CN105184394A - On-line data mining optimized control method based on cyber physical system (CPS) of power distribution network - Google Patents

Abstract

The invention provides an on-line data mining optimized control method based on the cyber physical system (CPS) of a power distribution network. The method comprises the steps of acquiring the data source of the CPS of the power distribution network and cleaning the data source; establishing an association relationship model between the data source and a controlled variable and conducting the data mining; verifying the correctness of the above association relationship with the actual power distribution network as an analysis example; on the condition that the above association relationship is correct, matching the above association relationship with association rules and outputting a matching result; on the condition that the above association relationship is not correct, correcting association rules; and controlling the power distribution network according to the output result. By adopting the above method, on the premise that the safe operation of the power distribution network is ensured, the cost of power enterprises is lowered. Therefore, the method facilitates the pioneering and developing process of the electricity market.

Description

Based on the optimal control method of power distribution network CPS online data mining

Technical field

The present invention relates to a kind of optimal control method, be specifically related to the optimal control method based on power distribution network CPS online data mining.

Background technology

Information physical system (CyberPhysicalSystems, CPS) information physical emerging system is again, it is the fast development along with Information & Communication Technology, one that emerged in large numbers in recent years new research field, proposed in 2006 years by American National fund committee (NSF, NationalScienceFoundation) the earliest.CPS is considered to the third wave being expected to become world information technology after computing machine, internet, and its core is the fusion of calculating, communication, control, as shown in Figure 2.CPS be one on the basis of environment sensing, the degree of depth has merged calculating, communication and control ability controlled, credible, extendible networking physical equipment system, it is by calculation procedure and the interactional feedback circulation of physics process realizes degree of depth fusion and real-time, interactive increases or expands new function, with safety, reliably, a physical entity is monitored or controlled to efficient and real-time mode, it contains ubiquitous environment sensing, embedding assembly, the systems engineering such as network service and network control, physical system is made to have calculating, communication, accurate control, remote collaboration and autonomy function.

Along with the development of intelligent grid, power distribution network forms day by day complicated, intelligent distribution network comprises electrical network one second part (distributed power source, feeder line, transformer, load), a large amount of data acquisition equipment (sensor, FTU, RTU and embedded data acquisition equipment), advanced high-speed communicating network (electric power dedicated fibre optical network, data dispatching communication network, public wired Internet and wireless network), control center and a large amount of computational analysis resources in institutional framework, is typical CPS.The information monitoring equipment that power distribution network CPS relies on real world abundant, and improve reliable communication network, realize the integrated fusion of the quantity of information such as power distribution network traffic control and the internal data involved by it, external data, mutually use.Using power distribution network as CPS, better describe real power distribution network, and more accurately effective control can be carried out to traffic control.

Under the background of intelligent grid, power distribution network CPS provides mass data, and these data have become the basis that electric power enterprise carries out decision-making.But, the accumulation of simple data can not bring benefit to electric power enterprise, only have and use relevant technological means, deep processing is carried out to the data of magnanimity, find implicit information and be used, and then instructing electric power enterprise to make correct decision-making, the effect competence exertion of such mass data is to ultimate attainment.

Human society has been brought into the information age by the fast development of computer technology and the communication technology.In order to improve the utilization factor of data message further, based on the Knowledge Discovery (KnowledgeDiscoveryinDatabase of database, be called for short KDD) and core technology---data mining (DataMining is called for short DM) has just been arisen at the historic moment.KDD finds that there is the whole process by knowledge from data, and data mining is a particular step in KDD process.The definition of KDD has several versions, is generally used now and is defined as follows: KDD refers to the non-trivial process extracting effective, novel, potentially useful and final intelligible pattern from mass data.Say in the narrow sense, data mining is a key step in knowledge excavation process.But the data mining in power distribution network generally represents whole knowledge excavation process, refer to from extracting data and lie in wherein people's the unknown in advance but the information of potentially useful and knowledge, and it is finally expressed as the level process that people can understand pattern.

Based on the magnanimity operation data that power distribution network CPS produces every day.How to utilize existing technology and means, excavate out potential value, and then better control power distribution network from so numerous data, improving the security of power distribution network operation, reliability, is the current problem needing solution badly.

The final goal realizing power distribution network CPS strengthens the control ability to physical system.Current power distribution network adopts relatively simple and fixing control model, the very flexible of control, and owing to being difficult to realize the overall operation efficiency of optimum control in system scope and sacrificial system of having to.

Constantly produce in operational process for power distribution network and accumulate a large amount of service data, existing many scholar's research maintenance data digging technologies therefrom disclose the equipment operation characteristic or rule that are hidden in its behind, find out the method making power distribution network safe and reliable operation more, improve the level of decision-making of operations staff.Data mining technology has five large tasks: prediction (Prediction), classification (Classification), Association Rule Analysis (AssociationRulesAnalysis), cluster analysis (Cluster), Outlier Analysis (OutlierAnalysis).At present, each generic task is applied all in electric system.But all power distribution network is not considered as CPS based on the types of applications of data mining, the real-time of data is not considered yet, data mining results fails to act on further various physical entity in power distribution network, fails the optimum control of the whole synthesis realizing power distribution network.

Summary of the invention

For solving the problem, the present invention proposes the optimal control method based on power distribution network CPS online data mining, comprise: the mass historical data provide power distribution network CPS and real time data are cleaned, excavate its potential value, find out incidence relation between data, the validity of analysis verification correlation rule is carried out by example, and constantly revise correlation rule, the real time data that power distribution network CPS provides is mated with Result, Output rusults is also fed back to power distribution network CPS control center, realizes the optimum control of power distribution network.

The object of the invention is to adopt following technical proposals to realize:

Based on the optimal control method of power distribution network CPS online data mining, comprising:

(1) obtain the data source of power distribution network CPS, data source is cleaned;

(2) set up the Association Rules Model between data source and controlled quentity controlled variable, and carry out data mining;

(3) using actual distribution network as example, whether validate association rule correct, if correctly, goes to step (4), otherwise revises correlation rule;

(4) receive Real-time Electrical Distribution Network Data to mate with correlation rule, output matching result.

Preferably, the acquisition methods obtaining the data source of power distribution network CPS in described step (1) comprises, in power distribution network CPS, embed sensor, for obtaining real time data in power distribution network CPS and historical data.

Preferably, cleaning is carried out to data source and comprises,

Step (1-1) adopts lagranges interpolation, by the Completing Missing Values of data source; y _n+1for missing data, then:

$y_{n+1}=\underset{i=0}{\overset{n}{\Σ}}{x}_i\·\frac{(x_{n+1}-x_0)\·(x_{n+1}-x_1)\·(x_{n+1}-x_2)...(x_{n+1}-x_n)}{(x_i-x_0)\·(x_i-x_1)\·(x_i-x_2)...(x_i-x_{i-1})\·(x_i-x_{i+1})...(x_i-x_n)}---\left(1\right)$

Wherein, x _ifor the data of i-th in ordered series of numbers, the data volume that n comprises for ordered series of numbers; If x ₀, x ₁..., x _n+1and y ₀, y ₁..., y _nfor known quantity.

Step (1-2) adopts statistic mixed-state method to calculate ordered series of numbers x respectively ₀, x ₁..., x _nmean value and standard deviation, obtain the doubtful abnormal data in data source, its expression formula is:

$\stackrel{\‾}{x}=\frac{\underset{i=0}{\overset{n}{\Σ}}{x}_i}{n+1}---\left(2\right)$

$\mathrm{\σ}=\sqrt{\frac{\underset{i=0}{\overset{n}{\Σ}}{\left(x_i-\stackrel{\‾}{x}\right)}^2}{n+1}}---\left(3\right)$

Based on chebyshev's theorem, judge current data x _iwhether be doubtful abnormal data; If or then x _ifor doubtful abnormal data; Wherein, ε is the self-defined factor.

Step (1-3) is to described doubtful abnormal data x _ianalyze, if described x _ito be broken down generation, then x by collecting device and/or acquisition channel _ifor misdata; If produced by line fault, then x _ifor correct data.

Further, whether check analysis result correctly comprises, by data x _ithe data x collected with next data collection cycle _i' contrast, if x _i' (1-λ)≤x _i≤ x _i' (1+ λ), then x _ifor correct data, otherwise x _ifor misdata; Wherein, λ is User Defined constant, and value is 5%.

Further, after having verified, misdata is revised; Described data correcting method is identical with method of Completing Missing Values.

Preferably, the Association Rules Model that described step (2) is set up between data source and controlled quentity controlled variable comprises, and definition data mining object is item set I={i ₁, i ₂..., i _m, it comprises m different pieces of information item; i _kfor a kth data item k=1,2 ..., m; Element number in described I and item set length, length is the item set of k is k dimension data item collection;

Event T is a subset of item set I, and each event is all carried a unique identification tid and is attached thereto, and is denoted as T1; 1 is tid value; Multiple different event forms event database D;

If X is the set of item set I middle term, if then presentation of events T comprises X.

Further, described correlation rule is model implication, have and comprise support, degree of confidence and the degree of correlation.

Preferably, by performing the data mining of described step (2), obtaining support, degree of confidence and the degree of correlation, specifically comprising:

If for item set, B is the event number comprising X in event set D, and A is the sum of all events in event set D, and the support of described item set X is Sup (X), then:

$Sup\left(X\right)=\frac{B}{A}---\left(4\right);$

The degree of confidence Conf (R) of definition correlation rule R, for the degree of reliability of description rule, wherein, and then:

$Conf\left(R\right)=\frac{Sup(X\∪Y)}{Sup\left(X\right)}---\left(5\right)$

The described degree of correlation is for characterizing the degree of correlation between X and Y, and its expression formula is:

$\mathrm{Re}l(X,Y)=\frac{Sup(X\∪Y)}{Sup\left(X\right)Sup\left(Y\right)}---\left(6\right)$

In formula, the ratio that Sup (Y) is the event number with all total number of events in event D that comprise Y in event D.

Preferably, the verification method of described step (3) specifically comprises, if Sup (X ∪ Y)=Sup (X) Sup (Y), represents that X is independent of Y; If the degree of correlation is greater than 1, then X and Y is positive correlation, otherwise if the degree of correlation is less than 1, pattern X and Y is negative correlation, by this redundant rule elimination.

Preferably, described step (4) output matching result comprises:

$\left\{\begin{array}{c}Sup\left(X\right)\≥{\mathrm{Sup}}_{min}\\ Conf\left(R\right)\≥{\mathrm{Conf}}_{min}\\ \mathrm{Re}l(X,Y)\≥{\mathrm{Rel}}_{\min }\end{array}\right.---\left(7\right)$

In formula, Sup _minfor minimum support, Sup _min> 0; If item set meets minimum support, then it is frequent item set; Conf _minfor min confidence, Conf _min> 0; Rel _minfor Minimum relevance weight, Rel _min> 1.

Compared with prior art, the beneficial effect that the present invention reaches is:

The overall situation controls to control to combine with Topical Dispersion by the power distribution network CPS of the method that the present invention proposes, data mining results is made to be transferred to the control center of power distribution network CPS, power distribution network CPS carries out Partial controll by various embedded control system to physical equipment, control center passes through the parameter of on-line tuning control system and directly controls physical equipment coordination optimization whole system where necessary, achieves the closed-loop control of power distribution network.

The mass historical data provide power distribution network CPS and real time data are cleaned, excavate its potential value, find out incidence relation between data, the validity of analysis verification correlation rule is carried out by example, and constantly revise correlation rule, the real time data that power distribution network CPS provides mated with Result, Output rusults is also fed back to power distribution network CPS, realizes the optimum control of power distribution network.

Make full use of historical data and the real time data of power distribution network CPS output, the degree of depth excavates that power distribution network is potential, useful information, accuracy is high, that promptness good, data validity is good Result feeds back to power distribution network CPS control center, realize the optimum control of the whole synthesis of power distribution network, will reduce at electric power enterprise cost, play significant role in power distribution network safe operation etc.

Accompanying drawing explanation

Fig. 1 is the optimal control method process flow diagram based on power distribution network CPS online data mining;

The core schematic diagram of the CPS that Fig. 2 provides for background technology;

Fig. 3 is the closed loop control process schematic diagram based on power distribution network CPS.

Embodiment

Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described in further detail.

As shown in Figure 1, under perfecting information condition for power distribution network, from the feature such as complicacy, paradox, redundancy perfecting information, propose the optimal control method based on power distribution network CPS online data mining, wherein, power distribution network is typical CPS;

Concrete steps comprise:

(1) obtain the data source of power distribution network CPS, data source is cleaned; Cleaning power distribution network Data Source, information status and data layout, is prerequisite of the present invention and basis.

The acquisition methods of step (1) comprises, in power distribution network CPS, embed sensor, for obtaining real time data in power distribution network CPS and historical data.Physical entity in power distribution network CPS all embedded in sensing equipment, to realize information acquisition to distribution system and behavior perception, for the invention provides sufficient data resource.

The data that power distribution network CPS provides are served as theme with the time, are automatically produced, but due to the problem of collecting device and acquisition channel, can produce shortage of data and data exception.Therefore step (1-1) adopts lagranges interpolation, by the Completing Missing Values of data source; As shown in the table, if x ₀, x ₁..., x _n+1and y ₀, y ₁..., y _nfor known quantity, y _n+1for missing data;

x

x 0

x 1

x 2

……

x n

x n+1

y

y 0

y 1

y 2

……

y n

y n+1(the unknown)

The table 1 a certain moment one group records data automatically

According to lagrange polynomial:

$L_n\left(x\right)=\underset{i=0}{\overset{n}{\Σ}}f\left(x_i\right)l_i\left(x\right)---\left(1\right)$

With Lagrange coefficient computing formula:

$l_i\left(x\right)=\frac{(x-x_0)(x-x_1)...(x-x_{i-1})(x-x_i)\·\·(x-x_n)}{(x_i-x_0)(x_i-x_1)...(x_i-x_{i-1})(x_i-x_{i+1})...(x_i-x_n)}---\left(2\right)$

Thus obtain:

$y_{n+1}=\underset{i=0}{\overset{n}{\Σ}}{x}_i\·\frac{(x_{n+1}-x_0)\·(x_{n+1}-x_1)\·(x_{n+1}-x_2)...(x_{n+1}-x_n)}{(x_i-x_0)\·(x_i-x_1)\·(x_i-x_2)...(x_i-x_{i-1})\·(x_i-x_{i+1})...(x_i-x_n)}---\left(1\right)$

Wherein, x _ifor the data of i-th in ordered series of numbers, the data volume that n comprises for ordered series of numbers;

Step (1-2) adopts statistic mixed-state method to calculate ordered series of numbers x respectively ₀, x ₁..., x _nmean value and standard deviation, obtain the doubtful abnormal data in data source, its expression formula is:

$\stackrel{\‾}{x}=\frac{\underset{i=0}{\overset{n}{\Σ}}{x}_i}{n+1}---\left(2\right)$

$\mathrm{\σ}=\sqrt{\frac{\underset{i=0}{\overset{n}{\Σ}}{\left(x_i-\stackrel{\‾}{x}\right)}^2}{n+1}}---\left(3\right)$

Based on chebyshev's theorem: any one data centralization, be positioned at its average m standard deviation scope ratio or part be at least 1-1/ ㎡, wherein m be greater than 1 positive count.There is following result for m=2 and m=3: in all data, have at least the data of 3/4 (or 75%) to be positioned at average 2 standard deviation scopes; In all data, the data of 8/9 (or 88.9%) are had at least to be positioned at average 3 standard deviation scopes; In all data, the data of 24/25 (or 96%) are had at least to be positioned at average 5 standard deviation scopes.Judge current data x _iwhether be doubtful abnormal data; For abnormal data, adopt statistic mixed-state method to carry out preliminary identification, then contrast with the data that next data collection cycle collects, carry out identification again;

If or then x _ifor doubtful abnormal data; Wherein, ε is the self-defined factor, can determine according to actual conditions, and general span is 0 < ε≤5.

Step (1-3) is to described doubtful abnormal data x _ianalyze, if described x _ito be broken down generation, then x by collecting device and/or acquisition channel _ifor misdata; If produced by line fault, then x _ifor correct data.

For avoiding causing x _imisjudge into abnormal data for correct data, whether check analysis result correctly comprises, by data x _ithe data x collected with next data collection cycle _i' contrast, if x _i' (1-λ)≤x _i≤ x _i' (1+ λ), then x _ifor correct data, otherwise x _ifor misdata; Wherein, λ is User Defined constant, and value is 5%.

After having verified, misdata is revised; Described data correcting method is identical with method of Completing Missing Values.

(2) set up the Association Rules Model between data source and controlled quentity controlled variable, and carry out data mining; Association rule mining, as important method a kind of in data mining technology, because finding the relevant valuable data pattern be hidden in mass data, has important practical value to the generation of decision-making.

The Association Rules Model that step (2) is set up between data source and controlled quentity controlled variable comprises, and definition data mining object is item set I={i ₁, i ₂..., i _m, it comprises m different pieces of information item; i _kfor a kth data item k=1,2 ..., m; Element number in described I and item set length, length is the item set of k is k dimension data item collection (k-Itemset);

Event (Transaction) T is a subset of item set I, and each event is all carried a unique identification tid and is attached thereto, and is denoted as T1; 1 is tid value; Multiple different event forms event database D (i.e. transaction database);

If X is the set of item set I middle term, if then presentation of events T comprises X.

Described correlation rule is model implication, have and comprise support, degree of confidence and the degree of correlation.

By performing the data mining of described step (2), obtaining support (Support), degree of confidence (Confidence) and the degree of correlation (Relevancy), specifically comprising:

If for item set, B is the event number comprising X in event set D, and A is the sum of all events in event set D, and the support of described item set X is Sup (X), then:

$Sup\left(X\right)=\frac{B}{A}---\left(4\right);$

The degree of confidence Conf (R) of definition correlation rule R, for the degree of reliability of description rule, wherein, and then:

$Conf\left(R\right)=\frac{Sup(X\&cup;Y)}{Sup\left(X\right)}---\left(5\right)$

The described degree of correlation is for characterizing the degree of correlation between X and Y, and its expression formula is:

$\mathrm{Re}l(X,Y)=\frac{Sup(X\&cup;Y)}{Sup\left(X\right)Sup\left(Y\right)}---\left(6\right)$

In formula, the ratio that Sup (Y) is the event number with all total number of events in event D that comprise Y in event D.

(3) using actual distribution network as example, whether validate association rule correct, if correctly, goes to step (4), otherwise revises correlation rule;

The verification method of described step (3) specifically comprises, if Sup (X ∪ Y)=Sup (X) Sup (Y), represents that X is independent of Y; If the degree of correlation is greater than 1, then X and Y is positive correlation, otherwise, if the degree of correlation is less than 1, pattern X and Y is negative correlation, namely the former piece (condition) of rule and the probability of occurrence of consequent (result) are reciprocal, and this rule-like does not meet objective logic, by this redundant rule elimination.

(4) receive Real-time Electrical Distribution Network Data to mate with correlation rule, output matching result.

Step (4) output matching result comprises:

$\left\{\begin{array}{c}Sup\left(X\right)\&GreaterEqual;{\mathrm{Sup}}_{min}\\ Conf\left(R\right)\&GreaterEqual;{\mathrm{Conf}}_{min}\\ \mathrm{Re}l(X,Y)\&GreaterEqual;{\mathrm{Rel}}_{\min }\end{array}\right.---\left(7\right)$

In formula, Sup _minfor minimum support, Sup _min> 0; If item set meets minimum support, then it is frequent item set;

Conf _minfor min confidence, Conf _min> 0; Rel _minfor Minimum relevance weight, Rel _min> 1.

Control power distribution network according to Output rusults, described control method data mining algorithm, includes but not limited to statistical analysis technique, decision tree method, neural network, rough set method etc.As shown in Figure 3, power distribution network controls control center data mining results being transferred to power distribution network CPS, power distribution network CPS carries out Partial controll by various embedded control system to physical equipment, control center can by the parameter of on-line tuning control system and where necessary directly control physical equipment carry out coordination optimization whole system, achieve the closed-loop control of power distribution network.

Finally should be noted that: above embodiment is only in order to illustrate the technical scheme of the application but not the restriction to its protection domain; although with reference to above-described embodiment to present application has been detailed description; those of ordinary skill in the field are to be understood that: those skilled in the art still can carry out all changes, amendment or equivalent replacement to the embodiment of application after reading the application; these change, amendment or equivalent to replace, and it is all within it applies for the right that awaits the reply.

Claims (10)

1. based on the optimal control method of power distribution network CPS online data mining, it is characterized in that, comprising:

(1) obtain the data source of power distribution network CPS, data source is cleaned;

(2) set up the Association Rules Model between data source and controlled quentity controlled variable, and carry out data mining;

(3) using actual distribution network as example, whether validate association rule correct, if correctly, goes to step (4), otherwise revises correlation rule;

(4) receive Real-time Electrical Distribution Network Data to mate with correlation rule, output matching result.

2. the method for claim 1, is characterized in that, the data source of the acquisition power distribution network CPS of described step (1), and its acquisition methods comprises, in power distribution network CPS, embed sensor, for obtaining real time data in power distribution network CPS and historical data.

3. the method for claim 1, is characterized in that, described step (1) is carried out cleaning to data source and comprised,

Step (1-1) adopts lagranges interpolation, by the Completing Missing Values of data source; y _n+1for missing data, then:

$y_{n+1}=\underset{i=0}{\overset{n}{\&Sigma;}}{x}_i\&CenterDot;\frac{(x_{n+1}-x_0)\&CenterDot;(x_{n+1}-x_1)\&CenterDot;(x_{n+1}-x_2)...(x_{n+1}-x_n)}{(x_i-x_0)\&CenterDot;(x_i-x_1)\&CenterDot;(x_i-x_2)...(x_i-x_{i-1})\&CenterDot;(x_i-x_{i+1})...(x_i-x_n)}---\left(1\right)$

Wherein, x _ifor the data of i-th in ordered series of numbers, the data volume that n comprises for ordered series of numbers; If x ₀, x ₁..., x _n+1and y ₀, y ₁..., y _nfor known quantity;

Step (1-2) adopts statistic mixed-state method to calculate ordered series of numbers x respectively ₀, x ₁..., x _nmean value and standard deviation, obtain the doubtful abnormal data in data source, its expression formula is:

$\stackrel{\&OverBar;}{x}=\frac{\underset{i=0}{\overset{n}{\&Sigma;}}{x}_i}{n+1}---\left(2\right)$

$\mathrm{\&sigma;}=\sqrt{\frac{\underset{i=0}{\overset{n}{\&Sigma;}}{\left(x_i-\stackrel{\&OverBar;}{x}\right)}^2}{n+1}}---\left(3\right)$

Based on chebyshev's theorem, judge current data x _iwhether be doubtful abnormal data; If or then x _ifor doubtful abnormal data; Wherein, ε is the self-defined factor;

Step (1-3) is to described doubtful abnormal data x _ianalyze, if described x _ito be broken down generation, then x by collecting device and/or acquisition channel _ifor misdata; If produced by line fault, then x _ifor correct data.

4. method as claimed in claim 3, it is characterized in that, whether check analysis result correctly comprises, by data x _ithe data x ' collected with next data collection cycle _icontrast, if x ' _i(1-λ)≤x _i≤ x ' _i(1+ λ), then x _ifor correct data, otherwise x _ifor misdata; Wherein, λ is User Defined constant, and value is 5%.

5. method as claimed in claim 4, is characterized in that, after having verified, revise misdata; Described data correcting method is identical with method of Completing Missing Values.

6. the method for claim 1, is characterized in that, the Association Rules Model that described step (2) is set up between data source and controlled quentity controlled variable comprises, and definition data mining object is item set I={i ₁, i ₂..., i _m, it comprises m different pieces of information item; i _kfor a kth data item k=1,2 ..., m; Element number in described I and item set length, length is the item set of k is k dimension data item collection;

Event T is a subset of item set I, and each event is all carried a unique identification tid and is attached thereto, and is denoted as Tl; L is tid value; Multiple different event forms event database D;

If X is the set of item set I middle term, if then presentation of events T comprises X.

7. method as claimed in claim 6, is characterized in that, described Association Rules Model as implication, have and comprise support, degree of confidence and the degree of correlation.

8. the method as described in claim 1 or 7, is characterized in that, by performing the data mining of described step (2), obtaining support, degree of confidence and the degree of correlation, specifically comprising:

If for item set, B is the event number comprising X in event set D, and A is the sum of all events in event set D, and the support of described item set X is Sup (X), then:

$Sup\left(X\right)=\frac{B}{A}---\left(4\right);$

The degree of confidence Conf (R) of definition correlation rule R, for the degree of reliability of description rule, wherein, $X\&Subset;I,Y\&Subset;I,$ And then:

$Conf\left(R\right)=\frac{Sup(X\&cup;Y)}{Sup\left(X\right)}---\left(5\right)$

The described degree of correlation is for characterizing the degree of correlation between X and Y, and its expression formula is:

$\mathrm{Re}l(X,Y)=\frac{Sup(X\&cup;Y)}{Sup\left(X\right)Sup\left(Y\right)}---\left(6\right)$

In formula, the ratio that Sup (Y) is the event number with all total number of events in event D that comprise Y in event D.

9. the method for claim 1, is characterized in that, the verification method of described step (3) specifically comprises, if Sup (X ∪ Y)=Sup (X) Sup (Y), represents that X is independent of Y; If the degree of correlation is greater than 1, then X and Y is positive correlation, otherwise if the degree of correlation is less than 1, pattern X and Y is negative correlation, by this redundant rule elimination.

10. the method for claim 1, is characterized in that, described step (4) output matching result comprises:

$\left\{\begin{array}{c}Sup\left(X\right)\&GreaterEqual;{\mathrm{Sup}}_{min}\\ Conf\left(R\right)\&GreaterEqual;{\mathrm{Conf}}_{min}\\ \mathrm{Re}l(X,Y)\&GreaterEqual;{\mathrm{Rel}}_{\min }\end{array}\right.---\left(7\right)$

In formula, Sup _minfor minimum support, Sup _min> 0; If item set meets minimum support, then it is frequent item set; Conf _minfor min confidence, Conf _min> 0; Rel _minfor Minimum relevance weight, Rel _min> 1.