CN113988441A - Power wireless network link quality prediction and model training method and device - Google Patents

Power wireless network link quality prediction and model training method and device Download PDF

Info

Publication number
CN113988441A
CN113988441A CN202111291478.0A CN202111291478A CN113988441A CN 113988441 A CN113988441 A CN 113988441A CN 202111291478 A CN202111291478 A CN 202111291478A CN 113988441 A CN113988441 A CN 113988441A
Authority
CN
China
Prior art keywords
link
samples
link quality
wireless network
group
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.)
Pending
Application number
CN202111291478.0A
Other languages
Chinese (zh)
Inventor
宋曦
马乐
李文辉
刘豆
詹文浩
李颖
王丽丹
宫皓泉
袁平亮
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.)
Information and Telecommunication Branch of State Grid Gansu Electric Power Co Ltd
Original Assignee
Information and Telecommunication Branch of State Grid Gansu Electric Power Co Ltd
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 Information and Telecommunication Branch of State Grid Gansu Electric Power Co Ltd filed Critical Information and Telecommunication Branch of State Grid Gansu Electric Power Co Ltd
Priority to CN202111291478.0A priority Critical patent/CN113988441A/en
Publication of CN113988441A publication Critical patent/CN113988441A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06395—Quality analysis or management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/40—Business processes related to the transportation industry

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Health & Medical Sciences (AREA)
  • Development Economics (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Game Theory and Decision Science (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Educational Administration (AREA)
  • Primary Health Care (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention provides a method and a device for predicting link quality of a power wireless network and training a model, wherein the method for training the model comprises the following steps: acquiring a plurality of groups of samples, wherein each group of samples comprises parameter values of a plurality of types of link parameters; calculating fuzzy evaluation subsets of each group of samples according to the membership of parameter values in each group of samples to each value range and the weight of each type of link parameters, and calculating link quality grade division ranges according to the closeness of the evaluation subsets of each group of samples and ideal target fuzzy subsets representing different link quality grades; acquiring a plurality of groups of training samples, wherein each group of training samples comprises parameter values of a plurality of types of link parameters; and inputting the link quality grade division range and parameter values in the training samples into a cyclic neural network, and training the cyclic neural network to obtain a power wireless network link quality prediction model. By the method, the quality of the power wireless network link can be quickly and accurately predicted by combining various types of link parameters.

Description

Power wireless network link quality prediction and model training method and device
Technical Field
The invention relates to the technical field of power system communication, in particular to a method and a device for predicting link quality of a power wireless network and training a model.
Background
The intelligent power grid establishes a wide and reliable communication network for public facilities and electric equipment, provides an effective information channel for information transmission and monitoring, and ensures the reliability of electric wireless communication. However, as the propagation channel is increasingly complex and dynamically changed in time and space, the link quality of the wireless communication network is reduced, and data loss is caused in the transmission process of the highly reliable power transmission service. Therefore, the prediction of the quality grade of the wireless communication link in the power network has important significance, and the reliability of service transmission in the smart grid can be effectively improved.
At present, researches for link quality prediction at home and abroad are mainly divided into two categories, one category is a link quality prediction method based on the traditional technology, single physical layer parameters are used for prediction, and link quality grade judgment is carried out based on evaluation indexes of a data Packet Receiving Rate (PRR) and a retransmission rate.
The other type is based on a new machine learning method, but most of the machine learning algorithms proposed at present are simple in structure, the prediction accuracy of the model is not particularly high, and the requirements of power services on the quality of a wireless link in a smart grid scene are not fully considered. Therefore, how to design a method which can comprehensively consider the link quality multi-attribute indexes, objectively and accurately divide the link quality grades, and improve the reliability of reliable power services in the data transmission process to a certain extent is an important problem to be faced in the wireless link quality evaluation prediction in the smart grid scene.
Disclosure of Invention
Therefore, the technical problem to be solved by the present invention is to overcome the defect in the prior art that the quality of the wireless link cannot be objectively and accurately evaluated, thereby providing a method and a device for predicting the quality of the power wireless network link and training a model.
The invention provides a power wireless network link quality prediction model training method in a first aspect, which comprises the following steps: acquiring a plurality of groups of samples, wherein each group of samples comprises parameter values of a plurality of types of link parameters; calculating fuzzy evaluation subsets of each group of samples according to the membership of parameter values in each group of samples to each value range and the weight of each type of link parameters, and calculating link quality grade division ranges according to the closeness of the evaluation subsets of each group of samples and ideal target fuzzy subsets for representing different link quality grades, wherein the link quality grade division ranges comprise the value ranges of each type of link parameters under different link quality grades; acquiring a training data set, wherein the training data set comprises a plurality of groups of training samples, and each group of training samples comprises parameter values of a plurality of types of link parameters; and inputting the link quality grade division range and parameter values in the training samples into a cyclic neural network, and training the cyclic neural network to obtain a power wireless network link quality prediction model.
Optionally, in the method for training a link quality prediction model of a power wireless network provided by the present invention, the fuzzy evaluation subset of each group of samples is calculated according to the membership of the parameter values in each group of samples to each value range and the weight of each type of link parameter, and the link quality grade division range is calculated according to the closeness of the evaluation subset of each group of samples and the ideal target fuzzy subset for representing different link quality grades, including: constructing a fuzzy evaluation subset of the ith group of samples according to the membership of the parameter values in the ith group of samples to each value range and the weight of each type of link parameters, wherein different value ranges correspond to different link quality grades; determining the link quality grade of the ith group of samples according to the closeness of the fuzzy evaluation subset of the ith group of samples and an ideal target fuzzy subset for representing different link quality grades, and updating the value range of each link parameter under different link quality grades according to the link quality grade of the ith group of samples; and if i is less than N, adding 1 to the value of i, returning to the step of constructing a fuzzy evaluation subset of the ith group of samples by using the membership of the parameter values in the ith group of samples to each value range and the weight of each type of link parameter until i is equal to N, and determining the value range of each link parameter under different current link quality levels as a level division range.
Optionally, in the power wireless network link quality prediction model training method provided by the present invention, the weights of various link parameters are calculated through the following steps: normalizing the parameter values in the level division data set to obtain relative values corresponding to the parameter values in the level division data set; calculating the proportion of each link parameter in the samples according to the relative values in each group of samples; calculating entropy values of various link parameters according to the specific weights of the various link parameters in the sample:
Figure BDA0003329267840000031
where m denotes the number of types of link parameters, n denotes the number of samples in the hierarchically divided data set, pijRepresenting the proportion of the j-th type link parameter in the ith sample; calculating the weight of each link parameter according to the entropy value of each link parameter:
Figure BDA0003329267840000032
wherein, wjWeight representing a class j link parameter, djRepresenting class j link parametersAnd (4) information benefit.
Optionally, in the method for training a power wireless network link quality prediction model provided by the present invention, constructing a fuzzy evaluation subset of the ith group of samples according to the membership of the parameter values in the ith group of samples to each value range and the weight of each type of link parameter, includes: and (3) constructing a fuzzy matrix according to the membership of the parameter values in the ith group of samples to each value range:
Figure BDA0003329267840000033
where m denotes the number of link parameter types, l denotes the number of link quality classes, rmlRepresenting the membership degree of the parameter value of the mth type link parameter to the value range corresponding to the quality grade of the lth type link; determining the product of the fuzzy matrix of the ith group of samples and the weight matrix of each type of link parameters as a fuzzy evaluation subset of the ith group of samples: and A is w multiplied by R, wherein w represents a weight matrix of each type of link parameter.
Optionally, in the power wireless network link quality prediction model training method provided by the present invention, the closeness of the evaluation subset of the samples to the ideal target fuzzy subset for characterizing different link quality levels is calculated by the following formula:
Figure BDA0003329267840000041
wherein the content of the first and second substances,
Figure BDA0003329267840000042
representing a fuzzy evaluation subset AiThe corresponding sample belongs to the class vkThe degree of membership of (a) is,
Figure BDA0003329267840000043
representing an ideal target fuzzy subset BjCorresponding sample to rank vkN represents the number of samples, and p is a predetermined constant.
The second aspect of the present invention provides a method for predicting link quality of a power wireless network, including: acquiring parameter values of parameters of multiple links of a link to be detected; and inputting the parameter values into a power wireless network link quality prediction model to obtain link quality grades corresponding to each group of samples, wherein the power wireless network link quality prediction model is obtained by executing the training method of the power wireless network link quality prediction model provided by the first aspect of the invention.
The third aspect of the present invention provides a power wireless network link quality prediction model training device, including: the system comprises a sample acquisition module, a link parameter analysis module and a link parameter analysis module, wherein the sample acquisition module is used for acquiring a plurality of groups of samples, and each group of samples comprises parameter values of a plurality of types of link parameters; the link quality grade dividing module is used for calculating fuzzy evaluation subsets of each group of samples according to the membership of parameter values in each group of samples to each value range and the weight of each type of link parameters, calculating link quality grade dividing ranges according to the closeness of the evaluation subsets of each group of samples and ideal target fuzzy subsets for representing different link quality grades, and the link quality grade dividing ranges comprise the value ranges of each type of link parameters under different link quality grades; the training set acquisition module is used for acquiring a training data set, wherein the training data set comprises a plurality of groups of training samples, and each group of training samples comprises parameter values of a plurality of types of link parameters; and the training module is used for inputting the link quality grade division range and parameter values in the training samples into the recurrent neural network, and training the recurrent neural network to obtain the power wireless network link quality prediction model.
The fourth aspect of the present invention provides a power wireless network link quality prediction device, including: the to-be-detected parameter acquisition module is used for acquiring parameter values of parameters of multiple links of the to-be-detected link; and the link quality prediction module is used for inputting the parameter values into the power wireless network link quality prediction model to obtain the link quality grade of the link to be detected.
A fifth aspect of the present invention provides a computer apparatus comprising: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to perform the method for training a power wireless network link quality prediction model as provided in the first aspect of the invention or the method for predicting power wireless network link quality as provided in the second aspect of the invention.
A sixth aspect of the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the power wireless network link quality prediction model training method as provided in the first aspect of the present invention or the power wireless network link quality prediction method as provided in the second aspect of the present invention.
The technical scheme of the invention has the following advantages:
the method and the device for predicting the link quality of the power wireless network and training the model provided by the invention have the advantages that the obtained multiple groups of samples all comprise parameter values of multiple types of link parameters, calculating the fuzzy evaluation subset of each group of samples according to the membership of the parameter values in each group of samples to each value range and the weight of each type of link parameters, and calculating a link quality grade division range according to the closeness of the evaluation subsets of each group of samples and ideal target fuzzy subsets for representing different link quality grades, wherein the link quality grade division range is obtained by calculating parameter values of various types of link parameters by adopting a closeness method, the closeness method has advantages in describing the closeness degree among the subsets, can better solve the multi-attribute decision problem, therefore, the invention can be used for more effectively and objectively dividing the link quality grades by combining various attributes of the links. The calculated link quality grade division range and the training data are input into a recurrent neural network, and the obtained power wireless network link quality prediction model can be used for quickly and accurately predicting the quality of the power wireless network link by combining various link parameters.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in 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 other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a specific example of a power wireless network link quality prediction model training method according to an embodiment of the present invention;
FIG. 2 is a schematic block diagram of a specific example of RNN architecture in an embodiment of the present invention;
FIG. 3 is a comparison graph of the link true level and the predicted results of the two models under the "distance 15" data set in the embodiment of the present invention;
FIG. 4 is a comparison graph of the link true level and the predicted results of the two models under the "distance 20" data set in the embodiment of the present invention;
fig. 5 is a flowchart illustrating a specific example of a method for predicting link quality of a power wireless network according to an embodiment of the present invention;
fig. 6 is a schematic block diagram of a specific example of a power wireless network link quality prediction model training apparatus according to an embodiment of the present invention;
fig. 7 is a schematic block diagram of a specific example of a power wireless network link quality prediction apparatus according to an embodiment of the present invention;
fig. 8 is a schematic block diagram of a specific example of a computer device in the embodiment of the present invention.
Detailed Description
The technical solutions of the present invention will be described clearly and completely with reference to the accompanying drawings, and it should be understood 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.
In the description of the present invention, it should be noted that the terms "first", "second", and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
In addition, the technical features involved in the different embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The embodiment of the invention provides a power wireless network link quality prediction model training method, as shown in fig. 1, comprising the following steps:
step S10: and acquiring a plurality of groups of samples, wherein each group of samples comprises parameter values of a plurality of types of link parameters.
In the electric power wireless network, link parameters are divided into two types, one type is parameters obtained in a software mode, and the other type is physical layer parameters based on hardware, wherein the parameters obtained in the software mode and the physical layer parameters comprise multiple types, and the link parameters contained in a sample can be the parameters obtained in the software mode or the physical layer parameters.
In an alternative embodiment, where physical layer parameters are more readily available and less complex to compute, hardware-based physical layer parameters may be selected to enable link quality prediction in power wireless networks. The physical parameters in the link are numerous and all affect the Packet Reception Rate (PRR) of the receiver to a greater or lesser extent.
In an alternative embodiment, three physical layer parameters, namely, a Received Signal Strength Indicator (RSSI), a Link Quality Indicator (LQI) and a signal-to-noise ratio (SNR), can be selected as input parameters of the link quality prediction model by comparing the correlation between each parameter and the PRR. Among the three parameters, RSSI and SNR can be directly measured by an instrument, and the calculation formula of LQI is: QI ═ 255(RSSI + 81))/91.
In an optional embodiment, when selecting RSSI, LQI, and SNR as the link parameters, the complete value ranges of the three link parameters are: RSSI ∈ 100, -75] dBm, LQI ∈ [75,100], SNR ∈ [0,30] dBm.
Step S20: and calculating fuzzy evaluation subsets of each group of samples according to the membership of the parameter values in each group of samples to each value range and the weight of each type of link parameters, and calculating a link quality grade division range according to the closeness of the evaluation subsets of each group of samples and ideal target fuzzy subsets for representing different link quality grades, wherein the link quality grade division range comprises the value ranges of each type of link parameters under different link quality grades.
In the embodiment of the invention, each group of samples comprises multiple types of link parameters, different types of link parameters correspond to different value ranges under one link quality level, and the membership calculated in the embodiment of the invention comprises the membership of each parameter value in the samples to the value ranges under different levels.
In an optional embodiment, the fuzzy matrix formed by the membership of each parameter value in the sample to the value ranges under different levels is:
Figure BDA0003329267840000091
where m denotes the number of link parameter types, l denotes the number of link quality classes, rmlThe parameter value representing the mth link parameter may be a membership degree of a value range corresponding to the lth link quality level, and m and l may be any integer values.
In an alternative embodiment, the evaluation subset of samples is the product of the ambiguity matrix for the sample and the weight matrix for each type of link parameter: and A is w multiplied by R, wherein w represents a weight matrix of each type of link parameter.
In an alternative embodiment, any different matrix may be used as the ideal target fuzzy subset for characterizing different link qualities, and the form of the target ideal fuzzy subset is not specifically limited in the embodiment of the present invention. Illustratively, the ideal target ambiguity subset may be composed of a plurality of 0 s and a 1 s, where 1 s in the ideal target ambiguity subset represent different link quality levels when located at different positions, for example, when 1 st column in the ideal target ambiguity subset is 1 and the rest columns are 0, the link quality represented by the ideal target ambiguity subset is a first level, and when 3 rd column in the ideal target ambiguity subset is 1 and the rest columns are 0, the link quality represented by the ideal target ambiguity subset is a third level.
Step S30: a training data set is obtained, the training data set includes a plurality of groups of training samples, each group of training samples includes parameter values of a plurality of types of link parameters, and details of the link parameters in the training samples refer to the description of the plurality of groups of samples in the hierarchical data set in step S10, which is not described herein again.
Step S40: and inputting the link quality grade classification range and parameter values in the training samples into a Recurrent Neural Network (RNN) to train the RNN so as to obtain a power wireless network link quality prediction model.
In an alternative embodiment, the structure of the RNN is shown in fig. 2, for example, in the embodiment of the present invention, the number of RNN layers may be set to 4, where the number of neurons in each layer is 20, and in the training process, the training data may be divided into 1024 batches (batch _ size), and training may be performed using Adam optimizer, and the training time (epoch) is 500 times.
When the RNN predicts the quality of the power wireless network link, the input of the current moment is considered, the previous information is memorized, and the accurate prediction of the link quality grade is effectively realized.
In an optional embodiment, to speed up the convergence speed of the algorithm, before inputting the parameter values in the training samples into the RNN, the parameter values x of various link parameters in the training samples are also inputiNormalization processing is performed so that the normalized value is [0,1 ]]In the meantime. The calculation formula is as follows:
Figure BDA0003329267840000101
wherein q is*Is the normalized parameter value, q is the original parameter value, q is the normalized parameter valueminAnd q ismaxRespectively the minimum and maximum of the original parameter values.
The method for training the link quality prediction model of the power wireless network comprises the steps of firstly obtaining a level division data set, wherein a plurality of groups of samples in the level division data set comprise parameter values of various link parameters, then calculating fuzzy evaluation subsets of various groups of samples according to membership of the parameter values in the various groups of samples to various value ranges and weights of the various link parameters, and calculating link quality level division ranges according to closeness of the evaluation subsets of the various groups of samples and ideal target fuzzy subsets for representing different link quality levels, wherein the link quality level division ranges are obtained by calculating the parameter values of the various types of link parameters by adopting a closeness method, and the closeness method has advantages in describing the closeness degree among the subsets and can better solve the multi-attribute decision problem And objectively dividing the link quality grade. The calculated link quality grade division range and the training data are input into the RNN, and the obtained power wireless network link quality prediction model can be used for quickly and accurately predicting the quality of the power wireless network link by combining various link parameters.
In an optional embodiment, the step S20 specifically includes the following steps:
step one, constructing a fuzzy evaluation subset of the ith group of samples according to the membership of the parameter values in the ith group of samples to each value range and the weight of each type of link parameters, wherein different value ranges correspond to different link quality grades.
And secondly, determining the link quality grade of the ith group of samples according to the closeness of the fuzzy evaluation subset of the ith group of samples and the ideal target fuzzy subset for representing different link quality grades, and updating the value range of each link parameter under different link quality grades according to the link quality grade of the ith group of samples. In the embodiment of the present invention, the process of calculating the link quality class division range is actually a process of continuously approximating the value ranges corresponding to different link quality classes.
In an alternative embodiment, the closeness of the evaluated subset of samples to the ideal target fuzzy subset for characterizing different link quality levels is calculated by the following formula:
Figure BDA0003329267840000111
wherein the content of the first and second substances,
Figure BDA0003329267840000112
representing a fuzzy evaluation subset AiThe corresponding sample belongs to the class vkThe degree of membership of (a) is,
Figure BDA0003329267840000113
representing an ideal target fuzzy subset BjCorresponding sample to rank vkN represents the number of samples, p is a preset constant, and p may be 1, 2, 3, etc. for example.
And if i is less than N, adding 1 to the value of i, returning to the first step, repeatedly executing the first step and the second step until i is equal to N, and determining the value range of each link parameter under different current link quality levels as the link quality level division range. In an optional embodiment, N may be the number of samples in the level-divided data set, and after all the samples in the level-divided data set participate in the operation, the value range of each link parameter under different current link quality levels is determined as the link quality level division range; in an optional embodiment, N may be a preset value, and when the number of times of updating the value range of each link parameter under different link quality levels reaches the preset value, it is determined that the value range of each link parameter under different link quality levels is a level division range.
In an optional embodiment, the step of updating the value ranges of the link parameters under different link quality levels according to the link quality level to which the ith group of samples belongs specifically includes:
determining a current value range corresponding to the link quality grade to which the ith group of samples belongs, and if the parameter values in the ith group of samples do not belong to the current value range, taking the parameter values in the ith group of samples as demarcation points to subdivide the value ranges corresponding to different link quality grades; and if the parameter values in the ith group of samples belong to the current value range, not executing the operation.
Illustratively, taking the division into two different link quality levels of RSSI, if it is required to divide (-85, -75) into two different link quality levels, it is assumed that the current value range is (-85, -80], (-80, -75), the RSSI in the i-th group of samples is-79, and the proximity of the i-th group of samples to (-85, -80) is greater than that of (-80, -75), but-79 does not belong to the range (-85, -80), and then the i-th group of samples can be subdivided into (-85, -79], (-79, -75) at-79 boundary point.
Illustratively, when the link parameters include RSSI, LQI, and SNR, and the link quality levels need to be divided into 5 levels, the obtained parameter ranges under different link quality levels are as shown in table 1 below:
TABLE 1
Figure BDA0003329267840000121
Figure BDA0003329267840000131
In an optional embodiment, even though the link quality grade division range can be calculated through steps S10 and S20, in actually acquired link parameters, a situation may occur in which the RSSI value belongs to the interval (-79, -75), the LQI value belongs to the interval (100, 107), and the SNR value belongs to the interval (12, 19), at this time, the link quality grade cannot be determined according to table 1, and therefore, the RNN is further introduced in the embodiment of the present disclosure, and link parameters with various different attributes are fused and integrated through the RNN to obtain the link quality grade.
In an optional embodiment, in the power wireless network link quality prediction model training method provided in the embodiment of the present invention, weights of various link parameters are calculated through the following steps:
firstly, carrying out normalization processing on parameter values in a level division data set to obtain a relative value corresponding to each parameter value in the level division data set:
Figure BDA0003329267840000132
wherein x isijThe parameter value representing the link parameter of the jth class in the ith sample, and n representing the number of samples.
Secondly, calculating the proportion of each type of link parameters in the samples according to the relative values in each group of samples.
Then, calculating entropy values of various link parameters according to the specific weights of the various link parameters in the sample:
Figure BDA0003329267840000133
where m denotes the number of types of link parameters, n denotes the number of samples in the hierarchically divided data set, pijRepresenting the proportion of the j-th type link parameter in the ith sample;
and finally, calculating the weight of each link parameter according to the entropy value of each link parameter:
Figure BDA0003329267840000141
wherein, wjWeight representing a class j link parameter, djIndicating the information benefit of the class j link parameters.
In the embodiment of the invention, the weight of each link parameter is calculated by adopting an entropy method, so that the interference of subjective factors is reduced, and the result obtained by model prediction obtained by training is closer to the real result.
In an optional embodiment, in order to evaluate the performance of the power wireless network link quality prediction model obtained by performing the power wireless network link quality prediction model training method provided in the above embodiment, in the implementation of the present invention, a Mean Absolute Percentage Error (MAPE) and a Root Mean Square Error (RMSE) are selected as error evaluation indexes of the link quality class prediction:
Figure BDA0003329267840000142
Figure BDA0003329267840000143
where N is the total number of predicted data, yiAnd
Figure BDA0003329267840000144
the true grade value and the predicted grade of the ith sample point are respectively.
In a specific embodiment, in order to verify that when the model obtained by using the power wireless network link quality prediction model training method provided by the embodiment of the invention is used for predicting the link quality grade, the obtained prediction result is better than the prediction result obtained by prediction in the prior art, the embodiment of the invention uses parameter data sets at two different distances to train and test the model, 66% of data in the whole data set is used for training, and the remaining 34% of data is used as a test set. The data in each data set are parameter values of three types of link parameters measured at the same distance, and in the data set acquisition process, the data are acquired once every 15 minutes and are continuously acquired for one month to obtain the data set. The present invention was implemented using an SVM model as the comparative model of the experiment, with the same data set as the input. The number of RNN layers is set to 4, the number of neurons in each layer is 20, training data is divided into 1024 batches (batch _ size), an Adam optimizer is used for training, the training frequency (epoch) is 500 times, and prediction error indexes of two models under two data sets are shown in Table 2:
TABLE 2
Figure BDA0003329267840000151
Compared with the SVM model, MAPE of the model obtained by training of the embodiment of the invention is superior to that of the SVM model under two data sets, and the MAPE is respectively improved by 6.0% and 5.4% under two conditions. Since the instability of the parameter values in the link increases with the distance, the prediction error of the "distance 20" data set is slightly larger than that of the "distance 15" data set.
FIGS. 3 and 4 show the comparison between the link true level and the predicted result of the two models under the data sets "distance 15" and "distance 20", respectively. For the convenience of observation, 48 sampling points in one day are selected for illustration. As can be seen from the figure, the prediction curve of the model obtained by training the power wireless network link quality prediction model provided by the embodiment of the invention is more fit with the real link quality grade curve. Even under the 'distance 20' data set with a plurality of catastrophe points in the real curve, the prediction result of the model obtained by the training method for the power wireless network link quality prediction model provided by the embodiment of the invention is still closer to the real curve.
The embodiment of the invention provides a method for predicting the link quality of a power wireless network, which comprises the following steps of:
step S21: and acquiring parameter values of multiple types of link parameters of the link to be detected. For details, reference is made to the description in the above embodiments, which are not repeated herein.
Step S22: and inputting parameter values of various link parameters of the link to be detected into the link quality prediction model of the power wireless network to obtain the link quality grade of the link to be detected. For details, reference is made to the description in the above embodiments, which are not repeated herein.
The embodiment of the invention provides a power wireless network link quality prediction model training device, as shown in fig. 6, comprising:
the sample obtaining module 31 is configured to obtain multiple sets of samples, where each set of samples includes parameter values of multiple types of link parameters, and details of the samples are described in the foregoing method embodiments and are not described herein again.
The link quality grade division module 32 is configured to calculate a fuzzy evaluation subset of each group of samples according to the membership of the parameter values in each group of samples to each value range and the weight of each type of link parameter, and calculate a link quality grade division range according to the closeness of the evaluation subset of each group of samples and an ideal target fuzzy subset for representing different link quality grades, where the link quality grade division range includes the value ranges of each type of link parameter at different link quality grades, and the details are described in the above method embodiment and are not described herein again.
The training set obtaining module 33 is configured to obtain a training data set, where the training data set includes multiple sets of training samples, and each set of training samples includes parameter values of multiple types of link parameters, and details of the training samples are described in the foregoing method embodiment and are not described herein again.
The training module 34 is configured to input parameter values in the link quality class classification range and the training samples into the RNN, and train the RNN to obtain a power wireless network link quality prediction model, where details are described in the foregoing method embodiment and are not described herein again.
An embodiment of the present invention provides a device for predicting quality of a power wireless network link, as shown in fig. 7, including:
the parameter obtaining module 41 to be detected is configured to obtain parameter values of multiple types of link parameters of the link to be detected, and details of the parameter obtaining module refer to the description in the foregoing method embodiment and are not described herein again.
And the link quality prediction module 42 is configured to input parameter values of multiple types of link parameters of the link to be detected into the link quality prediction model of the power wireless network to obtain the link quality grade of the link to be detected, where the detailed content is described in the foregoing method embodiment and is not described herein again.
An embodiment of the present invention provides a computer device, as shown in fig. 8, the computer device mainly includes one or more processors 51 and a memory 52, and one processor 51 is taken as an example in fig. 8.
The computer device may further include: an input device 53 and an output device 54.
The processor 51, the memory 52, the input device 53 and the output device 54 may be connected by a bus or other means, and fig. 8 illustrates the connection by a bus as an example.
The processor 51 may be a Central Processing Unit (CPU). The Processor 51 may also be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, or combinations thereof. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The memory 52 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created from use of the power wireless network link quality prediction model training apparatus, or the power wireless network link quality prediction apparatus, or the like. Further, the memory 52 may include high speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory 52 optionally includes a memory remotely located from the processor 51, and these remote memories may be connected over a network to a power wireless network link quality prediction model training device, or a power wireless network link quality prediction device. The input device 53 may receive a calculation request (or other numerical or character information) input by a user and generate a key signal input related to the power wireless network link quality prediction model training device, or the power wireless network link quality prediction device. The output device 54 may include a display device such as a display screen for outputting the calculation result.
Embodiments of the present invention provide a computer-readable storage medium, where the computer-readable storage medium stores computer instructions, and the computer-readable storage medium stores computer-executable instructions, where the computer-executable instructions may perform the power wireless network link quality prediction model training method or the power wireless network link quality prediction method in any of the above method embodiments. The storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory (Flash Memory), a Hard Disk (Hard Disk Drive, abbreviated as HDD) or a Solid State Drive (SSD), etc.; the storage medium may also comprise a combination of memories of the kind described above.
It should be understood that the above examples are only for clarity of illustration and are not intended to limit the embodiments. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. And obvious variations or modifications therefrom are within the scope of the invention.

Claims (10)

1. A power wireless network link quality prediction model training method is characterized by comprising the following steps:
acquiring a plurality of groups of samples, wherein each group of samples comprises parameter values of a plurality of types of link parameters;
calculating fuzzy evaluation subsets of each group of samples according to the membership of parameter values in each group of samples to each value range and the weight of each type of link parameters, and calculating link quality grade division ranges according to the closeness of the evaluation subsets of each group of samples and ideal target fuzzy subsets for representing different link quality grades, wherein the link quality grade division ranges comprise the value ranges of each type of link parameters under different link quality grades;
acquiring a training data set, wherein the training data set comprises a plurality of groups of training samples, and each group of training samples comprises parameter values of a plurality of types of link parameters;
and inputting the link quality grade division range and parameter values in the training samples into a cyclic neural network, and training the cyclic neural network to obtain the power wireless network link quality prediction model.
2. The power wireless network link quality prediction model training method according to claim 1, wherein the fuzzy evaluation subsets of each group of samples are calculated according to the membership of parameter values in each group of samples to each value range and the weight of each type of link parameters, and the link quality grade division range is calculated according to the closeness of the evaluation subsets of each group of samples and ideal target fuzzy subsets for representing different link quality grades, comprising:
constructing a fuzzy evaluation subset of the ith group of samples according to the membership of the parameter values in the ith group of samples to each value range and the weight of each type of link parameters, wherein different value ranges correspond to different link quality grades;
determining the link quality grade of the ith group of samples according to the closeness of the fuzzy evaluation subset of the ith group of samples and an ideal target fuzzy subset for representing different link quality grades, and updating the value range of each link parameter under different link quality grades according to the link quality grade of the ith group of samples;
and if i is less than N, adding 1 to the value of i, returning to the step of constructing a fuzzy evaluation subset of the ith group of samples by using the membership of the parameter values in the ith group of samples to each value range and the weight of each type of link parameter until i is equal to N, and determining the value range of each link parameter under different current link quality levels as a link quality level division range.
3. The power wireless network link quality prediction model training method according to claim 1 or 2, wherein the weights of the various types of link parameters are calculated by the following steps:
normalizing the parameter values in the level division data set to obtain relative values corresponding to the parameter values in the level division data set;
calculating the proportion of each link parameter in the samples according to the relative values in each group of samples;
calculating entropy values of various link parameters according to the specific weights of the various link parameters in the sample:
Figure FDA0003329267830000021
wherein m denotes the number of kinds of link parameters, n denotes the number of samples in the hierarchically divided data set, pijRepresenting the proportion of the j-th type link parameter in the ith sample;
calculating the weight of each link parameter according to the entropy value of each link parameter:
Figure FDA0003329267830000022
wherein, wjWeight representing a class j link parameter, djIndicating the information benefit of the class j link parameters.
4. The power wireless network link quality prediction model training method according to any one of claims 1 to 3, wherein the step of constructing the fuzzy evaluation subset of the ith group of samples according to the membership degree of the parameter values in the ith group of samples to each value range and the weight of each type of link parameters comprises the following steps:
and (3) constructing a fuzzy matrix according to the membership of the parameter values in the ith group of samples to each value range:
Figure FDA0003329267830000031
where m denotes the number of link parameter types, l denotes the number of link quality classes, rmlRepresenting the membership degree of the parameter value of the mth type link parameter to the value range corresponding to the quality grade of the lth type link;
determining the product of the fuzzy matrix of the ith group of samples and the weight matrix of each type of link parameters as a fuzzy evaluation subset of the ith group of samples:
A＝w×R，
wherein w represents a weight matrix of various link parameters.
5. The power wireless network link quality prediction model training method according to any one of claims 1 to 4, characterized in that the closeness of the evaluated subset of samples to the ideal target fuzzy subset for characterizing different link quality classes is calculated by the following formula:
Figure FDA0003329267830000032
wherein the content of the first and second substances,
Figure FDA0003329267830000033
representing a fuzzy evaluation subset AiThe corresponding sample belongs to the class vkThe degree of membership of (a) is,
Figure FDA0003329267830000034
representing an ideal target fuzzy subset BjCorresponding sample to grade vkN represents the number of samples, and p is a predetermined constant.
6. A method for predicting link quality of a power wireless network is characterized by comprising the following steps:
acquiring parameter values of parameters of multiple links of a link to be detected;
inputting the parameter values into a power wireless network link quality prediction model to obtain the link quality grade of the link to be detected, wherein the power wireless network link quality prediction model is obtained by executing the power wireless network link quality prediction model training method according to any one of claims 1 to 5.
7. A power wireless network link quality prediction model training device is characterized by comprising:
the system comprises a sample acquisition module, a link parameter analysis module and a link parameter analysis module, wherein the sample acquisition module is used for acquiring a plurality of groups of samples, and each group of samples comprises parameter values of a plurality of types of link parameters;
the link quality grade dividing module is used for calculating fuzzy evaluation subsets of each group of samples according to the membership degree of parameter values in each group of samples to each value range and the weight of each type of link parameters, calculating link quality grade dividing ranges according to the closeness of the evaluation subsets of each group of samples and ideal target fuzzy subsets for representing different link quality grades, and the link quality grade dividing ranges comprise the value ranges of each type of link parameters under different link quality grades;
the training set acquisition module is used for acquiring a training data set, wherein the training data set comprises a plurality of groups of training samples, and each group of training samples comprises parameter values of a plurality of types of link parameters;
and the training module is used for inputting the link quality grade division range and parameter values in the training samples into a cyclic neural network, and training the cyclic neural network to obtain the power wireless network link quality prediction model.
8. An apparatus for predicting link quality of a power wireless network, comprising:
the to-be-detected parameter acquisition module is used for acquiring parameter values of parameters of multiple links of the to-be-detected link;
the link quality prediction module is used for inputting the parameter values into a power wireless network link quality prediction model to obtain the link quality grade of the link to be detected, and the power wireless network link quality prediction model is obtained by executing the power wireless network link quality prediction model training method according to any one of claims 1 to 5.
9. A computer device, comprising:
at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to perform the power wireless network link quality prediction model training method of any one of claims 1-5 or the power wireless network link quality prediction method of claim 6.
10. A computer-readable storage medium storing computer instructions for causing a computer to execute the power wireless network link quality prediction model training method according to any one of claims 1 to 5 or the power wireless network link quality prediction method according to claim 6.
CN202111291478.0A 2021-10-29 2021-10-29 Power wireless network link quality prediction and model training method and device Pending CN113988441A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111291478.0A CN113988441A (en) 2021-10-29 2021-10-29 Power wireless network link quality prediction and model training method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111291478.0A CN113988441A (en) 2021-10-29 2021-10-29 Power wireless network link quality prediction and model training method and device

Publications (1)

Publication Number Publication Date
CN113988441A true CN113988441A (en) 2022-01-28

Family

ID=79745957

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111291478.0A Pending CN113988441A (en) 2021-10-29 2021-10-29 Power wireless network link quality prediction and model training method and device

Country Status (1)

Country Link
CN (1) CN113988441A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115037630A (en) * 2022-04-29 2022-09-09 电子科技大学长三角研究院(湖州) Weighted network link prediction method based on structural disturbance model
CN116437409A (en) * 2023-06-13 2023-07-14 微网优联科技(成都)有限公司 Channel switching method and device for wireless router
CN117156485A (en) * 2023-10-31 2023-12-01 西安明赋云计算有限公司 Link quality detection method
WO2024045576A1 (en) * 2022-08-30 2024-03-07 中兴通讯股份有限公司 Network link generation method, server and storage medium

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115037630A (en) * 2022-04-29 2022-09-09 电子科技大学长三角研究院(湖州) Weighted network link prediction method based on structural disturbance model
CN115037630B (en) * 2022-04-29 2023-10-20 电子科技大学长三角研究院(湖州) Weighted network link prediction method based on structure disturbance model
WO2024045576A1 (en) * 2022-08-30 2024-03-07 中兴通讯股份有限公司 Network link generation method, server and storage medium
CN116437409A (en) * 2023-06-13 2023-07-14 微网优联科技(成都)有限公司 Channel switching method and device for wireless router
CN116437409B (en) * 2023-06-13 2023-08-22 微网优联科技(成都)有限公司 Channel switching method and device for wireless router
CN117156485A (en) * 2023-10-31 2023-12-01 西安明赋云计算有限公司 Link quality detection method
CN117156485B (en) * 2023-10-31 2024-01-09 西安明赋云计算有限公司 Link quality detection method

Similar Documents

Publication Publication Date Title
CN113988441A (en) Power wireless network link quality prediction and model training method and device
CN106471851B (en) Learning model based device positioning
CN111526119B (en) Abnormal flow detection method and device, electronic equipment and computer readable medium
CN111294812B (en) Resource capacity-expansion planning method and system
CN109409628A (en) Acquisition terminal production firm evaluation method based on metering big data Clustering Model
CN111222776A (en) Satellite network coordination situation assessment method and system based on convolutional neural network
WO2021129464A1 (en) Channel identification method and apparatus, transmission method, transmission device, base station, and medium
CN103577876A (en) Credible and incredible user recognizing method based on feedforward neural network
CN116596095B (en) Training method and device of carbon emission prediction model based on machine learning
CN111460380B (en) Multi-working-condition driving range prediction method and system based on Gaussian process regression
CN111177500A (en) Data object classification method and device, computer equipment and storage medium
CN110751201B (en) SAR equipment task failure cause reasoning method based on textural feature transformation
CN111882119A (en) Battery SOH prediction optimization method based on SA-BP neural network
CN107066786A (en) Aerosol optical depth inversion algorithm based on neutral net
CN116186507A (en) Feature subset selection method, device and storage medium
CN115829209A (en) Environment-friendly intelligent warehouse environment-friendly quality analysis method and device based on carbon path
CN109389281A (en) A kind of acquisition terminal production firm evaluation method based on gauss hybrid models
CN107506824B (en) Method and device for detecting bad observation data of power distribution network
CN113516229A (en) Credible user optimization selection method facing crowd sensing system
CN111352813A (en) Distributed storage use capacity estimation method and device
Song et al. AI-Enabled Quality Prediction of 5G Wireless Network in Smart Grid
CN116188834B (en) Full-slice image classification method and device based on self-adaptive training model
CN116108974A (en) Demand response baseline load prediction method and device considering meteorological factors
Douak et al. Design of a multiblock general regression neural network for wind speed prediction in Algeria
CN117574080A (en) CEEMDAN-RF-LSTM-based traffic flow time series data prediction method and system

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