CN111144705B - Whole-network same-section data processing method based on information acquisition with time scale - Google Patents
Whole-network same-section data processing method based on information acquisition with time scale Download PDFInfo
- Publication number
- CN111144705B CN111144705B CN201911231245.4A CN201911231245A CN111144705B CN 111144705 B CN111144705 B CN 111144705B CN 201911231245 A CN201911231245 A CN 201911231245A CN 111144705 B CN111144705 B CN 111144705B
- Authority
- CN
- China
- Prior art keywords
- data
- points
- curve
- point
- time
- 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.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims description 10
- 238000000034 method Methods 0.000 claims abstract description 63
- 238000012545 processing Methods 0.000 claims abstract description 33
- 238000004364 calculation method Methods 0.000 claims abstract description 11
- 239000013598 vector Substances 0.000 claims description 33
- 238000005070 sampling Methods 0.000 claims description 20
- 230000008569 process Effects 0.000 claims description 10
- 230000015572 biosynthetic process Effects 0.000 claims description 7
- 238000005259 measurement Methods 0.000 claims description 7
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000007670 refining Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 3
- 238000013500 data storage Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 description 8
- 230000008859 change Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 238000013524 data verification Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06312—Adjustment or analysis of established resource schedule, e.g. resource or task levelling, or dynamic rescheduling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Complex Calculations (AREA)
Abstract
The method divides the data processing of the same section into two directions of a transformer substation and a dispatching master station, and the data processing of the same section in the transformer substation provides a basis for the dispatching master station to form a same-section database. In a station, a maximum value and minimum value calculation method of the acquisition amount under normal operation is given according to historical data of an acquisition device, and an improved cubic B-spline fitting algorithm is provided for curvilinearly processing the acquisition data aiming at the acquisition amount which does not exceed the range. And intercepting fitting curve data of the same time point by selecting a time section and setting a time interval, storing to obtain data of the same section in the plant station, and uploading the data of the same section to a dispatching master station by a plurality of plant stations to obtain information of the same section of the whole station. The method can improve the accuracy and the common-section property of the power grid data, form a data chain with the same section for the whole power grid, and improve the availability of the data.
Description
Technical Field
The invention belongs to the field of automation information of power systems, and particularly relates to a full-network same-section data processing method based on information acquisition with time scales.
Background
With the automation of the power grid and the improvement of the informatization level, the requirements of the power grid on the accuracy and the real-time performance of the power data are higher and higher. After the data collected by the power system is uploaded to the master station, the collected data does not have timeliness, the time uniformity of each plant station is not high, the time and section of the data are low, and the data cannot be directly applied to a power grid. At present, technical research on processing collected power grid data into data with the same section is few, and a state estimation method for power system data verification can only identify bad data, but because time references of collection points of various stations may be different, the accuracy of the whole evaluation value is not high due to the used data property, and a final processing result cannot form a data chain with the same section of the whole network, so that the data availability is low.
Disclosure of Invention
The invention aims to solve the technical problems and provides a whole-network same-section data processing method based on time scale information acquisition, which performs weighted interpolation processing on discrete time data of data acquired in a station and a main station to obtain continuous time data, and on the basis, the influence of time synchronization precision on the time section is considered, and a proper time section and a proper time interval are selected to form whole-network same-section data, so that the accuracy and the same section of power grid data are greatly improved, a whole-network same-section data chain is formed for a power grid, and the availability of the data is improved.
The technical solution of the invention is as follows:
a whole network same-section data processing method based on information acquisition with time marks is characterized by comprising the following steps:
step 1: a whole network same-section data processing method based on time scale information acquisition is characterized in that the implementation process is divided into two stages, the first stage is the formation of same-section data in a transformer substation, and the second stage is the formation of whole network same-section data between transformer substations;
firstly, a plurality of data acquisition measuring points exist in a station, meanwhile, an alternating current sampling device has a time check function, the time of the acquisition points is ensured to be synchronous with a main clock of the station, the acquired data are marked with acquisition time scales, and the sampled data with the time scales are sent to a background server of the station for data storage and processing;
step 2: determining the maximum value and the minimum value limit of the collected data quantity according to historical operating data of the collection point, including historical data within the last year; setting the upper limit of the normal operation interval of the collection amount as X max The lower limit of the normal operation interval is X min According to the maximum value and the minimum value of the collection amount within the last 12 months, the upper limit and the lower limit of the normal operation interval are calculated by using a weighting algorithm, and the calculation formula is as follows:
where i is the ith month, k, closest to the sampling time i Is the ith month upper limit value weight factor, h i Is the weight factor of the ith month floor value, x imax Maximum value of collected X at month i, X imin Is the minimum value of the collected amount X at month i; when the acquisition amount exceeds the upper limit and the lower limit of operation, the acquired data exceeding the limit does not participate in the subsequent operation process;
and step 3: dividing the collected data into two situations according to the reason that the collected data exceeds the normal operation range; one is that two points exceed the normal operating range after the single point, the number of erroneous acquisition points is very small, the situation is inferred to be caused by the measurement error of the acquisition device, and the processing mode is to delete data; the other is that the acquisition amount exceeds the normal operation range when multiple points occur, the error acquisition points are more than two, the condition is inferred that an accident occurs, the acquisition amount exceeds the normal operation range, and at the moment, the processing mode is to start fault filtering to record related operation data so as to facilitate the follow-up analysis of accident reasons;
and 4, step 4: carrying out interpolation fitting curve processing on the collected data in the normal operation range, and changing discrete collected data points into a continuous smooth time curve; the method comprises the steps of calculating the curvature of discrete points, setting a reasonable curvature threshold value to extract characteristic points, carrying out chord length parameterization on the characteristic points, constructing node vectors according to the parameters of the characteristic points, and solving a control vertex by using a least square method to fit discrete acquisition quantity data;
step 4.1: at a given acquisition quantity data point vector P k Wherein k = (0,1,2, … n, composition [ A, B]In the interval, every 4 points are sequentially connected into a polygon called a B characteristic polygon, and a fitting curve formed by fitting the B characteristic polygon by utilizing a cubic spline function is a cubic B spline curveA wire;
the matrix representation of the cubic B-spline curve is in the form:
the expression for the ith cubic B-spline is therefore expressed as:
wherein t is a parameter, P i ,P i+1 ,P i+2 ,P i+3 (i =0,1,2, …, n) as four vertices of a feature polygon; take j =3,N j,3 (t) is a basis function of the cubic B-spline curve, and a recurrence formula is as follows:
because the four basic functions are positioned in the same node interval, a finished B spline can be spliced;
and 4.2: in the actual fitting process of the sampled data, data points are often more, and if all the data participate in fitting, the fitting efficiency is possibly low, and the iteration times are more; the extraction of characteristic points of the curve is called as a key, and the key has a direct effect on the shape of the fitted curve; the curvature reflects the overall and local shape characteristics of the curve, and a method for selecting characteristic points by calculating the curvature is adopted;
for a cubic spline B (t), the parameter value t is obtained by differentiation i The curvature of (b) is calculated by the formula:
wherein B' (t) i ) Is the first derivative of curve B (t), B' (t) i ) Is the second derivative of curve B (t), k i Is a curvature;
step 4.3: extracting characteristic points, and obtaining curvature K = (K) of discrete points by using 0 ,k 1 ,…,k n ) The principle of extracting the feature points is as follows:
(1) For a curve that is not closed, its two endpoint values must be selected;
(2) Determining a set curvature threshold value selection characteristic point, and setting a point with the curvature larger than the threshold value as an initial characteristic point; the mean value of curvature is K avg The curvature threshold is set to K ths ＝α×K avg Alpha is a proportionality coefficient;
the selection requirement of the threshold value can not only reflect the overall shape of the curve, but also reduce the number of characteristic points as much as possible; when the address selection of the threshold is smaller, more control points are possibly caused; when the threshold value is too large, the overall shape of the curve may not be represented;
step 4.4: in the step, a chord length parameter method is adopted to carry out characteristic point parameterization, and a data point Q 0 ,Q 1 ,…Q n The parameter field t is the same as [0,1 ∈ ]]The nodes between the nodes have one-to-one correspondence; let d total chord length:
then
Wherein Q k -Q k-1 Is a chord edge vector; the parameterized method truly reflects the condition of data points of chord length distribution, and overcomes the problem of processing the condition of uneven chord length distribution;
step 4.5: constructing a node vector, namely constructing the node vector by adopting an average value method, and evenly distributing the node vector, namely:
t 0 ＝t 1 ＝…t p ＝0
t m-p ＝t m-p-1 ＝…＝t m ＝1
m +1 represents the total number of nodes, both ends of the node vector are respectively provided with p +1 same nodes so as to control the tangency of the head end point and the tail end point, j represents the ordinal number of the middle node, and the node vector can well reflect the parameter point vector by adopting an average value method
The distribution of (c);
step 4.6: fitting a least square cubic B spline curve, and constructing a cubic B spline curve approximate to the characteristic points by using a node vector and a least square method; as can be seen from the cubic B-spline expression, the feature point D is approximated in the least-squares sense, i.e.:
wherein D 0 ＝B(0)，D r = B (1), r is the number of feature points;
order to
f is with respect to n-1 variables P 1 ,…P n-1 A scalar value function of; to minimize f, let f pair control points P l The partial derivative of (a) is zero, the following equation is obtained:
this is an unknown quantity P 1 ,…,P n-1 Let l =1,2, …, n-1 then obtain a linear system of equations with n-1 unknowns and n-1 equations:
(N T N)P＝R
solving the linear equation set can obtain the value of the unknown quantity P, so that a fitting curve of cubic B-spline can be determined;
and 5: the fitting curve is refined, the curve only fitted by the characteristic points can not meet the requirement of approximate allowable error generally, and the number of the characteristic points needs to be increased in order to obtain a high-quality curve; calculating the deviation between the data point and the initial fitting curve, taking the point with the deviation exceeding a set deviation threshold value as a characteristic point, and inserting the characteristic point into the initial fitting curve for fitting again;
step 5.1: calculating approximation deviation, and calculating approximation deviation between a data point and a curve by adopting Hausdorff distance; suppose two sets of data sets a = { a = { (a) 1 ,a 2 ,…,a p }，B＝{b 1 ,b 2 ,…b q And the Hausdorff distance between the two point sets of A and B is:
H(A,B)＝max(h(A,B),h(B,A))
where | · | | is a distance model between the point set a and the point set B, such as a euclidean distance, H (a, B) is a bidirectional Hausdorff distance, and H (a, B) and H (B, a) are unidirectional Hausdorff distances from the set a to the set B and from the set B to the set a, respectively; the two-way Hausdorff distance H (A, B), which is the greater of the two singleton distances H (A, B) and H (B, A), measures the maximum degree of mismatch between the two sets of points;
step 5.2: refining the curve locally, calculating the deviation value between the initial curve and all corresponding acquisition points, and setting the deviation thresholdValue e ths Finding out all maximum value points exceeding the deviation threshold, and the specific operation is as follows:
(1) Calculating to obtain the approximation deviation of the data point and the curve according to a Hausdorff distance formula
(2) Solving the maximum approximation deviation h (Q, B);
(3) Setting a deviation threshold e ths = β × h (Q, B), β being a proportionality coefficient;
(4) Extracting all the deviation values larger than the deviation threshold value e on the deviation curve ths A point of (a);
(5) Re-fitting the points as new feature points, wherein one-time increased key point between two initial key points cannot exceed one;
(6) Repeating the steps until the fitting precision is met, and finishing the calculation;
since the cubic B-spline has locality, the newly inserted feature points can change the local knowledge quantity of the fitting curve without influencing the whole body; the fitting method greatly reduces the number of control points and improves the calculation efficiency on the premise of ensuring the precision;
step 5.3: and (5) evaluating the fitting precision, and after extracting the characteristic points of the previous stage, obtaining two adjacent characteristic points which are respectively (x) k ,y k ) And (x) k+1 ,y k+1 ) The point (x) on the curve segment between which the deviation from the two characteristic points is the greatest i ,y i ) By calculating the h value of a feature point, i.e. point (x) i ,y i ) To point (x) k ,y k ) And (x) k+1 ,y k+1 ) The vertical distance of the connected straight lines is calculated by the following formula:
calculating all h values except the end points; will be provided with
The integral square error is used as an index for measuring the fitting accuracy of the curve, so that the original shape of the curve can be ensured;
step 6: after the acquired data in the plant station is subjected to cubic B spline fitting, selecting an initial time section t 0 And a time interval Δ t; the sampling data is data with time mark, and the relation of the sampling quantity X with respect to time is set as X i (t), wherein i is the ith collection quantity number, and the collection quantity X is the operation parameters of the power grid such as voltage, current and the like; the acquired data becomes a continuous curve at discrete points after fitting, and X at any time can be obtained i (t) numerical value, cross-section through time t 0 And determining the time interval delta t, finally forming the time t in the plant station 0 ,t 0+Δt ,t 0+2Δt ,…t 0+nΔt All the acquired section data at the moment point are stored according to the time axis as scales to obtain a same-section database in the station;
and 7: different stations obtain databases of the same section of each station according to the same processing mode of the acquired data; and then, each plant station uploads the data of the same cross section to a scheduling master station, and a full-network database of the same cross section is formed at the scheduling master station. The same-section database is established, so that the accuracy of the real-time data of the power grid is improved, and reliable data support is provided for power grid dispatching operation decision.
The invention has the beneficial effects that:
the invention divides the formation of the database with the same section into two stages, and provides a calculation method of the collection quantity operation limit value and a processing method of the data exceeding the limit value by analyzing the normal operation interval limit value of the collected data with the time mark, thereby improving the time-same section property and the data processing accuracy of the data; a least square method is adopted to carry out cubic B spline fitting to change discrete power acquisition data into a continuous time curve, and approximation deviation and a deviation threshold value are introduced to achieve fitting curve refinement, so that the number of control points is greatly reduced on the premise of ensuring precision, and the calculation efficiency is improved; finally, a whole-network same-section database is established on the side of the dispatching master station, so that the accuracy of the real-time data of the power grid is improved, data support is provided for the analysis of advanced application software of the power grid and the dispatching operation decision of the power grid, and the application level of the real-time information of the power grid is improved; the method is suitable for supporting each system by high-precision electric power real-time same-section data, and has an important reference for construction of a digital and information intelligent power grid.
Drawings
FIG. 1 is a flow chart of a full-network co-section data processing method of the present invention;
FIG. 2 is a least squares curve fit plot of the present invention;
FIG. 3 is a Bezier curve fit of the present invention;
FIG. 4 is a graph of a Lagrangian fit of the present invention;
FIG. 5 is a graph of a cubic B-spline curve fit according to the present invention;
FIG. 6 is a graph of the results of a cubic B-spline initial fit of a period of sampled data according to the present invention;
FIG. 7 is a graph of a cubic B-spline multiple optimization fit of one cycle of sampled data according to the present invention;
Detailed Description
As shown in fig. 1 to 7, the method for processing the data of the whole network and the same section based on the information acquisition with the time scale includes the following steps:
step 1: a whole-network same-section data processing method based on time scale information acquisition is characterized in that the implementation process is divided into two stages, the first stage is the formation of same-section data in a transformer substation, and the second stage is the formation of same-section data between the transformer substations.
Firstly, a plurality of data acquisition measuring points exist in a station, meanwhile, the alternating current sampling device has a time check function, the time of the acquisition points is ensured to be synchronous with a main clock of the station, the acquired data are marked with the time marks of the acquisition time, and the sampled data with the time marks are sent to a background server of the station.
The method is verified by selecting a certain 220kV transformer substation as an example, the in-station time synchronization device adopts a Beidou/GPS double time-alignment mode, the time synchronization of all devices in the station is ensured to meet the measurement requirement of on-off surface data, the voltage of the high-voltage side of the transformer in the station is selected, and the measurement data of the sampling device is read when the transformer substation normally operates. The sampling frequency of the acquisition device is 80 points/period, and 80 sampling data in one period of 0.02s are taken as an example for analysis. Data collected are shown in table 1:
table 1:220kV station one-period voltage sampling data
Step 2: determining the maximum value and the minimum value limit of the collected data quantity according to historical operating data of the collection point, including historical data within the last year; the upper limit of the normal operation interval of the collection amount is set as X max The lower limit of the normal operation interval is X min According to the maximum value and the minimum value of the collection amount within the last 12 months, the upper limit and the lower limit of the normal operation interval are calculated by a weighting algorithm, and the calculation formula is as follows:
where i is the ith month, k, closest to the sampling time i Is the ith month upper limit value weight factor, h i As the weight factor x of the ith month lower limit imax Maximum value of collected X at month i, X imin Minimum value of X collected at month i. When the collection amount exceeds the upper limit and the lower limit of operation, the collected data exceeding the limit does not participate in the subsequent operation process. The last year of voltage operation upper limit and lower limit are U through big data analysis max ＝245kV，U min ＝-245kV。
And step 3: dividing the collected data into two situations according to the fact that the collected data exceed the operation range; one is that two points exceed the normal operation range after the single point, the number of erroneous acquisition points is very small, and the situation is inferred to be caused by the measurement error of the acquisition device, and the processing mode is to delete data; the other is that the acquisition amount exceeds the normal operation range at multiple points, the error acquisition points are more than two, and the condition is concluded that an accident occurs, so that the acquisition amount exceeds the normal operation range; at the moment, the processing mode is to start fault filtering to record relevant operation data, so that accident reasons can be conveniently analyzed subsequently. According to the comparison between the voltage data of the sampling points and the upper and lower operation limits, the serial numbers of the abnormal points are 34 and 75 (see table 2), which are caused by measurement errors, and the data of the two points are removed.
Table 2: outlier data
And 4, step 4: and carrying out interpolation fitting curve processing on the acquired data in the normal operation range, and changing discrete acquired data points into a continuous smooth time curve. The method comprises the steps of calculating the curvature of discrete points, setting a reasonable curvature threshold value to extract characteristic points, carrying out chord length parameterization on the characteristic points, constructing node vectors according to the parameters of the characteristic points, and solving a control vertex by using a least square method to fit discrete acquisition quantity data.
Step 4.1: the B spline curve is utilized to process the collection amount, a fitting mode of the B spline curve usually adopts k +1 vertexes to define a k-th-order polynomial, and the mathematical expression of the k-th-order polynomial is as follows:
in the formula P i Is a position vector of each vertex, N i,k Is a Bernstein-basis function, called a k-th order (k-1) th-order B-spline basis function, each of which is called a B-spline. To be counted atThe definition of the spline function is effectively realized in a computer, and a recursion formula is defined by adopting the following expression:
the following table i indicates the number and k the number. The recursion formula shows that t is needed to determine the ith k-th B spline i ,t i+1 ,…,t i+k+1 Total k +2 nodes, interval [ t ] i ,t i+k ]Is N i,k (t) a support interval. In the curve equation, with n +1 control vertices P i (i =0,1,2, …, N) corresponding to N +1 k-th-order B-spline basis functions N i,k (t), (i =0,1,2, …, n). Therefore, the high-order B-spline can be derived on the basis of the low-order B-spline.
The cubic B-spline fitting effect is illustrated by using discrete data points (see Table 3), and curve fitting comparative analysis is performed by using a least square method, a Bezier method, a Lagrangian method and a cubic B-spline method respectively.
Table 3: discrete point data
The results of fitting the data curves are shown in the attached drawings 2-5, and the least square method reflects the general trend of data points and eliminates local fluctuation by comparing and analyzing the fitting methods of the curves, but the fitting effect of the least square method on the discrete points reflecting the key change of the linear trend is not ideal; the Bezier curve fits the whole trend of the intelligent response curve, and the local fitting effect on the curve is not ideal; the Lagrange method has the Longge phenomenon when more data points exist, and the fitting error is larger; the cubic B-spline fitting overcomes the defects of the Bezier method, keeps some points, only two adjacent curve line segments can be changed by changing the position of a certain control point, the shapes of other curve line segments are not influenced, the local fitting effect is good, the fitting curve is smoother, and the fitting method is more suitable for fitting the station sampling data with more processing data.
Step 4.2: at a given acquisition quantity data point vector P k Wherein k = (0,1,2, …, n) for [ A, B]In the interval, every 4 points are connected into a polygon in sequence, and the polygon is called a B-feature polygon. And fitting the B characteristic polygon by utilizing a cubic spline function to form a fitting curve which is a cubic B spline curve.
The matrix representation of the cubic B-spline curve is in the form:
the expression for the ith cubic B-spline can be expressed as:
wherein t is a parameter, P i ,P i+1 ,P i+2 ,P i+3 (i =0,1,2, …, n) as four vertices adjacent to the feature polygon; take j =3,N j,3 (t) is the basis function of the cubic B-spline curve, and the recursion formula is as follows:
because the four basic functions are in the same node interval, the four basic functions can be spliced into a finished B spline.
Step 4.3: the feature points occupy important positions in the accuracy of the data fitting, which has a direct effect on the shape of the fitted curve. The curvature reflects the overall and local shape characteristics of the curve, and a method for selecting characteristic points by calculating the curvature is adopted.
For a cubic spline B (t), the parameter value t is obtained by differentiation i The curvature of (b) is calculated by the formula:
wherein B' (t) i ) Is the first derivative of curve B (t), B' (t) i ) Is the second derivative of curve B (t), k i Is a curvature.
Step 4.4: in the actual process of fitting the sampled data, there are often many data points, and if all the data participate in fitting, the fitting efficiency may be low and the number of iterations may be many. Therefore, it is very important to use fewer points to achieve the purpose of curve fitting on the premise of ensuring the accuracy, and the extraction of characteristic points of the curve is called as a key.
Curvature K = (K) of discrete points obtained by application 0 ,k 1 ,…,k n ) The principle of extracting the graph is as follows:
(1) For curves that are not closed, the two endpoint values must be selected.
(2) And determining a set curvature threshold value selection characteristic point, and setting a point with the curvature larger than the threshold value as an initial characteristic point. Mean value of curvature K avg The curvature threshold is set to K ths ＝α×K avg And (alpha is a proportionality coefficient)
The selection of the threshold is crucial, and the requirement is that the overall shape of the curve can be reflected, and the number of the characteristic points is as small as possible. When the address selection of the threshold is small, more control points are possibly caused; if the threshold value is too large, the entire shape of the curve may not be expressed.
According to the sampling data in a period, the curvature threshold value is 0.18, 16 selected feature points are selected, and the table 4 extracts the feature points for the sampling data.
Table 4: sampling data extraction feature point
It should be noted that, since the selection of the threshold causes the occurrence of an approximation error, the curve is subsequently locally optimized to achieve accurate fitting of the curve.
Step 4.5: in the step, a chord length parameter method is adopted to carry out feature point parameterization, and a data point Q 0 ,Q 1 ,…Q n The parameter field t is the same as [0,1 ∈ ]]There is a one-to-one correspondence between the nodes. Let d total chord length:
then
Wherein Q k -Q k-1 The vector of the chord edge is taken as the vector of the chord edge,
the method is a parameter point vector, truly reflects the condition of data points of chord length distribution, and overcomes the problem of processing the condition of uneven chord length distribution.
Step 4.6: constructing node vectors, namely constructing the node vectors by adopting an average value method, and evenly distributing the node vectors, wherein the node vectors are [ t ] 0 ,t 1 ,…,t p ,…,t m-p ,t m-p+1 ,…,t m ]Wherein:
t 0 ＝t 1 ＝…t p ＝0
t m-p ＝t m-p+1 ＝…＝t m ＝1
m +1 represents the total number of nodes, both ends of the node vector are respectively provided with p +1 same nodes so as to control the tangency of the head end point and the tail end point, j represents the ordinal number of the middle node, and the node vector can well reflect by adopting an average value method
(parameter point vector)Distribution of (2).
Step 4.7: and fitting the least square cubic B-spline curve, and constructing a cubic B-spline curve approximating the characteristic points by using a node vector and a least square method. As can be seen from the cubic B-spline expression, the feature point D is approximated in the least-squares sense, i.e.:
wherein D 0 ＝B(0)，D r = B (1), r is the number of feature points. Order:
f is about n-1 variables P 1 ,…P n-1 A scalar value function of. To minimize f, let f pair control points P l The partial derivative of (c) is zero.
Namely:
this is an unknown quantity P 1 ,…,P n-1 Let l =1,2, …, n-1 then obtain a system of line equations with n-1 unknowns and n-1 equations:
(N T N)P＝R
the value of the unknown quantity P can be solved by solving the system of linear equations, so that a fitting curve of a cubic B-spline can be determined. Extracting the characteristic points according to the step 4.4, and performing initial cubic B-spline curve fitting, wherein the results of the cubic B-spline initial fitting of the sampling data in one period are shown in an attached figure 6. The shape of the acquired volume data points is substantially fitted from the feature points, but refined locally.
And 5: the curve fitted by only using the feature points generally cannot meet the requirement of approximation tolerance, and the number of the feature points needs to be increased in order to obtain a high-quality curve. And calculating the deviation between the data point and the initial fitting curve, taking the point with the deviation exceeding a set deviation threshold value as a characteristic point, and inserting the characteristic point into the initial fitting curve for refitting.
Step 5.1: and calculating approximation deviation, namely calculating the approximation deviation between the data points and the curve by adopting Hausdorff distance. The Hausdorff distance is a measure describing the degree of similarity between two sets of points.
Suppose two sets of data sets a = { a = { (a) 1 ,a 2 ,…,a p }，B＝{b 1 ,b 2 ,…b q And the Hausdorff distance between the two point sets of A and B is:
H(A,B)＝max(h(A,B),h(B,A))
where | · | | is a distance range (e.g., euclidean distance) between the point set a and the point set B, H (a, B) is a bidirectional Hausdorff distance, and H (a, B) and H (B, a) are unidirectional Hausdorff distances from the set a to the set B and from the set B to the set a, respectively. The two-way Hausdorff distance H (A, B) is the larger of the two singleton distances H (A, B) and H (B, A), which measures the maximum degree of mismatch between the two point sets.
Step 5.2: refining the curve locally, calculating deviation values of the initial curve and all corresponding acquisition points, and setting a deviation threshold value e ths Finding out all maximum value points exceeding the deviation threshold, and the specific operation is as follows:
(1) Calculating to obtain a number according to a Hausdorff distance formulaDeviation of approximation of point to curve
(2) Solving the maximum approximation deviation h (Q, B);
(3) Setting a deviation threshold e ths = β × h (Q, B) (β is a proportionality coefficient);
(4) Extracting all the deviation values larger than the deviation threshold value e on the deviation curve ths A point of (a);
(5) Refit these points as new feature points (one-time addition of more than one keypoint between two initial keypoints)
(6) And repeating the steps until the fitting precision is met, and finishing the calculation.
Since cubic B-splines are local, the newly inserted feature points can change the local knowledge of the fitted curve without affecting the whole.
The results of the cubic B-spline multiple optimization fitting of the sampled data in one period are shown in the attached figure 7. The initial fitting approximation deviation of the sampled data is compared with the calculated value of the optimized approximation deviation in table 5.
Table 5: comparing initial fitting deviation of sampling data with fitting deviation after optimization
The comparison in the table shows that the optimized approximation deviation is obviously better than the initial fitting effect, which also shows that the fitting effect of the method provided by the invention is better than that of the traditional method.
Step 5.3: and (4) evaluating the fitting accuracy, wherein the fitted data are irregular discrete points, and a curve equation between the fits is unknown and is formed by fitting characteristic points. After extracting the feature points of the upper level, obtaining two adjacent feature points as (x) k ,y k ) And(x k+1 ,y k+1 ) The point (x) on the curve segment between which the deviation from the two characteristic points is the greatest i ,y i ) By calculating the h value of a feature point, i.e. point (x) i ,y i ) To point (x) k ,y k ) And (x) k+1 ,y k+1 ) The vertical distance of the connected straight lines is calculated by the following formula:
all h values except the end point are calculated, and
as the integral square error of the curve, the integral square error is used as an index for measuring the fitting accuracy of the curve. The integral square error is an index for measuring the fidelity of the curve at the characteristic point of curve extraction, and the original shape of the curve can be ensured.
Considering that the fitting curve of the initial characteristic points can not meet the requirement of approximation error allowance generally, hausdorff distance is used for calculating approximation deviation, and points exceeding the deviation threshold are used as new characteristic points until the fitting accuracy is met. The fitting method greatly reduces the number of control points and improves the calculation efficiency on the premise of ensuring the precision.
Step 6: selecting an initial time section t after the collection amount of a plurality of collection points in the plant station is subjected to the fitting processing curve 0 And a time interval. Since the collected data is data with time mark, X represents the relation of the collected quantity with respect to time i And (t), wherein i is the ith collection quantity number, and the collection quantity X is collected power grid operation parameters such as voltage, current and the like. The discrete points after fitting become a continuous curve, and X at any time can be obtained i (t) numerical value, cross-section through time t 0 And a time interval Δ t Can ultimately be formed within the plant at t 0 ,t 0+Δt ,t 0+2Δt ,…t 0+nΔt All the acquisition amount section data of the moment are obtained, and the station side section data are obtained according to a time axisAnd storing the scales to obtain a database of the same section of the plant side.
And 7: and different stations obtain the same-section database of each station according to the same processing mode of the acquired data. And then, each plant station uploads the same section to the scheduling master station, and a same-section database of the whole network is formed at the scheduling master station, so that the accuracy of real-time data of the power grid is improved, and reliable data support is provided for power grid scheduling operation decision.
Claims (10)
1. A full-network same-section data processing method based on information acquisition with time marks is characterized by comprising the following steps:
step 1: the realization process is divided into two stages, wherein the first stage is the formation of data with the same section in the transformer substation, and the second stage is the formation of data with the same section in the whole network between the transformer substations;
firstly, a plurality of data acquisition and measurement points exist in a station, meanwhile, an alternating current sampling device has a time check function, the time of the acquisition points is ensured to be synchronous with a main clock of the station, the acquired data are marked with acquisition time scales, and the sampled data with the time scales are sent to a station background server for data storage and processing;
step 2: according to historical operation data of the acquisition points, determining the maximum and minimum limits of the acquisition amount under normal operation, and setting the upper limit of the normal operation interval of the acquisition amount as X max The lower limit of the normal operation interval is X min According to the maximum value and the minimum value of the collection amount within the last 12 months, the upper limit and the lower limit of the normal operation interval are calculated by a weighting algorithm, and the calculation formula is as follows:
where i is the ith month, k, nearest to the sampling time i Is a firsti month upper limit weight factor, h i Is the weight factor of the ith month floor value, x imax Maximum value of collected X at month i, X imin Is the minimum value of the collected amount X at month i; when the acquisition amount exceeds the upper limit and the lower limit of operation, the acquired data exceeding the limit does not participate in the subsequent operation process;
and step 3: dividing the collected data beyond the normal operation range into two situations; one is that a single point or two points exceed the normal operation range, and the situation is caused by the measurement error of the acquisition device, and the processing mode is to delete data; the other is that the acquisition amount exceeds the normal operation range at multiple points, the condition is an accident, and the processing mode is to start fault filtering to record relevant operation data;
and 4, step 4: aiming at the acquisition quantity data in a normal operation range, a cubic B-spline curve fitting method suitable for dense data is adopted, the method comprises the steps of calculating the curvature of discrete points, setting a reasonable curvature threshold value to extract characteristic points, carrying out chord length parameterization on the characteristic points, constructing a node vector according to the parameters of the characteristic points, and solving a control vertex by utilizing a least square method to fit the discrete acquisition quantity data;
and 5: in order to obtain a high-quality curve, the number of characteristic points is increased, the deviation between the data point and the initial fitting curve is calculated, and the point with the deviation exceeding a set deviation threshold value is used as the characteristic point and is inserted into the initial fitting curve for re-fitting;
step 6: by selecting the time section t 0 And a time interval Δ t, finally forming t in the plant 0 ,t 0+Δt ,t 0+2Δt ,…t 0+nΔt All the acquisition quantity section data at the moment point, the sampling data are data with time marks, and the relation of the acquisition quantity X with respect to time is set as X i (t), wherein i is the ith collection quantity number, and the collection quantity X is the voltage and the current of the power grid; the acquired data becomes a continuous curve at discrete points after fitting, and X at any time can be obtained i (t) numerical value, cross-section through time t 0 And determining the time interval delta t, finally forming the time t in the plant station 0 ,t 0+Δt ,t 0+2Δt ,…t 0+nΔt All the acquisition quantity section data at the moment are stored according to a time axis as scales to obtain a same section database in the station;
and 7: and each plant station processes the acquired data to obtain an in-station same-section database and uploads the in-station same-section database to the dispatching master station, and the whole-network same-section database is formed at the dispatching master station.
2. The method for processing the whole-network same-section data based on the time-scale information acquisition as claimed in claim 1, wherein the step 4 comprises the following steps:
step 4.1: at a given acquisition data point vector P k Wherein k = (0,1,2, … n) to form [ A, B]In the interval, every 4 points are sequentially connected into a polygon called a B characteristic polygon, and a fitting curve formed by fitting the B characteristic polygon by utilizing a cubic spline function is a cubic B spline curve;
matrix representation of cubic B-spline curve:
the expression of the i-th cubic B-spline curve is therefore expressed as:
wherein t is a parameter, P i ,P i+1 ,P i+2 ,P i+3 (i =0,1,2, …, n) as four vertices adjacent to the feature polygon; take j =3,N j,3 (t) is the basis function of the cubic B-spline curve, and the recursion formula is as follows:
3. the method for processing the whole-network same-section data based on the time-scale information acquisition as claimed in claim 2, wherein the step 4 comprises the following steps:
and 4.2: for a cubic spline B (t), the parameter value t is obtained by differentiation i The curvature of (b) is calculated by the formula:
wherein B' (t) i ) Is the first derivative of curve B (t), B' (t) i ) Is the second derivative, k, of curve B (t) i Is a curvature.
4. The method for processing the whole-network co-section data based on the time-scale information acquisition as claimed in claim 3, wherein the step 4 comprises the following steps:
step 4.3: extracting characteristic points, and obtaining curvature K = (K) of discrete points by using 0 ,k 1 ,…,k n ) The principle of extracting the feature points is as follows:
(1) For a curve that is not closed, its two endpoint values must be selected;
(2) Determining a set curvature threshold value selection characteristic point, and setting a point with the curvature larger than the threshold value as an initial characteristic point; mean value of curvature K avg The curvature threshold is set to K ths ＝α×K avg And alpha is a proportionality coefficient.
5. The method for processing the whole-network co-section data based on the time-scale information acquisition as claimed in claim 4, wherein the step 4 comprises the following steps:
step 4.4: parameterizing characteristic points by adopting a chord length parameter method, and obtaining data points Q 0 ,Q 1 ,…Q n The parameter field t is the same as [0,1 ∈ ]]The nodes between the nodes have one-to-one correspondence; let d total chord length:
then the
Wherein Q k -Q k-1 Is a chord edge vector.
6. The method for processing the whole-network co-section data based on the time-scale information acquisition according to claim 5, wherein the step 4 comprises the following steps:
step 4.5: constructing a node vector by adopting an average value method, and evenly distributing the node vector, namely:
t 0 ＝t 1 ＝…t p ＝0
t m-p ＝t m-p-1 ＝…＝t m ＝1
m +1 represents the total number of nodes, both ends of the node vector are respectively provided with p +1 same nodes so as to control the tangency of the head end point and the tail end point, j represents the ordinal number of the middle node, and the node vector can well reflect the parameter point vector by adopting an average value method
Distribution of (2).
7. The method for processing the whole-network co-section data based on the time-scale information acquisition as claimed in claim 6, wherein the step 4 comprises the following steps:
step 4.6: constructing a cubic B spline curve approximating the characteristic points by using a node vector and a least square method; as can be seen from the cubic B-spline expression, the feature point D is approximated in the least-squares sense, i.e.:
wherein D 0 ＝B(0)，D r = B (1), r is the number of feature points;
order to
f is about n-1 variables P 1 ,…P n-1 For minimizing f, let f be for control point P l The partial derivative of (a) is zero, the following equation is obtained:
this is an unknown quantity P 1 ,…,P n-1 Let l =1,2, …, n-1 then obtain a linear system of equations with n-1 unknowns and n-1 equations:
(N T N)P＝R
the value of the unknown quantity P can be solved by solving a system of linear equations, so that a fitting curve of a cubic B-spline can be determined.
8. The method for processing the whole-network co-section data based on the time-scale information acquisition as claimed in claim 1, wherein the step 5 comprises the following steps:
step 5.1: calculating approximation deviation, namely calculating the approximation deviation between the data points and the curve by adopting Hausdorff distance; suppose two sets of data sets a = { a = { (a) 1 ,a 2 ,…,a p }，B＝{b 1 ,b 2 ,…b q And the Hausdorff distance between the two point sets of A and B is:
H(A,B)＝max(h(A,B),h(B,A))
wherein | | · | | is the euclidean distance between the point set a and the point set B, H (a, B) is a bidirectional Hausdorff distance, and H (a, B) and H (B, a) are unidirectional Hausdorff distances from the set a to the set B and from the set B to the set a, respectively; the two-way Hausdorff distance H (A, B) is the larger of the two singleton distances H (A, B) and H (B, A), which measures the maximum degree of mismatch between the two point sets.
9. The method for processing the whole-network co-section data based on the time-scale information acquisition as claimed in claim 8, wherein the step 5 comprises the following steps:
and step 5.2: refining the curve locally, calculating deviation values of the initial curve and all corresponding acquisition points, and setting a deviation threshold value e ths Finding out all maximum value points exceeding the deviation threshold, and the specific operation is as follows:
(1) According to a Hausdorff distance formula, calculating to obtain the approximation deviation of the data point and the curve
(2) Solving the maximum approximation deviation h (Q, B);
(3) Setting a deviation threshold e ths = β × h (Q, B), β being a proportionality coefficient;
(4) Extracting all the deviation values larger than the deviation threshold value e on the deviation curve ths A point of (a);
(5) Re-fitting the points as new feature points, wherein one-time increased key point between two initial key points cannot exceed one;
(6) And repeating the steps until the fitting precision is met, and finishing the calculation.
10. The method for processing the whole-network co-section data based on the time-scale information acquisition as claimed in claim 1, wherein the step 5 comprises the following steps:
step 5.3: and (5) evaluating the fitting precision, and after extracting the characteristic points of the previous stage, obtaining two adjacent characteristic points which are respectively (x) k ,y k ) And (x) k+1 ,y k+1 ) The point (x) on the curve segment between which the deviation from the two characteristic points is the greatest i ,y i ) By calculating the h value of a feature point, i.e. point (x) i ,y i ) To point (x) k ,y k ) And (x) k+1 ,y k+1 ) The vertical distance of the connected straight lines is calculated by the following formula:
all h values except the end points are calculated, and
as the integral square error of the curve, the integral square error is used as an index for measuring the fitting accuracy of the curve. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911231245.4A CN111144705B (en) | 2019-12-05 | 2019-12-05 | Whole-network same-section data processing method based on information acquisition with time scale |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911231245.4A CN111144705B (en) | 2019-12-05 | 2019-12-05 | Whole-network same-section data processing method based on information acquisition with time scale |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111144705A CN111144705A (en) | 2020-05-12 |
| CN111144705B true CN111144705B (en) | 2023-04-18 |
Family
ID=70517568
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911231245.4A Active CN111144705B (en) | 2019-12-05 | 2019-12-05 | Whole-network same-section data processing method based on information acquisition with time scale |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111144705B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113654457A (en) * | 2021-07-22 | 2021-11-16 | 太原理工大学 | Spectrum confocal measuring head wavelength and displacement mapping relation calibration device and fitting method |
| CN116362625B (en) * | 2023-05-26 | 2023-08-15 | 华南理工大学 | Multi-level verification system and method for energy resource consumption data |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101739018A (en) * | 2008-11-20 | 2010-06-16 | 上海申瑞电力科技股份有限公司 | Processing method for fitting and sampling telemetry data of data acquisition and monitoring control system |
| CN102361353A (en) * | 2011-10-26 | 2012-02-22 | 东北电网有限公司 | Method for aligning section raw data on basis of double time scale delay evaluation |
| WO2015070604A1 (en) * | 2013-11-12 | 2015-05-21 | 国电南瑞科技股份有限公司 | Interval-oriented multi-functional secondary device of intelligent substation, and sampling platform |
| CN108874859A (en) * | 2018-04-17 | 2018-11-23 | 中国电力科学研究院有限公司 | It is a kind of to dispatch real-time profile data generation method and system |
| CN109191543A (en) * | 2018-09-19 | 2019-01-11 | 国网辽宁省电力有限公司 | A kind of AC sampling is the same as profile data generation method |
-
2019
- 2019-12-05 CN CN201911231245.4A patent/CN111144705B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101739018A (en) * | 2008-11-20 | 2010-06-16 | 上海申瑞电力科技股份有限公司 | Processing method for fitting and sampling telemetry data of data acquisition and monitoring control system |
| CN102361353A (en) * | 2011-10-26 | 2012-02-22 | 东北电网有限公司 | Method for aligning section raw data on basis of double time scale delay evaluation |
| WO2015070604A1 (en) * | 2013-11-12 | 2015-05-21 | 国电南瑞科技股份有限公司 | Interval-oriented multi-functional secondary device of intelligent substation, and sampling platform |
| CN108874859A (en) * | 2018-04-17 | 2018-11-23 | 中国电力科学研究院有限公司 | It is a kind of to dispatch real-time profile data generation method and system |
| CN109191543A (en) * | 2018-09-19 | 2019-01-11 | 国网辽宁省电力有限公司 | A kind of AC sampling is the same as profile data generation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111144705A (en) | 2020-05-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110516388B (en) | Harmonic mapping-based curved surface discrete point cloud model circular cutter path generation method | |
| CN108564204B (en) | Least square support vector machine electricity quantity prediction method based on maximum correlation entropy criterion | |
| CN111144705B (en) | Whole-network same-section data processing method based on information acquisition with time scale | |
| CN110969290B (en) | Runoff probability prediction method and system based on deep learning | |
| CN111275579B (en) | Heat supply system thermal delay time identification method based on Pearson correlation coefficient and moving average method | |
| CN112396690B (en) | Curved surface high-precision reconstruction method based on improved centripetal parameterization method | |
| CN109683552B (en) | Numerical control machining path generation method on complex point cloud model guided by base curve | |
| CN106980910B (en) | Medium-and-long-term power load measuring and calculating system and method | |
| CN109992921A (en) | A kind of online soft sensor method and system of the coal-fired plant boiler thermal efficiency | |
| CN109002650A (en) | A kind of Wind turbines power curve modeling method | |
| CN111882114B (en) | Short-time traffic flow prediction model construction method and prediction method | |
| CN108764547B (en) | Medium-and-long-term load curve generation method meeting electric quantity constraint | |
| CN116089777B (en) | Intelligent new energy settlement method and system based on intelligent information matching | |
| CN112033338B (en) | Blade curved surface contact type scanning measurement probe radius surface compensation method | |
| CN111199549B (en) | Blade profile extraction method based on blade profile measurement point cloud | |
| CN109035311B (en) | Automatic registration and internal fixation steel plate pre-bending modeling method for curved bone fracture | |
| Velasco et al. | Day-ahead base, intermediate, and peak load forecasting using k-means and artificial neural networks | |
| CN115964787B (en) | Phase redistribution-based method for extracting and characterizing initial geometric defects of lasso-type spinal rod | |
| CN110555189B (en) | Spatial interpolation method based on reverse computing thinking | |
| CN116049941B (en) | Method for extracting and analyzing multidimensional state of assembled ring truss structural member before assembly | |
| CN108170903A (en) | A kind of hyperbolic surface parts springback compensation method based on curvature mapping | |
| CN114997470A (en) | Short-term power load prediction method based on LSTM neural network | |
| CN111368257A (en) | Method and device for analyzing and predicting coal-to-electricity load characteristics | |
| Lu et al. | An approximation method combined with conformal transformation for the fast measurement of blade section curves | |
| CN113553538A (en) | Recursive correction hybrid linear state estimation 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 |