JPH0677036B2 - Error measurement value detection method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a method for detecting an error measurement value, and particularly to an error measurement suitable for detecting a small number of error measurement values present in measurement data of a power system. Value detection method.

[Background of the Invention]

It is well known that the value measured from the measured object generally includes an error.

Among them, the one with the largest error (that is, the error measurement value)
It is indispensable to remove from the measured value beforehand in order to perform various monitoring and control. Even in the power system, it is indispensable to grasp the accurate state of the system when operating the system, and the measured data such as the power and voltage of each part of the system are transmitted and collected at the power supply command center. However, since this transmitted data contains errors due to measurement and transmission, it is necessary to perform some estimation calculation based on these measured data to obtain the system state with higher accuracy. There is. Here, in addition to the usual small error (noise), the error of the measurement data may include a large error (error measurement value) that is rarely overlooked as noise. The value significantly impairs the accuracy of the estimation result. Therefore, in the estimation calculation, it is necessary to remove the influence of this error measurement value in advance, and for that purpose, it is necessary to know what the error measurement value is.

The method of detecting such an error measurement value has been applied to a power system and is described in “Processing Error Measurement Data for Real-Time State Estimation of Power System” by Monticelli and Garcia. Volume 102, 1126-
1139, 1983 ('Reliable Bad Data Processing for Rea
l-Time State Estimation ′ A. Monticelli, A. Garcia, IE
EE Trans., PAS-102, 1126-1139, 1983) ”.

In this method, the influence (sensitivity) on the estimated residual (difference between the measured value and the corresponding estimated value) with respect to the measured value itself when each measured value fluctuates by a small amount is obtained, On the contrary, when the estimated residual of each estimated value is given, which measured value is incorrect is determined based on the sensitivity. If there is only one error in the observation data and the observation error (noise) that is not enough to consider it as an error is not included, the estimated residual is It is theoretically guaranteed that the measured value with the maximum value corrected by the sensitivity can be accurately determined as an erroneous measured value.

However, noise in observation data is generally unavoidable, and complicated calculations are required to obtain the above-mentioned sensitivity, and especially when the scale of the system increases, the amount of calculation increases significantly. It was not practical for surveillance etc. In addition, depending on the nature of the system, the estimated value could not be obtained with sufficiently high accuracy, and as a result, the estimated residual after correction was not always large with respect to the error measurement value.

In short, although the above-mentioned conventional method requires complicated calculation, there is no guarantee that the error measurement value can be adequately determined in practical use.

[Object of the Invention]

It is an object of the present invention to solve the above-mentioned disadvantages of the conventional method and to provide an efficient and highly reliable method of detecting an error measurement value.

[Outline of Invention]

In order to achieve the above object, the present invention estimates the difference between the estimated value derived from the measured value from the measured system consisting of each node of the power system and the branch connecting these nodes, and the measured value. On the basis of the residual, in the method of detecting an error measurement value for detecting an error measurement value from the measurement value,
The sum of the estimated residuals obtained based on the measured value at each node, or the sum of the estimated residuals at both ends of each branch obtained based on the measured value is detected to exceed a predetermined value, and the measured value The error measurement value is specified from among these.

In the present invention, by performing the rationality check only on the nodes or branches where the difference between the estimated value and the measured value (estimated residual) becomes large, and the measurement points in the vicinity thereof, the reliability of error detection can be improved. On the other hand, the efficiency of the process is improved, but this is performed based on the facts shown in the following items (1) to (3).

(1) Error measurement values are a very small number of exceptions and do not appear frequently in the measurement data used for estimation.

(2) Although the estimation residual is generated due to the error included in the measurement data, the estimation residual becomes large when the error measurement value is included, and particularly the estimation residual becomes large at the measurement point and its periphery. Conversely, if the estimated residual is not large for any of the estimated values, the error measurement value is not included. This will be described below using mathematical expressions.

In general, the measured value is represented by the equation (1).

y = f (xt) + ε (1) y: measured value (vector) xt: true value (vector) of the independent variable (for example, node complex voltage) f (•): theoretical formula (vector) of measured value ε: Error of measured value (vector) In the estimation calculation, when y is given, x′t is first obtained, and then as an estimated value Calculate with. Therefore, if the measurement data is completely correct (ε = 0 in the equation (1)) and the estimation calculation based on it is also performed properly, the measurement value and the estimation value are completely in agreement and there is no estimation residual.

However, the measured value actually contains noise (ε
≠ 0), the x't obtained by the estimation also differs from xt. The estimation method will be described later (in the examples). However, if ε is small, this difference is not so large that it becomes a problem, and the estimated residual is small.

However, if any one of ε, for example, the i-th measured value | εi | is larger than a predetermined value, xt obtained by estimation
And x't are large. In particular, the difference is significant in the strongly related measurement values yi with large errors, and as a result,
The estimated residual is large with respect to the estimated values around yi including yi.

(3) The estimated residuals that are strongly influenced by the error measurement value and are particularly large are the estimated values that are closely related to each other. For example,
When the error measurement value is active power, the residual of active power is large mainly. On the contrary, in the case of reactive power, the residual difference of reactive power becomes large. That is, active power and reactive power are
The residuals due to erroneous measurements tend to close together.

Therefore, according to the terms (1) and (2), the rationality check for detecting an erroneous measurement value is performed when the estimated residual becomes large. The goal may be to find an error measure. Further, according to the above item (3), the rationality check may be performed on the larger of the estimated residuals of the active power and the reactive power.

For example, consider a measurement value around the node N in the system as shown in FIG. Where y _{1 to} y ₃ and Are measured and estimated values of the active power injected into the node N or the active power of the branch, respectively.

Here, when i (i = 1,2,3) is the measured value of the electric power going in and out of the node N, and xt and x′t are the true value and the estimated value of the voltage vector, respectively, the following equations hold.

Here, fi (•) is mainly the expression of Kirchhoff's law using xt (or x′t). Therefore, since the estimated value expressed by the equation (3) naturally satisfies Kirchhoff's law, The sum of the estimated values of active power that goes in and out of the node N is 0
Becomes That is, On the other hand, regarding the measured value expressed by the equation (2), since εi is a random value, Also, δ _{N in the following} equation (6), which is the absolute value of the sum of the estimated residuals, does not normally become zero.

Here, if εi is small, δ _{N in the} equation (6) is small, but if any of εi is large (that is, if an error measurement value is included), δ _N is large. . That is, by checking the balance of active power entering and leaving the node N under a certain threshold based on the equation (6), it is possible to determine whether or not there is an error in the measurement value related to it.

Also, the validity of the estimated values at both ends can be evaluated from the "transmission loss" δm _O , δm _E obtained by adding the measured values and estimated values at both ends of the branch. For example, for branch (m), And Here, if δm (= δm _O −δm _E ) is small, then y _{1 at} both ends
It can be considered that there is no major contradiction in the value of y ₈ and no error is included. However, if δm is large, it can be said that one of the measured values at both ends is erroneous because the mutual contradiction is large.

In the present invention, the rationality check between the estimated residuals is appropriately performed based on the equations (6) and (9). Then, there is no error in the measured value corresponding to the part that does not cause a contradiction, and conversely, there is a possibility that the measured value may include an error in the part with a large contradiction.
The minimum necessary measurement data that can explain this contradiction is taken as an error measurement value. Here, the efficiency of the processing can be improved by limiting the target of the rationality check to the portion where the estimated residual is large and its periphery.

Example of Invention

 An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a flow chart shown to explain an embodiment of the present invention. In FIG. 1, data necessary for estimation calculation (system connection state, impedance and power measurement values, etc.) are fetched in the process of block A. Based on this, the estimation calculation is performed in block B. In the least-squares method often used for estimation calculation, an estimated value is obtained so that the sum of squares of residuals between each measured value and each estimated value becomes the minimum. The estimated value is normally represented by the (complex) voltage x't of the node as an independent variable as shown in the equation (3), and x't is first calculated at the time of estimation. Since the estimated value is calculated so as to match each measured value as much as possible, if there is a measured value (erroneous measured value) that greatly differs from the true value, it is necessary to appropriately detect and remove it during estimation. A broken line block C is a portion related to the present invention. First, in block D, the absolute value of the estimated residual (which is the difference between each measured value and the corresponding estimated value) is calculated. Each result is compared in block E with a preset threshold. If none of the residuals is greater than the threshold value, it is determined that there is no error measurement value, the estimation result of block B does not need to be corrected, and processing such as displaying the estimation result is performed in block T and the process ends.

If the residual is larger than the threshold value, substantial processing of error detection is performed. This will be described by taking the case where the measured value y ₂ of the system in FIG. 6 is incorrect as an example. When y ₂ is erroneous, the estimated residual is large not only for y ₂ but also for adjacent y ₄ and y ₃ . As a result, in block F, node N becomes (measured value of active power injection y ₃
Will be picked up as a large residual). Y ₁ , y ₂ and y ₃ enter and leave node N, but y ₂ is incorrect, so
As shown in the equation (6), the absolute value of the sum of the estimated residuals exceeds the preset threshold (block G,
H). That is, it can be seen that any one of y ₁ , y ₂ and y ₃ is incorrect. Next, of the l, m2 branches connected to the node N, the absolute value of the sum of residuals of y ₁ and y ₈ at both ends of the branch m is less than or equal to a threshold value, and y ₂ and y of the branch l. _Four
Since the absolute value of the sum of the residuals of exceeds the threshold (block K,
L), it turns out that y ₂ on the node N side is erroneous (block M). Therefore, after displaying the error measurement value y ₂ (block N), the estimation calculation (block B is performed again).

Here, it is obvious that even if the error measurement value is not y ₂ but y ₃ or y ₄ , for example, it can be determined by the processing procedure shown in FIG. Further, as shown in FIG. 1, in addition to checking the connection branch based on the node, a method of checking the nodes at both ends or one end from the branch, or a combination thereof can also be utilized as a modification of this embodiment.

FIG. 2 is a flow chart showing another embodiment of the present invention.

The flowchart shown in FIG. 2 is different from the flowchart shown in FIG. 1 in that instead of blocks K and L, K ′,
L ', and R is newly provided (the portion marked with "'" in the block symbol indicates that the processing is similar to the block formula in FIG. 1 but different).

Similar to the above-described embodiment, the processing when y _{2 in} FIG. 6 is erroneous will be described with reference to FIG. (Note that up to the block H, it is the same as in FIG. ₁ , and it is known that any one of y ₁ , y ₂ , and y ₃ is incorrect.) In the block K ′, first, the measured value y ₂ ( In order to remove the effect of (estimated residual) of node M and N and branch l, node (N1) set by integrating (6) is checked (in this case, only y ₂ is incorrect, so check). (Because the error is removed from the target of 1.), it is determined that there is no error (block L '). I.e. y ₁ , y ₃ ,
Since there is no error in y ₅ and y ₆ , it can be seen that y ₂ is an error among y ₁ , y ₂ and y ₃ which are candidates up to block H (block M).

Also in this embodiment, as in FIG. 1, the error is not y ₂ .
It is obvious that even y ₃ and y ₄ can be determined by the processing procedure shown in FIG. Of course, the embodiments shown in FIGS. 1 and 2 may be combined and utilized depending on the situation. Further, the grouping of nodes is not limited to those such as Nl and Nm shown in FIG.
May be one. In this case, it can be determined whether the error is inside the integrated node or on the side of the outer branch.

FIG. 4 shows auxiliary means for complementing the measured value used when the measured value is missing in the present invention, and by using this auxiliary means, the practicality can be further improved. That is, in (a), when there is no measured value on the N side of the branch l when obtaining the sum of electric power going in and out of the node N, y ₄ on the opposite (node M) side is substituted. Also, in (b),
If there is no measured value in the branch l, the node N and M are combined (shorted in the branch l) to check the virtual node NM.

Even if it is not possible to identify which measurement value is incorrect only by the above-mentioned rationality check, it is limited to a certain extent by the rationality check and one measurement value with a high possibility is temporarily removed, and estimation is performed again to find the residual error. It is possible by observing the change of. That is, if the erroneous measurement values are properly removed, the residuals will be small with respect to the other estimated values, and thus the sum, after the estimation again, otherwise they will not be reduced. If you make a mistake, you must restore the deleted measurement values and try another candidate. As a candidate of the error measurement value, as a result of the rationality check, it is possible to give priority to an error candidate having a large residual error. In the above-mentioned example, the case where there is only one error measurement value has been illustrated, but it is of course possible to limit or specify them even in the case of a plurality of cases. As one method, as in the case of only one, it is sufficient to sequentially remove the one determined as an error.

To exclude multiple measurements at once, it is useful to distinguish between residuals that are not related to each other. This will be described with reference to FIG.

In FIG. 5, it is assumed that any of the eight measured values (1) to (3) related to the rationality checks Z _{1 to} Z ₆ are candidates for error. As is clear from FIG. 5, checks Z _{1 to} Z ₄
The contradiction in and that in Z ₅ and Z ₆ are independent of each other (there are no common measurements). Therefore, at least 1 for each
You can see that there are two mistakes.

By the way, regarding Z ₅ and Z ₆ , it can be easily determined that the only common error value is the error measurement value (using the procedure of the above-described embodiment). On the other hand, regarding Z _{1 to} Z _4, it cannot be explained that there is one error, so two of them are regarded as errors. Here, the determination of and can be performed based on how many (residuals of) the respective measured values are related to the check related to the error. Therefore, the processing shown in the table below is performed.

The related check in the table shows in which check each measurement value that is an error candidate becomes an error candidate. These checks can group related items that have common observations. In the example of FIG. 5, Z _{1 to} Z ₄ and Z ₅ and Z ₆ are grouped as described above, and it is determined whether each measured value belongs to one of the groups 1 and 2. The contradiction found by checking (contradiction between measured values due to erroneous measured values) is caused by the smallest erroneous measured value, which is the basic policy of erroneous measured value judgment. Whether you got involved is important information when making a decision. As a result, since ,, and are associated with multiple checks, they are primary candidates for error measurements according to the above basic policy. Among these, since it is not related to other candidates, it can be determined to be an error without a check as described below only by the procedure introduced in the above-mentioned embodiment.

On the other hand, if there is one error on the side of ,,, it is not possible to explain the contradiction of the check of Z _{1 to} Z ₆ by any one, so at least two of them are errors. In order to choose two out of three candidates, we will examine the degree of overlap with each other. When the number of related checks is 2 and all are equal, the high degree of duplication of other leads to a strong possibility of being substituted by making an error in another measurement value. Therefore, is, as the degree of overlap 2 indicates,
It is excluded from the secondary candidates as it can be replaced by the other two measurements. The secondary candidates, (and), are judged to be erroneous because the contradiction of all checks can be explained by making them erroneous.

By the way, in the embodiments introduced so far, a threshold value (block E in FIGS. 1 and 2) for checking the magnitude of the estimated residual used for determining whether or not there is an error in the measured value.
Is not specified, but can be done as follows. That is, the threshold is made equal for all measured values, and each residual is normalized by dividing it by the measurement error range of the corresponding measured value. Therefore, the comparison with the threshold is
It suffices to perform this normalized residual.

There is also a method of obtaining a distribution of the magnitudes of all residuals or normalized residuals, calculating a new threshold value from this distribution, and making a determination based on this threshold value. At that time, in order to calculate a new threshold value, a times the maximum value (a is 0.
1) If the distribution is bimodal, it is possible to use it as the residual or normalized residual of the valley.

Although the residual of the active power has been described in the present embodiment, it can be applied to the reactive power by correcting the ground capacitance and the like as necessary. Further, it is needless to say that current can be used instead of electric power, and the present invention can be applied not only to electric power systems but also to general electric networks and fluid pipeline networks in which Kirchhoff's law or a physical law similar thereto is established. .

〔The invention's effect〕

As described above, according to the present invention, since it is possible to limit the measurement value that is less likely to be affected by the accuracy of the estimation calculation result and has a high possibility of error, it is possible to improve the reliability and efficiency of error determination. Has the effect of improving.

[Brief description of drawings]

FIG. 1 is a flow chart showing an embodiment of the present invention,
FIG. 4 is a flow chart showing another embodiment of the present invention, FIG. 3 is a system diagram shown for explaining the embodiment of FIG. 2, and FIG. 4 is a description of an application method of the present invention when there is no measured value. FIG. 5 is an explanatory diagram for explaining the sum (or sum) of a plurality of error measurement values and abnormal residuals caused thereby, and FIG. 6 is a system showing a system for giving the measurement values. It is a figure. A: Import input data, B: Estimate calculation, C:
Error measurement value detection process, “N”, “M”, “L” …… node, l,
n, m …… Branch.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Kenichi Morita 5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture, Ltd. Inside the Omika Plant, Hitachi, Ltd. (56) Reference JP-A-61-231837 (JP, A)

Claims (1)

[Claims] 1. Based on an estimated value derived from a measured value from a measured system including each node of a power system and a branch connecting these nodes, and an estimated residual which is a difference between the measured values, In the detection method of the residual measurement value that detects an error measurement value from among the measurement values, the sum of the estimated residuals obtained based on the measurement value at each node, or both ends of each branch obtained based on the measurement value A method for detecting an error measurement value, which comprises detecting that the total sum of estimated residuals exceeds a predetermined value, and specifying an error measurement value from the measurement values.