; Dividing all the traffic key modal components into a data set to obtain a first traffic key modal component and a second traffic key modal component; inputting the first traffic key modal component and the second traffic key modal component into a target communication traffic prediction model constructed by long short-term memory network technology in sequence to obtain a component prediction result of each of the second traffic key modal components; Based on the relevant weight of each of the second traffic key modal components and the corresponding component prediction results, a target traffic prediction result of the target power optical communication network is obtained.

2. A flow prediction method for a power optical communication network according to claim 1, characterized in that: The determining of the key flow modal component of each target flow related data based on the relevant weight of each flow related modal component of each target flow related data comprises: Sorting the relevant weights of all the flow-related modal components of each of the target flow-related data in descending order to obtain a component weight sequence of each of the target flow-related data; The relevant weight of each of the flow-related modal components in the component weight sequence is input into a cumulative contribution calculation expression to calculate a cumulative contribution sequence corresponding to the component weight sequence; wherein the cumulative contribution calculation expression is designed to sequentially accumulate the descending order results of the relevant weights of all the flow-related modal components of each of the target flow-related data; The first element in the cumulative contribution sequence that is greater than a preset contribution rate threshold is used as a segmentation point to obtain the key modal component of the flow of each target flow-related data.

3. The method for predicting flow rate of a power optical communication network according to claim 1, characterized in that: The external impact data at least includes power grid load data, optical fiber status data, electromagnetic field data, and temperature and humidity data.

4. The method for predicting flow rate of a power optical communication network according to claim 1, characterized in that: The determining of the corresponding relevant weight based on the importance analysis result of each of the flow-related modal components includes: Inputting all the flow-related modal components of each target flow-related data into a random forest algorithm, and calculating the importance score of each flow-related modal component; The importance scores of all the flow-related modal components of each of the target flow-related data are normalized to obtain relevant weights of the respective flow-related modal components.

5. The method for predicting flow rate of a power optical communication network according to claim 2, characterized in that: The cumulative contribution calculation expression is designed as: in, is the ith element in the cumulative contribution sequence, is the jth element in the component weight sequence.

6. The method for predicting flow rate of a power optical communication network according to claim 1, characterized in that: The step of sequentially inputting the first traffic critical modal component and the second traffic critical modal component into a target communication traffic prediction model to obtain a component prediction result of each of the second traffic critical modal components includes: Inputting the first traffic key modal component obtained from the historical target traffic related data into the first target communication traffic prediction model for prediction training; The second target communication traffic prediction model obtained by the prediction training is used to perform prediction analysis on the second traffic key modal components obtained from the real-time target traffic related data to obtain component prediction results of each of the second traffic key modal components.

7. A method for predicting flow in a power optical communication network according to claim 6, characterized in that: The method further comprises: The parameters of the first target communication flow prediction model are feedback optimized by using a sparrow search algorithm during the prediction training process to obtain a target parameter combination of the second target communication flow prediction model.

8. The method for predicting flow rate of a power optical communication network according to claim 7, characterized in that: The method further comprises: Calculating each performance evaluation index of the second target communication flow prediction model based on the difference characteristics between the target flow prediction result and the corresponding target monitoring result of the real-time target flow related data; The target parameter combination of the second target communication traffic prediction model is feedback optimized using all the performance evaluation indicators.

9. A method for predicting flow in a power optical communication network according to claim 8, characterized in that: The performance evaluation indexes include at least root mean square error, mean absolute error, mean absolute percentage error and determination coefficient.

10. A flow prediction system for an electric power optical communication network, characterized in that: The system comprises: A data modal decomposition module, used for obtaining target flow-related data of a target electric power optical communication network; wherein the target flow-related data includes flow data and external influence data; performing modal decomposition on each target flow-related sequence data constructed by the target flow-related data to obtain each flow-related modal component of each target flow-related data; wherein the flow-related modal component reflects each factor affecting the target flow-related data; A data modality screening module, configured to determine a corresponding relevant weight based on the importance analysis result of each of the flow-related modal components; and determine a flow key modal component of each of the target flow-related data based on the relevant weight of each of the flow-related modal components of each of the target flow-related data; A key modal prediction module is used to divide the data set of all the traffic key modal components to obtain a first traffic key modal component and a second traffic key modal component; the first traffic key modal component and the second traffic key modal component are sequentially input into a target communication traffic prediction model constructed by long short-term memory network technology to obtain a component prediction result of each second traffic key modal component; A prediction result reconstruction module is used to obtain a target flow prediction result of the target electric power optical communication network based on the relevant weight of each of the second flow key modal components and the corresponding component prediction result.

11. A flow prediction system for a power optical communication network according to claim 10, characterized in that: The determining of the key flow modal component of each target flow related data based on the relevant weight of each flow related modal component of each target flow related data comprises: Sorting the relevant weights of all the flow-related modal components of each of the target flow-related data in descending order to obtain a component weight sequence of each of the target flow-related data; The relevant weight of each of the flow-related modal components in the component weight sequence is input into a cumulative contribution calculation expression to calculate a cumulative contribution sequence corresponding to the component weight sequence; wherein the cumulative contribution calculation expression is designed to sequentially accumulate the descending order results of the relevant weights of all the flow-related modal components of each of the target flow-related data; The first element in the cumulative contribution sequence that is greater than a preset contribution rate threshold is used as a segmentation point to obtain the key modal component of the flow of each target flow-related data.

12. A flow prediction system for a power optical communication network according to claim 10, characterized in that: The external impact data at least includes power grid load data, optical fiber status data, electromagnetic field data, and temperature and humidity data.

13. The flow prediction system of the power optical communication network according to claim 10, characterized in that: The determining of the corresponding relevant weight based on the importance analysis result of each of the flow-related modal components includes: Inputting all the flow-related modal components of each target flow-related data into a random forest algorithm, and calculating the importance score of each flow-related modal component; The importance scores of all the flow-related modal components of each of the target flow-related data are normalized to obtain relevant weights of the respective flow-related modal components.

14. The flow prediction system of the power optical communication network according to claim 11, characterized in that: The cumulative contribution calculation expression is designed as: in, is the ith element in the cumulative contribution sequence, is the jth element in the component weight sequence.

15. The flow prediction system of the power optical communication network according to claim 10, characterized in that: The step of sequentially inputting the first traffic critical modal component and the second traffic critical modal component into a target communication traffic prediction model to obtain a component prediction result of each of the second traffic critical modal components includes: Inputting the first traffic key modal component obtained from the historical target traffic related data into the first target communication traffic prediction model for prediction training; The second target communication traffic prediction model obtained by the prediction training is used to perform prediction analysis on the second traffic key modal components obtained from the real-time target traffic related data to obtain component prediction results of each of the second traffic key modal components.

16. A flow prediction system for a power optical communication network according to claim 15, characterized in that: The system further comprises: The parameters of the first target communication flow prediction model are feedback optimized by using a sparrow search algorithm during the prediction training process to obtain a target parameter combination of the second target communication flow prediction model.

17. A flow prediction system for a power optical communication network according to claim 16, characterized in that: The system further comprises: Calculating each performance evaluation index of the second target communication flow prediction model based on the difference characteristics between the target flow prediction result and the corresponding target monitoring result of the real-time target flow related data; The target parameter combination of the second target communication traffic prediction model is feedback optimized using all the performance evaluation indicators.

18. A flow prediction system for a power optical communication network according to claim 17, characterized in that: The performance evaluation indexes include at least root mean square error, mean absolute error, mean absolute percentage error and determination coefficient.

19. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the device where the computer-readable storage medium is located executes the computer program, a traffic prediction method for an electric power optical communication network as described in any one of claims 1 to 9 is implemented.

Images