Power system line fault detection discrimination method

Abstract

The application relates to the field of power system line fault monitoring, in particular to a power system line fault detection and discrimination method, which comprises the following steps: step one: obtaining historical data X and online data Y of a power system; step two: respectively calculating a singular value matrix Sigma 2 of the historical data X and a singular value matrix Sigma 1 of the online data Y; step three: calculating divergence distance D of the historical data X and the online data Y; and step four: judging whether a high-resistance fault occurs in a line based on a discrimination condition. div The method is suitable for complex fault environments of actual power system lines, can realize detection and discrimination without understanding the actual fault point environment, and is cheap and efficient.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power system line fault monitoring, in particular to a power system line fault detection and discrimination method. BACKGROUND

[0002] The high-resistance fault of the power system is a kind of power system line fault which cannot be effectively detected and identified at present, which not only includes single-phase grounding fault caused by overhead line breakage and drop, tree contact fault and other overhead line faults, but also includes cable faults caused by insulation damage and aging, assembly defects and the like. The time-frequency characteristics of this kind of fault are weak, the transition resistance is very large, even greater than 20 kilo-ohms, so that the fault current is less than 10% of the load current, and the voltage mutation is not obvious, which is difficult to detect and identify from the voltage change, therefore the traditional power system protection cannot complete the detection and identification of this kind of fault.

[0003] At present, the detection and identification of this kind of fault is started from the time / frequency characteristics of fault electrical quantities (voltage or current), and through physical modeling of the characteristics of fault electrical quantities under various fault conditions, the detection and identification of this kind of line fault are realized, however, the current method has the following main problems:

[0004] 1. The time-frequency characteristics of high-resistance fault are affected by uncertain factors such as grounding medium, medium condition, fault line length and environment, fault point electromagnetic environment, the current time-frequency modeling method only models for specific fault scenarios, and the modeling accuracy is not high, so that the established model is only applicable to specific fault scenarios, and cannot meet the needs of actual situation;

[0005] 2. If a fault physical model with strong applicability is to be obtained, it is necessary to enumerate various conditions, carry out actual experiments, tests and iterations, which has problems of high cost, poor feasibility and low efficiency;

[0006] 3. A large number of distributed power sources and energy storage devices are connected to the power system, the normal switching of these devices will produce a time-frequency domain transient process similar to high-resistance fault, which reduces the reliability of detecting and identifying this kind of fault from the time / frequency characteristics of electrical quantities, and increases the false alarm rate.

[0007] 4. If the previous physical model is to be modified to adapt to the actual situation, it is necessary to enumerate, experiment, fit and iterate various fault conditions, which is impossible to realize in terms of actual cost.

[0008] Therefore, it is necessary to invent a technology for detecting and identifying line high-resistance fault from existing device data without a large number of experiments and accurate network topology, distributed power supply and energy storage operation plan. SUMMARY

[0009] The technical problem solved by the present application is to overcome the shortcomings of the prior art and provide a power system line fault detection and discrimination method.

[0010] The technical scheme adopted by the present application to solve the technical problem is a power system line fault detection and discrimination method, comprising the following steps:

[0011] Step one: obtaining historical data X and online data Y of the power system;

[0012] Step two: calculating the singular value matrix Σ2 of the historical data X and the singular value matrix Σ1 of the online data Y, respectively;

[0013] Step three: calculating the divergence distance D div of the historical data X and the online data Y;

[0014] Step four: determining whether a high resistance fault occurs in the line based on a discrimination condition.

[0015] The historical data X and the online data Y in the step one are obtained based on a data measurement device capable of measuring electrical quantities.

[0016] The data measurement device is one or more of a remote terminal unit, a data transmission terminal, a feeder terminal device, a distribution transformer monitoring terminal, a phasor measurement device, and a micro phasor measurement device.

[0017] The singular value matrix of the data matrix corresponding to the online data Y and the historical data X in the step two is expressed as follows:

[0018]

[0019] In the formula, Σ1 and Σ2 are the singular value matrices of the online data Y and the historical data X, respectively; and are the i-th singular values.

[0020] The calculation process in the step three is as follows:

[0021] 3-1: calculating the divergence distance D div of the online data Y and the historical data X, which is calculated according to the following formula:

[0022] D div =(infE||X-Y|| 2 ) 1 / 2 (1);

[0023] In the formula, inf represents the lower limit of the calculated value, E||X-Y|| 2 represents the covariance of the online data Y and the historical data X.

[0024]

[0025] wherein, and respectively represent the mean of the online data Y and the historical data X,

[0026] 3-2: The covariance matrix ∑ of the joint distribution of the online data Y and the historical data X can be constructed as:

[0027]

[0028] wherein, Z is a transition matrix, Z T is the transpose matrix of the transition matrix Z;

[0029] According to formula (1), we can get:

[0030]

[0031] Further decomposition of ∑ can be obtained:

[0032]

[0033] wherein, the generalized Schur complement L of the matrix ∑ can be expressed as: I is the unit matrix, since ∑1 and ∑2 are singular data sequences, therefore, the generalized inverse can be obtained:

[0034]

[0035] U, M, V are respectively the unitary matrix after singular value decomposition;

[0036] Further spectral decomposition of ∑2-L is obtained:

[0037]

[0038] Then:

[0039]

[0040] Let get:

[0041] F(ζ, v) = C-2tr(ξv) + tr[δ(ζ T ζ-I)] (10);

[0042] wherein, C = tr(∑1) + tr(∑2), δ is the Lagrange operator;

[0043] Further:

[0044]

[0045] Further, the divergence distance D between the online data Y and the historical data X is obtained:

[0046]

[0047] Therefore, the divergence distance D between the online data Y and the historical data X is obtained: div is expressed as:

[0048]

[0049] that is,

[0050] In step four, the discrimination condition is as follows:

[0051]

[0052] In the formula, t d , t w are the over-limit time length and the over-limit time length threshold value respectively, and Ψ is a threshold value.

[0053] The Ψ is obtained by the following formula:

[0054]

[0055] In the formula, and S respectively represent the mean and the standard deviation of the divergence distance vector, and τ is an adjustment coefficient.

[0056] The discrimination condition is revised as:

[0057]

[0058] In the formula, χ is a threshold value fluctuation degree.

[0059] The discrimination rule for determining whether a high-resistance fault occurs in a line is as follows:

[0060] 4-1: If D div > Ψ (1+χ) is satisfied, t d > t W is not satisfied, and the protection device in the line acts, it is determined that a non-high-resistance fault occurs;

[0061] 4-2: If D div > Ψ (1+χ) is satisfied, t d > t W is not satisfied, and the protection device in the line does not act, it is determined that no fault occurs at present, and the over-limit is caused by normal operation;

[0062] 4-3: If D div > Ψ (1+χ) is satisfied, and t d > t W is satisfied at the same time, it is determined that a high-resistance fault occurs in the line.

[0063] Compared with the prior art, the present application has the following beneficial effects:

[0064] 1. Only daily measurement data are needed to detect line faults;

[0065] 2. Suitable for different distributed power multi-point access conditions;

[0066] 3. No need to know the type, location and quantity of the distributed power connected to the system, and no need to enumerate and model the fault scenarios;

[0067] 4. No need to consider the actual fault point environment and conditions, and the method has strong adaptability;

[0068] 5. The detection and identification function can be completed based on the existing main station or intelligent terminal, without additional investment in equipment;

[0069] 6. Actual system operation is only needed to detect and identify the whole network fault according to the protected content;

[0070] 7. The present application is suitable for the complex fault environment of the actual power system line, and can realize detection and identification without understanding the actual fault point environment, which is convenient and efficient. BRIEF DESCRIPTION OF DRAWINGS

[0071] Figure 1 is a flowchart of the present application. DETAILED DESCRIPTION

[0072] The embodiments of the present application will be further described below in combination with the drawings:

[0073] EMBODIMENT

[0074] With the large-scale installation of real-time measurement devices, the power system measurement data generally meet the requirements of the central limit theorem, which means that the power data from the measurement equipment meets the Gaussian distribution, and has great redundancy, i.e. the measurement data is sparse. Moreover, when a high-impedance fault occurs, although the physical characteristics of the electrical quantities are small, the probability distribution of these electrical quantities before the fault and during the fault duration will show obvious differences, so the high-impedance fault and the fault phase can be detected and identified through the divergence distance (probability distribution difference) of the pre-fault (normal history) data and the fault period (online measurement) data.

[0075] As shown in Figure 1 , the power system line fault detection and discrimination method includes the following steps:

[0076] Step one: obtain historical data X and online data Y of the power system; the historical data X and online data Y in the step one are obtained based on data measurement equipment capable of measuring electrical quantities. The data measurement equipment is one or more of RTU (remote terminal unit), DTU (data transmission terminal), FTU (feeder terminal device), TTU (distribution transformer monitoring terminal), PMU (phasor measurement device), micro-PMU (micro-phasor measurement device). It can also be other IED (intelligent electronic device) with data measurement function. The data can be obtained from the substation or master station connected to the above-mentioned equipment, or directly obtained from the wave recording device.

[0077] Step two: calculate the singular value matrix ∑2 of the historical data X and the singular value matrix ∑1 of the online data Y respectively; the singular value matrix of the corresponding data matrix of the online data Y and the historical data X in the step two is expressed as follows:

[0078]

[0079] In the formula, ∑1 and ∑2 are the singular value matrices of the online data Y and the historical data X respectively; and are the i-th singular values.

[0080] Step three: calculate the divergence distance D of the historical data X and the online data Y div ; the calculation process in the step three is as follows:

[0081] 3-1: calculate the divergence distance D div of the online data Y and the historical data X, and the calculation formula is as follows:

[0082] D div = (infE||X-Y|| 2 ) 1 / 2 (1);

[0083] In the formula, inf represents the lower limit of the value, E||X-Y|| 2 represents the covariance of the online data Y and the historical data X;

[0084]

[0085] In the formula, and respectively represent the mean of the online data Y and the historical data X,

[0086] 3-2: the covariance matrix ∑ of the joint distribution of the online data Y and the historical data X can be constructed as

[0087]

[0088] where Z is the transition matrix, T denotes transpose, Z T is the transpose matrix of the transition matrix Z;

[0089] According to equation (1), we have

[0090]

[0091] Further decomposition of ∑ gives

[0092]

[0093] where the generalized Schur complement L of matrix ∑ can be expressed as I is the identity matrix. Since ∑1 and ∑2 are singular data sequences, the generalized inverse of ∑1 and ∑2 can be expressed as We have

[0094]

[0095] U, M, V are the unitary matrices after singular value decomposition, respectively;

[0096] Further spectral decomposition of ∑2-L gives

[0097]

[0098] Then we have

[0099]

[0100] Let We have

[0101] F(ζ, v) = c - 2tr(ζv) + tr[δ(ζ T ζ-I)] (10);

[0102] where C = tr(∑1) + tr(∑2), and δ is the Lagrange operator;

[0103] Further we have

[0104]

[0105] Further we have

[0106]

[0107] Therefore, the divergence distance D div between the online data Y and the historical data X is expressed as

[0108]

[0109] That is

[0110] Step four: determine whether the line has high resistance fault based on the discriminant condition. In the step four, the discriminant condition is as follows:

[0111]

[0112] wherein t d , t w are the over-limit time length and the over-limit time length threshold value respectively, and ψ is the threshold value.

[0113] The ψ is calculated by the following formula:

[0114]

[0115] wherein and S are the mean value and the standard deviation of the divergence distance vector respectively, and τ is the adjustment coefficient.

[0116] Since the current various types of distributed power supply, electric vehicles and the like are connected to the power system, and due to the error of various measuring instruments, the distribution of the measured data of the power system in normal operation has certain differences, that is, the divergence distance between the online data Y and the historical data X cannot be completely equal to zero in normal operation, therefore, the threshold fluctuation degree χ(%) can be set. According to the relevant standards of the state and industry related to power quality, harmonics, voltage and frequency fluctuation and the like, χ can be set according to the actual use of electrical quantities, and it is recommended that it is not greater than 90% of the maximum value of the relevant provisions. The discriminant condition is revised as follows:

[0117]

[0118] wherein χ is the threshold fluctuation degree.

[0119] The discriminant rule for determining whether the line has high resistance fault is as follows:

[0120] 4-1: if D div > ψ(1+χ) is satisfied, t d > t W is not satisfied, the protection device in the line acts (such as circuit breaker reclosing, fuse blowing, relay tripping and the like), then it is determined that there is a non-high resistance fault, at this time, the fault identification, isolation and recovery can be carried out in the traditional way;

[0121] 4-2: if D div > ψ(1+χ) is satisfied, t d > t W is not satisfied, the protection device in the line does not act, then it is determined that there is no fault at present, and the over-limit is caused by normal operation;

[0122] ​

[0123] 4-3: If D div > ψ(1+χ) is satisfied, and t d > t W If both are satisfied, it can be determined that a high resistance fault occurs on the line. Further, the fault phase can be determined by equation (17), i.e. if the phase determination condition is satisfied, D div > ψ(1+χ) is satisfied, and t d > t W , then the phase is the fault phase.

Claims

1. A method for detecting and identifying faults in power system lines, characterized in that, Includes the following steps: Step 1: Obtain historical data X and online data Y from the power system; Step 2: Calculate the singular value matrix of the historical data X respectively. and the singular value matrix of online data Y ; Step 3: Calculate the divergence distance D between historical data X and online data Y. div ; The calculation process in step three is as follows: 3-1: Calculate the divergence distance D between online data Y and historical data X. div The calculation formula is as follows: （1）； In the formula, inf represents the lower limit of the value to be calculated. This represents the covariance between online data Y and historical data X; （2）； In the formula, and Let Y and X represent the mean values ​​of the online data and the historical data, respectively. , ; 3-2: The covariance matrix ∑ of the joint distribution of online data Y and historical data X can be constructed as follows: （4）； In the formula, Z is the transition matrix, Z T Let Z be the transpose of the transition matrix; According to equation (1), we can obtain: （5）； Further decomposition of ∑ yields: （6）； In the formula, the matrix The generalized Schul complement L can be expressed as: I is the identity matrix, since and They are all singular data sequences, therefore derived from the generalized inverse. , We can obtain: （7）； U, M, and V are the unitary matrices after singular value decomposition; Further Spectral decomposition yields: （8）； but: （9）； make , ,get: （10）； In the formula, , For Lagrange operators; Furthermore: （11）； Therefore, we get: （12）； Therefore, the divergence distance D between online data Y and historical data X is... div Represented as: （13）； Right now (14); Step 4: Based on the discrimination criteria, determine whether a high-resistance fault has occurred in the line; In step four, the discrimination criteria are as follows: （15）； In the formula, t d , t w These are the duration of the over-limit and the threshold duration of the over-limit, respectively, where Ψ is the threshold. The Ψ is calculated using the following formula: （16）； In the formula, S and S represent the mean and standard deviation of the divergence distance vector, respectively. This is the adjustment coefficient; The judgment criteria are revised as follows: （17）； In the formula, The degree of threshold fluctuation; The rules for determining whether a line has a high-impedance fault are as follows: 4-1: If satisfy, If the conditions are not met, the protection device in the line will activate, indicating that a non-high resistance fault has occurred. 4-2: If satisfy, If the conditions are not met and the protection device in the line does not activate, it is determined that there is no fault at present, and the over-limit is caused by normal operation. 4-3: If Satisfied, and If all conditions are met, it can be determined that a high-resistance fault has occurred in the line.

2. The power system line fault detection and discrimination method according to claim 1, characterized in that, In step one, the historical data X and online data Y are acquired based on equipment capable of measuring electrical quantities.

3. The power system line fault detection and discrimination method according to claim 2, characterized in that, The data measurement equipment is one or more of the following: remote terminal unit, data transmission terminal, feeder terminal device, distribution transformer monitoring terminal, phasor measurement device, and micro phasor measurement device.

4. The power system line fault detection and discrimination method according to claim 2, characterized in that, The singular value matrix representation of the data matrices corresponding to the online data Y and historical data X in step two is as follows: （3）； In the formula, and These are the singular value matrices of online data Y and historical data X, respectively. and These are the i-th singular values.

Images