CN109101465B - A kind of electric switch event detecting method and system - Google Patents
A kind of electric switch event detecting method and system Download PDFInfo
- Publication number
- CN109101465B CN109101465B CN201810768660.2A CN201810768660A CN109101465B CN 109101465 B CN109101465 B CN 109101465B CN 201810768660 A CN201810768660 A CN 201810768660A CN 109101465 B CN109101465 B CN 109101465B
- Authority
- CN
- China
- Prior art keywords
- time window
- data
- power sequence
- general power
- value
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R21/00—Arrangements for measuring electric power or power factor
- G01R21/001—Measuring real or reactive component; Measuring apparent energy
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Algebra (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Power Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses a kind of electric switch event detecting method and system.This method comprises: obtaining the time window optimization length being trained using first-order linear regression function and Kriging regression method to the power sequence of marked switch events;Obtain the general power sequence to be detected acquired in chronological order;It establishes first time window and follows the second time window of first time window closely, the length of two time windows is time window optimization length;Data filling is carried out to the left and right side of general power sequence to be detected;Two time windows are slided from first data of the center of first time window alignment general power sequence, until the last one data of the center alignment general power sequence of first time window, the step-length of one data of every sliding, Chi-square Test is carried out to the data in two time windows, so that it is determined that whether switch events occur at each data in general power sequence.Method and system of the invention can effectively avoid influence of the noise to detection accuracy, improve accuracy in detection.
Description
Technical field
The present invention relates to electric switch event technical fields, more particularly to a kind of electric switch event detecting method and are
System.
Background technique
With the development of smart grid, the analysis of household electricity load is become more and more important.Pass through point of power load
Analysis, domestic consumer can obtain the power information of each electric appliance and the fining inventory of the electricity charge in time;Power department can obtain
More detailed user power utilization information is obtained, and the accuracy of electro-load forecast can be improved, provides overall planning for power department
Foundation.Meanwhile using the power information of each electric appliance, would know that the electricity consumption behavior of user, this for family's energy consumption assessment and
The research of Energy Saving Strategy has directive significance.
Current power load decomposition is broadly divided into two methods of intrusive load decomposition and non-intrusion type load decomposition.It is non-to invade
Enter formula load decomposition method not needing that monitoring device is installed in the power inside equipment of load, it is only necessary to total according to power load
Information can be obtained the information on load of each electrical equipment.Non-intrusion type load decomposition method has less investment, convenient to use etc.
Feature, therefore, this method are suitable for the decomposition of family's load electricity consumption.
In non-intrusion type load decomposition algorithm, the switch events detection of electrical equipment is most important one link.Initially
Switch events detect judgment basis using the changing value Δ P of active-power P as event detection, it is convenient and intuitively.This be because
It changes for the operating status of any one electrical equipment, consumed performance number also necessarily changes, and this changes
Change will also embody in the general power consumed by all electric appliances.Conjunction of this method in addition to needing to be arranged power change values
Manage threshold value, it is also necessary to solve the problems, such as that event detecting method exists in practical applications: the instantaneous function at certain appliance starting moment
Rate value will appear biggish spike (for example, motor start-up current is much larger than rated current), will cause the variation of electric appliance steady state power
Value inaccuracy, to influence the judgement to switch events;The transient process or length or short of different household electrical appliance, therefore changed power
The determination of value becomes more difficult;The case where mutation will appear due to variation (such as voltage die) active power of power quality,
It is likely to judge by accident in this way.
According to the active power data of the common brand of household electrical appliance, typical active power data, can be with as shown in table 1-1
Find out that only illumination class electrical appliance active power is smaller, other electrical appliances are substantially all in 50W or more, so what the event of determination occurred
Changed power threshold value is set as 50W.
The active power meter of the common household electrical appliance of table 1
Existing switch events detection method are as follows: calculate the active power that each moment reads and read with corresponding last moment
The difference of the active power taken, judges whether the absolute value of the difference is greater than or equal to 30W, and will continuously meet difference and be greater than or wait
It is poor to make in the head and the tail moment corresponding active power for the period that all moment of 30W form, and obtains head and the tail difference power, judgement should
Whether head and the tail power absolute value of the difference is greater than or equal to 50W, then determines if more than or equal to 50W and is switched in the period
Event determines if being less than 50W and power swing occurs in the period.
Existing switch events detection method can identify the switch thing of active power changing value bigger (being greater than 70W)
Part, since system is there are the reason of noise, the electric appliance of certain active power close to threshold value (i.e. 50W) is caused active on startup
Power increase may be weakened by system noise, thus can not accurately identify.
Summary of the invention
The object of the present invention is to provide a kind of electric switch event detecting method and systems, effectively avoid noise to detection essence
The influence of degree improves accuracy in detection.
To achieve the above object, the present invention provides following schemes:
A kind of electric switch event detecting method, comprising:
Obtain using first-order linear regression function and Kriging regression method to the power sequences of marked switch events into
The time window optimization length L that row training obtains;Wherein L is odd number;
Obtain the general power sequence to be detected of the electric appliance acquired in chronological order;
First time window and the second time window are established, when the length of the first time window and second time window is
Between window optimization length L;Second time window follows the first time window closely;
Data filling is carried out to the left and right side of the general power sequence, make the first time window and it is described second when
Between window from first data that the center of the first time window is directed at the general power sequence slide into the first time window
Center when being directed at the last one data of the general power sequence, filled out in the first time window and second time window
Full of L data;
First data for being directed at the general power sequence from the center of the first time window slide the first time
Window and second time window, until the center of the first time window is directed at the last one data of the general power sequence,
The step-length of one data of every sliding carries out card side to the data in the first time window and the data in second time window
It examines, so that it is determined that whether switch events occur at each data in the general power sequence;
Record serial number of all data that switch events occur in the general power sequence.
Optionally, the left and right side to the general power sequence carries out data filling, specifically includes:
At least by the left side supplement of the general power sequenceA data;Left side supplements at leastA data are equal
It is identical as the data of the general power sequence leftmost side;
At least by the right side supplement of the general power sequenceA data;Right side supplements at leastA data
It is identical as the data of the general power sequence rightmost side.
Optionally, the step-length of one data of every sliding, when to the data in the first time window with described second
Between data in window carry out Chi-square Test, so that it is determined that whether occurring to switch thing at each data in the general power sequence
Part specifically includes:
Obtain using first-order linear regression function and Kriging regression method to the power sequences of marked switch events into
The subsequence Optimal units K that row training obtains;
The maximum value and minimum value for seeking all data in the first time window, obtain value interval;
The value interval is divided into K sections, obtains K sections of minizones;
Calculate quantity F of the data in each minizone in the first time window(n) P(k) and described second
Quantity F of the data in each minizone in time window(n) Q(k);Wherein n indicates to be located in the first time window
Serial number of the data of the heart in the general power sequence;K is minizone serial number, k=1,2 ..., K;
Number of computations difference ratio:
According to parameter K and confidence alpha, chi-square value χ is obtained by inquirying card side's table2(α,K-1);
Judge whether the number differences ratio is greater than or equal to the chi-square value, obtains the first judging result;
If first judging result indicates that the number differences ratio is greater than or equal to the chi-square value, it is determined that in institute
State generation switch events at the nth data of general power sequence;If first judging result indicates the number differences ratio
Less than the chi-square value, it is determined that switch events do not occur at the nth data of the general power sequence.
Optionally, using first-order linear regression function and Kriging regression method to the power sequence of marked switch events
The process being trained are as follows:
Obtain the switch events physical location of the power sequence of marked switch events;
Using subsequence number as abscissa, time window length is that ordinate establishes two-dimensional coordinate system, according to time window length
Value range and the value range of subsequence number determine the value region in the two-dimensional coordinate system;
The point of preset quantity is randomly selected in the value region as interpolation point, with the corresponding vertical seat of each interpolation point
The length of first time window and the second time window when being denoted as establishing for time window, using the corresponding abscissa of each interpolation point as taking
It is worth the division number in section, calculates switch events predicted position corresponding to each interpolation point;
According to the prediction of switch events predicted position corresponding to each interpolation point of switch events actual calculation of location
Accuracy obtains the prediction accuracy of each interpolation point;
The value region is subjected to grid dividing, obtains each mesh point as interpolated point;
Linear fit is carried out to each interpolated point using the prediction accuracy of each interpolation point, obtains match value;
The corresponding interpolation point of maximum value in all match values is sought, best interpolation point is obtained;The best interpolation point pair
The abscissa answered is subsequence Optimal units, and the corresponding ordinate of the best interpolation point is time window optimization length.
Invention additionally discloses a kind of electric switch event detection systems, comprising:
Time window optimization length obtains module, utilizes first-order linear regression function and Kriging regression method pair for obtaining
The time window optimization length L that the power sequence of marked switch events is trained;Wherein L is odd number;
Power sequence obtains module, for obtaining the general power sequence to be detected of the electric appliance acquired in chronological order;
Time window establishes module, for establishing first time window and the second time window, the first time window and described
The length of two time windows is time window optimization length L;Second time window follows the first time window closely;
Database population module carries out data filling for the left and right side to the general power sequence, makes described first
Time window and second time window are directed at first data source of the general power sequence from the center of the first time window
It moves when being directed at the last one data of the general power sequence to the center of the first time window, the first time window and institute
It states and fills full L data in the second time window;
Switch events detection module, for being directed at first of the general power sequence from the center of the first time window
Data slide the first time window and second time window, until the center of the first time window is directed at the general power
The last one data of sequence, the step-length of one data of every sliding, when to the data in the first time window with described second
Between data in window carry out Chi-square Test, so that it is determined that whether occurring to switch thing at each data in the general power sequence
Part;
Logging modle, for recording serial number of all data that switch events occur in the general power sequence.
Optionally, the database population module, specifically includes:
The left fills unit of sequence, for supplementing at least the left side of the general power sequenceA data;Left side supplement
At leastA data are identical as the data of the general power sequence leftmost side;
The right fills unit of sequence, for supplementing at least the right side of the general power sequenceA data;Right side supplement
At leastA data are identical as the data of the general power sequence rightmost side.
Optionally, the switch events detection module, specifically includes:
Subsequence Optimal units acquiring unit utilizes first-order linear regression function and Kriging regression method pair for obtaining
The subsequence Optimal units K that the power sequence of marked switch events is trained;
Interval computation unit is taken for seeking the maximum value and minimum value of all data in the first time window
It is worth section;
Interval division unit obtains K sections of minizones for the value interval to be divided into K sections;
Interval censored data amount computing unit, for calculating the data in the first time window in each minizone
Quantity F(n) P(k) the quantity F with the data in second time window in each minizone(n) Q(k);Wherein n table
Show serial number of the data positioned at the first time window center in the general power sequence;K be minizone serial number, k=1,
2,…,K;
Computation unit is used for number of computations difference ratio:
Chi-square value query unit, for obtaining chi-square value χ by inquirying card side's table according to parameter K and confidence alpha2(α,K-
1);
Judging unit obtains first and sentences for judging whether the number differences ratio is greater than or equal to the chi-square value
Disconnected result;
Switch events determination unit, if indicating that the number differences ratio is greater than or equal to for first judging result
The chi-square value, it is determined that switch events occur at the nth data of the general power sequence;If first judging result
Indicate that the number differences ratio is less than the chi-square value, it is determined that do not occur at the nth data of the general power sequence
Switch events.
Optionally, which further includes training module, and the training module is used to utilize single order
Linear regression function and Kriging regression method are trained the power sequence of marked switch events;The training module packet
It includes:
Known array acquiring unit, the switch events physical location of the power sequence for obtaining marked switch events;
Establishment of coordinate system unit, for using subsequence number as abscissa, time window length to be that ordinate establishes two dimension seat
Mark system, determines the value in the two-dimensional coordinate system according to the value range of the value range of time window length and subsequence number
Region;
Interpolation point switch events predicted position computing unit, for randomly selecting preset quantity in the value region
Point is used as interpolation point, first time window and the second time window when being established using the corresponding ordinate of each interpolation point as time window
Length is calculated and is opened corresponding to each interpolation point using the corresponding abscissa of each interpolation point as the division number of value interval
Close event prediction position;
Accuracy computing unit is predicted, for according to corresponding to each interpolation point of switch events actual calculation of location
The prediction accuracy of switch events predicted position obtains the prediction accuracy of each interpolation point;
Grid dividing unit obtains each mesh point conduct and is interpolated for the value region to be carried out grid dividing
Point;
Linear fit unit carries out Linear Quasi to each interpolated point for the prediction accuracy using each interpolation point
It closes, obtains match value;
Best interpolation point computing unit obtains optimal for seeking the corresponding interpolation point of maximum value in all match values
Interpolation point;The corresponding abscissa of the best interpolation point is subsequence Optimal units, the corresponding vertical seat of the best interpolation point
Mark is time window optimization length.
The specific embodiment provided according to the present invention, the invention discloses following technical effects: electric switch of the invention
Event detecting method and system pay close attention to the number that performance number falls into some section using the method identification switch event of Chi-square Test
Amount, can be avoided the size dependent on power difference, to effectively avoid the influence of noise or power swing, it is accurate to improve detection
Degree.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will be to institute in embodiment
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is the flow chart of training process in electric switch event detecting method embodiment of the present invention;
Fig. 2 is the interpolation point distribution map chosen in coordinate system in electric switch event detecting method embodiment of the present invention;
Fig. 3 is the flow chart of detection process in electric switch event detecting method embodiment of the present invention;
Fig. 4 is the system construction drawing of electric switch event detection system embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The object of the present invention is to provide a kind of electric switch event detecting method and systems, effectively avoid noise to detection essence
The influence of degree improves accuracy in detection.
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
Sampled value in time window is fallen into the number in a certain section as the sum of index, statistics number difference by the present invention,
And the sum of this difference compares with chi-square value, the mark whether comparison result is occurred as electric switch event.
The electric switch event detecting method, comprising: training process and detection process.
Fig. 1 is the flow chart of training process in electric switch event detecting method embodiment of the present invention.
Referring to Fig. 1, training process is as follows:
Step 101: obtaining the switch events physical location of the power sequence of marked switch events.Marked switch events
Power sequence indicate are as follows:NTFor the length of the power sequence of marked switch events.
Step 102: being abscissa with subsequence number K ', time window length L ' is that ordinate establishes two-dimensional coordinate system, root
The value area in the two-dimensional coordinate system is determined according to the value range of time window length L ' and the value range of subsequence number K '
Domain.
The value range of subsequence number K ' is [1,51], and the value range of time window length L ' is [1,201].And K ' and
L ' is integer, and L ' is odd number.After establishing coordinate system, subsequence number K ' and time window length L ' composition parameter vector
Step 103: the point of preset quantity is randomly selected in the value region as interpolation point, with each interpolation point pair
The length of first time window and the second time window when the ordinate answered is established as time window, with the corresponding horizontal seat of each interpolation point
Be denoted as the division number for value interval, call switch events detection algorithm (the switch events detection algorithm and detection process
Detecting step is similar), calculate switch events predicted position corresponding to each interpolation point;
The essence of training algorithm is the data interpolating algorithm based on Kriging model.Therefore, preferred needs are explicitly known
Interpolation point and its corresponding functional value (present invention is using prediction accuracy corresponding to each interpolation point as functional value).With nine
For a interpolation point, Fig. 2 is the interpolation point point chosen in coordinate system in electric switch event detecting method embodiment of the present invention
Butut.Referring to fig. 2, round coordinate points, that is, interpolation point in Fig. 2,9 known interpolation points are as follows:
WhereinFor the abscissa of 9 interpolation points, the i.e. corresponding time window length of 9 interpolation points;It is 9
The ordinate of a interpolation point, i.e. 9 interpolation points correspond to subsequence number.
Step 104: according to switch events prediction bits corresponding to each interpolation point of switch events actual calculation of location
The prediction accuracy set obtains the prediction accuracy of each interpolation point;
The prediction accuracy set of 9 interpolation points are as follows:WhereinRespectively 9 are inserted
It is worth the prediction accuracy of point.
Step 105: the value region being subjected to grid dividing, obtains each mesh point as interpolated point.
Referring to fig. 2, rectangular coordinates point is interpolated point in Fig. 2, only illustrates four interpolated points in Fig. 2.Of the invention
In one embodiment, the transverse and longitudinal coordinate range in value region is divided into 10 mesh points.The points of division are more, then most
Whole result accuracy is higher.
Step 106: linear fit being carried out to each interpolated point using the prediction accuracy of each interpolation point, is fitted
Value.
Seek the formula of match value are as follows:
WhereinFor match value, C is fit metric.The calculation formula of fit metric are as follows:
C=R-1[r-Fλ]
Wherein λ is intermediate parameters, is a parameter related with fit variations spatial position, λ=[FTR-1F]-1[FTR-1r-
1];
Wherein F is regression matrix, and R is correlation matrix, and r is another intermediate parameters.Specifically:
Regression matrixIndicate the coefficient of linear regression.
Correlation matrix R=[Rij]9×9, whereinFor related coefficient, coherent function in Kriging model is indicated
Form;The citing of two parameters in expression parameter space.RijAnd dijIt is intermediate ginseng
Number.I=1,2 ..., 9;J=1,2 ..., 9.
Intermediate parameters r=[ri], whereinriIndicate 9 known
Coherent function between interpolation point and interpolated point (L', K'), diIndicate known 9 interpolation points and interpolated point (L', K')
Between space length, i=1,2 ..., 9.
Step 107: seeking the corresponding interpolation point of maximum value in all match values, obtain best interpolation point;It is described optimal
The corresponding abscissa of interpolation point is subsequence Optimal units, and the corresponding ordinate of the best interpolation point is that time window is optimal
Length.
Fig. 3 is the flow chart of detection process in electric switch event detecting method embodiment of the present invention.
Referring to Fig. 3, detection process is as follows:
Step 301: obtaining and utilize first-order linear regression function and Kriging regression method to the function of marked switch events
The time window optimization length L that rate sequence is trained;Wherein L is odd number;
Step 302: obtaining the general power sequence to be detected of the electric appliance acquired in chronological order.
General power sequence to be detected is p0,p1,…,pN-1, wherein N be general power sequence to be detected length, i.e., to
The quantity for the general power data that the general power sequence of detection is included.
Step 303: establishing first time window and the second time window, the length of the first time window and second time window
Degree is time window optimization length L;Second time window follows the first time window closely.
Step 304: data filling being carried out to the left and right side of the general power sequence, makes the first time window and institute
It states the second time window and slides into described from first data that the center of the first time window is directed at the general power sequence
When the center of one time window is directed at the last one data of the general power sequence, the first time window and second time
Full L data are filled in window.
Wherein, data filling is carried out to the left and right side of the general power sequence, specifically included: by the general power sequence
The left side supplement of column is at leastA data;Left side supplements at leastA data with the general power sequence leftmost side
Data it is identical;At least by the right side supplement of the general power sequenceA data;Right side supplements at leastNumber
According to equal identical as the data of the general power sequence rightmost side.
Step 305: being aligned described in first data sliding of the general power sequence from the center of the first time window
First time window and second time window, until the center of the first time window is directed at last of the general power sequence
A data, the step-length of one data of every sliding, to the data in the first time window and the data in second time window
Chi-square Test is carried out, so that it is determined that whether switch events occur at each data in the general power sequence.
Data sequence in the first time window isWherein n is indicated
Positioned at serial number of the data in the general power sequence of the first time window center;pnFor positioned at the first time window center
Data;P indicates the data sequence of first time window, and p indicates each data in general power sequence;Since the second time window follows closely
One time window, then the data sequence in second time window be
Wherein, the step-length of one data of every sliding, in the data and second time window in the first time window
Data carry out Chi-square Test, so that it is determined that switch events whether occur at each data in the general power sequence, specifically
Include:
Obtain using first-order linear regression function and Kriging regression method to the power sequences of marked switch events into
The subsequence Optimal units K that row training obtains;
Seek the first time window PnThe maximum value and minimum value of interior all data, obtain value interval;It is wherein minimum
Value is denoted as p(n) min=min [Pn], maximum value is denoted as p(n) max=max [Pn], then value interval is
The value interval is divided into K sections, obtains K sections of minizones.Wherein the value of K is generally 10;Each minizone
Serial number k, each minizone after dividing can indicate are as follows:
Calculate quantity F of the data in each minizone in the first time window(n) P(k) and described second
Quantity F of the data in each minizone in time window(n) Q(k)。
Number of computations difference ratio:
According to parameter K and confidence alpha (α is generally taken as 95%), chi-square value χ is obtained by inquirying card side's table2(α,K-1);
Judge whether the number differences ratio is greater than or equal to the chi-square value, obtains the first judging result;
If first judging result indicates that the number differences ratio is greater than or equal to the chi-square value, i.e. γ(n)≥χ2
(α, K-1), it is determined that switch events occur at the nth data of the general power sequence;If the first judging result table
Show that the number differences ratio is less than the chi-square value, i.e. γ(n)< χ2(α, K-1), it is determined that the of the general power sequence
Switch events do not occur at n data.
Step 306: recording serial number of all data that switch events occur in the general power sequence.
The greatest problem that electric switch event detection is faced is adverse effect of the noise to detection accuracy.Noise is more than one
Determine degree, will cause the very fast decline of detection accuracy.Existing incident Detection Algorithm depends on the size of power difference simultaneously.And this
The number that performance number falls into section is then investigated in invention, and this statistical method can avoid noise or power waves to a certain extent
Dynamic influence.
Therefore, the present invention is to carry out statistics to the distribution of adjacent two data sequences (P and Q) to portray, if the two is poor
The sum of value is less than chi-square value, then it is assumed that the two data sequence (P and Q) distributions having the same do not have switch events;Such as
Fruit is greater than chi-square value, then it is assumed that (distribution of P and Q) are different, it is meant that switch events have occurred for the two data sequences.Using point
Whether cloth is identical, can avoid the influence of noise to a certain extent, improves detection accuracy, and calculation method very simple,
Principle is very clear.
The present invention uses Kriging interpolation model to be trained in the training process.The Kriging interpolation model includes two
A part, first part utilize first-order linear regression function, can preferably reduce impulsive noise (especially flash noise)
Influence to parameter, so that method has better robustness.Second part use exponential function, exponential function emphasize data it
Between relationship, and have isotropism, just agree with power data handled by the present invention.
To sum up, method of the invention robustness with higher, practicability and adaptivity, can be effectively reduced noise pair
The influence of switch events detection.
Fig. 4 is the system construction drawing of electric switch event detection system embodiment of the present invention.
Referring to fig. 4, the electric switch event detection system, comprising:
Time window optimization length obtains module 401, utilizes first-order linear regression function and Kriging regression side for obtaining
The time window optimization length L that method is trained the power sequence of marked switch events;Wherein L is odd number;
Power sequence obtains module 402, for obtaining the general power sequence to be detected of the electric appliance acquired in chronological order.
Time window establishes module 403, for establishing first time window and the second time window, the first time window and described
The length of second time window is time window optimization length L;Second time window follows the first time window closely.
Database population module 404 carries out data filling for the left and right side to the general power sequence, makes described the
One time window and second time window are directed at first data of the general power sequence from the center of the first time window
When sliding into the center of the first time window and being directed at the last one data of the general power sequence, the first time window and
Full L data are filled in second time window.
The database population module 404, specifically includes: the left fills unit of sequence and the right fills unit of sequence.Wherein sequence
Left fills unit, for supplementing at least the left side of the general power sequenceA data;Left side supplements at leastIt is a
Data are identical as the data of the general power sequence leftmost side;The right fills unit of sequence, for by the general power sequence
Right side supplements at leastA data;Right side supplements at leastA data with the general power sequence rightmost side
Data are identical.
Switch events detection module 405, for being directed at the of the general power sequence from the center of the first time window
One data slides the first time window and second time window, until the center alignment of the first time window is described total
The last one data of power sequence, the step-length of one data of every sliding, to the data and described the in the first time window
Whether the data in two time windows carry out Chi-square Test, so that it is determined that switching at each data in the general power sequence
Event.
The switch events detection module 405, specifically includes:
Subsequence Optimal units acquiring unit utilizes first-order linear regression function and Kriging regression method pair for obtaining
The subsequence Optimal units K that the power sequence of marked switch events is trained;Interval computation unit, for seeking
The maximum value and minimum value for stating all data in first time window, obtain value interval;Interval division unit, for being taken described
Being worth interval division is K sections, obtains K sections of minizones;Interval censored data amount computing unit, for calculating the data in the first time window
Quantity F in each minizone(n) P(k) and the data in second time window are in each minizone
Quantity F(n) Q(k);Wherein n indicates serial number of the data for being located at the first time window center in the general power sequence;K is small
Section serial number, k=1,2 ..., K;Computation unit is used for number of computations difference ratio:
Chi-square value query unit, for obtaining chi-square value χ by inquirying card side's table according to parameter K and confidence alpha2(α,K-1);Judgement
Unit obtains the first judging result for judging whether the number differences ratio is greater than or equal to the chi-square value;Switch thing
Part determination unit, if indicating that the number differences ratio is greater than or equal to the chi-square value for first judging result,
It determines and switch events occurs at the nth data of the general power sequence;If first judging result indicates the quantity
Difference ratio is less than the chi-square value, it is determined that switch events do not occur at the nth data of the general power sequence.
Logging modle 406, for recording serial number of all data that switch events occur in the general power sequence.
The training module 407, for utilizing first-order linear regression function and Kriging regression method to marked switch
The power sequence of event is trained;The training module 407 includes:
Known array acquiring unit, the switch events physical location of the power sequence for obtaining marked switch events;
Establishment of coordinate system unit is used for using subsequence number as abscissa, and time window length is that ordinate establishes two-dimensional coordinate system, according to
The value range of time window length and the value range of subsequence number determine the value region in the two-dimensional coordinate system;Interpolation
Point switch events predicted position computing unit, for randomly selecting the point of preset quantity in the value region as interpolation
Point, the length of first time window and the second time window when being established using the corresponding ordinate of each interpolation point as time window, with every
Division number of the corresponding abscissa of a interpolation point as value interval, calculates the prediction of switch events corresponding to each interpolation point
Position;Accuracy computing unit is predicted, for opening according to corresponding to each interpolation point of switch events actual calculation of location
The prediction accuracy for closing event prediction position, obtains the prediction accuracy of each interpolation point;Grid dividing unit, being used for will be described
Value region carries out grid dividing, obtains each mesh point as interpolated point;Linear fit unit, for utilizing each interpolation
The prediction accuracy of point carries out linear fit to each interpolated point, obtains match value;Best interpolation point computing unit, for asking
The corresponding interpolation point of maximum value in all match values is taken, best interpolation point is obtained;The corresponding abscissa of the best interpolation point
As subsequence Optimal units, the corresponding ordinate of the best interpolation point is time window optimization length.
The specific embodiment provided according to the present invention, the invention discloses following technical effects: electric switch of the invention
Event detecting method and system pay close attention to the number that performance number falls into some section using the method identification switch event of Chi-square Test
Amount, can be avoided the size dependent on power difference, to effectively avoid the influence of noise or power swing, it is accurate to improve detection
Degree.
For the system disclosed in the embodiment, since it is corresponded to the methods disclosed in the examples, so the ratio of description
Relatively simple, reference may be made to the description of the method.
Used herein a specific example illustrates the principle and implementation of the invention, and above embodiments are said
It is bright to be merely used to help understand method and its core concept of the invention;At the same time, for those skilled in the art, foundation
Thought of the invention, there will be changes in the specific implementation manner and application range.In conclusion the content of the present specification is not
It is interpreted as limitation of the present invention.
Claims (6)
1. a kind of electric switch event detecting method characterized by comprising
Acquisition instructs the power sequence of marked switch events using first-order linear regression function and Kriging regression method
The time window optimization length L got;Wherein L is odd number;
Obtain the general power sequence to be detected of the electric appliance acquired in chronological order;
Establish first time window and the second time window, the length of the first time window and second time window is time window
Optimization length L;Second time window follows the first time window closely;
Data filling is carried out to the left and right side of the general power sequence, makes the first time window and second time window
First data for being directed at the general power sequence from the center of the first time window slide into the first time window
When the heart is directed at the last one data of the general power sequence, filling is full in the first time window and second time window
L data;
From first data sliding described first of the general power sequence before the alignment filling of the center of the first time window
Time window and second time window, until the general power sequence before the center alignment filling of the first time window is most
The latter data, the step-length of one data of every sliding, in the data and second time window in the first time window
Data carry out Chi-square Test, so that it is determined that whether switch events occur at each data in the general power sequence;
Record serial number of all data that switch events occur in the general power sequence;
The left and right side to the general power sequence carries out data filling, specifically includes:
At least by the left side supplement of the general power sequenceA data;Left side supplements at leastA data with it is described
The data of the general power sequence leftmost side are identical;
At least by the right side supplement of the general power sequenceA data;Right side supplements at leastA data are and institute
The data for stating the general power sequence rightmost side are identical.
2. a kind of electric switch event detecting method according to claim 1, which is characterized in that one number of every sliding
According to step-length, in the first time window data and second time window in data carry out Chi-square Test, thus really
Whether switch events occur at each data in the fixed general power sequence, specifically include:
Acquisition instructs the power sequence of marked switch events using first-order linear regression function and Kriging regression method
The subsequence Optimal units K got;
The maximum value and minimum value for seeking all data in the first time window, obtain value interval;
The value interval is divided into K sections, obtains K sections of minizones;
Calculate quantity F of the data in each minizone in the first time window(n) P(k) and second time
Quantity F of the data in each minizone in window(n) Q(k);Wherein n indicates to be located at the first time window center
Serial number of the data in the general power sequence;K is minizone serial number, k=1,2 ..., K;
Number of computations difference ratio:
According to parameter K and confidence alpha, chi-square value χ is obtained by inquirying card side's table2(α,K-1);
Judge whether the number differences ratio is greater than or equal to the chi-square value, obtains the first judging result;
If first judging result indicates that the number differences ratio is greater than or equal to the chi-square value, it is determined that described total
Switch events occur at the nth data of power sequence;If first judging result indicates that the number differences ratio is less than
The chi-square value, it is determined that switch events do not occur at the nth data of the general power sequence.
3. a kind of electric switch event detecting method according to claim 2, which is characterized in that returned using first-order linear
The process that function and Kriging regression method are trained the power sequence of marked switch events are as follows:
Obtain the switch events physical location of the power sequence of marked switch events;
Using subsequence number as abscissa, time window length is that ordinate establishes two-dimensional coordinate system, according to taking for time window length
The value range of value range and subsequence number determines the value region in the two-dimensional coordinate system;
The point of preset quantity is randomly selected in the value region as interpolation point, is made with the corresponding ordinate of each interpolation point
The length of first time window and the second time window when being established for time window, using the corresponding abscissa of each interpolation point as value area
Between division number, calculate switch events predicted position corresponding to each interpolation point;
It is correct according to the prediction of switch events predicted position corresponding to each interpolation point of switch events actual calculation of location
Rate obtains the prediction accuracy of each interpolation point;
The value region is subjected to grid dividing, obtains each mesh point as interpolated point;
Linear fit is carried out to each interpolated point using the prediction accuracy of each interpolation point, obtains match value;
The corresponding interpolation point of maximum value in all match values is sought, best interpolation point is obtained;The best interpolation point is corresponding
Abscissa is subsequence Optimal units, and the corresponding ordinate of the best interpolation point is time window optimization length.
4. a kind of electric switch event detection system characterized by comprising
Time window optimization length obtains module, for obtaining using first-order linear regression function and Kriging regression method to having marked
The time window optimization length L that the power sequence of Slate event is trained;Wherein L is odd number;
Power sequence obtains module, for obtaining the general power sequence to be detected of the electric appliance acquired in chronological order;
Time window establishes module, for establishing first time window and the second time window, the first time window and it is described second when
Between the length of window be time window optimization length L;Second time window follows the first time window closely;
Database population module carries out data filling for the left and right side to the general power sequence, makes the first time
Window and second time window slide into from first data that the center of the first time window is directed at the general power sequence
When the center of the first time window is directed at the last one data of the general power sequence, the first time window and described
Full L data are filled in two time windows;
Switch events detection module, for from the of the general power sequence before the alignment filling of the center of the first time window
One data slides the first time window and second time window, until before the center alignment filling of the first time window
The general power sequence the last one data, it is every sliding one data step-length, to the data in the first time window
Carry out Chi-square Test with the data in second time window, so that it is determined that at each data in the general power sequence whether
Switch events occur;
Logging modle, for recording serial number of all data that switch events occur in the general power sequence;
The database population module, specifically includes:
The left fills unit of sequence, for supplementing at least the left side of the general power sequenceA data;Left side supplements extremely
It is fewA data are identical as the data of the general power sequence leftmost side;
The right fills unit of sequence, for supplementing at least the right side of the general power sequenceA data;Right side supplements extremely
It is fewA data are identical as the data of the general power sequence rightmost side.
5. a kind of electric switch event detection system according to claim 4, which is characterized in that the switch events detection
Module specifically includes:
Subsequence Optimal units acquiring unit, for obtaining using first-order linear regression function and Kriging regression method to having marked
The subsequence Optimal units K that the power sequence of Slate event is trained;
Interval computation unit obtains value area for seeking the maximum value and minimum value of all data in the first time window
Between;
Interval division unit obtains K sections of minizones for the value interval to be divided into K sections;
Interval censored data amount computing unit, for calculating number of the data in the first time window in each minizone
Measure F(n) P(k) the quantity F with the data in second time window in each minizone(n) Q(k);Wherein n indicates position
In serial number of the data in the general power sequence of the first time window center;K is minizone serial number, k=1,2 ..., K;
Computation unit is used for number of computations difference ratio:
Chi-square value query unit, for obtaining chi-square value χ by inquirying card side's table according to parameter K and confidence alpha2(α,K-1);
Judging unit obtains the first judgement knot for judging whether the number differences ratio is greater than or equal to the chi-square value
Fruit;
Switch events determination unit, if it is described to indicate that the number differences ratio is greater than or equal to for first judging result
Chi-square value, it is determined that switch events occur at the nth data of the general power sequence;If first judging result indicates
The number differences ratio is less than the chi-square value, it is determined that does not switch at the nth data of the general power sequence
Event.
6. a kind of electric switch event detection system according to claim 5, which is characterized in that it further include training module,
The training module is used to utilize first-order linear regression function and Kriging regression method to the power sequence of marked switch events
Column are trained;The training module includes:
Known array acquiring unit, the switch events physical location of the power sequence for obtaining marked switch events;
Establishment of coordinate system unit is used for using subsequence number as abscissa, and time window length is that ordinate establishes two-dimensional coordinate system,
The value region in the two-dimensional coordinate system is determined according to the value range of the value range of time window length and subsequence number;
Interpolation point switch events predicted position computing unit, the point for randomly selecting preset quantity in the value region are made
For interpolation point, the length of first time window and the second time window when being established using the corresponding ordinate of each interpolation point as time window
Degree, using the corresponding abscissa of each interpolation point as the division number of value interval, calculates switch corresponding to each interpolation point
Event prediction position;
Accuracy computing unit is predicted, for the switch according to corresponding to switch events actual calculation of location each interpolation point
The prediction accuracy of event prediction position obtains the prediction accuracy of each interpolation point;
Grid dividing unit obtains each mesh point as interpolated point for the value region to be carried out grid dividing;
Linear fit unit carries out linear fit to each interpolated point for the prediction accuracy using each interpolation point, obtains
To match value;
Best interpolation point computing unit obtains best interpolation for seeking the corresponding interpolation point of maximum value in all match values
Point;The corresponding abscissa of the best interpolation point is subsequence Optimal units, and the corresponding ordinate of the best interpolation point is
For time window optimization length.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810768660.2A CN109101465B (en) | 2018-07-13 | 2018-07-13 | A kind of electric switch event detecting method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810768660.2A CN109101465B (en) | 2018-07-13 | 2018-07-13 | A kind of electric switch event detecting method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109101465A CN109101465A (en) | 2018-12-28 |
CN109101465B true CN109101465B (en) | 2019-11-08 |
Family
ID=64846327
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810768660.2A Expired - Fee Related CN109101465B (en) | 2018-07-13 | 2018-07-13 | A kind of electric switch event detecting method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109101465B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110133396B (en) * | 2019-04-09 | 2021-07-20 | 上海电力学院 | Method for detecting household electrical appliance switch event based on binary recursive SVD decomposition |
CN111161097B (en) * | 2019-12-31 | 2023-07-25 | 四川大学 | Method and device for detecting switch event by event detection algorithm based on hypothesis test |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101429397B1 (en) * | 2013-04-11 | 2014-08-14 | 전북대학교산학협력단 | Method and system for extracting core events based on message analysis in social network service |
CN104660467A (en) * | 2015-02-10 | 2015-05-27 | 深圳市科信通信技术股份有限公司 | Testing method for communication comprehensive test terminal |
-
2018
- 2018-07-13 CN CN201810768660.2A patent/CN109101465B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101429397B1 (en) * | 2013-04-11 | 2014-08-14 | 전북대학교산학협력단 | Method and system for extracting core events based on message analysis in social network service |
CN104660467A (en) * | 2015-02-10 | 2015-05-27 | 深圳市科信通信技术股份有限公司 | Testing method for communication comprehensive test terminal |
Non-Patent Citations (4)
Title |
---|
SMA局部场电在力竭运动过程中的特征分析及BIC干预实验研究;侯莉娟,等;《首都体育学院学报》;20160930;第455-458页 * |
基于功率谱分段对消算法的跳频信号检测;刘旭波;《仪器仪表与检测技术》;20171231;第69-73页 * |
射频仿真系统天线馈电通道幅相平衡优化设计;徐若曦,等;《计算机仿真》;20151031;第196-201页 * |
混沌序列在降低OFDM系统的峰均比中的应用;张鸷,等;《研究与探讨》;20091231;第5-8页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109101465A (en) | 2018-12-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Stephen et al. | Incorporating practice theory in sub-profile models for short term aggregated residential load forecasting | |
CN107861070B (en) | A kind of health state of lithium ion battery inline diagnosis method | |
CN109325545A (en) | Low-voltage network topological structure method of calibration, device, equipment and storage medium | |
CN106096726A (en) | A kind of non-intrusion type load monitoring method and device | |
CN104239963A (en) | Method for finding abnormal electric energy meter based on gray GM (1, 1) model | |
CN106296465B (en) | Method for detecting abnormal electricity utilization behavior of smart power grid | |
CN109101465B (en) | A kind of electric switch event detecting method and system | |
CN109557422A (en) | A kind of intelligent power distribution network short circuit fault localization method and system | |
KR20100133570A (en) | Method and system for selection of optimal location of distributed generation, method and system for selection of optimal size of distributed generation | |
US20150088441A1 (en) | Energy usage estimation device and energy usage estimation method | |
CN110716097A (en) | Power supply network topology identification method and device | |
CN108876476B (en) | A kind of electric switch event detecting method and system | |
TWI391675B (en) | Method and system for estimating use time of electric appliances, and computer program product thereof | |
Baran | Branch current based state estimation for distribution system monitoring | |
CN109902356A (en) | Electric power signal non-intrusion type loads decomposition method | |
Stefanidou-Voziki et al. | Feature selection and optimization of a ML fault location algorithm for low voltage grids | |
CN108491559A (en) | A kind of time series method for detecting abnormality based on normalized mutual information estimation | |
CN111666870A (en) | Power signal reconstruction method and system by utilizing quadratic constraint | |
CN106066415A (en) | For the method detecting the swindle in supply network | |
CN110780129A (en) | Electricity stealing and leakage positioning method based on current deviation analysis technology | |
CN110244115B (en) | Load switch event detection method and system based on signal connectivity | |
CN110133365B (en) | Method and device for detecting switching event of load | |
CN110221119B (en) | Load switch event detection method and system based on power and akie fusion information | |
CN115222018A (en) | Non-invasive load decomposition method and device | |
CN110658385B (en) | Power consumption classification metering method based on complex trapezoid method |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20191108 Termination date: 20200713 |
|
CF01 | Termination of patent right due to non-payment of annual fee |