T-type power transmission line parameter identification and fault location method based on Karenbauer transformation and physical information neural network

Abstract

The invention relates to a T-type power transmission line parameter identification and fault location method based on Karenbauer transformation and a physical information neural network. The method comprises the steps of obtaining a 1-mode equivalent circuit and a 0-mode equivalent circuit of a T-type power transmission line in a normal operation state and a 1-mode equivalent circuit and a 0-mode equivalent circuit when an asymmetric grounding fault occurs; a differential equation with the positive sequence parameters as coefficients is established, and the coefficients of the differential equation are solved through the PINN; establishing a differential equation taking the positive sequence parameter and the fault distance proportionality coefficient as coefficients, and solving the coefficient of the differential equation through the PINN; a differential equation with the zero-sequence parameters as coefficients is established, the coefficients of the differential equation are solved through the PINN, and identification of the zero-sequence parameters is achieved. According to the invention, the positive sequence parameters and zero sequence parameters of each branch of the T-type power transmission line can be accurately identified, and the fault branch can be effectively judged and the fault distance can be calculated under different fault types; karrenbauer transformation is carried out on the T-type power transmission line, so that the calculation process of parameter identification and fault location is simplified, and the calculation precision is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system line parameter identification and fault location, and particularly relates to a T-type transmission line parameter identification and fault location method based on Karenbauer transformation and physical information neural network. Background Art

[0002] In recent years, with the accelerated construction of new power systems, transmission line corridor resources have become increasingly scarce. T-type transmission lines, due to their advantages such as high transmission capacity, short construction period, small outgoing line corridor occupation, and small footprint, have gradually gained widespread application in AC transmission systems. However, as a multi-terminal transmission line structure, the use of T-type transmission lines brings new challenges to the safe and stable operation of power systems, especially in terms of protection and control. Accurate line parameters are the basis for analysis such as fault location.

[0003] Line parameter identification relies on obtaining the required data from data measurement devices. Data measurement devices in power systems include Supervisory Control and Data Acquisition (SCADA), synchronized phasor measurement units (PMUs), and fault recorders. SCADA systems can acquire data such as voltage, current amplitude, active power, and reactive power, but the accuracy of their measurement data is low, making identification methods less accurate. PMUs can provide highly accurate voltage, current amplitude, and phase angle information, but they are expensive and typically installed only at key nodes in the power system. Therefore, methods based on PMUs have a limited scope of application. Fault recorders are more widely installed and used in power systems and can provide voltage and current waveforms at their installation points before and after a fault. However, methods based on fault recorder data involve calculating voltage and current differential values. Existing methods typically use numerical differentiation, which introduces approximation errors and results in low parameter identification accuracy. Furthermore, due to the multi-terminal structure of T-type transmission lines, the calculation process of existing methods is also complex.

[0004] Therefore, there is an urgent need for a T-type transmission line parameter identification method with a wider scope of application, a simpler calculation process, and higher calculation accuracy, so as to achieve accurate fault location of the T-type transmission line on this basis. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a T-type transmission line parameter identification and fault location method based on Karenbauer transform and physical information neural network. The method uses fault recording data at the three ends of the T-type transmission line to perform parameter identification and fault location, which significantly broadens the application scenarios. By performing Karenbauer transform on the T-type transmission line, the subsequent analysis and calculation process is simplified. At the same time, with the help of PINN, the precise calculation of T-type transmission line parameters and the accurate location of faults are achieved.

[0006] The present invention solves the technical problem by the following technical solutions:

[0007] A method for parameter identification and fault location of T-type transmission lines based on Karenbauer transformation and physical information neural network, the method comprising the following steps:

[0008] S1. Obtain the 1-mode and 0-mode equivalent circuits of a T-type transmission line in normal operation and the 1-mode and 0-mode equivalent circuits when an asymmetric ground fault occurs through Karenbauer transformation;

[0009] S2. Based on the equivalent circuit of the T-type transmission line 1 under normal operation, a differential equation with positive sequence parameters as coefficients is established, and the coefficients of the differential equation are solved through PINN to realize the identification of the positive sequence parameters;

[0010] S3. Based on the 1-mode equivalent circuit under the asymmetric ground fault state, a differential equation with the positive sequence parameter and the fault distance proportional coefficient as coefficients is established, and the coefficients of the differential equation are solved through PINN to determine the fault section and the fault distance;

[0011] S4. Under the asymmetric ground fault state, a differential equation with zero-sequence parameters as coefficients is established based on the 0-module equivalent circuit, and the coefficients of the differential equation are solved through PINN to realize the identification of the zero-sequence parameters.

[0012] Moreover, the S2 is specifically:

[0013] From the equivalent circuit of the T-type transmission line under normal operation, we can get

[0014]

[0015]

[0016] Where: u (1) M 、u (1) N 、u (1) P They are the voltage 1-mode components of the M, N, and P terminals in normal conditions;

[0017] i (1) M 、i (1) N 、i (1) P They are the 1-mode components of the current injected into the T node from the three terminals in the normal state;

[0018] They are the current 1 mode components injected into the T node from the three terminals through the positive sequence grounding capacitor;

[0019] The parameters to be determined include the positive sequence resistance of the three branches Positive sequence inductance and positive sequence capacitor

[0020] Construct PINN1 for solving the positive sequence parameters of T-type transmission lines, where j = M, N, P, and the input of PINN1 is the sampling time series t under normal conditions. zc , the output is used to fit the voltage and current 1-mode components under normal conditions;

[0021] Based on formula (1), the physical information loss function is constructed as formula (5). By minimizing the loss function through the optimization algorithm, the positive sequence capacitance can be obtained. On the basis of obtaining the positive sequence capacitance identification value, the physical information loss function is obtained based on the differential equation group (4) as shown in formula (6), and the positive sequence resistance is further identified. and positive sequence inductance PINN1 can accurately identify all positive sequence parameters of T-type transmission lines in steps;

[0022]

[0023] Moreover, the S3 is specifically:

[0024] In the equivalent circuit of the T-type transmission line when an asymmetric ground fault occurs on the MT branch, f is the fault point, which is located at a distance of αl from the line end M. M Where, α is the fault distance proportional coefficient to be determined; is the 1-mode component of the fault point voltage; They are the voltage 1-mode components at the M, N, and P terminals under fault conditions; They are the current 1-mode components of the three terminals under fault conditions; and They are the 1-mode components of the current flowing to the T node after passing through the positive-sequence grounding capacitor at the three terminals;

[0025] Assume that the fault occurs on the MT, NT and PT branches respectively, and assume that the distance between the fault point f and the branch endpoint is a M l M 、a N l N and a N l P , we can get equations (7) to (9);

[0026]

[0027] Based on this, PINN2 is constructed for fault location of T-type transmission lines. The input of PINN2 is the sampling time series t under the asymmetric ground fault state. gz , the output is used to fit the voltage and current 1-mode components under the asymmetric ground fault state, and three physical information loss functions are constructed as shown in formula (10);

[0028]

[0029] By training PINN2 three times, the loss function containing the above three physical information is minimized in turn, and the fault distance proportional coefficient α is gradually calculated M , α N and α P ; When f is located near the T node, the calculated α M , α N , α P are all close to 1, and ε = 0.001 is set, then |α M -1|≤ε、|α N -1|≤ε、|α P -1|≤ε;

[0030] When f is located on the MT branch, there exists |α M -1|≤ε、|α N -1|>ε、|α P -1|>ε; when f is located in the NT branch, there exists |α M -1|≥ε、|α N -1|≤ε、|α P -1|>ε;

[0031] When f is located in the PT branch, there exists |α M -1|>ε、|α N -1|>ε、|α P -1|≤ε;

[0032] Based on this, the construction criteria P1, P2, P3 and P4 are shown in equations (11) to (15);

[0033] P1=P M &PN &P P (11)

[0034] P2=P1^P M (12)

[0035] P3=P1^P N (13)

[0036] P4=P1^P P (14)

[0037]

[0038] Where i = M, N, P;

[0039] The corresponding criterion value is calculated based on the result of PINN2, and then the specific branch where the fault occurs is determined and the fault distance is determined.

[0040] Moreover, the S4 is specifically:

[0041] In the 0-mode equivalent circuit when an asymmetric ground fault occurs on the MT branch, u (0) M 、u (0) N 、u (0) P are the zero-mode components of the voltage at the M, N, and P terminals under fault conditions; i (0) M 、i (0) N 、i (0) P They are the 0-mode components of the current flowing into the T node from the three terminals; and They are the zero-mode components of the current flowing from the three terminals to the T node after passing through the positive-sequence grounding capacitor;

[0042] The parameters to be determined in the circuit include the zero-sequence resistance of the three branches Zero-sequence inductance Zero-sequence capacitance

[0043] Equation (16) is obtained from the zero-mode equivalent circuit when an asymmetric ground fault occurs on the MT branch. Based on this, PINN3 is constructed to identify the zero-sequence parameters of the T-type transmission line. The input of PINN3 is t gz , the output is used to fit the voltage and current zero-mode components under the asymmetric ground fault state, and the physical information loss function is constructed as shown in formula (17). By minimizing the loss function through the optimization algorithm, all zero-sequence parameters of the T-type transmission line can be obtained; in order to improve the training effect, the zero-sequence capacitor The learning rate is set to 10 -8 , and the zero sequence resistance and zero-sequence inductance The learning rate is set to 0.1; this effectively balances the different parameters during training and improves the convergence and accuracy of the model;

[0044]

[0045] The advantages and beneficial effects of the present invention are:

[0046] 1. The present invention can not only accurately identify the positive sequence parameters and zero sequence parameters of each branch of the T-type transmission line, but also effectively determine the fault branch and calculate the fault distance under different fault types.

[0047] 2. The present invention simplifies the calculation process of parameter identification and fault location and improves calculation accuracy by performing Karrenbauer transformation on T-type transmission lines.

[0048] 3. The present invention uses PINN to realize line parameter identification and fault location, avoiding the approximate error caused by traditional numerical differential calculation and further improving the calculation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is the equivalent model of the T-type transmission line of the present invention;

[0050] Figure 2 It is the 1-module equivalent circuit of the T-type transmission line under normal operation of the present invention;

[0051] Figure 3 PINN1 schematic diagram for solving the positive sequence parameters of a T-type transmission line according to the present invention;

[0052] Figure 4 It is the 1-module equivalent circuit of the T-type transmission line when an asymmetric grounding fault occurs in the MT branch of the present invention;

[0053] Figure 5 Schematic diagram of PINN2 for T-type transmission line fault location according to the present invention;

[0054] Figure 6 It is the 0-mode equivalent circuit of the T-type transmission line when an asymmetric grounding fault occurs in the MT branch of the present invention;

[0055] Figure 7 PINN3 schematic diagram for solving the zero-sequence parameters of a T-type transmission line according to the present invention;

[0056] Figure 8 Schematic diagram of the T-type transmission line simulation model of the present invention. DETAILED DESCRIPTION

[0057] The present invention will be further described in detail below through specific examples. The following examples are only illustrative and not restrictive, and the scope of protection of the present invention cannot be limited thereto.

[0058] A method for parameter identification and fault location of T-type transmission lines based on Karenbauer transformation and physical information neural network, the innovation of which lies in the following steps:

[0059] S1, T-type transmission line equivalent model is as follows Figure 1 As shown. The line consists of three branches: MT, NT, and PT. The branch lengths are l M 、l N 、l P In the figure, u Ma 、u Mb 、u Mc 、u Na 、u Nb 、u Nc 、u Pa 、u Pb 、u Pc are the instantaneous values ​​of the three-phase voltages at terminals M, N, and P respectively; i Ma 、i Mb 、i Mc 、i Na 、i Nb 、i Nc 、i Pa 、i Pb 、i Pc are the instantaneous values ​​of the three-phase currents injected into the T node from terminals M, N, and P, respectively. To further simplify the calculation, the phase-to-phase capacitance and ground capacitance of each branch are concentrated at the branch endpoints, thereby reducing the order of the transmission line differential equation. Karenbauer transformations are performed on the T-type transmission line under normal operation and when an asymmetric ground fault occurs, resulting in the corresponding 0-mode and 1-mode equivalent circuits.

[0060] S2. The equivalent circuit of T-type transmission line 1 under normal operation is as follows Figure 2 As shown. Among them, u (1) M 、u (1) N 、u (1) P are the voltage 1-mode components of the M, N, and P terminals in normal state; i (1) M 、i (1) N 、i (1) P They are the 1-mode components of the current injected into the T node from the three terminals in the normal state; They are the 1 modulus components of the current injected into the T node after passing through the positive sequence grounding capacitor from the three terminals. The parameters to be determined in the line include the positive sequence resistance of the three branches Positive sequence inductance and positive sequence capacitor Depend on Figure 2 Based on this, PINN1 is constructed to solve the positive sequence parameters of T-type transmission lines, and its schematic diagram is shown as follows: Figure 3 As shown. Where j = M, N, P. The input of PINN1 is the sampling time series t under normal conditions. zc , the output is used to fit the voltage and current 1-mode components under normal conditions. First, based on equation (1), the physical information loss function is constructed as shown in equation (5). By minimizing the loss function through the optimization algorithm, the positive sequence capacitance can be obtained. On the basis of the positive sequence capacitance identification value, the physical information loss function is obtained based on the differential equation group (4) as shown in formula (6), and the positive sequence resistance is further identified. and positive sequence inductance In this way, PINN1 can accurately identify all positive sequence parameters of the T-type transmission line in steps.

[0061]

[0062] The equivalent circuit of the T-type transmission line when an asymmetric ground fault occurs on the S3 and MT branches is as follows: Figure 4 In the figure, f is the fault point, which is located at a distance of αl from the end of the line M. M Where, α is the fault distance proportional coefficient to be determined; is the 1-mode component of the fault point voltage; They are the voltage 1-mode components at the M, N, and P terminals under fault conditions; They are the current 1-mode components of the three terminals under fault conditions; and are the 1 modulus components of the current flowing to the T node after passing through the positive sequence grounding capacitor at the three terminals. Assume that the fault occurs in the MT, NT and PT branches respectively, and assume that the distance between the fault point f and the branch endpoint is a M l M 、a N l N and a N l P Based on this, PINN2 for fault location of T-type transmission line is constructed, and its schematic diagram is shown as follows: Figure 5 The input of PINN2 is the sampling time series t under the asymmetric ground fault state. gzThe output is used to fit the voltage and current 1-mode components under the asymmetric ground fault state, and three physical information loss functions are constructed as shown in formula (10). By training PINN2 three times, the loss functions containing the above three physical information are minimized in turn, and the fault distance proportional coefficient α is gradually calculated. M 、a N and a P When f is located near the T node, the calculated a M 、a N , α P are all close to 1, we can set ε=0.001, then |α M -1|≤ε、|α N -1|≤ε、|α P -1|≤ε; when f is located on the MT branch, there exists |α M -1|≤ε、|α N -1|>ε、|α P -1|>ε; when f is located in the NT branch, there exists |α M -1|>ε、|α N -1|≤ε、|α P -1|>ε; when f is located in the PT branch, there exists |α M -1|>ε、|α N -1|>ε、|α P -1|≤ε. Based on this, the criteria P1, P2, P3, and P4 are constructed as shown in equations (11)-(15). The values ​​of the corresponding criteria are calculated based on the results of PINN2, and the specific branch where the fault occurred is determined and the fault distance is determined.

[0063]

[0064] P1=P M &P N &P P (11)

[0065] P2=P1∧P M (12)

[0066] P3=P1∧P N (13)

[0067] P4=P1∧P P (14)

[0068]

[0069] Where i = M, N, P.

[0070] The 0-mode equivalent circuit when an asymmetrical ground fault occurs on the S4 and MT branches is as follows: Figure 6As shown in the figure, u (0) M 、u (0) N 、u (0) P are the zero-mode components of the voltage at the M, N, and P terminals under fault conditions; i (0) M 、i (0) N 、i (0) P They are the 0-mode components of the current flowing into the T node from the three terminals; and are the zero-modulus components of the current flowing from the three terminals through the positive-sequence grounding capacitor to the T node. The parameters to be determined in the line include the zero-sequence resistance of the three branches Zero-sequence inductance Zero-sequence capacitance Depend on Figure 6 Based on this, PINN3 is constructed to identify the zero-sequence parameters of T-type transmission lines, and its schematic diagram is shown as follows: Figure 7 As shown. The input of PINN3 is t gz , the output is used to fit the voltage and current zero-mode components under the asymmetric ground fault state, and the physical information loss function is constructed as shown in formula (17). By minimizing the loss function through the optimization algorithm, all zero-sequence parameters of the T-type transmission line can be obtained. However, due to the large order of magnitude difference between the zero-sequence capacitance value, the zero-sequence resistance value, and the zero-sequence inductance value, using the same learning rate may cause the model to fall into a local optimal solution. Therefore, in order to improve the training effect, the zero-sequence capacitance is set to zero. The learning rate is set to 10 -8 , and the zero sequence resistance and zero-sequence inductance The learning rate is set to 0.1. This setting can effectively balance the different parameters in the training process and improve the convergence and accuracy of the model.

[0071]

[0072] The following is the comparative verification result of the above embodiment:

[0073] In order to verify the accuracy and effectiveness of the present invention, a 110kV voltage level T-type transmission line simulation model was built in the MATLAB simulation environment. The specific structure of the simulation model is as follows Figure 8 As shown. The length of each branch is l M 、l N , are 75km, 45km and 105km respectively. The line parameters are shown in Table 1. The power sources connected to the three ends of the line are The fault is set at a distance of αl from the line end MM In addition, fault recording devices were simulated at the three ends of the line, and the data sampling frequency was set to 2kHz.

[0074] Table 1 Line parameters

[0075]

[0076] Table 2 lists the positive sequence parameter identification results of the present invention during normal line operation. As can be seen from Table 2, the absolute error between the identified and true positive sequence parameters for each branch remains within 1%. This demonstrates the high accuracy of the present invention in positive sequence parameter identification.

[0077] Table 2 Positive sequence parameter identification results

[0078]

[0079] Table 3 shows the fault branch determination results of the present invention when different types of faults occur near the T node or on each branch. The bold portion of the table also shows the final determined fault distance proportional coefficient. As can be seen from Table 3, the present invention can accurately distinguish whether the fault occurs near the T node or on a specific branch. Furthermore, when the fault occurs on a specific branch, the fault distance proportional coefficient can be effectively calculated, and the absolute error between the calculated value and the actual value is within 2%. This demonstrates that the present invention has high accuracy in fault location.

[0080] Table 3 Fault branch judgment and fault location results

[0081]

[0082] Tables 4 and 5 show the zero-sequence parameter identification results of the present invention for single-phase ground faults and two-phase ground faults, respectively. The results show that under different fault types, the absolute error between the identified and true zero-sequence parameters of each branch remains within 2%. This demonstrates the high accuracy of the present invention in zero-sequence parameter identification.

[0083] Table 4 Zero-sequence parameter identification results under AG fault

[0084]

[0085] Table 5 Zero-sequence parameter identification results under ABG fault

[0086]

[0087]

[0088] Gaussian noise with a mean of 0 and a standard deviation of 0.05%, 0.10%, 0.15% and 0.20% is added to the voltage and current fault recording data, and the line parameters and fault location are identified on this basis. Table 6 shows the positive-sequence parameter identification results of the present invention under different noise intensities. Table 7 shows the fault branch judgment results and fault distance measurement results of the present invention when the fault occurs on the MT branch under different noise intensities. Table 8 shows the zero-sequence parameter identification results of the present invention when a single-phase ground short circuit fault occurs on the line under different noise intensities. The results show that under different noise intensities, the absolute errors between the calculated values ​​and the true values ​​of the positive-sequence parameters, fault distance proportional coefficients and zero-sequence parameters are all within 2%, and the performance of the present invention is not greatly affected by noise.

[0089] Table 6 Positive sequence parameter identification results under different noise intensities

[0090]

[0091]

[0092] Table 7 Results of fault location methods under different noise intensities

[0093] Strong noise Fault branch judgment Absolute error 0.05％ 0.401 1.497 1.214 0 1 0 0 MT 0.26 0.10％ 0.397 1.496 1.217 0 1 0 0 MT 0.68 0.15％ 0.399 1.491 1.221 0 1 0 0 MT 0.20 0.20％ 0.398 1.488 1.220 0 1 0 0 MT 0.50

[0094] Table 8 Zero-sequence parameter identification results under different noise intensities

[0095]

[0096]

[0097] Although the embodiments and drawings of the present invention are disclosed for illustrative purposes, those skilled in the art will understand that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A method for parameter identification and fault location of T-type transmission lines based on Karenbauer transform and physical information neural network, characterized by: The steps of the method are: S1. Obtain the 1-mode and 0-mode equivalent circuits of a T-type transmission line in normal operation and the 1-mode and 0-mode equivalent circuits when an asymmetric ground fault occurs through Karenbauer transformation; S2. Based on the equivalent circuit of the T-type transmission line 1 under normal operation, a differential equation with positive sequence parameters as coefficients is established, and the coefficients of the differential equation are solved through PINN to realize the identification of the positive sequence parameters; S3. Based on the 1-mode equivalent circuit under the asymmetric ground fault state, a differential equation with the positive sequence parameter and the fault distance proportional coefficient as coefficients is established, and the coefficients of the differential equation are solved through PINN to determine the fault section and the fault distance; S4. Under the asymmetric ground fault state, a differential equation with zero-sequence parameters as coefficients is established based on the 0-module equivalent circuit, and the coefficients of the differential equation are solved through PINN to realize the identification of the zero-sequence parameters.

2. The method for parameter identification and fault location of a T-type transmission line based on Karenbauer transformation and physical information neural network according to claim 1 is characterized in that: The S2 is specifically: From the equivalent circuit of the T-type transmission line under normal operation, we can get in: They are the voltage 1-mode components of the M, N, and P terminals in normal conditions; They are the 1-mode components of the current injected into the T node from the three terminals in the normal state; They are the current 1 mode components injected into the T node from the three terminals through the positive sequence grounding capacitor; The parameters to be determined include the positive sequence resistance of the three branches Positive sequence inductance and positive sequence capacitor Construct PINN1 for solving the positive sequence parameters of T-type transmission lines, where j = M, N, P, and the input of PINN1 is the sampling time series t under normal conditions. zc , the output is used to fit the voltage and current 1-mode components under normal conditions; Based on formula (1), the physical information loss function is constructed as formula (5). By minimizing the loss function through the optimization algorithm, the positive sequence capacitance can be obtained. On the basis of obtaining the positive sequence capacitance identification value, the physical information loss function is obtained based on the differential equation group (4) as shown in formula (6), and the positive sequence resistance is further identified. and positive sequence inductance PINN1 can accurately identify all positive sequence parameters of T-type transmission lines in steps; 3. The method for parameter identification and fault location of a T-type transmission line based on Karenbauer transformation and physical information neural network according to claim 1 is characterized in that: The S3 is specifically: In the equivalent circuit of the T-type transmission line when an asymmetric ground fault occurs on the MT branch, f is the fault point, which is located at a distance of αl from the line end M. M Where, α is the fault distance proportional coefficient to be determined; is the 1-mode component of the fault point voltage; They are the voltage 1-mode components at the M, N, and P terminals under fault conditions; They are the current 1-mode components of the three terminals under fault conditions; and They are the 1-mode components of the current flowing to the T node after passing through the positive-sequence grounding capacitor at the three terminals; Assume that the fault occurs on the MT, NT and PT branches respectively, and assume that the distance between the fault point f and the branch endpoint is a M l M 、a N l N and a N l P , we can get equations (7) to (9); Based on this, PINN2 is constructed for fault location of T-type transmission lines. The input of PINN2 is the sampling time series t under the asymmetric ground fault state. gz , the output is used to fit the voltage and current 1-mode components under the asymmetric ground fault state, and three physical information loss functions are constructed as shown in formula (10); By training PINN2 three times, the loss function containing the above three physical information is minimized in turn, and the fault distance proportional coefficient α is gradually calculated M , α N and α P ; When f is located near the T node, the calculated α M , α N , α P are all close to 1, and ε = 0.001 is set, then |α M -1|≤ε、|α N -1|≤ε、|α P -1|≤ε; When f is located on the MT branch, there exists |α M -1|≤ε、|α N -1|>ε、|α P -1|>ε; when f is located in the NT branch, there exists |α M -1|>ε、|α N -1|≤ε、|α P -1|>ε; When f is located in the PT branch, there exists |α M -1|>ε、|α N -1|>ε、|α P -1|≤ε; Based on this, the construction criteria P1, P2, P3 and P4 are shown in equations (11) to (15); P1=P M &P N &P P (11) P2=P1^P M (12) P3=P1^P N (13) P4=P1^P P (14) Where i = M, N, P. The corresponding criterion value is calculated based on the result of PINN2, and then the specific branch where the fault occurs is determined and the fault distance is determined.

4. The method for parameter identification and fault location of a T-type transmission line based on Karenbauer transformation and physical information neural network according to claim 1, characterized in that: The S4 is specifically: In the 0-mode equivalent circuit when an asymmetric ground fault occurs on the MT branch, They are the zero-mode components of the voltage at the M, N, and P terminals under fault conditions; They are the 0-mode components of the current flowing into the T node from the three terminals; and They are the zero-mode components of the current flowing from the three terminals to the T node after passing through the positive-sequence grounding capacitor; The parameters to be determined in the circuit include the zero-sequence resistance of the three branches Zero-sequence inductance Zero-sequence capacitance Equation (16) is obtained from the zero-mode equivalent circuit when an asymmetric ground fault occurs on the MT branch. Based on this, PINN3 is constructed to identify the zero-sequence parameters of the T-type transmission line. The input of PINN3 is t gz , the output is used to fit the voltage and current zero-mode components under the asymmetric ground fault state, and the physical information loss function is constructed as shown in formula (17). By minimizing the loss function through the optimization algorithm, all zero-sequence parameters of the T-type transmission line can be obtained; in order to improve the training effect, the zero-sequence capacitor The learning rate is set to 10 -8 , and the zero sequence resistance and zero-sequence inductance The learning rate is set to 0.1; Effectively balance different parameters during training to improve the convergence and accuracy of the model;

Images