CN106600449A - Automatic power trend recognition method - Google Patents
Automatic power trend recognition method Download PDFInfo
- Publication number
- CN106600449A CN106600449A CN201510673427.2A CN201510673427A CN106600449A CN 106600449 A CN106600449 A CN 106600449A CN 201510673427 A CN201510673427 A CN 201510673427A CN 106600449 A CN106600449 A CN 106600449A
- Authority
- CN
- China
- Prior art keywords
- data
- trend
- active power
- footnote
- 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.)
- Granted
Links
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to an automatic power trend recognition method. The method comprises the steps of receiving original data of active power by a server in real time; pre-processing the original data; recognizing an oscillation state of the active power; and recognizing an active power trend. With technical scheme provided by the invention, automatic recognition on the trend and the oscillation of the active power is achieved by employing a high-frequency data smoothing technology and combining experience of people during the long-term calculation process and a special calculation method so that the purposes of helping signal trend and oscillation recognition of a power system and guiding actual running are achieved.
Description
Technical field
The present invention relates to a kind of recognition methodss of power system, and in particular to a kind of automatic power trend recognition methodss.
Background technology
During Operation of Electric Systems, natural environment, artificial, communication or internal system element is inevitably subject to hide hidden
The impact of trouble.After Power System Disturbances occur, system disturbance scene is timely and accurately obtained, accurately identifies disturbance type and position,
Reasonable, effective control measure could be implemented, so as to ensure power network safety operation.WAMS WAMS is with same
The whole network monitoring system of new generation of step phasor measurement unit PMU primary element, it can realize that the synchronization of whole network data is adopted in real time
Collection, remote transmission in real time, so that the whole network dynamic monitoring is possibly realized.When disturbance occurs, the PMU for being dispersed throughout the whole network will
The information for collecting sends in real time control centre, after operations staff to real-time synchronization data by being processed and being screened, obtains table
The effective information of perturbation features is levied, and disturbance is classified to determine the countermeasure of next step.However, automatic identification disturbance send out
Life is still highly difficult.Existing disturbance occurs in technology of identification scheme, mainly to be judged using threshold method.
The computing formula of power swing is as follows in existing method:
Relative to the intensity of variation of normal power before disturbance, Δ P is just reduction to power when power increases after Δ P is characterized and disturbed in above formula
Shi Zewei bears.
By arranging different threshold values to above-mentioned different measuring values, then think that disturbance occurs in system operation when observed quantity exceedes threshold value.
Though above-mentioned technical proposal calculating speed is very fast, also have the disadvantages that:
(1) due to the noise jamming of PMU metrical informations, sometimes measured value exceedes threshold value only because noise jamming is caused, because
This hard threshold method sometimes occurs erroneous judgement.
(2) rational threshold value setting is extremely difficult.Because with the adjustment of power system operating mode, threshold value also should be adjusted accordingly.
(3) each measured value of PMU is the performance of perturbation features in electrical network, and the similar electricity of difference PMU substations measurement
There is position difference in tolerance, therefore affected also different by disturbing due to it relative to disturbance.Distance disturbance electrical distance is nearer
It is impacted larger, farther out impacted less.And disturb generation position and do not know, therefore existing threshold value disturbance criterion is not
Rationally.
The content of the invention
The present invention is based on electric network data, a kind of vibration identification and Trend judgement problem for presence, there is provided automatic power trend
Recognition methodss, the method adopts high-frequency data antialiasing, with reference to the experience and specific meter of the people in long-term calculating process
Calculation method, realizes the automatic identification of power trend and vibration, to reach auxiliary power system signal trend, vibration identification and instruct
The purpose of actual motion.
The purpose of the present invention is realized using following technical proposals:
The present invention provides a kind of 1, automatic power trend recognition methodss, and it is theed improvement is that, methods described includes following steps
Suddenly:
(1) server end real-time reception active power initial data;
(2) initial data pretreatment;
(3) active power oscillations state recognition;
(4) active power trend identification.
Further, in the step (2), rolling average smooth manner is adopted to initial data pretreatment, reaches data smoothing,
Extract active power variation tendency;The initial data pretreatment comprises the steps:
1) meansigma methodss are extracted at equal intervals to initial data, including:Input initial data Dt, the initial data Dt is individual per T
Point extracts its meansigma methods, is designated as 1 point, and exists in A variables;The A variables include A (1), A (2), A (3), A (4) ...;
The A (1), A (2), A (3), A (4) ... respectively Dt (1:T)、Dt(T+1:2T)、Dt(2T+1:3T)、Dt(3T+1:4T)......;
2) average data to extracting moves average treatment, to reach the purpose of smoothed data, including:To A variables
Average treatment is moved using equation below;
Wherein:To have processed rear data, i represents data footnote parameter to B, and j is the loop parameter of cumulative sum, and D is represented between translation
Every length;
The length for making data A is Len, and N divides exactly the result of D for Len, then the span of footnote parameter i is 1,2 ..., N;
3) data after gliding smoothing are expanded, including:Each data point in B data is expanded into forward 24 numerical value,
Extended value is the numerical value of current point, and expansion is finished and is designated as C variables;The C variables include C (1:T)、C(T+1:2T)、
C(2T+1:3T)、C(3T+1:4T).......
Further, the step (3) comprises the steps:
1. according to different unit features, following parameter is initialized, including:
A, active power oscillations threshold percentage Per;The parameter is used to adjust the sensitivity of concussion feature identification, and parameter value is bigger
Then shake identification region less, on the contrary it is bigger then to shake identification region;
B, upper footnote recorded amounts U_ind;The footnote position of all maximum extreme points in for recording initial data;
C, mark recorded amounts D_ind of getting a foothold;The footnote position of all minimum extreme points in for recording initial data;
2. using pre-treatment step 1 in 2) and method 2) smooth pretreatment is carried out to active power initial data, return and do not expand
The pretreated active power data filled;
3. find oscillatory regime when coenvelope line Up and lower envelope line Down;
4. the coenvelope line number to finding carries out linear interpolation according to Up and lower envelope line number according to Down, obtains data U_A after interpolation
And D_A;Linear interpolation algorithm is as follows:
For envelope data Up, the first two data point coordinate information (x1, y1) and (x2, y2) is taken first;Wherein x coordinate
Represent moment, the active power numerical value at y coordinate representation correspondence moment;If x2-x1 > 1, the point x between x1 to x2 is entered
Row linear interpolation obtains corresponding y values, as follows:
Y=(y2-y1)/(x2-x1) * (x-x1)+y1
Then again linear interpolation is carried out to (x2, y2) and (x3, y3), goes down finally obtain U_A successively;Same method,
D_A is obtained by lower envelope line number according to Down;
5. the coenvelope line number after interpolation processing is reused respectively according to U_A and lower envelope line number according to D_A and initial data
Algorithm is smoothed described in pre-treatment step, and the data for finally giving are designated as respectively Temp_U, Temp_D, M;
6. according to active power oscillations decision rule, the detection of active power oscillations state is carried out;Active power oscillations decision rule
For:Active power oscillations are judged to as min ((Temp_U-M), (M-Temp_D)) > M*Per, are unsatisfactory for, terminated;
7. expand judged result, reach the purpose flux matched with former data.
Further, the active power oscillations threshold percentage Per value is between 0.005-0.01.
Further, coenvelope line when finding oscillatory regime includes:A () chooses continuous d from the starting point of active power data
Data, find the upper footnote of maximum data and correspondence in d data, if the maximum data of two and its above is then chosen
The big data of footnote, maximum data and the upper footnote of correspondence are respectively present in array Up and U_ind, and remember that footnote is variable i ndex;
(b) with index+1 as starting point, repeat step (a) and (b), the upper bag until having inquired about all data, during oscillatory regime
Winding thread is found and is terminated;
Lower envelope line when finding oscillatory regime includes:C () chooses continuous d data from the starting point of data, search out d numbers
According to middle minimum data and the lower footnote of correspondence, if the minimum data of two and its above then chooses the little data of lower footnote, will most
Small data and the lower footnote of correspondence are respectively present in array Down and D_ind, and remember that footnote is index ' ();D () is with index '+1
For starting point, repeat step (c) and (d), until having inquired about all data, the searching of lower envelope line during oscillatory regime terminates.
Further, the step (4) comprises the steps:
<1>According to different unit features, initiation parameter, including historical data fragment k, smoothed data spacing parameter D, have
Work(power trend threshold percentage Per and translational movement py;
<2>Receive 25*D untreated active power signal sampling value, using the step of preprocess method 1) and 2) to having
Work(power data is translated downwards and is smoothed and is designated as B, returns the data not expanded;
<3>It is average according to being weighted to the front k of current data B (j) according to active power preprocessed data, obtain comprehensively referring to
MarkWherein, weight meet w (j-k)≤w (j- (k-1))≤...≤w (j-1), threshold value
Thre=C*Per;J is currently processed data segment subscript, and m is loop parameter, and w (j-m) is corresponding weight parameter, B (k-m+1)
Represent the value of lower target array B of correspondence;
<4>With reference to the result of calculation of Trend judgement aggregative indicator, using Trend judgement rule, the inspection of active power Trend judgement is carried out
Survey;
<5>According to preliminary judged result, if active power trend is ascendant trend or downward trend, trend extension strategies are performed;
<6>Expand judged result, each data point in B data is expanded into forward 24, extended value is the numerical value of current point, expands
Fill the state that state is current point.
Further, the step<4>In, the detection of active power Trend judgement includes:
Trend judgement rule abs (B (j)-C) > thre are judged, if meet to continue to judge B (j) < C Down, if meet being determined with
Work(power trend is ascendant trend, is otherwise downward trend;If abs (B (j)-C) < thre, active power Trend Stationary is judged,
Proceed next step, wherein Down represents lower envelope line number evidence;Thre represents threshold value;C is that the data after gliding smoothing are entered
The variable that row expands after finishing;B (j) represents current data.
Further, the historical data fragment k span is 8~12, and smoothed data amount D span is 35~50;Have
Work(power trend threshold percentage Per value is between 1~5;Translational movement py values be chosen for correspondence unit normal work when reach
Minima.
The excellent effect that has of technical scheme that the present invention is provided is:
1. when system is disturbed impact, the generation for disturbing can as early as possible be found, this is for follow-up system operation control and thing
Therefore post analysis are respectively provided with significance.Under normal operating conditions, the present invention can automatically judge the Startup time for disturbing, and energy
Enough identifying system power swing trend and vibration, this has great importance for ensureing the safety of operation of power networks, reliability.
2. real time data of the method that the present invention is provided based on actual electric network, has carried out automatic identification oscillation of power, fluctuation tendency and has known
The research and trial of other method, the accumulation trend that will manually study for a long period of time observation experience, necessary computational methods and actual electric network system
The characteristics of combine, it is proposed that the automatic trend recognition methodss based on PMU.The method can utilize trend analysiss technology, automatically
Vibration, fluctuation characteristic of identification observed quantity etc., with higher calculating speed.Trend recognition methodss allow to be applied to actual electricity
In net, the requirement of actual electric network rapidity is met.
3. the method that the present invention is provided changes traditional dependence threshold method and starts the pattern that disturbance judges, realizes online meter in real time
Calculate and trend Intelligent Recognition, be the further lifting for starting decision procedure to tradition disturbance, provide for actual electric network operation important
Guidance effect, be favorably improved actual electric network operation safety.
Description of the drawings
Fig. 1 is the PMU measuring value trend identification model flow charts that the present invention is provided;
Fig. 2 is the data prediction flow chart that the present invention is provided;
Fig. 3 is the active power oscillations identification process figure that the present invention is provided;
Fig. 4 is the active power trend identification process figure that the present invention is provided;
Fig. 5 is the active power oscillations identification figure that the present invention is provided;
Fig. 6 is the active power trend identification figure that the present invention is provided.
Specific embodiment
The specific embodiment of the present invention is described in further detail below in conjunction with the accompanying drawings.
The following description and drawings fully illustrate specific embodiments of the present invention, to enable those skilled in the art to put into practice it
.Other embodiments can include structure, logic, it is electric, process and it is other changes.Embodiment only generation
The possible change of table.Unless explicitly requested, otherwise single component and function are optional, and the order for operating can change.
The part of some embodiments and feature can be included in or replace part and the feature of other embodiments.The enforcement of the present invention
The scope of scheme includes the gamut of claims, and all obtainable equivalent of claims.Herein,
These embodiments of the present invention individually or generally can be represented with term " invention " that this is used for the purpose of facilitating, and
And if in fact disclosing the invention more than, the scope for being not meant to automatically limit the application is any single invention or sends out
Bright design.
The present invention is based on electric network data, the vibration identification and Trend judgement problem for presence, using high-frequency data smoothing processing skill
Art, with reference to the experience and specific computational methods of the people in long-term calculating process, realizes the automatic identification of power trend and vibration,
With the purpose for reaching auxiliary power system signal trend, vibration identification and instructing actual motion.
Based on PMU measuring values data volume greatly, the characteristics of frequency is fast, the undulatory property in order to reduce data gives Trend judgement journey
Original data have been carried out smoothing processing by the huge puzzlement that sequence is brought first, obtain preprocessed data.And then carry for power
Corresponding trend and vibration recognizer are gone out.
Below pretreatment and power trend, vibration are recognized into totally 3 partial content, the careful explanation of piecemeal.
(1) server end real-time reception active power initial data;
(2) data prediction:
Process of data preprocessing before the identification of primary signal trend mainly to carrying out corresponding data processing.Main thought:Due to 1s
It is interior to produce the data that 50 numerical value is continually changing, data fluctuations phenomenon is caused, data fluctuations amplitude size is again by data type and product
Raw data environment affects, therefore takes data prediction means.Pretreatment stage adopts rolling average smoothing method, reaches data and puts down
It is sliding, extract the effect of overall trend.Its algorithm is comprised the following steps that:
1) data are taken average and are designated as 1 point by input data Dt per T spaced points, and are existed in A variables;
2) average treatment is moved using equation below to A, if translational movement is D;
Wherein:I is the subscript for having processed rear data B, j be it is cumulative and when loop parameter, T represents gap length;
3) each data point in B data is expanded into forward 24, extended value is the numerical value of current point.Expansion is finished and is designated as C changes
Amount, returns C.
It is as shown in Figure 2 that data prediction specifically performs flow process.
(3) active power oscillations identification:Mainly active power curves are carried out with the automatic identification of intellectual analysis and feature, including:
The detection of starting of oscillation point, vibration end point detection, change-detection of vibration trend etc., by adaptive algorithm wattful power is given
The identification feature of rate.
General active power monitoring signals have wave phenomenon, and the size of fluctuation amplitude is closely related with network system.Shake
Swing the wave phenomenon for referring to that power signal exists by a relatively large margin within the sampling period.Generally oscillation amplitude can be much larger than normal fluctuation width
Value.The correct detection that oscillator signal feature starts is to determine the key of starting of oscillation point, is also to determine the basis of vibration trend.
For the vibration identification of network re-active power need to determine the normal of active power in electrical network with reference to the concrete feature of current electric grid
Fluctuation threshold.Changing value size according to active power signal within the unit interval is compared with normal fluctuation threshold value, can sentence
The generation of disconnected oscillator signal.So as to identify starting of oscillation point and end point.When vibration occurs, given using envelope detected algorithm
Go out the dynamic change trend of hunting range.
The automatic identification of oscillation characteristicses is carried out using active power signal.Specific algorithm step is as follows:
1) initiation parameter threshold percentage Per, upper footnote recorded amounts U_ind, mark recorded amounts D_ind of getting a foothold;
2) input data, and carry out pretreatment;
3) coenvelope line is found:Continuous d data are chosen from the starting point of data.This section of maximum data and correspondence footnote are searched out,
In being respectively present array Up and U_ind, and remember that footnote is index.(if the maximum data of two and its above then chooses foot
The larger data of mark)
4) with index+1 as starting point, repeat step 3) and 4), until having inquired about all data, the searching of coenvelope line terminates;
5) lower envelope line is found:Continuous d data are chosen from the starting point of data.This section of minimum data and correspondence footnote are searched out,
In being respectively present array Down and D_ind, and remember footnote for index (if the minimum data of two and its above is then chosen
The less data of footnote);
6) with index+1 as starting point, repeat step 5) and 6).Until having inquired about all data, the searching of lower envelope line terminates;
7) linear interpolation is carried out to Up, Down, U_A, D_A is denoted as after interpolation;Linear interpolation algorithm is as follows:
For envelope data Up, the first two data point coordinate information (x1, y1) and (x2, y2) is taken first;Wherein x coordinate
Represent moment, the active power numerical value at y coordinate representation correspondence moment;If x2-x1 > 1, the point x between x1 to x2 is entered
Row linear interpolation obtains corresponding y values, as follows:
Y=(y2-y1)/(x2-x1) * (x-x1)+y1
Then again linear interpolation is carried out to (x2, y2) and (x3, y3), goes down finally obtain U_A successively;Same method, by
Lower envelope line number obtains D_A according to Down;
8) the coenvelope line number after interpolation processing is reused respectively according to U_A and lower envelope line number according to D_A and initial data
Algorithm is smoothed described in pre-treatment step, and the data for finally giving are designated as respectively Temp_U, Temp_D, M;
9) according to active power oscillations decision rule, the detection of active power oscillations state is carried out;Active power oscillations decision rule
For:Active power oscillations are judged to as min ((Temp_U-M), (M-Temp_D)) > M*Per, are unsatisfactory for, terminated;
Expand judged result, reach the purpose flux matched with former data.
Active power signal oscillating Automatic feature recognition result is shown in Fig. 5, and the wherein part of dotted line is envelope and smooth intermediate value,
The terminal of dotted line is starting over a little for vibration.If increasing threshold value, it is judged as that the region for vibrating narrows dotted line in i.e. Fig. 5
Envelope part can tail off;Threshold value is reduced, the starting point of vibration can accurately judge, while being likely to cause the erroneous judgement of state.
So under considering, arrange parameter value makes result as shown in Figure 5.
In specific implementation process, there are several parameters to influence whether the result of algorithm.
1) power normal fluctuation range threshold value.
When actually judging, threshold value is measured using the percentage ratio of the meansigma methodss after smoothing.Therefore, the size of percent value according to
Rely in the size of index itself.In general, Per values are less, and detecting the raising and lowering part come can be more, while also can
Detect more compared with minor swing interval extremely.When Per values are bigger, the abnormality for detecting is less, it is also possible to there is missing inspection and shows
As.Per value general recommendations value is between 0.005-0.01.
Active power oscillations differentiate that concrete execution flow process is as shown in Figure 3.
(4) active power trend identification:
Active power signal is in normal fluctuation range in more times in addition to it there is oscillatory occurences.In normal fluctuation model
In enclosing, active power signal has three kinds of variation tendencies:Normally, raising and lowering.Because trend-monitoring is to active power
The analysis of the overall trend change of signal, it is therefore desirable to carried out the smoothing processing of large scale to its signal before trend analysiss,
Eliminate the impact to power signal overall trend such as burr signal noise.Due to the power signal order of magnitude it is inconsistent, firstly, it is necessary to
Initial data is translated.Simultaneously as power signal sample frequency higher (50Hz), before overall trend change is judged,
It is 1 equivalent point that initial data is chosen per 25 points, then carries out smoothing processing to equivalent point.When Long-term change trend is judged, extract
The index of current data and front k continuous data carries out trend automatic identification.
The automatic identification of trend is changed using active power signal.Specific algorithm step is as follows:
1) initiation parameter:It is related to historical data fragment k, smooth spaces point D;Threshold percentage Per;Translational movement py;
2) input initial data is translated downwards and is smoothed and is designated as B.Smoothing method see data prediction step (2) and
(3);(notice that power pretreatment stage does not carry out step (4), but directly return it is smooth after the data that do not expand).
3) it is average according to being weighted to the front k of current data B (j), obtain indexWherein, weigh
Meet again w (j-k)≤w (j- (k-1))≤...≤w (j-1);Threshold value thre=C*Per;
4) automatic trend identification is carried out:
5) trend extension strategies:Power swing is judged as under abnormal conditions in step 4, and abnormality is reasonably extended.
To drop to example, by adjusting parameter (hupu_down) value, can reach on the premise of with general trend as decline, but cannot
Trend extension is carried out by the part that step 4 is accurately identified to decline.
6) data extending.Each data point in B data is expanded into forward 24, extended value is the numerical value of current point, expands shape
State is the state of current point.
It is as shown in Figure 4 that active power Trend judgement specifically performs flow process.
Active power signal trend automatic identification result is shown in Fig. 6, it is easy to see that the present invention makes to the abnormality for declining and its rise
Rational judgement.The previous segment data for being judged as downward trend is normal condition, is not combining future time data real-time judge
In, artificially also it is difficult to determine whether downward trend.Traditional method described in background technology can not necessarily reach the effect of the present invention
Really.
In specific implementation process, there are several parameters to influence whether the result of algorithm.
1) active power normal fluctuation range threshold value.
When actually judging, threshold value is measured using the percentage ratio of the meansigma methodss after smoothing.Therefore, the size of percent value according to
Rely in the size of data itself.In general, Per values are less, and detecting the raising and lowering part come can be more, while also can
Detect more compared with minor swing interval extremely.When Per values are bigger, the abnormality for detecting is less, it is also possible to there is missing inspection and shows
As.Per value general recommendations value is between 1~5.
2) smoothed data amount D
D initial data is extracted every time as the data in an interval, average treatment is moved to D data, to reach
To the purpose of smoothed data.The value size of D influences whether the accuracy and time delay of judged result.In general D takes
Value is bigger, and smooth effect is better, and the continuity of recognition result abnormal conditions is more preferable, while also bringing along bigger time delay.
The value of D is less, and smooth effect is poorer, and recognition result has more discrete anomalies intervals, but time delay can be less.
Therefore, the suggestion value of general D is 40 (35-50) left and right.
3) it is related to historical data fragment k
I.e. with reference to length is judged, the parameter main purpose is, in order to calculate the size cases of historical data, to be easy to calculate to historical data
The change size cases of current data and historical data.In general, the value of k is 10 or so proper.
4) selection of shift value size
Because the size of the possible corresponding data of difference units has obvious difference, in order to reduce the data volume size and threshold of data
The interference that value judges, before pretreatment is carried out to data, carries out integral translation, the general root of size of translational movement in data first
Chosen according to historical data, the size of translational movement typically chooses the minima being likely to be breached during correspondence unit normal work.
Above example is only to illustrate technical scheme rather than a limitation, although reference above-described embodiment is to the present invention
Be described in detail, those of ordinary skill in the art still can to the present invention specific embodiment modify or
Person's equivalent, these any modifications or equivalent without departing from spirit and scope of the invention are applying for pending this
Within bright claims.
Claims (8)
1. a kind of automatic power trend recognition methodss, it is characterised in that methods described comprises the steps:
(1) server end real-time reception active power initial data;
(2) initial data pretreatment;
(3) active power oscillations state recognition;
(4) active power trend identification.
2. automatically power trend recognition methodss as claimed in claim 1, it is characterised in that right in the step (2)
Initial data pretreatment adopts rolling average smooth manner, reaches data smoothing, extracts active power variation tendency;It is described original
Data prediction comprises the steps:
1) meansigma methodss are extracted at equal intervals to initial data, including:Input initial data Dt, the initial data Dt is individual per T
Point extracts its meansigma methods, is designated as 1 point, and exists in A variables;The A variables include A (1), A (2), A (3), A (4) ...;
The A (1), A (2), A (3), A (4) ... respectively Dt (1:T)、Dt(T+1:2T)、Dt(2T+1:3T)、Dt(3T+1:4T)......;
2) average data to extracting moves average treatment, to reach the purpose of smoothed data, including:To A variables
Average treatment is moved using equation below;
Wherein:To have processed rear data, i represents data footnote parameter to B, and j is the loop parameter of cumulative sum, and D is represented between translation
Every length;
The length for making data A is Len, and N divides exactly the result of D for Len, then the span of footnote parameter i is 1,2 ..., N;
3) data after gliding smoothing are expanded, including:Each data point in B data is expanded into forward 24 numerical value,
Extended value is the numerical value of current point, and expansion is finished and is designated as C variables;The C variables include C (1:T)、C(T+1:2T)、
C(2T+1:3T)、C(3T+1:4T).......
3. automatically power trend recognition methodss as claimed in claim 1, it is characterised in that under the step (3) includes
State step:
1. according to different unit features, following parameter is initialized, including:
A, active power oscillations threshold percentage Per;The parameter is used to adjust the sensitivity of concussion feature identification, and parameter value is bigger
Then shake identification region less, on the contrary it is bigger then to shake identification region;
B, upper footnote recorded amounts U_ind;The footnote position of all maximum extreme points in for recording initial data;
C, mark recorded amounts D_ind of getting a foothold;The footnote position of all minimum extreme points in for recording initial data;
2. using pre-treatment step 1 in 2) and method 2) smooth pretreatment is carried out to active power initial data, return and do not expand
The pretreated active power data filled;
3. find oscillatory regime when coenvelope line Up and lower envelope line Down;
4. the coenvelope line number to finding carries out linear interpolation according to Up and lower envelope line number according to Down, obtains data U_A after interpolation
And D_A;Linear interpolation algorithm is as follows:
For envelope data Up, the first two data point coordinate information (x1, y1) and (x2, y2) is taken first;Wherein x coordinate
Represent moment, the active power numerical value at y coordinate representation correspondence moment;If x2-x1 > 1, the point x between x1 to x2 is entered
Row linear interpolation obtains corresponding y values, as follows:
Y=(y2-y1)/(x2-x1) * (x-x1)+y1
Then again linear interpolation is carried out to (x2, y2) and (x3, y3), goes down finally obtain U_A successively;Same method,
D_A is obtained by lower envelope line number according to Down;
5. the coenvelope line number after interpolation processing is reused respectively according to U_A and lower envelope line number according to D_A and initial data
Algorithm is smoothed described in pre-treatment step, and the data for finally giving are designated as respectively Temp_U, Temp_D, M;
6. according to active power oscillations decision rule, the detection of active power oscillations state is carried out;Active power oscillations decision rule
For:Active power oscillations are judged to as min ((Temp_U-M), (M-Temp_D)) > M*Per, are unsatisfactory for, terminated;
7. expand judged result, reach the purpose flux matched with former data.
4. automatically power trend recognition methodss as claimed in claim 3, it is characterised in that the active power oscillations threshold value
Percentage ratio Per values are between 0.005-0.01.
5. automatically power trend recognition methodss as claimed in claim 3, it is characterised in that find upper bag during oscillatory regime
Winding thread includes:A () chooses continuous d data from the starting point of active power data, maximum data in d data of searching and right
Footnote should be gone up, if the maximum data of two and its above then chooses the big data of upper footnote, by maximum data and the upper foot of correspondence
Mark is respectively present in array Up and U_ind, and remembers that footnote is variable i ndex;(b) with index+1 as starting point, repeat step
A () and (b), until having inquired about all data, coenvelope line during oscillatory regime is found and is terminated;
Lower envelope line when finding oscillatory regime includes:C () chooses continuous d data from the starting point of data, search out d numbers
According to middle minimum data and the lower footnote of correspondence, if the minimum data of two and its above then chooses the little data of lower footnote, will most
Small data and the lower footnote of correspondence are respectively present in array Down and D_ind, and remember that footnote is index ' ();D () is with index '+1
For starting point, repeat step (c) and (d), until having inquired about all data, the searching of lower envelope line during oscillatory regime terminates.
6. automatically power trend recognition methodss as claimed in claim 1, it is characterised in that under the step (4) includes
State step:
<1>According to different unit features, initiation parameter, including historical data fragment k, smoothed data spacing parameter D, have
Work(power trend threshold percentage Per and translational movement py;
<2>Receive 25*D untreated active power signal sampling value, using the step of preprocess method 1) and 2) to having
Work(power data is translated downwards and is smoothed and is designated as B, returns the data not expanded;
<3>It is average according to being weighted to the front k of current data B (j) according to active power preprocessed data, obtain comprehensively referring to
MarkWherein, weight meet w (j-k)≤w (j- (k-1))≤...≤w (j-1), threshold value
Thre=C*Per;J is currently processed data segment subscript, and m is loop parameter, and w (j-m) is corresponding weight parameter, B (k-m+1)
Represent the value of lower target array B of correspondence;
<4>With reference to the result of calculation of Trend judgement aggregative indicator, using Trend judgement rule, the inspection of active power Trend judgement is carried out
Survey;
<5>According to preliminary judged result, if active power trend is ascendant trend or downward trend, trend extension strategies are performed;
<6>Expand judged result, each data point in B data is expanded into forward 24, extended value is the numerical value of current point, expands
Fill the state that state is current point.
7. automatically power trend recognition methodss as claimed in claim 6, it is characterised in that the step<4>In, it is active
The detection of power Trend judgement includes:
Trend judgement rule abs (B (j)-C) > thre are judged, if meet to continue to judge B (j) < C Down, if meet being determined with
Work(power trend is ascendant trend, is otherwise downward trend;If abs (B (j)-C) < thre, active power Trend Stationary is judged,
Proceed next step, wherein Down represents lower envelope line number evidence;Thre represents threshold value;C is that the data after gliding smoothing are entered
The variable that row expands after finishing;B (j) represents current data.
8. automatically power trend recognition methodss as claimed in claim 6, it is characterised in that historical data fragment k takes
Value scope is 8~12, and smoothed data amount D span is 35~50;Active power trend threshold percentage Per value is 1~5
Between;The minima for being chosen for being reached during correspondence unit normal work of translational movement py values.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510673427.2A CN106600449B (en) | 2015-10-16 | 2015-10-16 | Automatic power trend identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510673427.2A CN106600449B (en) | 2015-10-16 | 2015-10-16 | Automatic power trend identification method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106600449A true CN106600449A (en) | 2017-04-26 |
CN106600449B CN106600449B (en) | 2021-03-16 |
Family
ID=58553766
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510673427.2A Active CN106600449B (en) | 2015-10-16 | 2015-10-16 | Automatic power trend identification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106600449B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110489810A (en) * | 2019-07-24 | 2019-11-22 | 西安理工大学 | A kind of trend extraction method based on data block |
CN113075180A (en) * | 2021-03-24 | 2021-07-06 | 临海市鸥巡电子科技有限公司 | Method and system for detecting change trend of fluorescence data |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1314054A (en) * | 1998-06-23 | 2001-09-19 | 汤姆森许可公司 | Inter polation method for smooth conveygence correction |
CN1968228A (en) * | 2006-11-23 | 2007-05-23 | 中兴通讯股份有限公司 | Waveform recognition method in signal processing |
US20140297606A1 (en) * | 2013-03-28 | 2014-10-02 | Nec (China) Co., Ltd. | Method and device for processing a time sequence based on dimensionality reduction |
CN104485666A (en) * | 2014-12-22 | 2015-04-01 | 大连理工大学 | Trend recognition method of total regulation power curve in automatic power generation control |
-
2015
- 2015-10-16 CN CN201510673427.2A patent/CN106600449B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1314054A (en) * | 1998-06-23 | 2001-09-19 | 汤姆森许可公司 | Inter polation method for smooth conveygence correction |
CN1968228A (en) * | 2006-11-23 | 2007-05-23 | 中兴通讯股份有限公司 | Waveform recognition method in signal processing |
US20140297606A1 (en) * | 2013-03-28 | 2014-10-02 | Nec (China) Co., Ltd. | Method and device for processing a time sequence based on dimensionality reduction |
CN104485666A (en) * | 2014-12-22 | 2015-04-01 | 大连理工大学 | Trend recognition method of total regulation power curve in automatic power generation control |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110489810A (en) * | 2019-07-24 | 2019-11-22 | 西安理工大学 | A kind of trend extraction method based on data block |
CN110489810B (en) * | 2019-07-24 | 2022-11-25 | 上海金润联汇数字科技有限公司 | Automatic trend extraction method based on data blocks |
CN113075180A (en) * | 2021-03-24 | 2021-07-06 | 临海市鸥巡电子科技有限公司 | Method and system for detecting change trend of fluorescence data |
Also Published As
Publication number | Publication date |
---|---|
CN106600449B (en) | 2021-03-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110311376B (en) | Dynamic safety assessment comprehensive model and space-time visualization method for power system | |
CN110718910B (en) | Transient stability evaluation method for Bayesian optimization LightGBM | |
CN104091070B (en) | Rail transit fault diagnosis method and system based on time series analysis | |
CN103903408B (en) | Method for early warning and system are investigated in equipment fault | |
CN106597136B (en) | A kind of warping apparatus intelligent polling method based on Intelligent Mobile Robot | |
CN106809251B (en) | The monitoring of magnetic suspension train rail irregularity and prediction technique, device and system | |
CN102163261A (en) | Case-reasoning-based molten steel temperature prediction method | |
CN109209790A (en) | A kind of wind electricity blade icing estimation method based on data modeling | |
CN105932774A (en) | Device state early warning method in smart transformer substation based on ICA algorithm | |
CN103323702A (en) | Composite power quality disturbing signal identifying method | |
CN102841131A (en) | Intelligent steel cord conveyer belt defect identification method and intelligent steel cord conveyer belt defect identification system | |
CN108876163A (en) | The transient rotor angle stability fast evaluation method of comprehensive causality analysis and machine learning | |
CN111898669A (en) | Machine learning-based direct-current submerged arc furnace abnormal event early warning system | |
Zhao et al. | On-line least squares support vector machine algorithm in gas prediction | |
CN111628494A (en) | Low-voltage distribution network topology identification method and system based on logistic regression method | |
CN105005822A (en) | Optimal step length and dynamic model selection based ultrahigh arch dam response prediction method | |
CN106600449A (en) | Automatic power trend recognition method | |
CN114429189A (en) | High-universality landslide displacement rate early warning method | |
CN105974223B (en) | A kind of method and system for on-line checking electrical equipment working condition | |
CN108168924A (en) | A kind of reciprocating compressor life-span prediction method based on VMD and MFSS models | |
CN110108979A (en) | The recognition methods of transmission line lightning stroke flashover and non-flashover based on OPGW | |
CN113641486A (en) | Intelligent turnout fault diagnosis method based on edge computing network architecture | |
CN110378624A (en) | Fired power generating unit primary frequency modulation index calculating method and system based on trend abstraction | |
Tiwari | Wind speed forecasting methods for wind energy generation | |
CN104517034B (en) | Fired power generating unit return passage identification Method and system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |