A digital-twin-based direct-current filter protection method

Abstract

The present application relates to the technical field of circuit protection, in particular to a DC filter protection method based on digital twinning, comprising the following steps: S1, setting current and voltage measuring devices in the filter circuit, collecting current signals and voltage signals; S2, constructing a digital twinning model according to the filter circuit, wherein the digital twinning model comprises time domain and frequency domain models established according to KCL and KVL laws; S3, calculating the predicted values of the current and voltage of the measuring point by using the digital twinning model of step S2, and performing normalized residual sum of squares on the measured current and voltage values collected in step S1; S4, judging the internal and external faults of the circuit according to the chi-square value of the normalized residual sum of squares of step S3, and determining the circuit protection action. The present application simultaneously uses time domain and frequency domain to distinguish faults, the frequency domain model is used to determine whether the filter has failed, the time domain model is used to distinguish internal faults and external disturbances, and the protection misoperation is prevented, so that the filter can reliably act and protect the safe and stable operation of the filter when the filter harmonic impedance changes, short circuit faults and other conditions occur.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of circuit protection, in particular to a DC filter protection method based on digital twinning. BACKGROUND

[0002] High-voltage direct current transmission technology is widely used in China due to its low loss, large transmission capacity, and long transmission distance. However, the harmonic voltage and current generated during the operation of the converter can cause additional heating of the DC side devices such as the smoothing reactor and the DC line, which can accelerate the aging of the insulation devices and affect the operation performance of the system. If the harmonic current flows into the overhead line, it can also affect the communication equipment near the line. Therefore, a DC filter is required in high-voltage direct current transmission projects to filter out harmonic components and ensure the normal and stable operation of the direct current transmission system. However, if the DC filter fails, the harmonic current will flow into the line and affect the power supply quality, especially when the high-voltage capacitor fails, which can generate a very strong discharge current and endanger the operation of the system. According to the investigation, the correct action rate of the DC filter protection is only 66.7%. Therefore, it is of great significance to improve the DC filter protection.

[0003] Currently, scholars at home and abroad have conducted some exploration and research on the protection of DC filters. The main protection includes differential protection, high-voltage capacitor unbalance protection, and harmonic impedance protection.

[0004] For differential protection, some inventions point out that the existence of harmonics and energy storage elements may cause the traditional differential protection to malfunction. Some inventions propose a new criterion using a constant plus ratio restraint to solve the problem of traditional differential protection. However, both the traditional and improved differential protection have the problem that when an internal short-circuit fault occurs in the element, the difference between the two-sided mutual inductors is almost the same, and the fault cannot be well detected. Moreover, the current flow of the protection is often selected as the characteristic current flow. When a major fault occurs, the harmonic impedance of the filter changes greatly, resulting in a decrease in the characteristic current flow flowing into the protection, which leads to insufficient protection sensitivity. For unbalance protection, the current ratio method is used to obtain the unbalance current by subtracting the upper and lower arm currents. However, when a low unbalance degree fault occurs, there is almost no unbalance current, resulting in insufficient sensitivity. The characteristic harmonic impedance protection uses Fourier decomposition to extract the first-end characteristic voltage and current to calculate the harmonic impedance of the filter for fault judgment. However, Fourier decomposition is easily affected by external disturbances, which may cause protection malfunctions.

[0005] In summary, differential protection is based on KCL model to distinguish faults, but due to the use of only current, the model is not comprehensive and accurate enough, resulting in insufficient protection sensitivity. Unbalanced protection distinguishes according to the characteristics of current at the time of fault, but the consideration is not comprehensive enough, and there is a protection blind area. The characteristic harmonic impedance method uses the rising rule of harmonic impedance at the time of fault to distinguish, but it will be affected by external disturbance, so that the harmonic impedance obtained is not accurate, which may lead to protection misoperation. Therefore, it is necessary to continue to study the main protection of the DC filter. SUMMARY

[0006] In view of the defects in the prior art, the purpose of the present application is to provide a DC filter protection method based on digital twinning, which can reliably act when the filter harmonic impedance changes, short circuit faults and the like occur, and protect the safe and stable operation of the filter.

[0007] To achieve the above purpose, the technical scheme adopted by the present application is:

[0008] A DC filter protection method based on digital twinning, comprising the following steps:

[0009] S1, setting current and voltage measuring devices in the filter circuit, collecting current signals and voltage signals;

[0010] S2, constructing a digital twinning model according to the filter circuit, wherein the digital twinning model comprises time domain and frequency domain models established according to KCL and KVL laws;

[0011] S3, using the digital twinning model of step S2 to calculate the predicted value of the current and voltage of the measurement point, and making the normalized residual sum of squares of the measured current value and voltage value collected in step S1;

[0012] S4, judging the internal and external faults of the circuit according to the chi-square value of the normalized residual sum of squares of step S3 to determine the protection action of the circuit.

[0013] On the basis of the above technical scheme, the step S3 specifically comprises:

[0014] S31, establishing a measurement equation:

[0015] z=h(x)+v=Yx+C+v (1)

[0016] In the formula, z represents the measurement quantity column vector, which includes two parts: the actual measurement quantity and the virtual measurement quantity. The actual measurement quantity part is actually measured by the measuring device, and the virtual measurement quantity part represents the relevant physical laws satisfied and is represented by zero; x represents the state quantity column vector, which is the electrical quantity that is not directly measured inside the filter; Y is the relation matrix, which is derived from the relationship between each measurement quantity and the state quantity in the digital twin model; C is the historical value matrix; and v is the error column vector between the actual measurement value and the estimated measurement value.

[0017] S32. Minimize the normalized sum of squared residuals to obtain the estimated value x' of the state variable x. The objective function is as follows:

[0018]

[0019] In the formula, σ is the standard deviation of the measured quantity, W is the weighted diagonal matrix, which reflects the error of each measured quantity, and each value is obtained by the reciprocal of the square of the standard deviation of the corresponding measured quantity;

[0020] Setting MinJ(x) = 0 yields the estimated value x' of the state variable x:

[0021] x'=(H T WH) -1 H T W(zC) (3)

[0022] In the formula, H is the Jacobian matrix of h(x);

[0023] S33. Calculate the estimated value z' of the measured quantity and the normalized sum of squared residuals:

[0024] z'=h(x')=Hx'+C (4)

[0025]

[0026] Based on the above technical solution, step S4 specifically includes:

[0027] S41. If the chi-square value of the normalized residual sum of squares obtained from the frequency domain operation is greater than the set value, it is determined that the circuit has a fault and enters the fault discrimination between the internal and external zones.

[0028] S42. If the chi-square value of the normalized residual sum of squares obtained from the time-domain operation is greater than the set value for 5 consecutive time points, the fault outside the zone is ruled out and it is determined to be a fault within the zone.

[0029] Based on the above technical solution, the measuring device is a current transformer, used to measure the voltage or current value at the setting point.

[0030] This invention also claims protection for DC filter protection devices that utilize the above-described method.

[0031] The DC filter protection method based on digital twinning has the following beneficial effects:

[0032] 1. The method can effectively judge the symmetrical fault with poor protection effect of high-voltage capacitor protection and differential protection, and perfects the current protection for the DC filter.

[0033] 2. The method simultaneously uses the full current instantaneous value for protection, avoiding the problem of protection failure or misoperation caused by harmonic impedance change.

[0034] 3. The method simultaneously uses the time domain and the frequency domain, combines the advantages of both, takes the advantages and makes up for the disadvantages, and effectively protects the filter under the condition of preventing external faults and lightning interference. BRIEF DESCRIPTION OF DRAWINGS

[0035] The application has the following drawings:

[0036] Figure 1 Structure diagram of a double-tuned filter;

[0037] Figure 2 Time domain twin model diagram of a double-tuned filter;

[0038] Figure 3 Frequency domain twin model diagram of a double-tuned filter;

[0039] Figure 4 Chi-square distribution probability density curve diagram;

[0040] Figure 5 Flowchart of the algorithm of the application. DETAILED DESCRIPTION

[0041] The application will be further described in detail below in combination with the drawings.

[0042] The application is based on digital twinning technology, and identifies faults according to the matching degree of real-time electrical physical quantities and twin model prediction quantities on the basis of accurate modeling and full use of physical laws.

[0043] The DC filter is one of the key devices in high-voltage direct current transmission projects, and guarantees the normal and stable operation of the DC transmission system. The most commonly used DC filter structure in the current high-voltage direct current transmission project is a double-tuned DC filter, which can filter out 12th and 24th harmonics. The method of the application will be described below in combination with the double-tuned DC filter. The structure of the double-tuned filter in the high-voltage direct current system is as follows: Figure 1The overall system adopts the CIGRE standard high-voltage direct-current model, and the direct-current filter is located at the sending end of the system. C1 is the high-voltage capacitor of the filter, which is connected by multiple capacitor units in actual engineering, and is replaced by four identical capacitors in series and parallel in the model. PT1 and CT1 are the voltage transformer and current transformer assembled at the first section of the filter, CT2 is the current transformer assembled in the high-voltage capacitor to connect the left and right bridge arm circuits. CT3 is the current transformer assembled at the tail end of the filter. The positions of the transformers are positions convenient for measuring data, and the measured values are measurement quantities. The electrical quantities inside the electrical elements are difficult to measure directly, and are state quantities. Generally, the voltage and current at the first and last ends are selected for the establishment of the digital twin model.

[0044] After setting the current and voltage transformers at appropriate positions in the circuit, the time-domain and frequency-domain models are established according to the KCL and KVL laws. Here, only the expressions of the KCL and KVL laws are given due to the large number of related formulas. To better illustrate, the high-voltage capacitor C1 and the internal capacitor C n1 For example, the related model equations (6)-(15) are given. The internal electrical quantities such as Figure 2 As shown in the figure. Each electrical quantity is the instantaneous value of the total current to prevent the influence of the change of the filter harmonic impedance.

[0045] i1=i c1 (6)

[0046] 0=-i c1 +i cn1 +i cn2 (7)

[0047]

[0048] i l1 +i R1 =i c1 (9)

[0049] 0=-u c1 +u cn1 +u cn3 (10)

[0050] u1=u c1 (11)

[0051] 0=-i c1 +i cn3 +i cn4 (12)

[0052] i p =i cn1 -i cn3 (13)

[0053]

[0054] 0 = -u cn1 +u cn2 (15)

[0055] The frequency domain model is also based on the high voltage capacitor C1 and the internal capacitor C n1 For example, the relevant model equations (11)-(20) are given. Except for the energy storage element, the imaginary part equation is the same as the real part equation, so only the real part equation is given here. Both the real and imaginary parts are obtained by Fourier decomposition. The various electrical quantities used are shown in Figure 3

[0056]

[0057]

[0058]

[0059]

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066] After the digital twin model of the filter is established, the present application determines whether the physical structure of the filter has changed by comparing whether the actual measured quantity matches the predicted quantity of the twin model. The predicted quantity obtained by the digital twin model is calculated by the dynamic state estimation method.

[0067] Taking the time domain digital twin model of the filter as an example, the measurement equation of the system is established. The measurement quantity is divided into two parts: real measurement quantity and virtual measurement quantity. The real measurement quantity part is measured by the mutual inductor, and the virtual measurement quantity part represents the satisfaction of the relevant physical law, represented by zero. The overall measurement quantity is represented by z, specifically z = [i1, i p , 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, i4, u1, 0], a total of 24 measurement quantities. The state quantity is the electrical quantity inside the filter that does not need to be measured or cannot be directly measured, represented by x, specifically x = [u c1 , i c1 , u c2 , i c2 ​,u l1 i l1 ,u l2 i l2 ,u4,u R i R ,i2,i cn1 i cn2 i cn3 i cn4 ,u cn1 ,u cn2 ,u cn3 ,u cn4 There are a total of 20 measured quantities. Based on the above measured quantities and state quantities, the following measurement equations are established:

[0068] z=h(x)+v=Yx+C+v (1)

[0069] In the formula, z is the column vector of measured quantities; x is the column vector of state quantities; Y is the relation matrix, which is derived from the relationship between each measured quantity and state quantity in the twin model; C is the historical value matrix; and v is the column vector of error between the actual measured value and the estimated measured value.

[0070] This section uses a high-voltage capacitor as an example to establish a specific expression of the relevant measurement equations according to the above method, where the measured quantity is z=[i1,0,0,i l1 +i r ,0,i p ,0,0,0,0,0,0,0,u1],The state variable is x=[i c1 i cn1 i cn2 i cn3 i cn4 ,u c1 ,u cn1 ,u cn2 ,u cn3 ,u cn4 It includes 14 measured quantities and 10 state quantities. The measurement equations are expressed as follows:

[0071]

[0072] For the physical relationships in the time-domain model, we can rewrite the above measurement equations as follows:

[0073]

[0074] The linear formula obtained from the method of quadratic integration:

[0075]

[0076]

[0077] In formula 27, 28, h represents the interval between two time points, which refers to the sampling interval of the transformer.

[0078] The z(t) on both sides of equation (26) and the part with x(t) are brought into formula 27, 28, and linearization can be arranged into the following matrix form:

[0079]

[0080] In the above formula, the Jacobian matrix H of h(x) is: The matrix C is:

[0081] The construction of the matrix form of the measurement equation of the frequency domain model does not require linearization, and the remaining steps are completely consistent with the time domain model, which will not be repeated here.

[0082] In order to obtain the optimal estimation value x' of x(t), it is necessary to minimize the normalized residual sum of squares, according to the optimization theory:

[0083]

[0084] In the formula, σ is the standard deviation of the measurement, W is the weight diagonal matrix, which reflects the error of each measurement, and each value is obtained by taking the reciprocal of the square value of the standard deviation of the corresponding measurement.

[0085] In the twin model of the filter, h(x) is a linear function, and in the ideal state, when the model is completely satisfied, the normalized residual sum is 0, so MinJ(x) = 0 can be used to obtain the estimated state quantity:

[0086] x' = (H T WH) -1 H T W(z-C) (3)

[0087] According to the optimal state quantity estimation value, the estimated value z' of the measurement and the normalized residual sum of squares are calculated:

[0088] z' = h(x') = Hx' + C (4)

[0089]

[0090] When no fault occurs inside the filter, each measurement satisfies the internal physical law of the twin model, so the measurement residual caused by the measurement error of the transformer should conform to the Gaussian model, and the normalized residual sum of squares should conform to the chi-square distribution. The chi-square distribution probability density curve is shown in Figure 4 .

[0091] From Figure 4It can be seen that the greater the normalized residual sum of squares, the smaller the probability of conforming to the chi-square distribution, thereby determining whether a fault occurs inside the filter.In the present application, there are 8 more measured quantities than state quantities, so the chi-square distribution probability density curve with n=8 is selected here.

[0092] When judging the fault, the chi-square critical value can be selected according to the required accuracy, and the chi-square value 13.6 corresponds to an accuracy of 90%.

[0093] As shown in the algorithm flowchart: Figure 5

[0094] First, the current signal and the voltage signal are collected.

[0095] Second, the collected voltage and current signals are input in the program.

[0096] Third, frequency domain operation is performed, and when the criterion is greater than the setting value, it is determined that a fault occurs, and the in-zone and out-of-zone fault discrimination is entered.

[0097] Fourth, time domain operation is performed to prevent the influence of out-of-zone disturbance.When the criterion calculated by 5 data samples is greater than the setting value, the out-of-zone fault disturbance is excluded, and it is determined as an in-zone fault.

[0098] Fifth, the operation is ended, and the next judgment is waited.

[0099] It should be noted that: because the frequency domain discriminates the result once every 167 time points, when the frequency domain judgment has a fault, the fault may occur at any time in the 167 time points, so from the first time point, the time domain discrimination is performed once every 5 time points to judge whether it is an out-of-zone fault.If the discrimination result of the next frequency domain is continuously 5 time points in the time domain model, the residual sum is less than 13.36, then it is proved that the fault is an out-of-zone fault, and if there are 5 continuous points in the time domain model with residual sum greater than 13.36, it is an in-zone fault.

[0100] The present application discriminates the fault by using the time domain and the frequency domain at the same time, the frequency domain model is used to discriminate whether the filter has a fault, and the time domain model is used to distinguish the in-zone fault and the out-of-zone disturbance, thereby preventing the protection misoperation.

[0101] The method of the present application is not limited to the protection of the double-tuned filter, and as long as the digital twin model is established for the related power device, the protection can be performed by using the present method.

[0102] The contents not described in detail in the present specification belong to the prior art known to those skilled in the art.​

Claims

1. A DC filter protection method based on digital twin, characterized in that, Includes the following steps: S1. Set up current and voltage measuring devices in the filter circuit to collect current and voltage signals; S2. Construct a digital twin model based on the filter circuit, wherein the digital twin model includes time-domain and frequency-domain models established based on KCL and KVL laws; S3. Calculate the predicted values ​​of current and voltage at the measurement point using the digital twin model described in step S2, and perform a normalized sum of squared residuals with the measured current and voltage values ​​collected in step S1. S4. Determine the fault inside or outside the circuit area based on the chi-square value of the normalized sum of squared residuals described in step S3, and decide on the circuit protection action. The time-domain model described in S2 is: ; ; ; ; ; ; ; ; ; ; In the above time-domain model equations, all electrical quantities are instantaneous values ​​of the total current to prevent the influence of changes in filter harmonic impedance; the parameters are as follows: i1 is the total current flowing through the filter in the time-domain model; i c1 This is the current flowing through one end of the high-voltage capacitor C1 in the filter of the time-domain model; i cn1 i cn2 i cn3 i cn4 These are the currents flowing through one end of the four internal capacitors Cn1, Cn2, Cn3, and Cn4 that make up the high-voltage capacitor C1 in the time-domain model; u c1 This represents the voltage across C1 in the time-domain model. i l1 i R1 These are the currents flowing through inductor L1 and resistor R1 in the time-domain model, respectively. u cn1 u cn2 u cn3 These are the voltages across the internal capacitors Cn1, Cn2, and Cn3 in the time-domain model, respectively. u1 is the potential at the beginning of the filter in the time-domain model; i p This refers to the current flowing through the bridge circuit between the four internal capacitors Cn1, Cn2, Cn3, and Cn4 in the time-domain model; The frequency domain model described in S2 is: ; ; ; ; ; ; ; ; ; ； In the frequency domain model above, I1 + I c1 + I cn1 + I cn2 + I cn3 + I cn4 + U c1 + c1, I c1 - I l1 + I R1 + U cn1 + U cn2 + U cn3 + U1 + I p + The parameters are as follows: I1 + This represents the total current flowing through the filter in the frequency domain model. I c1 + This is the current flowing through one end of the high-voltage capacitor C1 in the filter of the frequency domain model; I cn1 + I cn2 + I cn3 + I cn4 + These are the currents flowing through one end of the four internal capacitors Cn1, Cn2, Cn3, and Cn4 that make up the high-voltage capacitor C1 in the frequency domain model; U c1 + This represents the voltage across C1 in the frequency domain model. I l1 + I R1 + These are the currents flowing through inductor L1 and resistor R1 in the frequency domain model, respectively. U cn1 + U cn2 + U cn3 + These are the voltages across the internal capacitors Cn1, Cn2, and Cn3 in the frequency domain model, respectively. U1 + This represents the potential at the beginning of the filter in the frequency domain model. I p + This represents the current flowing through the bridge circuit between the four internal capacitors Cn1, Cn2, Cn3, and Cn4 in the frequency domain model.

2. The DC filter protection method based on digital twin as described in claim 1, characterized in that, Step S3 specifically includes: S31. Establish the measurement equation: ; In the formula, z The measurement quantity column vector consists of two parts: real measurement quantity and virtual measurement quantity. The real measurement quantity part is actually measured by the measuring device, and the virtual measurement quantity part represents the relevant physical laws that are satisfied, and is represented by zero. x This represents a column vector of state quantities, which are electrical quantities that are not directly measured inside the filter. Y This is a relation matrix, derived from the relationships between each measured quantity and state quantity in the digital twin model. C This is a historical value matrix. v This is a column vector of errors between the actual measured values ​​and the estimated measured values; S32. Minimize the normalized sum of squared residuals to obtain the state variables. x The estimated value The objective function is as follows: ; In the formula, σ The standard deviation of the measured quantity W This is a weighted diagonal matrix that reflects the error of each measurement. Each value is obtained from the reciprocal of the squared standard deviation of the corresponding measurement. make MinJ(x)=0 Determine the state variables x The estimated value : ； In the formula H for h(x) The Jacobian matrix; S33. Calculate the estimated value of the measured quantity. Sum of squared normalized residuals: ; 。 3. The DC filter protection method based on digital twin as described in claim 1, characterized in that, Step S4 specifically includes: S41. If the chi-square value of the normalized residual sum of squares obtained from the frequency domain operation is greater than the set value, it is determined that the circuit has a fault and enters the fault discrimination between the internal and external zones. S42. If the chi-square value of the normalized residual sum of squares obtained from the time-domain operation is greater than the set value for 5 consecutive time points, the fault outside the zone is ruled out and it is determined to be a fault within the zone.

4. The DC filter protection method based on digital twin as described in claim 1, characterized in that, The measuring device is a current transformer, used to measure the voltage or current value at the setting point.

5. A DC filter protection device, using the method described in any one of claims 1-4.

Images