CN104134999B - Distribution network based on multi-data source measures the practical method of calculation of efficiency analysis - Google Patents

Abstract

The present invention is that a kind of distribution network based on multi-data source measures the practical method of calculation of efficiency analysis, and its step is (one) static near-realtime data calibration analyte, and the non-real-time measurement data of other system acquisition are carried out reasonableness check analysis; (2) near-realtime data is utilized to carry out real-time measurement completion; (3) distribution switch measures efficiency analysis step; The multi-data sources such as the comprehensive distribution SCADA real time data of its method, distribution transforming non real time information, typical load characteristic, telemetry data is carried out Effective judgement, invalid signals object is positioned, the bad switch of Detection and identification measures and switch state, therefore state of electric distribution network is estimated the normal operation and management of distribution network, ensures that distribution management system normally plays its function and is significant.

Description

Distribution network based on multi-data source measures the practical method of calculation of efficiency analysis

Technical field

The present invention relates to distribution scheduling automation system technology, in particular a kind of senior application numerical evaluation of distribution scheduling, namely based on the distribution network measurement practical method of calculation of efficiency analysis of multi-data source.

Background technology

The general thought of state of electric distribution network analysis first by message bus system, the distribution network model of production management system or GIS system is imported to electrical power distribution automatization system, the major network model of dispatch automated system is imported to electrical power distribution automatization system, undertaken being spliced to form the topological main distribution integration model being connected of main distribution by major network model and distribution network model by electrical power distribution automatization system, the distribution transforming electric current of using electricity system acquisition system, power measurement are imported to electrical power distribution automatization system simultaneously. Distribution state estimation is based on main distribution integration model, major network part measures, the whole SCADA of distribution measures, and all distribution transformings of electricity consumption acquisition system comprise public change and the survey of special variable, utilize topological analysis, numerical analysis relative theory to realize state of electric distribution network analysis, main functional modules comprises distribution switch remote measurement pre-treatment, distribution switch state pre-treatment, network distribution transformer measurement calibration and measurement completion, observability analysis, bad data detection and identification, bad data statistics and topology information are shown.

Although the measurement level of distribution network is improving constantly, distribution local there are a large amount of measurement data, but from whole distribution network the overall situation angle, Real-Time Monitoring data can observation scope proportion less. On the one hand, part region causes part Data duplication due to the situation of multi-data source, and on the other hand, part region then exists the situation of disappearance.

Summary of the invention

For the deficiency existed in prior art, it is an object of the present invention to provide one and bad data can be carried out Detection and identification, and there are the means revising bad data, remove bad data, calculate the distribution network based on multi-data source more accurate than SCADA telemetry data, operation scheme comprehensively and measure the practical method of calculation of efficiency analysis, provide operation scheme more accurately and monitor that distribution network runs for management and running personnel, and be other application software complete real-time system cloud gray model mode of offer.

In order to realize above-mentioned purpose, the present invention realizes by the following technical solutions:

A kind of distribution network based on multi-data source measures the practical method of calculation of efficiency analysis, its feature exists, its method is: according to distribution SCADA (SupervisoryControlAndDataAcquisition, i.e. data gathering and Monitor and Control system) real time data, distribution transforming non real time information, typical load characteristic the accurate actual quantities of the distribution of multi-data source survey, telemetry data is carried out Effective judgement, being positioned by invalid signals object, the bad switch of Detection and identification measures and switch state;

The accurate real-time measurement of above-mentioned distribution transforming, carries out Effective judgement to telemetry data, is positioned by invalid signals object, and the bad switch of Detection and identification measures and switch state.

Its method steps is as follows:

(1) the non-real-time measurement data of other system acquisition are carried out reasonableness check analysis by static near-realtime data calibration analyte, it is intended that remove the great unreasonable data of error; The history real time data that load node in distribution network is gathered by power information acquisition system, and it is forwarded to power distribution automation main station system by information interactive bus; The history real time data forwarded is carried out the static load calibration process based on numerical method simultaneously;

Described is that the history real time data state analysis to load node and Load flow calculation carry out load calibration based on numerical method, described history real time data state analysis is that the load node historical data by gathering utilizes this circuit basic norm formula, judge that the history real time data of this load node of all collections is whether within the scope of permissible error, if history real time data is within the scope of permissible error, limit of error regulation electric current is at the 3% of rated value, and power is the 2% of rated value. Then as the historical load value of this load node, otherwise, then no;

To not having the load node of current measurement value after completing above-mentioned static load and calibrating, real-time measurement completion step should be carried out;

(2) near-realtime data is utilized to carry out real-time measurement completion, pass through topology analyzing method, when utilizing the non-solid of other system forwards and historical data estimate out the real time data of current time of each load node, it is intended that carry out the load not collecting real-time measurement measuring completion;

According to network topology structure and existing history real time data, the static load calibration value of above-mentioned steps () is revised by topology analyzing method, this topology analyzing method carries out modification method: the history curve data utilizing distribution network switch, pass through Topologically ergodic, and predict gained value according to after the historical data static load calibration of this load node by load, i.e. topology calibration load value, verify the current near-realtime data of the load node of substation transformer, and the current load data to the load node not collected, by load, historical data according to this load node predicts that gained value carries out real-time measurement completion as the real time data of the current time of the load node not collected, finally obtain complete distribution profile data, as the input data of Load flow calculation,

(3) distribution switch measures efficiency analysis step; The preload of working as obtained by above-mentioned steps (2) predicts that gained value and the current load data value recorded carry out measurement rough detection, distribution switch state identification and distribution switch bad data detection and identification, invalid signals object is positioned, the bad switch of Detection and identification measures and switch state, it is intended that real-time switch measures the Detection and identification carrying out bad data.

The above-mentioned historical data according to this load node predicts gained value by load, its concrete grammar is as follows: (a) determines network calibration group, utilize topological analysis, using state as distribution switch that is that divide and that have measurement as edge device, distribution network feeder is divided into some switch segments, and a switch segments is a topological calibration group. Each group can calculate the current value flowing into this group and flow out the current value of this group, and the electric current of inflow subtracts the electric current that the electric current flowed out is this group internal loading and consumes. Total load electric current in group can be distributed to each load by distribution factor. The load distribution factor calculates according to the non-real time data of other system forwards.

B () measures according to the static load verified is that each group of networks is calculated each burden apportionment factors A F by topology calibration load value respectively; The universal calculation equation of distribution factor AF is:

$\mathrm{AF}=\frac{\mathrm{\Σ}{L}_{\mathrm{inp}-\mathrm{mea}}+\mathrm{\Σ}{L}_{g-\mathrm{mea}}-\mathrm{\Σ}{L}_{\mathrm{out}-\mathrm{mea}}}{\mathrm{\Σ}{L}_{\mathrm{ca}}}$

In formula: �� L_inp-meaThe summation of input measurement value in group of networks, comprising: the observed value of 10kV feeder line outlet, the observed value that other group of networks input to this group,

��L_g-meaGroup of networks Small Power exports summation. The summation of the observed value that little power supply inputs to this group of networks,

��I_out-meaThe summation of outputting measurement value in group of networks. Group of networks has the observed value summation of the load node of observed value;

��I_caThe summation of all load non-solid observed values in group of networks.

C static load value is carried out topology by following formula and calibrates by (): the preload of working as being calculated each load node by above-mentioned distribution factor predicts gained value, i.e. topology calibration load value,

Topology calibration load value=current load data total value * AF;

(d) consistency check; According to Logic judgment, (basis for estimation has: distribution switch PQI does not mate; Feeder line section two ends are meritorious, idle, electric current conflicts mutually; Distribution bus measures inflow and outflow and does not mate; Distribution load measures with on-load switch and does not mate; Distribution switch remote measurement is not corresponding with remote signalling. ) find out wrong or suspicious measurement data, or the defective part of measurement system is supplemented automatically. The scope of rough detection: remote measurement, remote signalling. The result of rough detection: to determine mistake: revise or filter; To suspicious: point out in the table; To what lack: automatic makeup is neat. Unbalanced remote measurement can not be automatically modified, it is necessary to manually according to circumstances revises, it has been found that remote signalling information errors is also corrected; Because telemetering state directly has influence on topology search.

E the calibration of () topology should change when network topology, or carry out when yardman asks.

In described step (three), the method of distribution switch bad data detection and identification is: namely estimate when calculating converges to iteration convergence error value in state estimation iterative computation that (iteration convergence error is manual maintenance, can revise), to a certain group of suspicious data, first get the maximum measurement of one of them residual error and carry out identification, forecast that this residual error changes, judge whether it is bad data, when the residual values of the measurement calculated is greater than manual maintenance permission residual values, then it is judged as bad data, otherwise, then it is normal data; If after detecting out bad data, first estimate that it is correctly worth, and revise up-to-date residual error, again residual error is ranked, again the bad data of identification; When preload predicts by weighting measures residual values, gained value and the current load data value recorded judge whether measurement is suspicious data, and its determination methods is as follows:

Mistake measures data and is included in state estimation solving equation, make calculation result by deviation system time of day, reject maximum absolute value value and it is greater than the quantity of state that Operation system setting allows worst error value, if mistake measures after calculating terminates, between quantity of state and side value, absolute value is less than manual maintenance permission residual values, according to the bad data of the size detection of residual values.

In above-mentioned steps (2), point switch include measurement point switch and without measure point switch. Suspicious switch: the charged the other end in switch one end is not charged, the switch having the state of measurement to be point is recognized as conjunction.

The charged the other end in switch one end is not charged, and without the state measured for dividing, not charged end but has the switch of measurement to be recognized as conjunction with load and load.

Loop measures the minimum switch of electric current be recognized as point, on loop without the switch measured and this switch first and last end from electric point away from switch be recognized as point. Note identification switch state, and need after changing switch state to carry out local topology search. The electrical island that switch changed position may have influence on is re-started painted by the electrical island that search switch changed position affects.

State of electric distribution network is estimated when real-time measurement information deficiency, in conjunction with data sources such as power information and typical load characteristics, telemetry data is carried out Effective judgement, is positioned by invalid signals object, and the bad switch of Detection and identification measures and switch state.

Load calibration normally runs with the state circulated, and also can be activated when switch state changes He after receiving operator's instruction. Whole calibration be generally one hour once, and topology calibration and every five minutes of state estimation are once.

Bad data can be carried out Detection and identification by the present invention, and there are the means revising bad data, remove bad data, it is more accurate than SCADA telemetry data to calculate, comprehensive operation scheme, provide operation scheme more accurately and monitor that distribution network runs for management and running personnel, and be other application software complete real-time system cloud gray model mode of offer, good raw data detection result can be obtained, on the key positions such as substation's outlet and section switch, generally can obtain and the result that actual trend matches, substantially the needs of distribution network actual motion are met, solve electricity distribution network model amount big, wiring diagram is many, topology is complicated, distribution network model maintenance workload is big, and measure configuration is weak, lack voltage and the power measurement of a large amount of collection point, real-time property is not strong, data do not refresh, measure imperfect, the practical situation problems such as state is suspicious, state of electric distribution network is estimated the normal operation and management of distribution network, ensure that distribution management system normally plays its function and is significant.

Accompanying drawing explanation

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail;

Fig. 1 is that the present invention measures efficiency analysis functional structure chart based on multi-data source distribution.

Embodiment

For the present invention is realized technique means, creation characteristic, reach object and effect is easy to understand, below in conjunction with embodiment, set forth the present invention further.

See Fig. 1, the distribution network of the comprehensive multi-data source of the present invention measures the practical method of calculation of efficiency analysis, it is when studying real-time measurement information deficiency, in conjunction with the state of electric distribution network analysis of power information acquisition system, metering system, marketing system and typical load characteristic, for distribution Dispatching Control System provides the state profile data of comparatively accurately real-time full. Its research contents realizes function main as follows based on the accurate real-time measurement of distribution transforming:

S1: the non-real-time measurement data of other system acquisition are carried out reasonableness check analysis by static near-realtime data calibration analyte, it is intended that remove the great unreasonable data of error;

S2: utilize near-realtime data to estimate measurement and carry out real time data completion, pass through topology analyzing method, the near-realtime data of other system forwards is utilized to estimate out the real time data of current time of each equipment, it is intended that to carry out the load not collecting real-time measurement measuring completion;

S3: the distribution switch based on numerical analysis measures and state validity analysis, it is intended that real-time switch measures the Detection and identification carrying out bad data.

For better setting forth the present invention, specific embodiment of the invention method is as follows:

1. static near-realtime data calibration analyte

(1). screen based on capacity of distribution transform.

Circuit is filled the power supply capacity that distribution transformer capacity characterizes circuit. The load theoretical upper limit that can reach on circuit can be judged according to filled distribution transformer capacity:

$I_M=\frac{C}{10\sqrt{3}K}$

In formula: K is the theoretical upper limit of circuit load current; C is filled distribution transformer capacity by circuit; Rate when K is different. , when line current exceedes the line theory load upper limit, can be used as bad data and process. K has fluctuation within the specific limits.

(2). static accurate real-time measurement consistency desired result.

Static load measures and comprises wattful power, wattless power, electric current, voltage, power factor etc. Carry out check analysis according to the basic norm such as circuit KVL and KCL, PQI coupling, the measurement not meeting circuit and substantially retraining relation is removed. The content of consistency check also comprises: distribution switch PQI does not mate; Feeder line section two ends (middle without bifurcated) are meritorious, idle, electric current conflicts mutually; Distribution bus measures inflow and outflow and does not mate; Distribution load measures with on-load switch and does not mate; Distribution switch remote measurement is not corresponding with remote signalling.

2. utilize near-realtime data to carry out real-time measurement completion

(1). utilize Real-Time Switch to measure and load distribution is measured

Power direction defines: flowing into node power is just, it is negative for flowing out node power. Treatment step is as follows:

If a () distribution switch does not have power measurement value, then adopt the power factor of 10kV outlet isolating switch, otherwise adopt the power factor of acquiescence.

B () seeks node power value: node flows into the power that power is more than or equal to outflow, then this node has remaining power to distribute to load; If flowing out power to be greater than inflow power, then this node has remaining power to distribute to distributed power source.

C () obtains the total surplus power of away minor segment, inject node injecting power and branch road loss power that total power subtracts each injection element, and last difference is just for load or the power distribution of this away minor segment obtain performance number.

D surplus power in away minor segment, for without the load measured, is distributed to load cell by (). If away minor segment does not have load, then surplus power is distributed to source element. Load power distribution prerequisite is: all distribution switches all calculate the performance number that switch flows through. Skill is: can distinguish those switches in section is the switch that power flows into, and those switches are the switches that power flows out.

(e) all inflows section switch power all outflows section switch power section interior nodes injecting power (node injecting power refers to the power of this section internal loading and the power sum of power supply)=imbalance power. Then imbalance power is distributed in this section without the load measured or power supply.

(2). utilize Real-time Load always to measure and according to load accurate real-time measurement distribution factor, load distribution is measured

The universal calculation equation of distribution factor is

$\mathrm{AF}=\frac{\mathrm{\Σ}{L}_{\mathrm{inp}-\mathrm{mea}}+\mathrm{\Σ}{L}_{g-\mathrm{mea}}-\mathrm{\Σ}{L}_{\mathrm{out}-\mathrm{mea}}}{\mathrm{\Σ}{L}_{\mathrm{ca}}}$

In formula: �� L_inp-meaThe summation of input measurement value in group of networks, comprising: the observed value of 10kV feeder line outlet, the observed value that other group of networks input to this group.

��L_g-meaGroup of networks Small Power exports summation. The summation of the observed value that little power supply inputs to this group of networks.

��I_out-meaThe summation of outputting measurement value in group of networks. Group of networks has the observed value summation of the load node of observed value.

3. distribution switch measures state validity analysis;

Distribution switch measures and state validity analysis mainly comprises: measure rough detection; Distribution switch state identification; Distribution switch bad data detection and identification. Measuring the measurement that rough detection function can only count on multiple switch within the scope of not mate or uneven or conflict mutually, but it is suspicious to navigate to that measurements concrete, its result can only be statistics displaying, can not data be revised, it is necessary to manually revise. Bad data detection and identification function is that to navigate to certain concrete measurement on the basis of rough detection function suspicious, and its result is the measurement of program auto modification mistake, and identification goes out correct measurement.

(1). bad data detection and identification

Data are measured when system can read comparatively accurately comparatively real-time and complete distribution transforming, the injection rate measured value that so distribution state analysis calculates is just relatively accurately, and the injection rate measured value of the overwhelming majority is all based on the amount side value of distribution transforming in distribution, several points that amount measured value is only little. Such state analysis value just has bigger redundancy, just can judge value and the state of switch more accurately.

State analysis program pin is relatively weak to distribution network measure configuration, and most collection point does not configure the practical situation of voltage and power measurement, tie point through connecting without the switch measured and branch road is merged into a section, propose taking the electric current of ribbon amount slowdown monitoring switch as quantity of state, the switch current estimation model that current measurement forms is injected, it is achieved that to estimation and the bad data recognition of the electric current of ribbon amount slowdown monitoring switch by switch current measurement and section.

Bad data detection and identification main contents are as follows:

1) create section: intend estimating switch in search electrical island, create and intend estimating to measure.

Intending estimating that switch state is necessary for conjunction, the first and last end section of switch belongs to different sections and has measuring value. Intend estimating that switch must be the boundary switch having measurement closed. Measurement comprises switch and measures and section injection measurement.

Switch measures: No. 1, switch node thinks first section, and No. 2, node is end, is positive direction from first section to end, and end is to the negative power direction that first section is this branch road. Power should be flow to from the node away from power supply from from the node close to power supply, and we define the end that electric current flows to branch road from the first end of branch road, and therefore electric current flows to No. 2, node from No. 1, node, and electric current is just, electric current flows to No. 1, node from No. 2, node, and electric current is negative.

Section measures: section injects and measures, and power direction definition is inflow is negative, flows out for just. Reading the measuring value of power supply, current value is negative, reads the measuring value of load, and current value is just. Section injects to measure and just adds all lifting capacity measured values for all power supplys measure. As long as there being a power supply not measure, or having a load not measure, section injects and is just set to 0.

2) the accurate real-time measurement of distribution transforming improves data redundancy

For distribution state analysis, the present invention adopts the electric current of distribution switch to calculate as quantity of state. Calculating is had extremely important meaning by the redundancy value therefore obtaining distribution switch electric current. Utilize network distribution transformer power data to increase and measure redundancy.

3) weight analysis is measured

All measurements are arranged identical weighted value and there will be bigger error by system default, need the accuracy measured according to difference that the precision that different weights improves calculating is set, but how coming by the accuracy that automatic program identification measures for the distribution amount system that survey redundancy is not high will be a problem. The weight of all outlet isolating switchs is floated, through an analytical calculation, finds out the measurement switch that residual error is maximum, therefore reduce the weight of the maximum measurement of residual error, then re-start and once calculate.

Program analyzes the automatic weighted value rule analyzing different measurement by topological analysis with measuring:

[principle 1] is if each switch being connected with a bus collects load data all effectively, and the load data gathered meets the constraint of formula (2), then the confidence level that trend flows into switch adds 2 point, and the confidence level of rest switch respectively adds 1 point.

[principle 2] is if two switches being directly connected collect load data all effectively, and the load data gathered is substantially equal, then the confidence level of these two switches adds 2 point, and:

If a enters the constraint that the load data that each switch of the bus a little connected gathers meets formula (2), then the confidence level of the rest switch being connected with this bus is respectively added 1 point.

If b goes out the constraint that the load data that each switch of the bus a little connected gathers meets formula (2), then the confidence level that the trend being connected with this bus flows into switch adds 1 point.

[principle 3] is designated the collection load data of " old data " for quality, by its etc. be all and do not gather load data and treat.

4) distribution switch state identification

The uneven section of appointment is carried out switch identification. System obtains the adjacent switch divided of this section. Point switch have following several situation: have measurement point switch and without measure point switch. Suspicious switch: the charged the other end in switch one end is not charged, the switch having the state of measurement to be point is recognized as conjunction. The charged the other end in switch one end is not charged, and without the state measured for dividing, not charged end but has the switch of measurement to be recognized as conjunction with load and load. Loop measures the minimum switch of electric current be recognized as point, on loop without the switch measured and this switch first and last end from electric point away from switch be recognized as point. Note identification switch state, and need after changing switch state to carry out local topology search. The electrical island that switch changed position may have influence on is re-started painted by the electrical island that search switch changed position affects.

5) bad data detection and identification

It is when state estimation iterative computation converges to a certain degree, for a certain group of suspicious data, first gets the maximum measurement of one of them residual error and carries out identification, forecast that this residual error changes, judges whether it is bad data. If after identification goes out bad data, first estimate that it is correctly worth, and revise up-to-date residual error, again residual error is ranked, again the bad data of identification.

General by weighting measurement residual values is judged whether measurement is suspicious:

Mistake measures data and is included in state estimation solving equation, makes calculation result by deviation system time of day, it is necessary to reject the measurement that deviation is excessive. It is relatively big that mistake measures its measurement residuals after calculating terminates, can according to the bad data of the size detection of residual values.

V=Zhx

Successively type estimates that identification method makes after eliminating bad data state estimation no longer again calculate, and utilize residual sensitivity matrix and Jacobian matrix directly to revise residual error and unknown quantity to be asked, greatly reduced so that it is the ability of accurate recognition suspicious data is played and practicality the residual error search procedure time.

The residual error forecast is the residual error after correct estimation suspicious data I or group of suspicious data, as shown in the formula:

R=ZHX

Wherein, r_w ^(I)Weighted residual phasor for identification and after estimating suspicious data; Z_wFor weighting measures phasor; X^(I)Identification amount to be asked after estimating suspicious data, comprise each node voltage amplitude U and angle, ��.

The correction formula of amount unknown quantity to be asked is:

X^(I)=X-(H_w ^TH_w)^-1H_wiW_wii ^-1r_wi

Wherein, H_wFor weighting Jacobian matrix; H_w ^TFor weighting Jacobian matrix transposed matrix; H_wiThe weighting Jacobian matrix part relevant to suspicious data; W_wWeighted residual sensitivity matrix; W_wiThe weighted residual sensitivity matrix of corresponding suspicious data part; r_wiThe weighted residual of corresponding suspicious data part; W_wii ^-1The inverse matrix of the weighted residual sensitivity matrix of corresponding suspicious data part, X is unknown quantity, H_w ^TFor weighting Jacobian matrix transposed matrix.

Thus can release identification suspicious data rear weight residual error forecast formula:

r_w ^(I)=r_w+H_w(H_w ^TH_w)^-1H_wiW_wii ^-1r_wi

And other measure weighted residual forecast formula:

r_wj ^(I)=r_wj-W_wijW_wii ^-1r_wi(j �� I, j �� I)

Corresponding weighted residual variance is changed to:

Var(r_w ^(I))=1

Var(r_wj ^(I))=(W_wij-W_wjj ²)/W_wii

W_wjj��W_wiiThe diagonal element of weighted residual sensitivity matrix; W_wijI-th row jth column element of weighted residual sensitivity matrix.

Successively type estimates that identification method has played the accuracy of the successively property exploration identification suspicious data of residual error search procedure, and in state estimation procedure identification method suspicious data, it is a kind of success suspicious data discrimination method, it is used widely in the state estimation software of reality.

Successively type estimates that the discrimination method of identification method is as follows:

A) suspicious data scope is determined;

B) copy matrix of coefficients, remove the residual vector after bad data, matrix of coefficients and right-hand vector are carried out this orthogonal transformation of propitious essay;

State estimation software adopts the mixing algorithm that normal equation combines with orthogonal transformation, only Jacobian matrix Hp and Hq is converted, it is necessary to retain sparse matrix Hp and Hq, and the matrix L after orthogonal transformation;

Gain matrix Hp^TRp^-1Hp and Hq^TRq^-1Hq can be write as (Hpw)^THpw and (Hqw)^THqw form:

(Hpw)^THpw=(QpHpw)^T(QpHpw)=Lp^TLp

(Hqw)^THqw=(QqHqw)^T(QqHqw)=Lq^TLq

State estimation solves iterative equation below:

Lp^TLp����_(l)=a_(l)

Lq^TLq��U_(l)=b_(l)

Orthogonal transformation becomes and is converted by Hpw and Hqw:

$\mathrm{QpHpw}=\begin{array}{c}\mathrm{Lp}\\ 0\end{array}$ With $\mathrm{QqHqw}=\begin{array}{c}\mathrm{Lq}\\ 0\end{array}$

Lp��Lp^TFor the upper triangular matrix Lp after orthogonal transformation, Qp and Qq is orthogonal matrix, and Hpw is above the matrix that a nxn ties up after conversion, is 0 below entirely, Hqw through conversion after above be the matrix that (n-1) x (n-1) is tieed up, be 0 below entirely. After trying to achieve Lp and Lq matrix, other computation processes are consistent with quick decoupling zero weighted least-squares method; Orthogonal transformation is exactly mainly that Hpw and Hqw adopts this conversion of propitious essay, from cancellation non-zero entry below;

C) successively type bad data identification is started; Circulation measurement matrix corresponds to the part of bad measurement corresponding to the row of bad data, sensitivity matrix, residual vector, temporarily vector, and sensitivity matrix is corresponding to the diagonal angle part of bad measurement;

D) the suspicious residual error r measuring correspondence that can detect_s, by the relevant part W of the residual sensitivity matrix of residual error equation r=Wv_s, corresponding error in measurement v can be tried to achieve_s, obtain estimating the identification formula of identification method:

v_s=(W_s ^TG^-1W_s)W_s ^TG^-1r_s

In formula: v_sThe error in measurement vector that can detect; W_sThe relevant part of corresponding suspicious measurement residuals sensitivity matrix; G^-1Weighting diagonal matrix, G^�C1=(H^TR^-1H)^-1��

Estimate the weighted type formula of identification method:

v_ws=(W_ws ^TG^-1W_ws)W_ws ^TG^-1r_ws

v_wsSuspicious measurement weighted error vector; r_wsSuspicious measurement weighted residual vector; W_wsThe relevant part of corresponding suspicious measurement weighted residual sensitivity matrix;

W_s ^TFor W_sTransposed matrix, W_ws ^TFor W_wsTransposed matrix.

E) weighted residual vector is upgraded;

Identification suspicious data rear weight residual error forecast formula:

r_w ^(I)=r_w+H_w(H_w ^TH_w)^-1H_wiW_wii ^-1r_wi

And other measure weighted residual forecast formula:

r_wj ^(I)=r_wj-W_wijW_wii ^-1r_wi(j �� i, j �� i)

Wherein, H_wiThe weighting Jacobian matrix part relevant to suspicious data; W_wWeighted residual sensitivity matrix; W_wiThe weighted residual sensitivity matrix of corresponding suspicious data part; r_wiThe weighted residual of corresponding suspicious data part. W_wjj��W_wiiThe diagonal element of weighted residual sensitivity matrix; W_wijI-th row j column element of weighted residual sensitivity matrix.

F) next raw data detection, until the maximum residual values in all suspicious datas is less than the threshold value that weighting measures residual detection.

Adopt above-mentioned distribution network switch to measure efficiency analysis method and generally can obtain good raw data detection result, between the actual motion state of this possibility of result and distribution network, there is some difference, particularly in the inside of section, may there is the situation bigger with virtual condition deviation in the load of each distribution transforming. But on the key positions such as substation's outlet and section switch, generally can obtain and the result that actual trend matches, substantially meet the needs of distribution network actual motion.

More than show and describe the ultimate principle of the present invention and the advantage of main characteristic sum the present invention. The technician of the industry should understand; the present invention is not restricted to the described embodiments; the principle that the present invention is just described described in above-described embodiment and specification sheets; without departing from the spirit and scope of the present invention; the present invention also has various changes and modifications, and these changes and improvements all fall in the claimed scope of the invention. The claimed scope of the present invention is defined by appending claims and equivalent thereof.

Claims (10)

1. the distribution network based on multi-data source measures the practical method of calculation of efficiency analysis, its feature exists, its method is: survey according to the accurate actual quantities of the distribution of the multi-data source of the non real time information of distribution SCADA system real time data, substation transformer, typical load characteristic, telemetry data is carried out Effective judgement, being positioned by invalid signals object, the bad switch of Detection and identification measures and switch state;

Its method steps is as follows:

(1) the non-real-time measurement data of other system acquisition are carried out reasonableness check analysis by static near-realtime data calibration analyte; The history real time data that load node in distribution network is gathered by power information acquisition system, and it is forwarded to power distribution automation main station system by information interactive bus; The history real time data forwarded is carried out the static load calibration process based on numerical method simultaneously;

Described is that the history real time data state analysis to load node and Load flow calculation carry out load calibration based on numerical method, described history real time data state analysis is that the load node historical data by gathering utilizes circuit basic norm formula, judge that the history real time data of this load node of all collections is whether within the scope of permissible error, if history real time data is within the scope of permissible error, then as the historical load value of this load node, otherwise, then no;

To not having the load node of current measurement value after completing above-mentioned static load and calibrating, real-time measurement completion step should be carried out;

(2) utilize near-realtime data to carry out real-time measurement completion, by topology analyzing method, utilize the non-real-time history data of other system forwards; Estimate out the real time data of the current time of each load node;

According to network topology structure and existing history real time data, the static load calibration value of above-mentioned steps () is revised by topology analyzing method; This topology analyzing method carries out modification method: the history curve data utilizing distribution network switch, by Topologically ergodic, and predicts gained value according to after the historical data static load calibration of this load node by load, as the instantaneous value of load current time; To the load data of the load node not collected, by load, historical data according to this load node predicts that gained value carries out real-time measurement completion as the real time data of current time, finally obtain complete distribution profile data, measure the input data of efficiency analysis calculating as switch;

(3) distribution switch measures efficiency analysis step; The preload of working as obtained by above-mentioned steps (two) predicts that gained value and the current load data value recorded carry out measurement rough detection, distribution switch state identification and distribution switch bad data detection and identification, being positioned by invalid signals object, the bad switch of Detection and identification measures and switch state.

2. the distribution network based on multi-data source according to claim 1 measures the practical method of calculation of efficiency analysis, it is characterized in that, in described step (), the step that the history real time data forwarded carries out the static load calibration process based on numerical method is as follows:

(1) screen based on capacity of distribution transform; Electric transformer capacity according to assembling judges the load theoretical upper limit I that can reach on circuit_M, its formula is as follows:

In formula: K is the theoretical upper limit of circuit load current; C is filled distribution transformer capacity by circuit;

(2) static accurate real-time measurement consistency desired result; Gather load node historical data be called that static load measures, this static load measure carry out check analysis according to the basic norm of this load node circuit, by do not meet circuit substantially retrain relation measurement removal.

3. the distribution network based on multi-data source according to claim 2 measures the practical method of calculation of efficiency analysis, it is characterised in that, the content of described consistency desired result also comprises: distribution switch PQI does not mate; Feeder line section two ends are meritorious, idle, electric current conflicts mutually; Distribution bus measures inflow and outflow and does not mate; Distribution load measures with on-load switch and does not mate; Distribution switch remote measurement is not corresponding with remote signalling.

4. the distribution network based on multi-data source according to claim 2 measures the practical method of calculation of efficiency analysis, it is characterised in that, described static load measures and comprises wattful power, wattless power, electric current, voltage, power factor.

5. the distribution network based on multi-data source according to claim 1 measures the practical method of calculation of efficiency analysis, it is characterised in that, in above-mentioned steps (two),

Historical data according to this load node predicts gained value by load, and its concrete grammar is as follows:

A () determines network calibration group; Utilizing topological analysis, using state as distribution switch that is that divide and that have measurement as edge device, distribution network feeder is divided into some switch segments, a switch segments is a topological calibration group; Each group calculates the current value flowing into this group and flows out the current value of this group, and the electric current of inflow subtracts the electric current that the electric current flowed out is this group internal loading and consumes; By distribution factor, total load electric current in group is distributed to each load; The load distribution factor calculates according to the non-real time data of other system forwards;

B each group of networks is calculated each burden apportionment factors A F according to the static load calibration value verified by () respectively; The universal calculation equation of distribution factor AF is:

In formula: �� L_inp-meaThe summation of input measurement value in group of networks, comprising: the observed value of 10kV feeder line outlet, the observed value that other group of networks input to this group,

��L_g-meaGroup of networks Small Power exports summation; The summation of the observed value that little power supply inputs to this group of networks;

��I_out-meaThe summation of outputting measurement value in group of networks; Group of networks has the observed value summation of the load node of observed value;

��I_caThe all lifting capacity measured value summations of group of networks;

C the current near-realtime data of load node is undertaken measuring completion by () by following formula, calculate the electricity of the current time of each load node according to the power energy allocation factors A F in historical data same moment, thus the measurement of all devices on completion circuit; The electricity of the current time of this each load node is topology calibration load value, and calculation formula and the state estimation solving equation of this topology calibration load value are as follows:

Topology calibration load=real-time measurement total amount * AF;

Real-time measurement total amount is total electricity that current time all loads node is surveyed;

(d) consistency check; Find out obviously wrong or suspicious measurement data according to Logic judgment, or the defective part of measurement system is supplemented automatically;

E the calibration of () topology should change when network topology, or carry out when yardman asks.

6. the distribution network based on multi-data source according to claim 5 measures the practical method of calculation of efficiency analysis, it is characterized in that, in above-mentioned steps (d), previously described consistence basis for estimation is again utilized to check the real time data predicting load data out and distribution switch; The data determining mistake are revised or filtered; Suspicious data are pointed out in the table, composition suspicious data group; To the data of disappearance, calibrate load value automatic makeup by topology neat;

In above-mentioned steps (d), the basis for estimation of described Logic judgment has: distribution switch PQI does not mate; Feeder line section two ends are meritorious, idle, electric current conflicts mutually; Distribution bus measures inflow and outflow and does not mate; Distribution load measures with on-load switch and does not mate; Distribution switch remote measurement is not corresponding with remote signalling.

7. the distribution network based on multi-data source according to claim 1 measures the practical method of calculation of efficiency analysis, it is characterized in that, in described step (three), according to present node load circuit basic norm, what measure that rough detection refers to each node load within the specific limits judges that the measurement of multiple switch is not mated or uneven or conflict mutually when preload prediction gained value and the current load data value that records.

8. the distribution network based on multi-data source according to claim 7 measures the practical method of calculation of efficiency analysis, it is characterized in that, in described step (three), distribution switch state identification is that the measurement of switch after measurement rough detection is carried out switch identification, obtains the switch divided adjacent in this certain limit;

The switch of described point is divided into the switch divided having measurement and the switch divided without measurement;

Suspicious switch discrimination method is as follows:

The charged the other end in switch one end is not charged, and the switch having the state of measurement to be point is recognized as conjunction; The charged the other end in switch one end is not charged, and without the state measured for dividing, not charged end but has the switch of measurement to be recognized as conjunction with load and load; Loop measures the minimum switch of electric current be recognized as point, on loop without the switch measured and this switch first and last end from electric point away from switch be recognized as point;

Identification switch state, and need after changing switch state to carry out local topology search.

9. the distribution network based on multi-data source according to claim 8 measures the practical method of calculation of efficiency analysis, it is characterized in that, in described step (three), the method of distribution switch bad data detection and identification is: state estimation iterative computation namely estimate calculating converge to iteration convergence error value time, to a certain group of suspicious data, first get the maximum measurement of one of them residual error and carry out identification, forecast that this residual error changes, judge whether it is bad data, when the residual values of the measurement calculated is greater than manual maintenance permission residual values, then it is judged as bad data, otherwise, it it is then normal data, if after detecting out bad data, first estimate that it is correctly worth, and revise up-to-date residual error, again residual error is ranked, again the bad data of identification,

The calculation formula of residual vector is as follows:

V=Z-h (x) ... (1)

Wherein, v is residual vector, and Z is the observed value vector of value to be estimated, and h (x) is non-linear measurement function;

The residual error forecast is the residual error after correct estimation suspicious data I or group of suspicious data, as shown in the formula:

Wherein, r_w ^(I)Weighted residual phasor for identification and after estimating suspicious data; Z_wFor weighting measures phasor; X^(I)Identification amount to be asked after estimating suspicious data, comprise the current value of each distribution switch; H_WIt it is weighting Jacobian matrix;

Correction formula is:

X^(I)=X-(H_w ^TH_w)^-1H_wiW_wii ^-1r_wi

Wherein, H_wiThe weighting Jacobian matrix part that i-th row is relevant to suspicious data; W_wWeighted residual sensitivity matrix; W_wiI-th row represents the weighted residual sensitivity matrix of corresponding suspicious data part; r_wiRepresent the weighted residual of the corresponding suspicious data part of the i-th row; X is unknown quantity;

When preload predicts by weighting measures residual values, gained value and the current load data value recorded judge whether measurement is suspicious data, and its determination methods is as follows:

Mistake measures data and is included in state estimation solving equation, make calculation result by deviation system time of day, reject maximum absolute value value and it is greater than the quantity of state that Operation system setting allows worst error value, if mistake measures after calculating terminates, between quantity of state and side value, absolute value is less than manual maintenance permission residual values, according to the bad data of the size detection of residual values.

10. the distribution network based on multi-data source according to claim 9 measures the practical method of calculation of efficiency analysis, it is characterized in that, the bad data of the size detection of described residual values, it is that after getting rid of bad data by successively type estimation identification method, state estimation calculates no longer again, and is directly revised residual error and unknown quantity to be asked by residual sensitivity matrix and Jacobian matrix;

Described successively type estimates that identification method puts into discrimination method step specific as follows:

A) suspicious data scope is determined;

B) copy matrix of coefficients, remove the residual vector after bad data, matrix of coefficients and right-hand vector are carried out this orthogonal transformation of propitious essay;

State estimation software adopts the mixing algorithm that normal equation combines with orthogonal transformation, only Jacobian matrix Hp and Hq is converted, it is necessary to retain sparse matrix Hp and Hq, and the matrix L after orthogonal transformation;

Gain matrix Hp^TRp^-1Hp and Hq^TRq^-1Hq can be write as (Hpw)^THpw and (Hqw)^THqw form:

(Hpw)^THpw=(QpHpw)^T(QpHpw)=Lp^TLp

(Hqw)^THqw=(QqHqw)^T(QqHqw)=Lq^TLq

State estimation solves iterative equation below:

Lp^TLp����_(l)=a_(l)

Lq^TLq��U_(l)=b_(l)

Orthogonal transformation becomes and is converted by Hpw and Hqw:

With

Qp and Qq is orthogonal matrix, and Hpw is above the matrix of a nxn dimension after conversion, be entirely below 0, Hqw after converting above be the matrix that (n-1) x (n-1) is tieed up, be 0 below entirely; After trying to achieve Lp and Lq matrix, other computation processes are consistent with quick decoupling zero weighted least-squares method; Orthogonal transformation is exactly mainly that Hpw and Hqw adopts this conversion of propitious essay, from cancellation non-zero entry below;

C) successively type bad data identification is started; Circulation measurement matrix corresponds to the part of bad measurement corresponding to the row of bad data, sensitivity matrix, residual vector, temporarily vector, and sensitivity matrix is corresponding to the diagonal angle part of bad measurement;

D) the suspicious residual error r measuring correspondence that can detect_s, by the relevant part W of the residual sensitivity matrix of residual error equation r=Wv_s, corresponding error in measurement v can be tried to achieve_s, obtain estimating the identification formula of identification method:

v_s=(W_s ^TG^-1W_s)W_s ^TG^-1r_s

In formula: v_sThe error in measurement vector that can detect; W_sThe relevant part of corresponding suspicious measurement residuals sensitivity matrix; G^-1Weighting diagonal matrix, G^�C1=(H^TR^-1H)^-1,

Estimate the weighted type formula of identification method:

v_ws=(W_ws ^TG^-1W_ws)W_ws ^TG^-1r_ws

v_wsSuspicious measurement weighted error vector; r_wsSuspicious measurement weighted residual vector; W_wsThe relevant part of corresponding suspicious measurement weighted residual sensitivity matrix;

E) weighted residual vector is upgraded;

Identification suspicious data rear weight residual error forecast formula:

r_w ^(I)=r_w+H_w(H_w ^TH_w)^-1H_wiW_wii ^-1r_wi

And other measure weighted residual forecast formula:

r_wj ^(I)=r_wj-W_wijW_wii ^-1r_wi(j �� I, j �� I)

Wherein, H_wiThe weighting Jacobian matrix part that i-th row is relevant to suspicious data; W_wWeighted residual sensitivity matrix; W_wiThe weighted residual sensitivity matrix of corresponding suspicious data part; r_wiThe weighted residual of corresponding suspicious data part; W_wijRepresent the element value of power residual sensitivity matrix i-th row jth row; I representation unit matrix; I, j belong to natural number;

F) next raw data detection, until the maximum residual values in all suspicious datas is less than the threshold value that weighting measures residual detection.