CN109325594A - A kind of power communication scene O&M aided diagnosis method improving CBR - Google Patents
A kind of power communication scene O&M aided diagnosis method improving CBR Download PDFInfo
- Publication number
- CN109325594A CN109325594A CN201811147460.1A CN201811147460A CN109325594A CN 109325594 A CN109325594 A CN 109325594A CN 201811147460 A CN201811147460 A CN 201811147460A CN 109325594 A CN109325594 A CN 109325594A
- Authority
- CN
- China
- Prior art keywords
- layer
- output
- neuron
- input
- cbr
- 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
Links
Classifications
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses a kind of power communication scene O&M aided diagnosis method for improving CBR, comprising the following steps: (1) the problem of input encounters during the O&M of power communication scene, and issue retrieval request；(2) client receives the retrieval request instruction, extracts the characteristic condition in the problem；(3) it is retrieved in case library, and from case similar with problem in the step (1) is retrieved in case library, if search result is undesirable, is then continued in next step；(4) using all cases obtained in the step (3) as candidate casebook, then based on BP neural network progress CBR retrieval, and retrieval data are transferred from database；(5) it obtains search result to be assessed, final definitive result.Only related to the parameter of input and neural network itself using the CBR retrieval of BP neural network, these parameters are available by training, are a kind of adaptive retrievings, and the CBR retrieval based on BP neural network has great advantages.
Description
Technical field
The present invention relates to technical field of electric power communication, and in particular to a kind of power communication scene O&M for improving CBR
Aided diagnosis method.
Background technique
In recent years, with the all-round construction of extra-high voltage grid and interconnected power network, the scale of powerline network is obtained fastly
Speed increases, and network topology structure is increasingly complicated, requires to be continuously increased to the reliability and safety of communication network, gives power communication
Daily operation management brings very big pressure.Bottom grade of the O&M of communication field as whole service maintenance system framework,
It is mainly responsible for the live on duty, maintenance and inspection of infrastructure and the network equipment, and the instruction that receive upper layer regulation personnel is real
The work such as the relevant fault restoration in the scene of applying, resource allocation.Therefore, the O&M of communication field is entire communication network maintenance work
Important component, the safety, quality and efficiency of work will be directly related to the effect of electric communication operation and maintenance work.When
Before, in-situ processing relies on personal experience and local information more, and the expertise of these accumulation is difficult to be shared, and cannot achieve homogeneity
The live O&M of change, so that live O&M quality ginseng time is uneven.
Management for power communication scene O&M needs, on the basis of to case-based reasioning (CBR) technical research,
It is necessary to propose a kind of new O&M aided diagnosis method, realizes the self study of site problems diagnosis and from growing up, effectively avoid
The problems such as Similarity matching degree is not high, convergence rate is slow existing for traditional CBR algorithm.
Summary of the invention
The problem to be solved in the present invention is to provide a kind of Similarity matching degree, and high, fast convergence rate the electric power for improving CBR is logical
Believe live O&M aided diagnosis method.
In order to solve the above technical problems, the scheme that the present invention uses is: the power communication scene O&M auxiliary of improvement CBR
Diagnostic method, comprising the following steps:
(1) the problem of input encounters during the O&M of power communication scene, and issue retrieval request；
(2) client receives the retrieval request instruction, extracts the characteristic condition in the problem；
(3) it is retrieved in case library, and from retrieving similar with problem in the step (1) case in case library,
If search result is undesirable, then continue in next step；
(4) using all cases obtained in the step (3) as candidate casebook, then based on BP neural network progress
CBR retrieval, and retrieval data are transferred from database；
(5) it obtains search result to be assessed, final definitive result.
Above-mentioned technical proposal is to propose on the basis of to case-based reasioning (CBR) technical research and utilize BP nerve
The optimization method of network improvement CBR recall precision, CBR is as Knowledge based engineering problem solving a kind of in artificial intelligence and study side
Method, originating from the research done of Schank and Abelson in 1977 and they to derived from the research of script and come, and
The cognitive model of CBR is proposed on the basis of this.It becomes the basis of subsequent CBR technical research and application.One typical CBR is pushed away
Reason solution procedure be generally divided into Case Retrieval (retrieval), case reuse (reuse), solution amendment (revise) and
Case saves (retain) four-stage；BP (back propagation) neural network be 1986 by Rumelhart and
The concept that scientist headed by McClelland proposes is a kind of multilayer feedforward mind according to the training of error backpropagation algorithm
Through network, it is mainly characterized by before signal to transmitting, and error back-propagating, by constantly regulate network weight weight values, so that net
The final output and desired output of network are as close possible to trained to achieve the purpose that；For the management of power communication scene O&M
It needs to set out, on the basis of to case-based reasioning (CBR) technical research, proposes and improve CBR inspection using BP neural network
The optimization method of rope efficiency, the improved CBR searching algorithm of the present invention have biggish advantage in terms of recall ratio and precision ratio,
With very strong practicability；It is only related to the parameter of input and neural network itself using the CBR retrieval of BP neural network, these
Parameter is available by training, is a kind of adaptive retrieving, simultaneously as having very strong non-linear mapping capability
With network structure flexible, compared with traditional CBR retrieval, the CBR retrieval based on BP neural network is had great advantages.
Preferably, in the step (4), the network structure of the BP neural network is by input layer, hidden layer, output layer
Composition；
Wherein, input layer: each neuron is responsible for receiving from extraneous input information, and passes to each nerve of middle layer
Member；
Hidden layer: being internal information process layer, is responsible for information transformation, the last one hidden layer is transmitted to each neuron of output layer
Information, after further treatment after, complete the forward-propagating treatment process that once learns；
Output layer: outwardly output information processing result.
Wherein, hidden layer may be designed as single hidden layer or more hidden layer configurations according to the demand of information change ability, last
A hidden layer is transmitted to the information of each neuron of output layer, after further treatment after, it is processed to complete the forward-propagating that once learns
Journey.
Preferably, in the network structure of the BP neural network, there is n neuron in input layer, hidden layer has p mind
Through member, output layer has q neuron, then its BP algorithm realizes to include the following steps:
S1 defined variable: input vector is set as x=(x1,x2,…,xn), hidden layer input vector is hi=(hi1,
hi2,…,hip), hidden layer output vector is ho=(ho1,ho2,…,hop), output layer input vector is yi=(yi1,
yi2,…,yiq), output layer output vector is yo=(yo1,yo2,…,yoq), desired output vector is d0=(d1,d2,…,
dq), the connection weight of input layer and middle layer is wih, the connection weight of hidden layer and output layer is who, each neuron of hidden layer
Threshold values be bh, the threshold values of each neuron of output layer is bo；The sample k=1,2 ..., m of acquisition, it is f that function is taken out in activation
(), error function are
The initialization of S2 BP neural network: to random number of one section of each connection weight assignment between (- 1,1), if
Error function is e, and given computational accuracy value is ε and maximum study number is M；
S3 randomly selects k-th of input sample and corresponding desired output:
X (k)=(x1(k),x2(k),…,xn(k)) (1)；
S4 calculates outputting and inputting for each neuron of hidden layer:
hoh(k)=f (hih(k)) (3)
yoo(k)=f (yio(k)) (5)
Wherein, h=1,2 ..., p, o=1,2 ..., q；
S5 utilizes network desired output and reality output, calculates error function to the partial derivative of each neuron of output layer:
According to formula (5)-(7), partial derivative δ is obtainedo(k)；
S6 utilizes hidden layer to the connection weight of output layer, the δ of output layero(k) and the output of hidden layer calculates error letter
The partial derivative δ of several pairs of each neurons of hidden layerh(k):
S7 corrects connection weight w using the output of 60 (k) and each neuron of hidden layer of each neuron of output layerho
(k):
S8 utilizes the δ of each neuron of hidden layerh(k) and the Introduced Malaria connection weight of each neuron of input layer:
S9 calculates global error:
S10 judges whether BP network error meets the requirements: when the error of network output reaches default precision or reaches preparatory
After the study number of setting, then terminate neural network learning training；Otherwise, it chooses next learning sample and corresponding expectation is defeated
Out, third step is returned to, into next round learning training.
Preferably, in the step (4), BP algorithm includes two mistakes of backpropagation of the propagated forward and error of signal
It is carried out when journey, i.e. calculating error output by from the direction for being input to output, and adjusts weight and threshold value then from being output to input
Direction carries out.
When propagated forward, input signal acts on output node by hidden layer, by nonlinear transformation, generates output letter
Number, if reality output is not consistent with desired output, it is transferred to the back-propagation process of error.Error-duration model is to lead to output error
Hidden layer is crossed to the layer-by-layer anti-pass of input layer, and gives error distribution to all units of each layer, is made with the error signal obtained from each layer
For the foundation for adjusting each unit weight.It is saved by adjusting the linking intensity and hidden node and output of input node and hidden node
The linking intensity and threshold value of point, decline error along gradient direction, and by repetition learning training, determination is opposite with minimal error
The network parameter (weight and threshold value) answered, training stop stopping.Trained neural network can be to similar sample at this time
Information is inputted, the smallest information by non-linear conversion of output error is voluntarily handled.
Detailed description of the invention
It is further described with reference to the accompanying drawing with embodiments of the present invention:
Fig. 1 is existing CBR course of work flow chart；
Fig. 2 is BP neural network basic structure schematic diagram；
Fig. 3 is that the present invention is based on the CBR of BP neural network to retrieve improved flow chart.
Specific embodiment
As shown in figure 3, the power communication scene O&M aided diagnosis method of improvement CBR of the invention the following steps are included:
(1) the problem of input encounters during the O&M of power communication scene, and issue retrieval request；
(2) client receives the retrieval request instruction, extracts the characteristic condition in the problem；
(3) it is retrieved in case library, and from retrieving similar with problem in the step (1) case in case library,
If search result is undesirable, then continue in next step；
(4) using all cases obtained in the step (3) as candidate casebook, then based on BP neural network progress
CBR retrieval, and retrieval data are transferred from database；Wherein traditional CBR course of work flow chart is as shown in Figure 1；
(5) it obtains search result to be assessed, final definitive result.
In the step (4), the network structure of the BP neural network is by input layer, hidden layer, output layer composition, BP
Neural network basic structure is as shown in Figure 2；
Wherein, input layer: each neuron is responsible for receiving from extraneous input information, and passes to each nerve of middle layer
Member；
Hidden layer: being internal information process layer, is responsible for information transformation, the last one hidden layer is transmitted to each neuron of output layer
Information, after further treatment after, complete the forward-propagating treatment process that once learns；
Output layer: outwardly output information processing result.
In the network structure of the BP neural network, there is n neuron in input layer, hidden layer has p neuron, defeated
Layer has q neuron out, then its BP algorithm realizes to include the following steps:
S1 defined variable: input vector is set as x=(x1,x2,…,xn), hidden layer input vector is hi=(hi1,
hi2,…,hip), hidden layer output vector is ho=(ho1,ho2,…,hop), output layer input vector is yi=(yi1,
yi2,…,yiq), output layer output vector is yo=(yo1,yo2,…,yoq), desired output vector is d0=(d1,d2,…,
dq), the connection weight of input layer and middle layer is wih, the connection weight of hidden layer and output layer is who, each neuron of hidden layer
Threshold values be bh, the threshold values of each neuron of output layer is bo；The sample k=1,2 ..., m of acquisition, it is f that function is taken out in activation
(), error function are
The initialization of S2 BP neural network: to random number of one section of each connection weight assignment between (- 1,1), if
Error function is e, and given computational accuracy value is ε and maximum study number is M；
S3 randomly selects k-th of input sample and corresponding desired output:
X (k)=(x1(k),x2(k),…,xn(k)) (1)；
S4 calculates outputting and inputting for each neuron of hidden layer:
hoh(k)=f (hih(k)) (3)
yoo(k)=f (yio(k)) (5)
Wherein, h=1,2 ..., p, o=1,2 ..., q；
S5 utilizes network desired output and reality output, calculates error function to the partial derivative of each neuron of output layer:
According to formula (5)-(7), partial derivative δ is obtainedo(k)；
S6 utilizes hidden layer to the connection weight of output layer, the δ of output layero(k) and the output of hidden layer calculates error letter
The partial derivative δ of several pairs of each neurons of hidden layerh(k):
S7 corrects connection weight w using the output of 60 (k) and each neuron of hidden layer of each neuron of output layerho
(k):
S8 utilizes the δ of each neuron of hidden layerh(k) and the Introduced Malaria connection weight of each neuron of input layer:
S9 calculates global error:
S10 judges whether BP network error meets the requirements: when the error of network output reaches default precision or reaches preparatory
After the study number of setting, then terminate neural network learning training；Otherwise, it chooses next learning sample and corresponding expectation is defeated
Out, third step is returned to, into next round learning training.
In the step (4), BP algorithm includes two processes of backpropagation of the propagated forward and error of signal, that is, is counted
Carried out when calculating error output by from the direction for being input to output, and adjust weight and threshold value then from be output to the direction of input into
Row.
The power communication scene O&M aided diagnosis method of the improvement CBR of the present embodiment is for power communication scene O&M
Management needs to set out, and on the basis of to case-based reasioning (CBR) technical research, proposes and is improved using BP neural network
The optimization method of CBR recall precision, improved CBR searching algorithm have biggish advantage in terms of recall ratio and precision ratio,
With very strong practicability.
Particular embodiments described above has carried out further in detail the purpose of the present invention, technical scheme and beneficial effects
It describes in detail bright, it should be understood that the above is only a specific embodiment of the present invention, is not intended to restrict the invention；It is all
Within the spirit and principles in the present invention, any modification, equivalent substitution, improvement and etc. done should be included in guarantor of the invention
Within the scope of shield.
Claims (4)
1. a kind of power communication scene O&M aided diagnosis method for improving CBR, which comprises the following steps:
(1) the problem of input encounters during the O&M of power communication scene, and issue retrieval request；
(2) client receives the retrieval request instruction, extracts the characteristic condition in the problem；
(3) it is retrieved in case library, and from retrieval case similar with problem in the step (1) in case library, if
Search result is undesirable, then continues in next step；
(4) using all cases obtained in the step (3) as candidate casebook, then based on BP neural network progress CBR inspection
Rope, and retrieval data are transferred from database；
(5) it obtains search result to be assessed, final definitive result.
2. the power communication scene O&M aided diagnosis method according to claim 1 for improving CBR, which is characterized in that
In the step (4), the network structure of the BP neural network is by input layer, hidden layer, output layer composition；
Wherein, input layer: each neuron is responsible for receiving from extraneous input information, and passes to each neuron of middle layer；
Hidden layer: being internal information process layer, is responsible for information transformation, the last one hidden layer is transmitted to the letter of each neuron of output layer
Breath, after further treatment after, complete the forward-propagating treatment process that once learns；
Output layer: outwardly output information processing result.
3. the power communication scene O&M aided diagnosis method according to claim 2 for improving CBR, which is characterized in that
In the network structure of the BP neural network, there is n neuron in input layer, hidden layer has p neuron, and output layer has q
A neuron, then its BP algorithm realizes to include the following steps:
S1 defined variable: input vector is set as x=(x1,x2,…,xn), hidden layer input vector is hi=(hi1,hi2,…,
hip), hidden layer output vector is ho=(ho1,ho2,…,hop), output layer input vector is yi=(yi1,yi2,…,yiq),
Output layer output vector is yo=(yo1,yo2,…,yoq), desired output vector is d0=(d1,d2,…,dq), input layer is in
The connection weight of interbed is wih, the connection weight of hidden layer and output layer is who, the threshold values of each neuron of hidden layer is bh, output
The threshold values of each neuron of layer is bo；The sample k=1,2 ..., m of acquisition, it is f () that function is taken out in activation, and error function is
S2 BP neural network initialization: to random number of one section of each connection weight assignment between (- 1,1), if error
Function is e, and given computational accuracy value is ε and maximum study number is M；
S3 randomly selects k-th of input sample and corresponding desired output:
X (k)=(x1(k),x2(k),…,xn(k)) (1)；
S4 calculates outputting and inputting for each neuron of hidden layer:
hoh(k)=f (hih(k)) (3)
yoo(k)=f (yio(k)) (5)
Wherein, h=1,2 ..., p, o=1,2 ..., q；
S5 utilizes network desired output and reality output, calculates error function to the partial derivative of each neuron of output layer:
According to formula (5)-(7), partial derivative δ is obtainedo(k)；
S6 utilizes hidden layer to the connection weight of output layer, the δ of output layero(k) and the output of hidden layer calculates error function to hidden
The partial derivative δ of each neuron containing layerh(k):
S7 corrects connection weight w using the output of 60 (k) and each neuron of hidden layer of each neuron of output layerho(k):
S8 utilizes the δ of each neuron of hidden layerh(k) and the Introduced Malaria connection weight of each neuron of input layer:
S9 calculates global error:
S10 judges whether BP network error meets the requirements: presetting when the error of network output reaches default precision or reaches
Study number after, then terminate neural network learning training；Otherwise, next learning sample and corresponding desired output are chosen,
Back to third step, into next round learning training.
4. the power communication scene O&M aided diagnosis method according to claim 3 for improving CBR, which is characterized in that
In the step (4), BP algorithm includes two processes of backpropagation of the propagated forward and error of signal, i.e. calculating error is defeated
It is carried out when out by from the direction for being input to output, and adjusts weight and threshold value and then carried out from the direction for being output to input.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811147460.1A CN109325594A (en) | 2018-09-29 | 2018-09-29 | A kind of power communication scene O&M aided diagnosis method improving CBR |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811147460.1A CN109325594A (en) | 2018-09-29 | 2018-09-29 | A kind of power communication scene O&M aided diagnosis method improving CBR |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109325594A true CN109325594A (en) | 2019-02-12 |
Family
ID=65266402
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811147460.1A Pending CN109325594A (en) | 2018-09-29 | 2018-09-29 | A kind of power communication scene O&M aided diagnosis method improving CBR |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109325594A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110825074A (en) * | 2019-12-04 | 2020-02-21 | 江苏方天电力技术有限公司 | Phase modulator fault diagnosis system and working method thereof |
| CN112365014A (en) * | 2020-11-11 | 2021-02-12 | 重庆邮电大学 | GA-BP-CBR-based industrial equipment fault diagnosis system and method |
-
2018
- 2018-09-29 CN CN201811147460.1A patent/CN109325594A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110825074A (en) * | 2019-12-04 | 2020-02-21 | 江苏方天电力技术有限公司 | Phase modulator fault diagnosis system and working method thereof |
| CN112365014A (en) * | 2020-11-11 | 2021-02-12 | 重庆邮电大学 | GA-BP-CBR-based industrial equipment fault diagnosis system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109948000B (en) | Abnormal target detection method, device, equipment and storage medium for heterogeneous network | |
| CN110738010A (en) | Wind power plant short-term wind speed prediction method integrated with deep learning model | |
| CN111488946B (en) | Radar servo system fault diagnosis method based on information fusion | |
| CN110070228B (en) | BP neural network wind speed prediction method for neuron branch evolution | |
| CN110212551B (en) | Micro-grid reactive power automatic control method based on convolutional neural network | |
| CN114091615B (en) | Electric energy metering data complement method and system based on generation countermeasure network | |
| Cai et al. | An efficient approach for electric load forecasting using distributed ART (adaptive resonance theory) & HS-ARTMAP (Hyper-spherical ARTMAP network) neural network | |
| Duan et al. | Decentralized adaptive NN state-feedback control for large-scale stochastic high-order nonlinear systems | |
| CN107133684A (en) | A kind of random matrix construction method towards GA for reactive power optimization | |
| CN109325594A (en) | A kind of power communication scene O&M aided diagnosis method improving CBR | |
| CN115632406B (en) | Reactive voltage control method and system based on digital-mechanism fusion driving modeling | |
| CN116663419A (en) | Sensorless equipment fault prediction method based on optimized Elman neural network | |
| CN113902207A (en) | Short-term load prediction method based on TCN-LSTM | |
| CN114336632A (en) | Method for correcting alternating current power flow based on model information assisted deep learning | |
| CN113422371A (en) | Distributed power supply local voltage control method based on graph convolution neural network | |
| Prema et al. | LSTM based Deep Learning model for accurate wind speed prediction | |
| CN117057623A (en) | Comprehensive power grid safety optimization scheduling method, device and storage medium | |
| Shayanfar et al. | Wind power prediction model based on hybrid strategy | |
| Hoballah et al. | Transient stability assessment using ANN considering power system topology changes | |
| Fan et al. | Ultra-short-term bus load forecasting method based on multi-source data and hybrid neural network | |
| CN109245182A (en) | A kind of distributed photovoltaic maximum capacity appraisal procedure based on parametric programming | |
| CN115660893A (en) | Transformer substation bus load prediction method based on load characteristics | |
| CN107025497A (en) | A kind of electric load method for early warning and device based on Elman neutral nets | |
| CN112199980A (en) | Overhead line robot obstacle identification method | |
| CN113515890A (en) | Renewable energy day-ahead scene generation method based on federal learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190212 |
|
| RJ01 | Rejection of invention patent application after publication |