CN104537271A - Power distribution network bad data identification method based on mass tags - Google Patents

Abstract

The invention discloses a method for identifying bad data in distribution networks based on quality labels, which belongs to the field of calculation and analysis of electric power systems. Under the condition of obtaining multi-source data, voltage measurement U, current measurement I, active power measurement P, and active power measurement are used. The constraint rules of the 6 influencing factors of electricity and reactive power are used to evaluate the corresponding data quality score score. After the evaluation, the quality score value of each quantity measurement is used to calculate the quality label value. According to the calculation results, the defective data. It overcomes the shortcomings of the traditional data identification method that the local data errors are spread to all nodes, and comprehensively utilizes the redundant measurement data of the upper and lower voltage levels, combined with the quality label, to identify bad data and improve the virtual quantity of the distribution network. The quality of the input data for measurement and state estimation.

Description

A method for identifying bad data in distribution network based on quality label

technical field

The invention relates to a method for identifying bad data of a distribution network based on a quality label, and belongs to the field of power system calculation and analysis. More specifically, through a distribution network data identification method based on quality labels, it is a practical data identification method to detect and identify bad data in the distribution network. State estimation provides data support.

Background technique

In the past, the distribution network lacked data measurement equipment and monitoring means. In this case, the data quality identification of the distribution network was often transformed into a power flow matching problem based on a series of assumptions. With the continuous implementation of power system automation, there are new requirements for the quality identification of distribution network data. Distribution network data quality identification requires two types of data, namely measurement data and line parameter data. The measurement data includes remote signaling data and telemetry data. The former mainly comes from distribution network telesignaling information, which is the basis for distribution network topology analysis. The latter mainly comes from distribution network telemetry data. Redundant data such as historical data and electricity measurement data are used as the basis for data quality identification to meet the requirements for data redundancy. Since the measurement data of distribution network is huge and contains bad data, corresponding data filtering technology is needed for data preprocessing, that is, to realize a practical method for identifying bad data of distribution network, so as to provide virtual measurement and State estimation provides data support.

In view of this, the inventor conducted research on this, and specially developed a method for identifying bad data in distribution networks based on quality labels, and this case arose from this.

Contents of the invention

The purpose of the present invention is to provide a method for identifying bad data in distribution networks based on quality labels. In view of the current distribution automation pilot area measurement configuration and the current status of multiple data sources, the measurement data of the distribution network can be identified before estimation. It overcomes the shortcomings of the traditional data identification method that the local data errors are spread to all nodes, and comprehensively utilizes the redundant measurement data of the upper and lower voltage levels, combined with the quality label, to identify bad data and improve the virtual quantity of the distribution network. The quality of the input data for measurement and state estimation.

In order to achieve the above object, the solution of the present invention is:

A method for identifying bad data in distribution networks based on quality labels, comprising the following steps:

1) Acquisition of multiple data sources: Under the condition of knowing the network structure CIM of the entire power grid, combined with the real-time power and voltage obtained from the feeder FTU and the bus RTU, the load data of different types of users are derived from the marketing system (CIS) and the centralized copying system, including Voltage, current, instantaneous active power, instantaneous reactive power, active power measurement and reactive power measurement form the data source for subsequent quality label formulation;

2) Quality score evaluation of each quantity measurement: Under the condition of obtaining multi-source data, the constraint rules of six influencing factors of voltage measurement U, current measurement I, active power measurement P, active power measurement and reactive power measurement are used to The corresponding data quality score is evaluated;

3) Quality label formulation: use the quality score value measured by each quantity to calculate the quality label value, and identify bad data according to the calculation results. The larger the calculated value, the better the data, and vice versa. .

The constraint rules described in the above step 2) are specifically:

i. Voltage U constraint:︱

Rule 1: Data continuity: the current voltage does not exceed the threshold compared with the voltage in the first 15 minutes and the next 15 minutes, that is,︱UU _-15︱≤δ , and︱UU ₊ 15︱≤δ;

Rule 2: Periodic law: the comparison between the same time within a week does not exceed the threshold (the error between the current voltage and the voltage average value at the same time within a week cannot exceed the threshold), that is

Rule 3: Rated voltage range comparison: meet within the set ratio of the rated voltage (this ratio can be adjusted according to the actual situation, generally within 10% of the rated voltage), that is, U∈[a,b];

Rule 4: Compared with the voltage of the previous device: all voltages are less than the bus voltage, that is, ︱UU _b ︱≤σ;

ii. Current I constraint:

Rule 1: Data continuity: Compared with the current current in the first 15 minutes and the next 15 minutes, the current current does not exceed the threshold, ︱II _-15 ︱≤δ, and ︱II ₊₁₅ ︱≤δ;

Rule 2: Periodic law: the comparison of the same time within a week does not exceed the threshold (the error between the current current and the current average value at the same time within a week cannot exceed the threshold), that is

Rule 3: Compared with the outlet circuit breaker: the current value on each switch is less than the current value of the outlet circuit breaker, that is, ︱II _b ︱≤σ;

Rule 4: The KCL law verification of the current, after the topology shrinks, assign the current value I to each line, select the current value of the switch adjacent to the line (not exceeding the T node), and then traverse each CN node for KCL verification , judge twice for each current I, if both times are correct between these two times, then the assignment is 1, one-to-one wrong assignment is 0.5, and both are wrong assignment is 0;

iii. Active power P constraints:

Rule 1: Data continuity: the current active power does not exceed the threshold compared with the active power of the first 15 minutes and the next 15 minutes, that is,︱PP _-15︱≤δ , and︱PP ₊ 15︱≤δ;

Rule 2: Periodic law: the comparison of the same time within a week does not exceed the threshold (the error between the current active power and the average value of active power at the same time within a week cannot exceed the threshold), that is

iv. Constraints on active energy:

Rule 1: |active energy*time-electricity|≤δ;

v. Reactive power Q constraint:

Rule 1: Data continuity: The comparison between the current reactive power and the reactive power of the first 15 minutes and the next 15 minutes does not exceed the threshold, that is,︱QQ _-15︱≤δ , and︱QQ _+15︱≤δ ;

Rule 2: Periodic law: the comparison of the same time within a week does not exceed the threshold (the error between the current reactive power and the average value of reactive power at the same time within a week cannot exceed the threshold), that is

vi. Constraints on reactive energy:

Rule 1: |reactive energy*time-electricity|≤δ;

Judging by the classification of the constraint rules of the above six factors, different quality scores are calculated for different quantity measurements. The quality scores obtained by each constraint rule are 0 to 1. The specific calculation formula is:

[1] Data continuity calculation score

Formula: score＝1-(|measurement-measurement in the first 15 minutes|/measurement in the first 15 minutes*threshold+|measurement-measurement in the next 15 minutes|/measurement in the last 15 minutes*threshold)/2;

If the calculated quality score score is not in the range [0,1], the score is calculated as 0;

Among them, the threshold is 0.5;

[2] Historical law calculation score:

Formula: score=1-|quantity measurement-average|/(average*threshold);

If the calculated quality score score is not in the range [0,1], the score is calculated as 0;

Among them, the active power P, the reactive power Q, and the threshold value of the current I are all taken as 0.5, and the voltage U is 0.1.

[3] Calculate the fractional drop in voltage along the line:

Formula: score=1-|measurement-voltage of the previous device|/(voltage of the previous device*threshold);

Among them, the threshold is 0.1, and the previous device can be considered as a device closer to the power supply in the radial network;

If the calculated quality score score is not in the range [0,1], the score is calculated as 0;

[4] Comparing the voltage with the rated voltage to calculate the score:

Formula: score=1-|measurement-rated voltage|/(rated voltage*threshold);

Among them, the threshold value is 0.1, and the rated voltage is the rated voltage of the voltage level where the quantity is measured;

If the calculated quality score score is not in the range [0,1], the score is calculated as 0;

[5] Current vs outlet breaker comparison (range) calculation score:

Formula: A score greater than the outlet circuit breaker is 0, and a score smaller than the outlet circuit breaker is 1.

[6] Calculation score of current KCL law verification:

Note: The KCL law of each current needs to be tested twice, so the upper limit of each score is 0.5 points;

Calculation formula for each KCL calibration: mark=0.5-|current-reference current|/(reference current*threshold);

If the calculated score is not in the range [0,0.5], the score will be calculated as 0;

Then add the scores on both sides to get the score of KCL law;

The calculation of the reference current is as follows: if the measurement of the quantity to be checked is the largest among all the measurements of the node, the reference current is the sum of all other currents; if it is not the largest, the reference current = the maximum current - the sum of all other currents (except nuclear current);

The threshold is 0.1.

[7] Calculation score of active and reactive energy:

Formula: score＝1-|P(Q)*time-with (without) power level|/(with (without) power level)*threshold);

Among them, the threshold is 0.2, and the time varies according to the number of measurement points per hour;

If the calculated quality score score is not in the range [0,1], the score is calculated as 0.

The calculation formula of the quality tag value described in the above step 3) is:

The quality score score with a value range of [0,1] is obtained through each constraint rule. Each constraint rule is weighted by the ID3 decision tree classification algorithm according to the sample. The sum of the weights of each constraint rule measured by the same quantity is 1, and the quantity Measured quality label value: Q(X)=∑each (weight corresponding to quality score score*), and its score is also in the range of [0,1].

The method for identifying bad data in distribution networks based on quality labels in the present invention uses voltage measurement U, current measurement I, active power measurement P, active power measurement and reactive power measurement under the condition of acquiring multi-source data The constraint rules of the six influencing factors evaluate the corresponding data quality score score. After the evaluation, the quality score score value measured by each quantity is used to calculate the quality label value. According to the calculation result, bad data is identified, and the calculated value is higher. Larger means better data, otherwise, worse. Has the following advantages:

1) It overcomes the residual submersion problem of existing bad data detection and identification, and greatly reduces the probability of misjudgment of data;

2) It overcomes the drawbacks of the traditional data identification method, which spreads local data errors to all nodes, and comprehensively utilizes redundant measurement data of upper and lower voltage levels, combined with quality labels, to identify bad data and greatly improve power distribution. Quality of input data for network virtual measurement and state estimation.

Detailed ways

A method for identifying bad data in distribution networks based on quality labels, comprising the following steps:

Step 1: Acquisition of multiple data sources: Under the condition of knowing the network structure CIM of the entire power grid, combined with the real-time power P and voltage U obtained from the feeder FTU and bus RTU, the loads of different types of users are derived from the marketing system (CIS) and centralized copying system Data, including voltage U, current I, instantaneous active power P, instantaneous reactive power Q, active energy measurement and reactive energy measurement, form the data source for subsequent quality label formulation;

Step 2: Evaluation of the quality scores of each quantity measurement: Under the condition of obtaining multi-source data, use the constraint rules of six influencing factors: voltage measurement U, current measurement I, active power measurement P, active power measurement and reactive power measurement Evaluate the corresponding data quality score score;

The constraint rules are specifically:

i. Voltage U constraints:

Rule 1: Data continuity: the current voltage does not exceed the threshold compared with the voltage in the first 15 minutes and the next 15 minutes, that is,︱UU _-15︱≤δ , and︱UU ₊ 15︱≤δ;

Rule 2: Periodic law: the comparison between the same time within a week does not exceed the threshold (the error between the current voltage and the voltage average value at the same time within a week cannot exceed the threshold), that is

Rule 3: Rated voltage range comparison: meet within the set ratio of the rated voltage (this ratio can be adjusted according to the actual situation, generally within 10% of the rated voltage), that is, U∈[a,b];

Rule 4: Compared with the voltage of the previous device: all voltages are less than the bus voltage, that is, ︱UU _b ︱≤σ;

ii. Current I constraint:

Rule 1: Data continuity: the current current does not exceed the threshold compared with the current in the first 15 minutes and the next 15 minutes, that is,︱II _-15︱≤δ , and︱II ₊ 15︱≤δ;

Rule 2: Periodic law: the comparison of the same time within a week does not exceed the threshold (the error between the current current and the current average value at the same time within a week cannot exceed the threshold), that is

Rule 3: Compared with the outlet circuit breaker: the current value on each switch is less than the current value of the outlet circuit breaker, that is, ︱II _b ︱≤σ;

Rule 4: The KCL law verification of the current, after the topology shrinks, assign the current value I to each line, select the current value of the switch adjacent to the line (not exceeding the T node), and then traverse each CN node for KCL verification , judge twice for each current I, if both times are correct between these two times, then the assignment is 1, one-to-one wrong assignment is 0.5, and both are wrong assignment is 0;

iii. Active power P constraints:

Rule 1: Data continuity: the current active power does not exceed the threshold compared with the active power of the first 15 minutes and the next 15 minutes, that is,︱PP _-15︱≤δ , and︱PP ₊ 15︱≤δ;

Rule 2: Periodic law: the comparison of the same time within a week does not exceed the threshold (the error between the current active power and the average value of active power at the same time within a week cannot exceed the threshold), that is

iv. Constraints on active energy:

Rule 1: |active energy*time-electricity|≤δ;

v. Reactive power Q constraint:

Rule 1: Data continuity: The comparison between the current reactive power and the reactive power of the first 15 minutes and the next 15 minutes does not exceed the threshold, that is,︱QQ _-15︱≤δ , and︱QQ _+15︱≤δ ;

Rule 2: Periodic law: the comparison of the same time within a week does not exceed the threshold (the error between the current reactive power and the average value of reactive power at the same time within a week cannot exceed the threshold), that is

vi. Constraints on reactive energy:

Rule 1: |reactive energy*time-electricity|≤δ

Judging by the classification of the constraint rules of the above six factors, different quality scores are calculated for different quantity measurements. The quality scores obtained by each constraint rule are 0 to 1. The specific calculation formula is:

[1] Data continuity calculation score

Formula: score＝1-(|measurement-measurement in the first 15 minutes|/measurement in the first 15 minutes*threshold+|measurement-measurement in the next 15 minutes|/measurement in the last 15 minutes*threshold)/2;

If the calculated quality score score is not in the range [0,1], the score is calculated as 0;

Among them, the threshold value is 0.5.

[2] Historical law calculation score:

Formula: score=1-|quantity measurement-average|/(average*threshold);

If the calculated quality score score is not in the range [0,1], the score is calculated as 0;

Among them, the active power P, the reactive power Q, and the threshold value of the current I are all taken as 0.5, and the threshold value of the voltage U is 0.1.

[3] Calculate the fractional drop in voltage along the line:

Formula: score=1-|measurement-voltage of the previous device|/(voltage of the previous device*threshold);

Among them, the threshold is 0.1, and the previous device can be considered as a device closer to the power supply in the radial network;

If the calculated quality score score is not in the range [0,1], the score is calculated as 0;

[4] Comparing the voltage with the rated voltage to calculate the score:

Formula: score=1-|measurement-rated voltage|/(rated voltage*threshold);

Among them, the threshold value is 0.1, and the rated voltage is the rated voltage of the voltage level where the quantity is measured;

If the calculated quality score score is not in the range [0,1], the score is calculated as 0;

[5] Current vs outlet breaker comparison (range) calculation score:

Formula: A score greater than the outlet circuit breaker is 0, and a score smaller than the outlet circuit breaker is 1.

[6] Calculation score of current KCL law verification:

Note: The KCL law of each current needs to be tested twice, so the upper limit of each score is 0.5 points;

Calculation formula after each KCL calibration: mark=0.5-|current-reference current|/(reference current*threshold);

If the calculated score is not in the range [0,0.5], the score will be calculated as 0;

Then add the scores on both sides to get the score of KCL rule;

The calculation of the reference current is as follows: if the measurement of the quantity to be checked is the largest among all the measurements of the node, the reference current is the sum of all other currents; if it is not the largest, the reference current = the maximum current - the sum of all other currents (except nuclear current);

The threshold is 0.1.

[7] Calculation score of active and reactive energy:

Formula: score＝1-|P(Q)*time-with (without) power level|/(with (without) power level)*threshold);

The threshold is 0.2, and the time varies according to the number of measurement points per hour;

If the calculated quality score score is not in the range [0,1], the score is calculated as 0;

Step 3: Formulate quality label: Use the quality score score of each quantity measurement to calculate the quality label value. According to the calculation result, identify bad data. The smaller the calculated value, the worse the data. In this embodiment, the voltage measurement U is taken as an example to calculate the quality label value:

Through the four detection rules of the voltage measurement U, four quality scores with a value range of [0,1] are obtained, and each rule is weighted a1, a2, a3, a4 by the ID3 decision tree classification algorithm according to the sample, where a1+a2+a3+a4=1,

The quality label value of the voltage quantity measurement U: Q(U)=rule 1 score*a1+rule 2 score*a2+rule 3 score*a3+rule 4 score*a4. The quality label value of the voltage measurement U is also in the range of [0,1], which is used to set the judging criteria according to the requirements. As shown in Table 1, the larger the quality label value, the better the measured value data, and vice versa. is worse.

Table 1: Judging criteria for the voltage quantity measurement U quality label value of the present embodiment:

excellent good generally poor extremely bad 0.8-1.0 0.6-0.8 0.4-0.6 0.2-0.4 0.0-0.2

The quality label value of the remaining quantity measurement can be similarly obtained according to the above steps.

The above-mentioned embodiments do not limit the form and style of the product of the present invention, and any appropriate changes or modifications made by those skilled in the art should be considered as not departing from the patent scope of the present invention.

Claims (4)

1., based on a power distribution network bad data recognition method for quality tab, it is characterized in that comprising the steps:

1) multi-data source obtains: knowing under full electric network network structure CIM condition, in conjunction with the realtime power obtained from feeder line FTU and bus RTU and voltage, the load data of dissimilar user is derived from marketing system, centralized meter-reading system, comprise voltage, electric current, instantaneous active, instantaneous reactive, active electrical degree amount and idle electricity, form the data source that postorder quality tab is formulated;

2) each measurement amount massfraction evaluation: under the condition obtaining multi-source data, utilizes voltage measurement U, the magnitude of current measures I, the constraint rule of gain merit measurement amount P, active electrical degree amount and idle electricity 6 factors of influence is evaluated corresponding quality of data mark score;

3) quality tab is formulated: utilize the massfraction score value of each measurement amount to carry out the calculating of quality tab value, according to result of calculation, pick out bad data, the value calculated is larger, illustrates that these data are better, otherwise, then poorer.

2. a kind of power distribution network bad data recognition method based on quality tab as claimed in claim 1, is characterized in that: step 2) described in constraint rule be specially:

Voltage U retrains:

Rule 1: data continuity: current voltage was no more than threshold value, i.e. ︱ U-U with first 15 minutes with the voltage compare of latter 15 minutes _-15︱≤δ, and ︱ U-U _{+ 15}︱≤δ;

Rule 2: periodic law: one week interior synchronization compares and is no more than threshold value, namely

Rule 3: the range of nominal tension compares: meet within the setting ratio of rated voltage, i.e. U ∈ [a, b];

Rule 4: with the voltage compare of previous equipment: all voltage is less than busbar voltage, i.e. ︱ U-U _b︱≤σ;

Electric current I retrains:

Rule 1: data continuity: current flow is no more than threshold value, ︱ I-I with comparing with the electric current of latter 15 minutes for first 15 minutes _-15︱≤δ, and ︱ I-I _{+ 15}︱≤δ;

Rule 2: periodic law: one week interior synchronization compares and is no more than threshold value, namely

Rule 3: compare with outlet breaker: the current value on each switch is less than the current value of outlet breaker, i.e. ︱ I-I _b︱≤σ; Rule 4: the KCL rule verification of electric current, after topology is shunk, for every bar circuit composes current value I, the current value of the switch that access line is adjacent, travels through each CN node afterwards and carries out KCL verification, judge twice to each electric current I, if be all for twice right between this twice, then assignment is 1, and wrong assignment is 0.5 one to one, and all wrong assignment is 0;

Active-power P retrains:

Rule 1: data continuity: current active power is no more than threshold value, i.e. ︱ P-P with comparing with the active power of latter 15 minutes for first 15 minutes _-15︱≤δ, and ︱ P-P _{+ 15}︱≤δ;

Rule 2: periodic law: one week interior synchronization compares and is no more than threshold value, namely

Active electrical degree amount retrains:

M-electricity during rule 1:| active electrical degree amount * |≤δ;

Reactive power Q retrains:

Rule 1: data continuity: current reactive power is no more than threshold value, i.e. ︱ Q-Q with comparing with the reactive power of latter 15 minutes for first 15 minutes _-15︱≤δ, and ︱ Q-Q _{+ 15}︱≤δ;

Rule 2: periodic law: one week interior synchronization compares and is no more than threshold value, namely

Idle electricity constraint:

M-electricity during rule 1:| idle electricity * |≤δ;

Judged by the constraint rule classification of above 6 factors, calculate different massfraction score to different measuring meters, the massfraction score that each constraint rule obtains is 0 ~ 1, and specific formula for calculation is:

[1] data continuity calculates mark

Formula: score=1-(| measurement amount-front 15 minute volume (MV)s are measured |/front 15 minute volume (MV)s measurement * threshold values+| measurement amount-rear 15 minute volume (MV)s are measured | and/rear 15 minute volume (MV)s measure * threshold values)/2;

If the massfraction score calculated is not in [0,1] scope, then score calculates according to 0;

Wherein, threshold value is 0.5;

[2] historical law calculates mark:

Formula: score=1-| measurement amount-mean value |/(mean value * threshold value);

If the massfraction score calculated is not in [0,1] scope, then score calculates according to 0;

Wherein, active-power P, reactive power Q, the threshold value of electric current I is all taken as 0.5, and voltage U threshold values is 0.1;

[3] voltage reduces calculating mark along circuit:

Formula: the voltage of score=1-| measurement amount-previous equipment |/(the voltage * threshold value of previous equipment);

Wherein, threshold value is 0.1, and previous equipment can think the equipment closer to power supply in radial net;

If the massfraction score calculated is not in [0,1] scope, then score calculates according to 0;

[4] voltage compares calculating mark with rated voltage:

Formula: score=1-| measurement amount-rated voltage |/(rated voltage * threshold value);

Here wherein, threshold value is 0.1, and rated voltage is the rated voltage of the electric pressure at this measurement amount place;

If the massfraction score calculated is not in [0,1] scope, then score calculates according to 0;

[5] electric current compares (scope) and calculates mark with outlet breaker:

Formula: be greater than outlet breaker and must be divided into 0, is less than outlet breaker and must be divided into 1;

[6] the KCL rule verify calculation mark of electric current:

Illustrate: the KCL rule of each magnitude of current needs detection twice, so the upper limit of each score is 0.5 point;

Computing formula after each KCL verification: mark=0.5-| electric current-reference current |/(reference current * threshold value);

The mark calculated is not in [0,0.5] scope, then score calculates according to 0;

Then the score of both sides is added the score obtaining KCL rule;

Being calculated as follows of reference current: if it is maximum to treat that check amount is measured as in the middle of node all measurements amount, then reference current be other all electric current and; If not maximum, reference current=maximum current-other all electric current and (remove wait check electric current); Threshold value is 0.1;

[7] electricity of active reactive calculates mark:

M-there is (no) merit electric degree during formula: score=1-|P (Q) * |/(there is (no) merit electric degree) * threshold value);

Wherein, threshold value is 0.2, and the time is different according to the number of the gauge point of each hour;

If the massfraction score calculated is not in [0,1] scope, then score calculates according to 0.

3. a kind of power distribution network bad data recognition method based on quality tab as claimed in claim 1, is characterized in that: step 3) computing formula of described quality tab value is:

It is [0 that each constraint rule obtains span, 1] massfraction score, each constraint rule obtains its weight by ID3 Decision Tree Algorithm according to sample, each constraint rule weight sum of same measurement amount is 1, the quality tab value of measurement amount: Q (X)=∑ each (weight that massfraction score* is corresponding), its score is also in [0,1] scope.

4. a kind of power distribution network bad data recognition method based on quality tab as claimed in claim 2, is characterized in that: the range of nominal tension of described voltage U constraint rule 3 is more specific is: meet within 10% of rated voltage.