CN110070118A - A kind of multi-space data fusion method - Google Patents
A kind of multi-space data fusion method Download PDFInfo
- Publication number
- CN110070118A CN110070118A CN201910286285.2A CN201910286285A CN110070118A CN 110070118 A CN110070118 A CN 110070118A CN 201910286285 A CN201910286285 A CN 201910286285A CN 110070118 A CN110070118 A CN 110070118A
- Authority
- CN
- China
- Prior art keywords
- data
- evidence
- formula
- probability
- naive bayes
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to the method for evaluating state of power equipment in electric system, more specifically, it is related to the multi-space data fusion method based on naive Bayesian and D-S evidence theory, first to a variety of controller switching equipment monitoring data extracted in same time, with these data training Naive Bayes Classifier, obtain different regions, basic probability assignment value in different time periods, then multi-space data are merged with D-S evidence fusion theory, obtain distribution system condition evaluation results, improve the ability of distribution network system processing multi-source data, repair based on condition of component, which is carried out, for power distribution network provides reference and foundation.The present invention allows the different location of power equipment monitoring device distribution in a power distribution system in training sample and new samples, and the device data sampling period is different, adapts to the continuous development of power equipment monitoring technology.
Description
Technical field
The present invention relates to the method for evaluating state of power equipment in electric system, more particularly, to based on simple pattra leaves
The multi-space data fusion method of this and D-S evidence theory.
Background technique
Ubiquitous electric power Internet of Things (electric Internet of Things, eIoT) is sufficiently with generation information
Mechanics of communication, realization electric system links interknit and human-computer interaction, have state complete perception, efficient information
The features such as processing.EIoT development key, be by advanced sensing measurement technology, Information and Communication Technology, analysis decision technology and
Automatic control technology combines.The sensor that different regions are mounted in ubiquitous electric power Internet of Things produces the data of magnanimity, from
Feature extraction is carried out in the data of these magnanimity and data fusion is the key that realize " transparent power grid ".
To multi-source information carry out fusion excavation method mainly have comprehensive assessment, machine learning, indetermination theory etc.,
Comprehensive evaluation theory includes analytic hierarchy process (AHP), expert system etc., relies on subjective experience, does not have promotional value, machine learning reason
By including deepness belief network, support vector machines, random forest etc., model is trained by inputting mass data, finally
The complex mapping relation between multi-source data and output is obtained, but often due to the isomerism of data leads to model performance not
Ideal, indetermination theory include Bayesian network and D-S evidence theory, " Yuan Jinsha, He Yajun, Qin Ying one kind are based on
Improve [J] the electrical measurement of basic trust distribution configuration method and instrument .2014,51 (18) of Bayes classifier: 34-38. " is used
Improve Bayes classifier give transformer equipment carry out failure modes carry out brief inference, effectively improve reliability precision and
Accuracy rate of diagnosis." Gao Zhanjun, Li Siyuan, Peng Zhengliang wait based on network tree figure and improve the power distribution network of D-S evidence theory
Fault Locating Method [J] Electric Power Automation Equipment .2018,38 (6): 65-71. " integrates various faults information, with changing
Fault location is realized into D-S evidence theory.However, all kinds of monitoring devices are widely distributed in practical controller switching equipment system, type is numerous
More, sampling period etc. is different, and the data fusion of multi-space monitoring device is still without proposing a preferable solution party
Case.
Summary of the invention
In order to solve widely distributed, all kinds of monitoring data that sample frequency is different in practical distribution system, the present invention
A kind of multi-space data fusion method combined based on naive Bayesian and D-S evidence theory is proposed, is preferably solved
There is the problem of interference data in single monitoring device, improves the precision of distribution system state classification.
The technical scheme is that
A kind of multi-space data fusion method, comprising the following steps:
Step 1: collecting the historical data of n monitoring device acquisition, the training sample as Naive Bayes Classifier；
Step 2: training sample being input in Naive Bayes Classifier and is trained, for discrete data and continuously
Data calculate likelihood probability using different Bayesian formulas, and training obtains final Naive Bayes Classifier；
Step 3: being divided into for each real-time controller switching equipment status assessment period, n monitoring of acquisition is set m period
Monitoring data in the standby m period, are input in Naive Bayes Classifier and carry out classification prediction, with Bayes classifier
It exports basic probability assignment value (Basic Probability Assignment, BPA)；
Step 4: merging the BPA of all monitoring devices in same a period with D-S evidence theory；
Step 5: again merging the BPA in m period with D-S evidence theory, the classification knot finally obtained
Fruit is the corresponding class of BPA after multi-space data fusion, wherein finding out the corresponding class of maximum BPA, such is exactly final assessment
The condition evaluation results of system in period.
Wherein detailed step is as follows:
Step 1: the historical data for collecting the acquisition of n monitoring device is denoted as { S1_data, S2_data ..., Sn_data },
As the training sample of Naive Bayes Classifier；
Bayes decision theory is in a kind of method based on implementation categorised decision under probabilistic framework, according to Bayes' theorem
Available following posterior probability calculation formula:
In formula: P (c) is class prior probability；P (x | c) it is that value is the general of x in the case where belonging to c classification in sample
Rate, referred to as likelihood probability；When x attribute is multi-source, it is difficult to the joint probability of P (x | c) is calculated, in order to avoid this obstacle, Piao
Plain Bayes classifier, which is adopted, assumes that all properties are mutually indepedent, independently classification results are had an impact between each attribute and
It is interrelated, it may be assumed that
In formula: d indicates the number of attribute, xiFor value of the x under ith attribute.
Step 2: training sample being input in Naive Bayes Classifier and is trained, wherein class prior probability calculates
Formula is as follows:
In formula: NtcIndicate the set for belonging to type c in training sample, NtIndicate all training sets in training sample；
For different types of data set, Naive Bayes Classifier has different calculation methods；
For discrete data, likelihood probability calculation formula is as follows:
In formula: Dtc,xiIt indicates in training set DtcMiddle ith attribute value is xiThe sample set constituted；
For connection attribute, it is assumed that its probability density approximation meets normal distribution, and replaces likelihood letter using probability density
Number is calculated:
In formula: μc,iAnd σc,iIt is the mean value and variance of c class sample value on i-th of sample respectively；
For different data types, final Naive Bayes Classifier is obtained according to formula (4) and (5) training respectively.
Step 3: for each real-time controller switching equipment status assessment period, the prison of the m period of n monitoring device of acquisition
Measured data, i.e., by T1Period is to TmThe data that n equipment acquires in period are (if T1There are multiple values in period, then takes T1
Average value in period) it is input in Naive Bayes Classifier and carries out classification prediction；Assuming that condition evaluation results have D kind
Situation, for algorithm of seeking of BPA, including fuzzy membership calculating, grey relational grade, neural network algorithm etc., before two kinds
Algorithm is easy to be influenced by subjective experience, and neural network algorithm is substantially general due to may cause in the case where lacking mass data
Rate allocation result is unstable, unreliable.
BPA is effectively sought in order to more objective, it is of good performance since Naive Bayes Classifier has structure simple
Characteristic, therefore the present invention exports BPA using Naive Bayes Classifier, calculation method is as follows:
M (c in formulaj) indicate that in known attribute be x1,x2,…,xnIn the case where be classified as cjReliability, with formula
(6) available n × m × D basic probability assignment value.
Step 4:D-S evidence theory merges the belief function of different information, and D-S evidence theory hypothesis has one
Non-empty identification framework set Θ, Θ={ A of mutual exclusion1,A2,...,ANIt is the exclusive subsets containing N number of element, define m:2Θ→
[0,1], if met:
In formula:Indicate impossible event, m is the basic probability assignment value on identification framework Θ, and this method sets distribution
Standby state is divided into three classes, and is normal, early warning, failure respectively, then corresponding to the identification framework of evidence theory then by three subsets
It constitutes；
The belief function on identification framework Θ are as follows:
Above formula indicates that the belief function of certain event is all the sum of subset probability of the event；
The likelihood function on identification framework Θ are as follows:
Above formula indicates the likelihood function of certain event and the event intersection is not empty the sum of probability；
After obtaining basic reliability, needs to merge different evident information sources, be carried out by taking two evidence fusions as an example
The case where illustrating, then being generalized to n evidence fusion；
D-S evidence fusion rule are as follows: assuming that two evidence M at same identification framework Θ1、M2, then its belief function divides
It Wei not m1、m2, it is as follows to define conflict coefficient K:
A in formula1∩L∩An≠ φ indicates the part to conflict in information, then n data carry out evidence fusion result are as follows:
A in formula1∩A2∩L∩An=A indicates consistent part in information, ⊕ indicate it is orthogonal and, 1-K is that reject conflict dry
The normalization factor disturbed；
The case where merging for n evidence only need to carry out merging finally obtaining commenting for n kind data fusion two-by-two
Valence as a result, the BPA of n monitoring device in same a period is merged with D-S evidence fusion rule again, obtain m ×
D BPA, this step have merged the data of different type monitoring device.
Step 5: the BPA in m different time sections being merged with D-S evidence fusion rule again, is obtained final
D BPA of the system within assessment cycle, each BPA expression take maximum BPA corresponding the degree of belief of every kind of classification results
State of the classification results as system in the final assessment cycle.
The present invention combines final realization distribution system multi-space data with naive Bayesian theory and D-S evidence theory
State evaluation and classification under fusion.
Many kinds of in view of all kinds of monitoring devices are widely distributed in practical controller switching equipment system, sampling period etc. is each not
It is identical, in order to solve the problem of data fusion of multi-space monitoring device, the present invention propose it is a kind of based on Naive Bayes Classification and
D-S evidence manages the multi-space data fusion method combined, monitors number to a variety of controller switching equipments extracted in same time first
According to, with these data training Naive Bayes Classifier, obtain different regions, basic probability assignment value in different time periods, so
Multi-space data are merged with D-S evidence fusion theory afterwards, distribution system condition evaluation results is obtained, improves and match
Network system handles the ability of multi-source data, carries out repair based on condition of component for power distribution network and provides reference and foundation.The present invention allows to instruct
Practice the different location of power equipment monitoring device distribution in a power distribution system in sample and new samples, device data sampling week
Phase is different, adapts to the continuous development of power equipment monitoring technology.
The invention has the advantages that solving the status assessment of multiple widely distributed monitoring devices in a power distribution system, energy
The data of multiple monitoring devices are enough integrated, the multi-space data of distribution system in assessment cycle are merged, are avoided with single prison
Measurement equipment carries out the interference item data being likely to occur when status assessment and misclassification occurs, simultaneously because Naive Bayes Classification
Device principle is simple, calculates efficiently, when being classified with Naive Bayes Classifier with stable classification effectiveness, for
Interference, missing data are insensitive, with the continuous application development of controller switching equipment monitoring technology, attribute value missing in data sample,
The problem of interference meeting self-assembling formation, the sample frequency of variety classes monitoring device also must be different, side proposed by the present invention
Method can improve the utilization efficiency to these data to a certain extent.
Detailed description of the invention
Fig. 1 is flow chart of the invention.
Fig. 2 is the relation schematic diagram of likelihood function and belief function.
Specific embodiment
The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent；In order to better illustrate this embodiment,
The certain components of attached drawing have omission, zoom in or out, and do not represent the size of actual product；Those skilled in the art are come
It says, the omitting of some known structures and their instructions in the attached drawings are understandable.Positional relationship is described in attached drawing to be only used for showing
Example property explanation, should not be understood as the limitation to this patent.
Embodiment 1:
As shown in Figure 1, the multi-space data fusion side proposed by the present invention based on naive Bayesian and D-S evidence theory
Method step is poly- are as follows:
Step 1: collecting the historical data of n monitoring device acquisition, the training sample as Naive Bayes Classifier；
Step 2: training sample being input in Naive Bayes Classifier and is trained, for discrete data and continuously
Data calculate likelihood probability using different Bayesian formulas, and training obtains final Naive Bayes Classifier；
Step 3: being divided into for each real-time controller switching equipment status assessment period, n monitoring of acquisition is set m period
Monitoring data in the standby m period, are input in Naive Bayes Classifier and carry out classification prediction, with naive Bayesian point
Class device exports basic probability assignment value (Basic Probability Assignment, BPA)；
Step 4: merging the BPA of all monitoring devices in same a period with D-S evidence theory；
Step 5: again merging the BPA in m period with D-S evidence theory, the classification knot finally obtained
Fruit is the corresponding class of BPA after multi-space data fusion, wherein finding out the corresponding class of maximum BPA, such is exactly final assessment
The condition evaluation results of system in period.
Wherein detailed step is as follows:
Step 1: the historical data for collecting the acquisition of n monitoring device is denoted as { S1_data, S2_data ..., Sn_data },
As the training sample of Naive Bayes Classifier；
Bayes decision theory is in a kind of method based on implementation categorised decision under probabilistic framework, according to Bayes' theorem
Available following posterior probability calculation formula:
In formula: P (c) is class prior probability；P (x | c) it is that value is the general of x in the case where belonging to c classification in sample
Rate, referred to as likelihood probability；When x attribute is multi-source, it is difficult to the joint probability of P (x | c) is calculated, in order to avoid this obstacle, Piao
Plain Bayes classifier, which is adopted, assumes that all properties are mutually indepedent, independently classification results are had an impact between each attribute and
It is interrelated, it may be assumed that
In formula: d indicates the number of attribute, xiFor value of the x under ith attribute.
Step 2: training sample being input in Naive Bayes Classifier and is trained, wherein class prior probability calculates
Formula is as follows:
In formula: NtcIndicate the set for belonging to type c in training sample, NtIndicate all training sets in training sample；
For different types of data set, Naive Bayes Classifier has different calculation methods；
For discrete data, likelihood probability calculation formula is as follows:
In formula: Dtc,xiIt indicates in training set DtcMiddle ith attribute value is xiThe sample set constituted；
For connection attribute, it is assumed that its probability density approximation meets normal distribution, and replaces likelihood letter using probability density
Number is calculated:
In formula: μc,iAnd σc,iIt is the mean value and variance of c class sample value on i-th of sample respectively；
For different data types, final Naive Bayes Classifier is obtained according to formula (4) and (5) training respectively.
Step 3: for each real-time controller switching equipment status assessment period, the prison of the m period of n monitoring device of acquisition
Measured data, i.e., by T1Period is to TmThe data that n equipment acquires in period are (if T1There are multiple values in period, then takes T1When
Between average value in section) be input in Naive Bayes Classifier and carry out classification prediction；Assuming that condition evaluation results have D kind feelings
Condition, for algorithm of seeking of BPA, including fuzzy membership calculating, grey relational grade, neural network algorithm etc., before two kinds of calculations
Method is easy to be influenced by subjective experience, and neural network algorithm is due to may cause elementary probability in the case where lacking mass data
Allocation result is unstable, unreliable.
BPA is effectively sought in order to more objective, it is of good performance since Naive Bayes Classifier has structure simple
Characteristic, therefore the present invention exports BPA using Naive Bayes Classifier, calculation method is as follows:
M (c in formulaj) indicate that in known attribute be x1,x2,…,xnIn the case where be classified as cjReliability, with formula
(6) available n × m × D basic probability assignment value.
Step 4:D-S evidence theory merges the belief function of different information, and D-S evidence theory hypothesis has one
Non-empty identification framework set Θ, Θ={ A of mutual exclusion1,A2,...,ANIt is the exclusive subsets containing N number of element, define m:2Θ→
[0,1], if met:
In formula:Indicate impossible event, m is the basic probability assignment value on identification framework Θ, and this method sets distribution
Standby state is divided into three classes, and is normal, early warning, failure respectively, then corresponding to the identification framework of evidence theory then by three subsets
It constitutes；
The belief function on identification framework Θ are as follows:
Above formula indicates that the belief function of certain event is all the sum of subset probability of the event；
The likelihood function on identification framework Θ are as follows:
Above formula indicates the likelihood function of certain event and the event intersection is not empty the sum of probability；
The relationship of above-mentioned likelihood function and belief function can indicate as shown in Figure 2.
After obtaining basic reliability, needs to merge different evident information sources, be carried out by taking two evidence fusions as an example
The case where illustrating, then being generalized to n evidence fusion；
D-S evidence fusion rule are as follows: assuming that two evidence M at same identification framework Θ1、M2, then its belief function divides
It Wei not m1、m2, it is as follows to define conflict coefficient K:
A in formula1∩L∩An≠ φ indicates the part to conflict in information, then n data carry out evidence fusion result are as follows:
A in formula1∩A2∩L∩An=A indicates consistent part in information, ⊕ indicate it is orthogonal and, 1-K is that reject conflict dry
The normalization factor disturbed；
The case where merging for n evidence only need to carry out merging finally obtaining commenting for n kind data fusion two-by-two
Valence as a result, the BPA of n monitoring device in same a period is merged with D-S evidence fusion rule again, obtain m ×
D BPA, this step have merged the data of different type monitoring device.
Step 5: the BPA in m different time sections being merged with D-S evidence fusion rule again, is obtained final
D BPA of the system within assessment cycle, each BPA expression take maximum BPA corresponding the degree of belief of every kind of classification results
State of the classification results as system in the final assessment cycle.
Embodiment 2:
The following examples further illustrate concrete application method of the invention.
In practical distribution system, all kinds of monitoring device infields are different, also there is sampling number in assessment cycle
Institute is different, in order to verify the historical data for needing to collect a certain number of monitoring devices set forth herein algorithm model to algorithm mould
Type is trained.It is mutually indepedent before assuming each monitoring device herein, meet the basic assumption of naive Bayesian.
Monitoring data (the monitoring data packet of 350 known state types is had collected from practical intelligent distribution system herein
Voltage change ratio is included, electric field strength change rate, cable temperature, three kinds of monitoring devices are positioned at the difference in status assessing system
Position), wherein monitoring data (50 failures, 100 alarms, 200 normal monitoring data comprising three kinds of Status Types
Collection), three kinds of Status Types can be marked with different colours.
In experiment, the identification framework of D-S evidence theory is Θ={ c1,c2,c3, c1Indicate that evaluation status is normal, c2Table
Show evaluation status for alarm, c3Expression evaluation status is failure.Continuous 3 of 3 monitoring devices within assessment cycle are chosen below
Monitoring data in a period, as shown in table 1.According to fault detection address, its correct assessment result should be failure.Then it transports
With the data fusion proposed in this paper based on naive Bayesian and D-S evidence theory, solving result is as shown in table 2 and table 3.
13, table different monitoring device measurement data
BPA calculated result of the table 2 based on naive Bayesian
D-S evidence fusion result in 3 period of table
After D-S evidence fusion in the period, then D-S evidence fusion is carried out with formula (11) and is distributed in different location
Monitoring device, obtain final status monitoring result three reliability be respectively { 0,0,0.99999 }, it is found that for
Containing distracter in period, can preferably cover these interference item datas with the algorithm of this paper, finally obtain reliability compared with
High classification results, namely obtain accurate distribution system assessment result.This is because Naive Bayes Classifier principle letter
It is single, it calculates efficiently, when being classified with Naive Bayes Classifier with stable classification effectiveness, for interfering, lacking
Data are insensitive.
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
Made any modifications, equivalent replacements, and improvements etc., should be included in the claims in the present invention within the spirit and principle of invention
Protection scope within.
Claims (6)
1. a kind of multi-space data fusion method, which comprises the following steps:
Step 1: collecting the historical data of n monitoring device acquisition, the training sample as Naive Bayes Classifier；
Step 2: training sample being input in Naive Bayes Classifier and is trained, discrete data and continuous data are adopted
Likelihood probability is calculated with different Bayesian formulas, training obtains final Naive Bayes Classifier；
Step 3: being divided into for each real-time controller switching equipment status assessment period, n monitoring device m of acquisition is a m period
Monitoring data in period are input in Naive Bayes Classifier and carry out classification prediction, defeated with Naive Bayes Classifier
Basic probability assignment value out；
Step 4: melting the basic probability assignment value of all monitoring devices in same a period with D-S evidence theory
It closes；
Step 5: the basic probability assignment value in m period being merged with D-S evidence theory again, is finally obtained
Classification results are the corresponding class of basic probability assignment value after multi-space data fusion, wherein finding out maximum basic probability assignment
It is worth corresponding class, such is exactly the condition evaluation results of system in final assessment cycle.
2. a kind of multi-space data fusion method according to claim 1, which is characterized in that collect n monitoring in step 1
The historical data of equipment acquisition is denoted as { S1_data, S2_data ..., Sn_data }, as Naive Bayes Classifier
Training sample；
Bayes decision theory is can be obtained based on a kind of method for implementing categorised decision under probabilistic framework according to Bayes' theorem
To following posterior probability calculation formula:
In formula: P (c) is class prior probability；P (x | c) it is the probability that value is x in the case where belonging to c classification in sample, referred to as
Likelihood probability；When x attribute is multi-source, it is difficult to the joint probability of P (x | c) is calculated, in order to avoid this obstacle, naive Bayesian
Classifier, which is adopted, assumes that all properties are mutually indepedent, independently has an impact classification results without mutually closing between each attribute
Connection, it may be assumed that
In formula: d indicates the number of attribute, xiFor value of the x under ith attribute.
3. a kind of multi-space data fusion method according to claim 2, which is characterized in that by training sample in step 2
Being input to the process being trained in Naive Bayes Classifier is；
Class prior probability calculation formula is as follows:
In formula: NtcIndicate the set for belonging to type c in training sample, NtIndicate all training sets in training sample；
For different types of data set, Naive Bayes Classifier has different calculation methods；
For discrete data, likelihood probability calculation formula is as follows:
In formula: Dtc,xiIt indicates in training set DtcMiddle ith attribute value is xiThe sample set constituted；
For connection attribute, it is assumed that its probability density approximation meets normal distribution, and using probability density replace likelihood function into
Row calculates:
In formula: μc,iAnd σc,iIt is the mean value and variance of c class sample value on i-th of sample respectively；
For different data types, final Naive Bayes Classification is obtained according to above-mentioned different calculation method training respectively
Device.
4. a kind of multi-space data fusion method according to claim 3, which is characterized in that for each reality in step 3
When the controller switching equipment status assessment period, the monitoring data of the m period of n monitoring device of acquisition, i.e., by T1Period is to TmTime
The data of n equipment acquisition, which are input in Naive Bayes Classifier, in section carries out classification prediction；
Basic probability assignment value is exported using Naive Bayes Classifier, calculation method is as follows:
M (c in formulaj) indicate that in known attribute be x1,x2,…,xnIn the case where be classified as cjReliability, can be with above formula
Obtain n × m × D basic probability assignment value.
5. a kind of multi-space data fusion method according to claim 4, which is characterized in that the D-S evidence reason in step 4
It being merged by the belief function to different information, D-S evidence theory assumes the non-empty identification framework set Θ for having a mutual exclusion,
Θ={ A1,A2,...,ANIt is the exclusive subsets containing N number of element, define m:2Θ→ [0,1], if met:
In formula:Indicate impossible event, m is the basic probability assignment value on identification framework Θ, and this method is by the shape of controller switching equipment
State is divided into three classes, and is normal, early warning, failure respectively, is then then made of three subsets corresponding to the identification framework of evidence theory；
The belief function on identification framework Θ are as follows:
Above formula indicates that the belief function of certain event is all the sum of subset probability of the event；
The likelihood function on identification framework Θ are as follows:
Above formula indicates the likelihood function of certain event and the event intersection is not empty the sum of probability；
After obtaining basic reliability, needs to merge different evident information sources, be illustrated by taking two evidence fusions as an example,
The case where being generalized to n evidence fusion again；
D-S evidence fusion rule are as follows: assuming that two evidence M at same identification framework Θ1、M2, then its belief function be respectively
m1、m2, it is as follows to define conflict coefficient K:
A in formula1∩L∩An≠ φ indicates the part to conflict in information, then n data carry out evidence fusion result are as follows:
A in formula1∩A2∩L∩An=A indicates consistent part in information,Indicate it is orthogonal and, 1-K be reject conflict interfere
Normalization factor；
The case where merging for n evidence need to only carry out merging the evaluation knot for finally obtaining n kind data fusion two-by-two
Fruit, then merge the basic probability assignment value of n monitoring device in same a period with D-S evidence fusion rule,
M × D basic probability assignment value is obtained, this step has merged the data of different type monitoring device.
6. a kind of multi-space data fusion method according to claim 5, which is characterized in that use D-S in step 5 again
Evidence fusion rule merges the basic probability assignment value in m different time sections, obtains final system in assessment cycle
D interior basic probability assignment value, each basic probability assignment value indicate the degree of belief for every kind of classification results, take maximum base
This probability assignments is worth state of the corresponding classification results as system in the final assessment cycle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910286285.2A CN110070118A (en) | 2019-04-10 | 2019-04-10 | A kind of multi-space data fusion method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910286285.2A CN110070118A (en) | 2019-04-10 | 2019-04-10 | A kind of multi-space data fusion method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110070118A true CN110070118A (en) | 2019-07-30 |
Family
ID=67367455
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910286285.2A Pending CN110070118A (en) | 2019-04-10 | 2019-04-10 | A kind of multi-space data fusion method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110070118A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110555477A (en) * | 2019-08-30 | 2019-12-10 | 青岛海信网络科技股份有限公司 | municipal facility fault prediction method and device |
| CN110955811A (en) * | 2019-09-12 | 2020-04-03 | 国网浙江省电力有限公司杭州供电公司 | Power data classification method and system based on naive Bayes algorithm |
| CN111582364A (en) * | 2020-05-06 | 2020-08-25 | 智慧航海（青岛）科技有限公司 | Step-by-step fusion system and method based on ship-side and shore-based perception information |
| CN111860613A (en) * | 2020-06-30 | 2020-10-30 | 江苏科技大学 | Multi-target tracking and state prediction method based on multi-source heterogeneous signals |
| CN112232553A (en) * | 2020-09-27 | 2021-01-15 | 同济大学 | Bayesian network-based high-speed rail train late influence factor diagnosis method |
| CN112348399A (en) * | 2020-11-24 | 2021-02-09 | 华东理工大学 | Boiler drum water level control system performance evaluation method based on naive Bayes |
| CN112395558A (en) * | 2020-11-27 | 2021-02-23 | 广东电网有限责任公司肇庆供电局 | Improved unbalanced data hybrid sampling method suitable for historical fault data of intelligent electric meter |
| CN112507290A (en) * | 2020-12-07 | 2021-03-16 | 国电南瑞科技股份有限公司 | Distribution equipment fault probability prejudging method and device and storage medium |
| CN113378654A (en) * | 2021-05-23 | 2021-09-10 | 南京理工大学 | Lane-level positioning method and system based on integration of positioning and vision technology |
| CN113609360A (en) * | 2021-08-19 | 2021-11-05 | 武汉东湖大数据交易中心股份有限公司 | Scene-based multi-source data fusion analysis method and system |
| CN113627544A (en) * | 2021-08-16 | 2021-11-09 | 沈阳工业大学 | Machine tool milling cutter state identification method based on multi-source heterogeneous data fusion |
| CN116388401A (en) * | 2023-05-31 | 2023-07-04 | 国网吉林省电力有限公司辽源供电公司 | Substation equipment control method and system |
| CN116738333A (en) * | 2023-06-09 | 2023-09-12 | 北京航空航天大学 | Electrical signal multi-classification and prediction method for naive Bayes of small sample of aircraft |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102802182A (en) * | 2012-08-09 | 2012-11-28 | 无锡赛睿科技有限公司 | Fault diagnosis device and method for wireless sensor network |
| CN108226711A (en) * | 2018-01-05 | 2018-06-29 | 广东电网有限责任公司电力科学研究院 | A kind of electric network failure diagnosis method and device based on information fusion |
| CN108763793A (en) * | 2018-06-01 | 2018-11-06 | 电子科技大学 | A kind of Weighted Fuzzy type D-S evidence theory frame |
| CN108846511A (en) * | 2018-06-04 | 2018-11-20 | 国家电网公司 | A kind of defect of transformer equipment trend analysis based on regulation big data platform |
-
2019
- 2019-04-10 CN CN201910286285.2A patent/CN110070118A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102802182A (en) * | 2012-08-09 | 2012-11-28 | 无锡赛睿科技有限公司 | Fault diagnosis device and method for wireless sensor network |
| CN108226711A (en) * | 2018-01-05 | 2018-06-29 | 广东电网有限责任公司电力科学研究院 | A kind of electric network failure diagnosis method and device based on information fusion |
| CN108763793A (en) * | 2018-06-01 | 2018-11-06 | 电子科技大学 | A kind of Weighted Fuzzy type D-S evidence theory frame |
| CN108846511A (en) * | 2018-06-04 | 2018-11-20 | 国家电网公司 | A kind of defect of transformer equipment trend analysis based on regulation big data platform |
Non-Patent Citations (4)
| Title |
|---|
| 李茹 等: "D-S证据理论在时-空信息融合中的应用", 《计算机工程与应用》 * |
| 苏敏邦: "《概率论与数理统计学习辅导与提高》", 31 August 2012 * |
| 苑津莎: "一种基于改进贝叶斯分类器的基本信任分配构造方法", 《电测与仪表》 * |
| 黄勤龙: "《云计算数据安全》", 30 March 2018 * |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110555477A (en) * | 2019-08-30 | 2019-12-10 | 青岛海信网络科技股份有限公司 | municipal facility fault prediction method and device |
| CN110955811A (en) * | 2019-09-12 | 2020-04-03 | 国网浙江省电力有限公司杭州供电公司 | Power data classification method and system based on naive Bayes algorithm |
| CN110955811B (en) * | 2019-09-12 | 2023-09-22 | 国网浙江省电力有限公司杭州供电公司 | Power data classification method and system based on naive Bayes algorithm |
| CN111582364A (en) * | 2020-05-06 | 2020-08-25 | 智慧航海（青岛）科技有限公司 | Step-by-step fusion system and method based on ship-side and shore-based perception information |
| CN111860613A (en) * | 2020-06-30 | 2020-10-30 | 江苏科技大学 | Multi-target tracking and state prediction method based on multi-source heterogeneous signals |
| CN111860613B (en) * | 2020-06-30 | 2023-10-27 | 江苏科技大学 | Multi-target tracking and state predicting method based on multi-source heterogeneous signals |
| CN112232553A (en) * | 2020-09-27 | 2021-01-15 | 同济大学 | Bayesian network-based high-speed rail train late influence factor diagnosis method |
| CN112232553B (en) * | 2020-09-27 | 2023-04-07 | 同济大学 | Bayesian network-based high-speed rail train late influence factor diagnosis method |
| CN112348399A (en) * | 2020-11-24 | 2021-02-09 | 华东理工大学 | Boiler drum water level control system performance evaluation method based on naive Bayes |
| CN112395558A (en) * | 2020-11-27 | 2021-02-23 | 广东电网有限责任公司肇庆供电局 | Improved unbalanced data hybrid sampling method suitable for historical fault data of intelligent electric meter |
| CN112507290B (en) * | 2020-12-07 | 2024-04-30 | 国电南瑞科技股份有限公司 | Power distribution equipment fault probability pre-judging method, device and storage medium |
| CN112507290A (en) * | 2020-12-07 | 2021-03-16 | 国电南瑞科技股份有限公司 | Distribution equipment fault probability prejudging method and device and storage medium |
| CN113378654B (en) * | 2021-05-23 | 2022-08-16 | 南京理工大学 | Lane-level positioning method and system based on integration of positioning and vision technology |
| CN113378654A (en) * | 2021-05-23 | 2021-09-10 | 南京理工大学 | Lane-level positioning method and system based on integration of positioning and vision technology |
| CN113627544A (en) * | 2021-08-16 | 2021-11-09 | 沈阳工业大学 | Machine tool milling cutter state identification method based on multi-source heterogeneous data fusion |
| CN113627544B (en) * | 2021-08-16 | 2024-03-08 | 沈阳工业大学 | Machine tool milling cutter state identification method based on multi-source heterogeneous data fusion |
| CN113609360A (en) * | 2021-08-19 | 2021-11-05 | 武汉东湖大数据交易中心股份有限公司 | Scene-based multi-source data fusion analysis method and system |
| CN116388401B (en) * | 2023-05-31 | 2023-08-25 | 国网吉林省电力有限公司辽源供电公司 | Substation equipment control method and system |
| CN116388401A (en) * | 2023-05-31 | 2023-07-04 | 国网吉林省电力有限公司辽源供电公司 | Substation equipment control method and system |
| CN116738333A (en) * | 2023-06-09 | 2023-09-12 | 北京航空航天大学 | Electrical signal multi-classification and prediction method for naive Bayes of small sample of aircraft |
| CN116738333B (en) * | 2023-06-09 | 2023-12-05 | 北京航空航天大学 | Electrical signal multi-classification and prediction method for naive Bayes of small sample of aircraft |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110070118A (en) | A kind of multi-space data fusion method | |
| Belagoune et al. | Deep learning through LSTM classification and regression for transmission line fault detection, diagnosis and location in large-scale multi-machine power systems | |
| Cai et al. | Bayesian networks in fault diagnosis | |
| CN105515184B (en) | Multisensor many reference amounts distribution synergic monitoring system based on wireless sensor network | |
| Juarez et al. | Characterization of multimachine system behavior using a hierarchical trajectory cluster analysis | |
| CN109768877A (en) | A kind of electric network failure diagnosis method based on space optimum code collection and DHNN error correction | |
| CN114689965A (en) | Power equipment online monitoring method and system based on embedded intelligent sensor | |
| CN116050888A (en) | Method applied to intelligent high-voltage switch cabinet sensor health state assessment | |
| Jaiswal et al. | Distributed fog computing architecture for real-time anomaly detection in smart meter data | |
| Kumar et al. | Cloud-based electricity consumption analysis using neural network | |
| Liu et al. | A Dual-spine Approach to Load Error Repair in a HEMS Sensor Network | |
| CN113922412A (en) | Panorama evaluation method and system for new energy multi-station short circuit ratio, storage medium and computing device | |
| Wang | Research on real-time reliability evaluation of CPS system based on machine learning | |
| Harish et al. | Fault detection and classification for wide area backup protection of power transmission lines using weighted extreme learning machine | |
| CN105353270B (en) | Consider the grid-connected fault-tolerant localization method of power quality disturbance of distributed generation resource | |
| CN104764979A (en) | Virtual information fusion power grid alarming method based on probabilistic reasoning | |
| CN113220751A (en) | Metering system and evaluation method for multi-source data state quantity | |
| Shu et al. | Link Prediction Model for Opportunistic Networks Based on Feature Fusion | |
| Liu | Intelligent detection and diagnosis of power failure relying on BP neural network algorithm | |
| Duan et al. | TCAD: unsupervised anomaly detection based on global local representation differences | |
| Strachan et al. | Intelligent diagnosis of defects responsible for partial discharge activity detected in power transformers | |
| Wang et al. | Real-time fault diagnosis method for low-voltage power line carrier communication network based on network topology | |
| Yuan | Power Failure Prediction Method Based on Data Analysis | |
| Pan et al. | A Model for Detecting False Data Injection Attacks in Smart Grids Based on the Method Utilized for Image Coding | |
| Chen et al. | Artificial Intelligence/Machine Learning Technology in Power System Applications |
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: 20190730 |
|
| RJ01 | Rejection of invention patent application after publication |