CN106529816B - The method of adjustment and system of power line channel - Google Patents
The method of adjustment and system of power line channel Download PDFInfo
- Publication number
- CN106529816B CN106529816B CN201611025645.6A CN201611025645A CN106529816B CN 106529816 B CN106529816 B CN 106529816B CN 201611025645 A CN201611025645 A CN 201611025645A CN 106529816 B CN106529816 B CN 106529816B
- Authority
- CN
- China
- Prior art keywords
- power line
- line channel
- channel quality
- sample
- quality
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- G06Q10/06—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
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of method of adjustment of power line channel and system, it is first to carry out entry evaluation with multiple power line channel parameter evaluation models to the current power line relevant parameter for influenceing power line channel quality, assessment result is assessed using power line channel Evaluation Model on Quality again, multiple power line channel parameter evaluation models and power line channel Evaluation Model on Quality are combined, assessed by two-wheeled, make the judgement to power line channel quality more accurate, the degree of accuracy and intellectuality of power line channel Quality estimation can be effectively improved, power line channel is adjusted on this basis, the Adjustment effect of power line channel can be improved, be advantageous to build the good power circuit of channel quality.
Description
Technical field
The present invention relates to power-line carrier communication field, more particularly to a kind of method of adjustment of power line channel and
System.
Background technology
At present, power line carrier communication has by the use of existing low-voltage network circuit as the transmission medium of communication channel
Effect reduces the erection cost and later maintenance expense at communication line initial stage, reduces the difficulty of equipment installation and debugging.But by
It is not the special circuit set up for communication in distribution network, its line environment opened, complex network structures and changeable
Load characteristic so that channel circumstance has the characteristics such as strong noise, highly attenuating and impedance mismatch.Build the good electricity of channel quality
Line of force road, it is significant for improving power line carrier communication quality.
Channel mathematical modeling mainly is carried out according to power line channel noise in conventional art, utilizes the channel model of foundation
The quality of power line channel is judged, power line channel is adjusted further according to judged result, due to power line channel
Environment diversity, the quality of power line channel is affected by various factors, according to power line channel noise establish letter
Road model is low to the accuracy of judgement degree of power line channel quality, makes the effect of adjustment power line channel reduce.
The content of the invention
Based on this, it is necessary to for conventional art adjust power line channel effect it is low the problem of, there is provided a kind of power line
The method of adjustment and system of channel.
A kind of method of adjustment of power line channel, comprises the following steps：
Obtain the current power line relevant parameter for influenceing power line channel quality；
According to multiple different power line channel parameter evaluation models respectively on the current electricity for influenceing power line channel quality
Line of force relevant parameter is assessed, and obtains multiple assessments on the current power line relevant parameter for influenceing power line channel quality
As a result；
Multiple assessment results are assessed according to power line channel Evaluation Model on Quality, obtain goal-based assessment result；
If goal-based assessment result is different from default assessed value, power line channel is adjusted.
A kind of adjustment system of power line channel, including with lower unit：
Acquiring unit, for obtaining the current power line relevant parameter for influenceing power line channel quality；
Entry evaluation unit, for influenceing electricity to current respectively according to multiple different power line channel parameter evaluation models
The power line relevant parameter of line of force channel quality is assessed, and is obtained multiple on the current electric power for influenceing power line channel quality
The assessment result of line relevant parameter；
Comprehensive assessment unit, for being assessed according to power line channel Evaluation Model on Quality multiple assessment results, obtain
Obtain goal-based assessment result；
Adjustment unit, in goal-based assessment result and default assessed value difference, being adjusted to power line channel.
According to the method for adjustment and system of the power line channel of the invention described above, it is first with multiple power line channel parameters
Assessment models carry out entry evaluation to the current power line relevant parameter for influenceing power line channel quality, then using power line channel
Evaluation Model on Quality is assessed assessment result, and multiple power line channel parameter evaluation models and power line channel quality are commented
Estimate model to be combined, assessed by two-wheeled, make the judgement to power line channel quality more accurate, power line can be effectively improved
The degree of accuracy and intellectuality that channel quality judges, are adjusted to power line channel on this basis, can improve power line letter
The Adjustment effect in road, be advantageous to build the good power circuit of channel quality.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of the method for adjustment of the power line channel of one of embodiment；
Fig. 2 is the schematic diagram of power line channel quality evaluation process in one of specific embodiment；
Fig. 3 is the topology diagram of recurrent neural network in one of specific embodiment；
Fig. 4 is the schematic diagram calculation of AdaBoost algorithms in one of specific embodiment；
Fig. 5 is the structural representation of the adjustment system of the power line channel of one of embodiment.
Embodiment
For the objects, technical solutions and advantages of the present invention are more clearly understood, below in conjunction with drawings and Examples, to this
Invention is described in further detail.It should be appreciated that embodiment described herein is only to explain the present invention,
Do not limit protection scope of the present invention.
It is shown in Figure 1, for the schematic flow sheet of the method for adjustment of the power line channel of the present invention.Electricity in the embodiment
The method of adjustment of line of force channel, comprises the following steps：
Step S101：Obtain the current power line relevant parameter for influenceing power line channel quality；
In this step, polytype data can be included by influenceing the power line relevant parameter of power line channel quality.
Step S102：Power line channel is influenceed according to multiple different power line channel parameter evaluation models on current respectively
The power line relevant parameter of quality is assessed, and is obtained multiple on the related ginseng of the current power line for influenceing power line channel quality
Several assessment results；
Step S103：Multiple assessment results are assessed according to power line channel Evaluation Model on Quality, target is obtained and comments
Estimate result；
Step S104：If goal-based assessment result is different from default assessed value, power line channel is adjusted.
In this step, it is to be determined whether according to the comparative result of goal-based assessment result and default assessed value to power line
Channel is adjusted；If goal-based assessment result is different from default assessed value, show power line channel existing defects, it is necessary to electric power
Line channel is adjusted；If goal-based assessment result is identical with default assessed value, show that power line channel quality is good, can not be right
Power line channel is adjusted.
In the present embodiment, first with multiple power line channel parameter evaluation models on current influence power line channel quality
Power line relevant parameter carries out entry evaluation, then assessment result is assessed using power line channel Evaluation Model on Quality, will
Multiple power line channel parameter evaluation models and power line channel Evaluation Model on Quality are combined, and are assessed, are made to electricity by two-wheeled
The judgement of line of force channel quality is more accurate, can be effectively improved the degree of accuracy and intellectuality of power line channel Quality estimation, with
This is that foundation is adjusted to power line channel, can improve the Adjustment effect of power line channel, is advantageous to build channel quality
Good power circuit.
In one of the embodiments, obtain it is current influence power line channel quality power line relevant parameter the step of
Comprise the following steps before：
Obtain the multiple historical samples for the power line relevant parameter for influenceing power line channel quality；
It is utilized respectively each power line channel parameter evaluation model to assess multiple historical samples one by one, obtains each
The assessment result of power line channel parameter evaluation model, wherein, each power line channel parameter evaluation model corresponds to multiple assessments
As a result；
The assessment models based on Adaboost algorithm are established, by what is obtained according to all power line channel parameter evaluation models
The assessment result of all historical samples is combined into training sample set, according to assessment mould of the training sample set pair based on Adaboost algorithm
Type is trained, and obtains power line channel Evaluation Model on Quality.
In the present embodiment, first with multiple power line channel parameter evaluation models to influence power line channel quality electric power
Multiple historical samples of line relevant parameter are assessed, and obtain multiple historical samples in different power line channel parameter evaluation models
In assessment result, be trained as assessment models of the training sample set pair based on Adaboost algorithm, obtain electric power
Line channel quality assessment model, power line channel Evaluation Model on Quality is assessed power line channel parameter evaluation model and obtain
Assessment result, so as to when being adjusted to power line channel use the power line channel Evaluation Model on Quality to electric power
Line channel quality is assessed；Meanwhile the assessment models of the invention based on Adaboost algorithm can be according to influence power line
The new historical sample constantly self-teaching of the power line relevant parameter of channel quality, power line channel quality evaluation is adjusted in real time
Model, the adjustment for power line channel under continually changing power line channel environment have very strong adaptability.
In one of the embodiments, instructed according to assessment models of the training sample set pair based on Adaboost algorithm
The step of practicing, obtaining the power line channel Evaluation Model on Quality comprises the following steps：
Initialize the probability that training sample concentrates all training samples；
Concentrated in training sample and randomly select several training samples, be based on according to the training of the training sample of extraction
Grader in the assessment models of Adaboost algorithm；
The error rate of grader is calculated, if error rate is less than error rate threshold, the power of grader is calculated according to error rate
Weight；If error rate is more than or equal to error rate threshold, the probability increase by first that will appear from the training sample of classification error is default
Step-length, the probability for not occurring the training sample of classification error is reduced into the second default step-length, corresponding instruct is concentrated to training sample
The probability for practicing sample is updated, and is back to and is concentrated the step of randomly selecting several training samples in training sample, until
Error rate is less than error rate threshold；
It is back to and concentrates the step of randomly selecting several training samples in training sample, until obtains point of predetermined number
The weight of class device；
According to the weight of the grader of predetermined number, the grader of predetermined number is overlapped, obtains power line channel
Evaluation Model on Quality.
In the present embodiment, several training samples training commenting based on Adaboost algorithm that training sample is concentrated is extracted
The grader in model is estimated, using the error rate of grader as training sample probability regularization condition, when the error rate of grader is more than
Or occur the probability of the training sample of classification error equal to error rate threshold, then increase, reduce the training for not occurring classification error
The probability of sample, and extract several training samples again and the grader in the assessment models based on Adaboost algorithm is carried out
Training, the focus of training can so be concentrated on the training sample for being difficult to classify, obtain multiple graders afterwards, folded
Add, obtaining power line channel Evaluation Model on Quality, this can possess the strong classifier of adaptive learning characteristic, it is possible to achieve to electricity
The correct classification of line of force channel parameter, so as to greatly improve the degree of accuracy of power line channel quality evaluation.
First default step-length and the second default step-length can freely be set as needed, and qualifications are adjusting training samples
Probability after, training sample concentrates the probability sum of all training samples to remain as 1.
The weight of grader is related to the error rate of grader, and the power of grader can be calculated according to the error rate of grader
Weight.
In one of the embodiments, the training sample after initialization concentrates the probability of all training samples identical.
In the present embodiment, the training sample after initialization concentrates all training samples for based on Adaboost algorithm
All it is brand-new for grader in assessment models, grader concentrates all training samples to the training sample after initialization
Classification results are unknown before training, therefore randomly select training sample to based on Adaboost algorithm with identical probability
Assessment models in grader be trained, grader to all training samples with identical probability carry out classification learning,
The training of grader can be made more effective.
In one of the embodiments, power line channel parameter evaluation model is two, respectively based on multi-variable decision
The analysis and evaluation model of tree and the failure predication model based on recurrent neural network；
The step of each power line channel parameter evaluation model assesses multiple historical samples one by one is utilized respectively to wrap
Include following steps：
Multiple historical samples are assessed respectively using the analysis and evaluation model based on multivariable decision tree；
Multiple historical samples are assessed respectively using the failure predication model based on recurrent neural network.
In the present embodiment, power line channel parameter evaluation model is two, point respectively based on multivariable decision tree
Analyse assessment models and the failure predication model based on recurrent neural network；The nicety of grading of multivariable decision tree is higher, Er Qiesheng
Pattern into decision tree is simple, there is good robustness to noise data；Recurrent neural network have very strong dynamic behaviour and
Computing capability, the analysis and evaluation model based on multivariable decision tree and the failure predication model based on recurrent neural network both
Model is of a relatively high to the accuracy rate for influenceing the assessment of the power line relevant parameter of power line channel quality.
In one of the embodiments, using the analysis and evaluation model based on multivariable decision tree respectively to multiple history samples
The step of this is assessed comprises the following steps：
According to electric power of the power line relevant parameter type to influence power line channel quality for influenceing power line channel quality
Multiple historical samples of line relevant parameter are divided, on the related ginseng of power line of each type of influence power line channel quality
Number carries out data analysis, it is determined that influenceing the feature samples of the power line relevant parameter type of power line channel quality；
Built according to the feature samples for the power line relevant parameter type for influenceing power line channel quality and determined based on multivariable
The analysis and evaluation model of plan tree, multiple historical samples are commented respectively according to the analysis and evaluation model based on multivariable decision tree
Estimate.
In the present embodiment, the analysis and evaluation model based on multivariable decision tree is according to influence power line channel quality
The feature samples structure of power line relevant parameter type, it can effectively judge the electric power of the influence power line channel quality of input
The degree of optimization of line relevant parameter, and then realize the assessment of the power line relevant parameter to influenceing power line channel quality.
In one of the embodiments, using the failure predication model based on recurrent neural network respectively to multiple history samples
The step of this is assessed comprises the following steps：
It is determined that the influent factor in the power line relevant parameter of influence power line channel quality, is carried from multiple historical samples
Influent factor sample is taken, is established using recurrent neural network between influent factor sample and default power line channel quality coded
Mapping relations；
The failure predication model based on recurrent neural network is built according to mapping relations, according to based on recurrent neural network
Failure predication model is assessed multiple historical samples respectively.
In the present embodiment, the failure predication model based on recurrent neural network is according to influence power line channel quality
What the mapping relations between the influent factor sample and default power line channel quality coded in power line relevant parameter were built, base
It can effectively judge that the power line of the influence power line channel quality of input is related in the failure predication model of recurrent neural network
Relation between parameter and power line channel quality, and then realize the power line relevant parameter to influenceing power line channel quality
Assess.
In one of the embodiments, recurrent neural network includes input layer, hidden layer and output layer；
The mapping established using recurrent neural network between influent factor sample and default power line channel quality coded is closed
The step of being comprises the following steps：
Output layer is built using influent factor sample structure input layer, using default power line channel quality coded, to passing
Return neutral net to be trained, obtain the mapping relations between influent factor sample and default power line channel quality coded.
In the present embodiment, influent factor sample structure input layer need to be only utilized, is compiled using default power line channel quality
Code structure output layer, because recurrent neural network has very strong dynamic behaviour and computing capability, it is determined that input layer and output
After layer, foundation can be provided with quick obtaining mapping relations for follow-up assessment by training.
In one of the embodiments, influence power line channel quality power line relevant parameter include power line length,
Branch power line number, branch power line length, power line load impedance characteristic, power line types and power line noise intensity.
In the present embodiment, influenceing the power line relevant parameter of power line channel quality includes power line length, power line
The polytypes such as branch's number, branch power line length, power line load impedance characteristic, power line types and power line noise intensity
Parameter, the method for adjustment of power line channel of the invention considered the parameter of the various influence quality of power line channel,
Make the assessment to power line channel quality more comprehensive, so as to improve the degree of accuracy of power line channel Quality estimation, be advantageous to electricity
The adjustment of line of force channel.
In a specific embodiment, power line channel quality evaluation process is as shown in Figure 2.First, by gathering electricity
A variety of shadows such as line of force length, branch power line number, branch length, load impedance characteristic, power line types, power line noise intensity
The power line relevant parameter of power line channel quality is rung, establishing magnanimity influences the power line relevant parameter number of power line channel quality
Include the different historical data samples of power line according to storage and shared platform, the inside, neutral net and depth are based on for application
Learn isotype recognizer and power line channel quality evaluation offer data support is provided.System evaluation method, which specifically utilizes, to be had
The integrated study model based on Adaboost algorithm of adaptive characteristic is to a variety of different power line channel parameter evaluation models
(the expert diagnosis assessment models such as based on multivariable decision tree and the failure predication model based on recurrent neural network) is commented
Estimate result to be assessed, including evaluation, weight regulation, summation final decision etc., obtain the final of power line channel quality evaluation
As a result (such as three assessment results of power line channel quality excellent, good, poor).Power line channel quality evaluation system utilizes big data meter
Calculate platform realization " the expert diagnosis assessment models based on multivariable decision tree ", " failure based on recurrent neural network (RNN) is pre-
Survey model " and the calculating of " based on the adaptive set of Adaboost algorithm into learning model " analyze, to a variety of different appraisal procedures
Result carry out classified calculating, realize with power line channel quality visualization and Intellisense.
In the present invention, power line channel parameter evaluation model can its quantity is not limited, below with two
The particular content of the present invention is described exemplified by individual power line channel parameter evaluation model：
Preferably, two power line channel parameter evaluation models are the expert diagnosis assessment models based on multivariable decision tree
With the failure predication model based on recurrent neural network；
Based on the power line associated parameter data storehouse for influenceing power line channel quality, analysis power line length, power line point
The data sets such as number, branch length, load impedance characteristic, power line types, power line noise intensity, are studied and are determined based on multivariable
The expert evaluation model of plan tree, using the data set of rough set theory division power line collection gained, select best variable
Set, construct multivariable decision tree；Then, failure predication model of the research based on recurrent neural network (RNN), power line is extracted
Feature and channel quality influent factor, the mapping relations established between influent factor and channel quality, electric power is predicted according to mapping
Line channel quality；Finally, on the basis of analysis expert assessment result and RNN prediction results is obtained, integrated using AdaBoost
Learning method, using error rate as qualifications, using the decision model that Nearest Neighbor with Weighted Voting mechanism construction is final, realize and power line is believed
The final assessment of road quality, obtain the assessment final result of power line channel quality.
Power line channel is adjusted according to the assessment final result can of obtained power line channel quality.Can be with
Final result will be assessed compared with default assessed value, default assessed value herein can represent that power line channel quality is good,
If it is different from default assessed value to assess final result, show current power line bad channel quality, then power line channel is carried out
Adjustment；If it is identical with default assessed value to assess final result, show that current power line channel quality is good, then can not be to electric power
Line channel is adjusted.The assessment final result of power line channel quality can represent with different quality codeds, can be pre-
If wherein the good quality coded of representation quality represents default assessed value.
Analysis expert assessment models based on multivariable decision tree to establish process as follows：
Using the power line associated parameter data for influenceing power line channel quality collected as target data set, utilization is coarse
Collection is theoretical, and according to power line parameters Type division target data set, multigroup experiment sample analysis is carried out to every kind of power line parameters
Afterwards, power line parameters feature samples are determined, obtain the variables collection most beneficial for classification, construct multivariable decision tree.Changeable
Multiple attributes can be examined simultaneously on a certain node of decision tree by measuring, and produce new, more relevant attribute, and change or remove
The incoherent attribute initially provided.Using relatively core and relative generalization theory, for the selection of decision attribute, construction and optimization
Method is studied, and based on this, extracting rule, establishes expert's power line channel parameter evaluation based on multivariable decision tree
Model.
It is the algorithm that specific decision-tree model is constructed from the power line channel supplemental characteristic of gained below：In form,
One information system S is defined as a four-tuple S=<U,A,V,f>.Wherein U is domain；A is the set of all properties, and it enters
One step can be divided into conditional attribute C and decision attribute D, V=UP∈AVp, VpIt is attribute P codomain, f:U×A→V,VpReferred to as one letter
Cease function.
Constructing the step of multivariable is examined is:
(1) design conditions property set C is relative to decision kind set D core, i.e. CORED (C).If CORED (C)=C ∩ D,
Then turn (2)；Otherwise, CORED (C)={ a might as well be set1,a2,...,ak, turn (3).
(2) with a kind of ID3 (ID3 algorithms are greedy algorithm, for constructing decision tree) one best attributes of method choice,
Inspection as the node.
(3) P=a is made1∧a2∧...∧ak(a1,…,akThe sample of structure decision tree is represented, ∧ represents sample synthesis), meter
P is calculated relative to D extensive GEND (P), the inspection using it as the node.
Root using GEND (P) as decision tree then according to the value of attribute, is divided into object different sons by this algorithm
Collection will export one tree in a similar way to each subset.
Failure predication model based on recurrent neural network (RNN) to establish process as follows：
Collecting influences the power line associated parameter data of power line channel quality, and channel matter is determined using factorial analysis algorithm
Influent factor is measured, is extracted per class channel quality influent factor sample, is designed preset channel quality coded, utilize recurrent neural network
The mapping relations established between influent factor and preset channel quality coded, and on the basis of gained mapping relations, to power line
Channel quality is assessed.
Fig. 3 is three layers of RNN topology diagrams, and it is by input layer (N1Individual node), hidden layer (N2Individual node) and output layer
(N3Individual node) composition.Wherein, hidden node not only receives the output signal from input layer, also receives the output of itself delay
Signal.
Xi(h) be time h i-th of hidden node input, Bj(h) it is output in time h j-th of hidden node,
Y (h) is N3Tie up output vector.The network can be described as：
Bj(h)=f (Sj(h)) (2)
Wherein:WI, WR, WORespectively from input layer to hidden layer, return signal, the weight coefficient from hidden layer to output layer；The weight coefficient of deviation unit respectively on hidden layer and output layer；F () is sigmoid functions.T, U distinguishes
For the deviation of output layer set in advance and hidden layer in network.
In this patent, input layer, preset channel quality coded conduct are built using power line channel quality influent factor
Output layer, RNN is formed by training, establishes channel quality assessment model.
Based on the adaptive set of Adaboost algorithm into learning model to establish process as follows：
The model that analysis expert assessment models and RNN failure predications model are trained as early stage, to influence electric power
Based on the power line associated parameter data of line channel quality, the assessment result of various historical data samples is obtained respectively, by institute
Some assessment results randomly select the weak of assessment models of the training sample training based on Adaboost algorithm as training sample set
Grader, using each error rate as weighting qualifications, the training sample that the training data chosen after weighting is replaced randomly selecting
This, focus is concentrated in the training sample data of difficult point of comparison, after successive ignition circulates, obtains multiple Weak Classifiers,
The weight of each Weak Classifier is calculated, by all Weak Classifier weighted superpositions, the strong classifier that can learn automatically is built, carries significantly
High power line channel quality evaluation system accuracy rate.
Adaboost is a kind of iterative algorithm, and its core concept is that different graders is trained for same training set
(Weak Classifier), then these weak classifier sets are got up, form a stronger final classification device (strong classifier).
AdaBoost algorithms comprise the following steps that：
Step 1. gives comprising the analysis expert power line channel parameter evaluation model based on multivariable decision tree and is based on
The instruction of the power line channel quality assessment result data of the power line channel quality error assessment models of recurrent neural network (RNN)
Practice sample set S, wherein A and B and correspond respectively to positive example sample and negative example sample；T is the maximum cycle of training；
Step 2. initialization sample weight is 1/n, as training sample initial probability distribution；
Step 3. first time iteration：
(1) under the conditions of the probability of training sample, several training samples training Weak Classifier is extracted；
(2) error rates of weak classifiers ε is calculatedi；
(3) suitable error rate threshold M is chosen (M is constantly adjusted in the training process) so that training sample error is minimum；If
Error rate is less than error rate threshold, then the weight of Weak Classifier is calculated according to error rate；If error rate is more than error rate threshold, turn
(4)；
(4) training sample probability ω is updated；The rule of renewal is：Reduce Weak Classifier and do not occur the training sample of classification error
There is the probability of the training sample of classification error in this probability, increase Weak Classifier；
(5) training sample training Weak Classifier is washed away again.
After successive ignition, T Weak Classifier is obtained, by Weak Classifier weight αiSuperposition, finally gives strong classifier.
Above-mentioned detailed process is as shown in Figure 4.
According to the method for adjustment of above-mentioned power line channel, the present invention also provides a kind of adjustment system of power line channel, with
The embodiment of the lower just adjustment system of the power line channel of the present invention is described in detail.
It is shown in Figure 5, for the embodiment of the adjustment system of the power line channel of the present invention.Power line in the embodiment
The adjustment system of channel includes acquiring unit 210, entry evaluation unit 220, comprehensive assessment unit 230 and adjustment unit 240；
Acquiring unit 210, for acquiring unit, for obtaining the related ginseng of the current power line for influenceing power line channel quality
Number；
Entry evaluation unit 220, for according to multiple different power line channel parameter evaluation models respectively to current shadow
The power line relevant parameter for ringing power line channel quality is assessed, and is obtained multiple on currently influenceing power line channel quality
The assessment result of power line relevant parameter；
Comprehensive assessment unit 230, for being assessed according to power line channel Evaluation Model on Quality multiple assessment results,
Obtain goal-based assessment result；
Adjustment unit 240, in goal-based assessment result and default assessed value difference, being adjusted to power line channel
It is whole.
In one of the embodiments, acquiring unit 210 is additionally operable to obtain the power line phase for influenceing power line channel quality
Multiple historical samples of related parameter；
Entry evaluation unit 220 is additionally operable to be utilized respectively each power line channel parameter evaluation model to multiple historical samples
Assessed one by one, obtain the assessment result of each power line channel parameter evaluation model, wherein, each power line channel parameter
Assessment models correspond to multiple assessment results；
Comprehensive assessment unit 230 is additionally operable to establish the assessment models based on Adaboost algorithm, will be according to all power lines
The assessment result for all historical samples that channel parameter assessment models obtain is combined into training sample set, according to training sample set pair base
It is trained in the assessment models of Adaboost algorithm, obtains power line channel Evaluation Model on Quality.
In one of the embodiments, comprehensive assessment unit 230 initializes training sample and concentrates the general of all training samples
Rate；Concentrated in training sample and randomly select several training samples, calculated according to the training of the training sample of extraction based on Adaboost
Grader in the assessment models of method；The error rate of grader is calculated, when error rate is less than error rate threshold, according to error rate
Calculate the weight of grader；When error rate is more than or equal to error rate threshold, will appear from classification error training sample it is general
The first default step-length of rate increase, the probability for not occurring the training sample of classification error is reduced into the second default step-length, to training sample
The probability of training sample corresponding to this concentration is updated, and is concentrated again in training sample and randomly selected several training samples
This, until error rate is less than error rate threshold；Again concentrated in training sample and randomly select several training samples, until obtaining
The weight of the grader of predetermined number；According to the weight of the grader of predetermined number, the grader of predetermined number is weighted
Superposition, obtain power line channel Evaluation Model on Quality.
In one of the embodiments, the training sample after initialization concentrates the probability of all training samples identical.
In one of the embodiments, power line channel parameter evaluation model is two, respectively based on multi-variable decision
The analysis and evaluation model of tree and the failure predication model based on recurrent neural network；
Entry evaluation unit 220 is entered to multiple historical samples respectively using the analysis and evaluation model based on multivariable decision tree
Row is assessed；
Entry evaluation unit 220 is entered to multiple historical samples respectively using the failure predication model based on recurrent neural network
Row is assessed.
In one of the embodiments, entry evaluation unit 220 is related according to the power line for influenceing power line channel quality
Parameter type divides to the multiple historical samples for influenceing the power line relevant parameter of power line channel quality, to each type
Influence power line channel quality power line relevant parameter carry out data analysis, it is determined that influence power line channel quality electric power
The feature samples of line relevant parameter type；According to the feature samples for the power line relevant parameter type for influenceing power line channel quality
The analysis and evaluation model based on multivariable decision tree is built, is gone through according to the analysis and evaluation model based on multivariable decision tree to multiple
History sample is assessed respectively.
In one of the embodiments, entry evaluation unit 220 determines that the power line for influenceing power line channel quality is related
Influent factor in parameter, power line channel quality influent factor sample is extracted from multiple historical samples, utilizes recurrent neural
The mapping relations that network is established between influent factor sample and default power line channel quality coded；Base is built according to mapping relations
In the failure predication model of recurrent neural network, according to the failure predication model based on recurrent neural network to multiple historical samples
Assessed respectively.
In one of the embodiments, recurrent neural network includes input layer, hidden layer and output layer；
Entry evaluation unit 220 builds input layer using influent factor sample, utilizes default power line channel quality coded
Build output layer, recurrent neural network be trained, obtain influent factor sample and default power line channel quality coded it
Between mapping relations.
In one of the embodiments, influence power line channel quality power line relevant parameter include power line length,
Branch power line number, branch power line length, power line load impedance characteristic, power line types and power line noise intensity.
The adjustment system of the power line channel of the present invention and the method for adjustment of the power line channel of the present invention correspond,
The technical characteristic and its advantage that the embodiment of the method for adjustment of above-mentioned power line channel illustrates are applied to power line channel
Adjustment system embodiment in.
Each technical characteristic of embodiment described above can be combined arbitrarily, to make description succinct, not to above-mentioned reality
Apply all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, the scope that this specification is recorded all is considered to be.
Embodiment described above only expresses the several embodiments of the present invention, and its description is more specific and detailed, but simultaneously
Can not therefore it be construed as limiting the scope of the patent.It should be pointed out that come for one of ordinary skill in the art
Say, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the protection of the present invention
Scope.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (8)
1. a kind of method of adjustment of power line channel, it is characterised in that comprise the following steps：
Obtain the current power line relevant parameter for influenceing power line channel quality；
According to multiple different power line channel parameter evaluation models respectively on the current electricity for influenceing power line channel quality
Line of force relevant parameter is assessed, and obtains multiple power line relevant parameters on the current influence power line channel quality
Assessment result；
Multiple assessment results are assessed according to power line channel Evaluation Model on Quality, obtain goal-based assessment result；Its
In, the power line channel Evaluation Model on Quality is the housebroken assessment models based on Adaboost algorithm；
If the goal-based assessment result is different from default assessed value, power line channel is adjusted；
Comprise the following steps before the step of power line relevant parameter of the current influence power line channel quality of the acquisition：
Obtain the multiple historical samples for the power line relevant parameter for influenceing power line channel quality；
It is utilized respectively each power line channel parameter evaluation model to assess the multiple historical sample one by one, described in acquisition
The assessment result of each power line channel parameter evaluation model, wherein, each power line channel parameter evaluation model corresponds to multiple
Assessment result；
The assessment models based on Adaboost algorithm are established, it is all by being obtained according to all power line channel parameter evaluation models
The assessment result of historical sample is combined into training sample set, according to commenting based on Adaboost algorithm described in the training sample set pair
Estimate model to be trained, obtain the power line channel Evaluation Model on Quality；
It is described to be trained according to the assessment models of the training sample set pair based on Adaboost algorithm, obtain the power line
The step of channel quality assessment model, comprises the following steps：
Initialize the probability that the training sample concentrates all training samples；
Concentrated in the training sample and randomly select several training samples, be based on according to the training of the training sample of extraction
Grader in the assessment models of Adaboost algorithm；
The error rate of the grader is calculated, if the error rate is less than error rate threshold, institute is calculated according to the error rate
State the weight of grader；
If the error rate is more than or equal to error rate threshold, the probability increase by first of the training sample of classification error will appear from
Default step-length, the probability for not occurring the training sample of classification error is reduced into the second default step-length, the training sample is concentrated
The probability of corresponding training sample is updated, and is back to described concentrated in the training sample and is randomly selected several training
The step of sample, until the error rate is less than the error rate threshold；
The step of randomly selecting several training samples is concentrated in the training sample described in being back to, until obtaining predetermined number
Grader weight；
According to the weight of the grader of the predetermined number, superposition is weighted to the grader of the predetermined number, obtains institute
State power line channel Evaluation Model on Quality.
2. the method for adjustment of power line channel according to claim 1, it is characterised in that the training sample set after initialization
In all training samples probability it is identical.
3. the method for adjustment of power line channel according to claim 1, it is characterised in that the power line channel parameter is commented
Model is estimated for two, respectively the analysis and evaluation model based on multivariable decision tree and the failure predication based on recurrent neural network
Model；
The step for being utilized respectively each power line channel parameter evaluation model and being assessed one by one the multiple historical sample
Suddenly comprise the following steps：
The multiple historical sample is assessed respectively using the analysis and evaluation model based on multivariable decision tree；
The multiple historical sample is assessed respectively using the failure predication model based on recurrent neural network.
4. the method for adjustment of power line channel according to claim 3, it is characterised in that based on changeable described in the utilization
The step of analysis and evaluation model of amount decision tree is assessed the multiple historical sample respectively comprises the following steps：
According to electric power of the power line relevant parameter type on the influence power line channel quality for influenceing power line channel quality
Multiple historical samples of line relevant parameter are divided, on the related ginseng of power line of each type of influence power line channel quality
Number carries out data analysis, it is determined that influenceing the feature samples of the power line relevant parameter type of power line channel quality；
Built according to the feature samples of the power line relevant parameter type of the influence power line channel quality and determined based on multivariable
The analysis and evaluation model of plan tree, according to the analysis and evaluation model based on multivariable decision tree to the multiple historical sample point
Do not assessed.
5. the method for adjustment of power line channel according to claim 3, it is characterised in that recurrence is based on described in the utilization
The step of failure predication model of neutral net is assessed the multiple historical sample respectively comprises the following steps：
The influent factor in the power line relevant parameter of the influence power line channel quality is determined, from the multiple historical sample
Middle extraction influent factor sample, using recurrent neural network establish influent factor sample and default power line channel quality coded it
Between mapping relations；
According to the failure predication model based on recurrent neural network described in mapping relations structure, according to described based on recurrence god
Failure predication model through network is assessed the multiple historical sample respectively.
6. the method for adjustment of power line channel according to claim 5, it is characterised in that the recurrent neural network includes
Input layer, hidden layer and output layer；
The mapping established using recurrent neural network between influent factor sample and default power line channel quality coded is closed
The step of being comprises the following steps：
The input layer is built using the influent factor sample, using described in the default power line channel quality coded structure
Output layer, the recurrent neural network is trained, obtain influent factor sample and default power line channel quality coded it
Between mapping relations.
7. the method for adjustment of power line channel as claimed in any of claims 1 to 6, it is characterised in that the shadow
Ringing the power line relevant parameter of power line channel quality includes power line length, branch power line number, branch power line length, electricity
Line of force load impedance characteristic, power line types and power line noise intensity.
8. the adjustment system of a kind of power line channel, it is characterised in that including with lower unit：
Acquiring unit, for obtaining the current power line relevant parameter for influenceing power line channel quality；
Entry evaluation unit, for electric on the current influence respectively according to multiple different power line channel parameter evaluation models
The power line relevant parameter of line of force channel quality is assessed, and acquisition is multiple currently to influence power line channel quality on described
The assessment result of power line relevant parameter；
Comprehensive assessment unit, for being assessed according to power line channel Evaluation Model on Quality multiple assessment results, obtain
Obtain goal-based assessment result；Wherein, the power line channel Evaluation Model on Quality is housebroken commenting based on Adaboost algorithm
Estimate model；
Adjustment unit, in the goal-based assessment result and default assessed value difference, being adjusted to the power line channel
It is whole；
The acquiring unit obtains the multiple historical samples for the power line relevant parameter for influenceing power line channel quality；
The entry evaluation unit is utilized respectively each power line channel parameter evaluation model and multiple historical samples is carried out one by one
Assess, obtain the assessment result of each power line channel parameter evaluation model, wherein, each power line channel parameter evaluation model
Corresponding multiple assessment results；
The comprehensive assessment unit establishes the assessment models based on Adaboost algorithm, will be commented according to all power line channel parameters
The assessment result for estimating all historical samples of model acquisition is combined into training sample set, and initialization training sample concentrates all training samples
This probability；Concentrated in training sample and randomly select several training samples, be based on according to the training of the training sample of extraction
Grader in the assessment models of Adaboost algorithm；The error rate of grader is calculated, when error rate is less than error rate threshold,
The weight of grader is calculated according to error rate；When error rate is more than or equal to error rate threshold, the instruction of classification error will appear from
Practice the first default step-length of probability increase of sample, the probability for not occurring the training sample of classification error is reduced into the second default step
It is long, the probability of training sample corresponding to training sample concentration is updated, and if being randomly selected again in training sample concentration
Dry training sample, until error rate is less than error rate threshold；Again concentrated in training sample and randomly select several training samples
This, until obtaining the weight of the grader of predetermined number；According to the weight of the grader of predetermined number, the classification to predetermined number
Device is weighted superposition, obtains power line channel Evaluation Model on Quality.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611025645.6A CN106529816B (en) | 2016-11-15 | 2016-11-15 | The method of adjustment and system of power line channel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611025645.6A CN106529816B (en) | 2016-11-15 | 2016-11-15 | The method of adjustment and system of power line channel |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106529816A CN106529816A (en) | 2017-03-22 |
| CN106529816B true CN106529816B (en) | 2018-03-30 |
Family
ID=58352678
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611025645.6A Active CN106529816B (en) | 2016-11-15 | 2016-11-15 | The method of adjustment and system of power line channel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106529816B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110324170B (en) * | 2018-03-30 | 2021-07-09 | 华为技术有限公司 | Data analysis equipment, multi-model co-decision system and method |
| CN113489514B (en) * | 2021-07-05 | 2022-07-26 | 国网湖南省电力有限公司 | Power line communication noise identification method and device based on self-organizing mapping neural network |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104535865A (en) * | 2014-12-30 | 2015-04-22 | 西安工程大学 | Comprehensive diagnosing method for operation troubles of power transformer based on multiple parameters |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7099880B2 (en) * | 2002-01-31 | 2006-08-29 | International Business Machines Corporation | System and method of using data mining prediction methodology |
-
2016
- 2016-11-15 CN CN201611025645.6A patent/CN106529816B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104535865A (en) * | 2014-12-30 | 2015-04-22 | 西安工程大学 | Comprehensive diagnosing method for operation troubles of power transformer based on multiple parameters |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106529816A (en) | 2017-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Wind speed forecasting using deep neural network with feature selection | |
| CN109272146B (en) | Flood prediction method based on deep learning model and BP neural network correction | |
| CN106650767B (en) | Flood forecasting method based on cluster analysis and real-time correction | |
| Kisi et al. | River suspended sediment estimation by climatic variables implication: Comparative study among soft computing techniques | |
| CN109933881A (en) | A kind of Fault Diagnosis of Power Electronic Circuits method based on optimization deepness belief network | |
| CN106656357B (en) | Power frequency communication channel state evaluation system and method | |
| CN104408562A (en) | Photovoltaic system generating efficiency comprehensive evaluation method based on BP (back propagation) neural network | |
| CN106198551A (en) | The detection method of a kind of transmission line of electricity defect and device | |
| CN108876163A (en) | The transient rotor angle stability fast evaluation method of comprehensive causality analysis and machine learning | |
| CN115099500B (en) | Water level prediction method based on weight correction and DRSN-LSTM model | |
| CN113536509B (en) | Micro-grid topology identification method based on graph convolution network | |
| CN103793887A (en) | Short-term electrical load on-line predicting method based on self-adaptation enhancing algorithm | |
| CN108595803A (en) | Shale gas well liquid loading pressure prediction method based on recurrent neural network | |
| CN116821774B (en) | Power generation fault diagnosis method based on artificial intelligence | |
| CN114492922A (en) | Medium-and-long-term power generation capacity prediction method | |
| CN110837915A (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
| CN113033081A (en) | Runoff simulation method and system based on SOM-BPNN model | |
| CN106529816B (en) | The method of adjustment and system of power line channel | |
| CN114021836A (en) | Multivariable reservoir water inflow amount prediction system based on different-angle fusion, training method and application | |
| Vafakhah et al. | Application of intelligent technology in rainfall analysis | |
| CN109035223A (en) | A kind of intelligent evaluation method for satellite remote sensing images availability | |
| CN117272102A (en) | Transformer fault diagnosis method based on double-attention mechanism | |
| CN112307410A (en) | Seawater temperature and salinity information time sequence prediction method based on shipborne CTD measurement data | |
| CN116611580A (en) | Ocean red tide prediction method based on multi-source data and deep learning | |
| Doan et al. | Derivation of effective and efficient data set with subtractive clustering method and genetic algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |