Double-end hybrid line asynchronous fault location method based on BP neural network
The fault data of the cable-overhead line hybrid line is processed through the BP neural network and the least squares method, and high-precision fault ranging is realized without relying on line parameters, solving the problem of large fault ranging error in the existing technology, and is suitable for actual power grids. There is less historical failure data during operation.
Patent Information
- Application Number
- CN202410859288.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-28
- Publication Date
- 2024-11-15
AI Technical Summary
The fault ranging method of existing cable-overhead line hybrid circuits is easily affected by line modeling and parameter errors, resulting in large distance measurement errors and high equipment requirements, making it difficult to achieve fast and accurate fault location in practical applications.
Using the double-ended hybrid line asynchronous fault ranging method based on BP neural network, the fault ranging function, data preprocessing and BP neural network model training can be used to realize fault ranging without relying on line parameters, using the least squares method and the full The Bofu algorithm processes electrical quantity data to reduce the number of fault data samples required for network training.
Achieve high-precision fault ranging accuracy, and the distance measurement accuracy is not affected by factors such as fault location, fault type, transition resistance, fault initial phase angle, load current and asynchronous data, reducing the dependence on line parameters and is suitable for actual conditions. There is less historical fault data during power grid operation.
Smart Images

Figure CN118962322A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of relay protection of electric power systems, and in particular relates to a double-terminal hybrid line asynchronous fault distance measurement method based on BP neural network. Background Art
[0002] With the development of my country's power system, urban power transmission lines have been gradually changed from the previous overhead open-wire method that occupied more space to the buried cable method due to the limitation of land resources and the influence of surrounding buildings. Compared with overhead lines, high-voltage power cables take up less space, are less susceptible to environmental impacts, and are safer and more stable to operate. Since cable lines are expensive and difficult to construct, transmission lines have developed cable-overhead hybrid lines based on the original cables and overhead lines, and their application has become more and more widespread, gradually becoming a common transmission method. When a hybrid line fails, a fast and accurate fault distance measurement method can greatly reduce the burden of line inspection, which is of great significance to the safety, stability and economic operation of the power system.
[0003] The existing fault distance measurement algorithms for cable-overhead line hybrid lines can be mainly divided into traveling wave method and fault analysis method according to different principles. The traveling wave method not only has the problem of traveling wave head identification, but is also easily affected by the multiple refraction and reflection of the traveling wave at the junction of the overhead line and the cable line. It needs to be equipped with high-speed sampling equipment, and the technical implementation is complex and the investment is large. In contrast, the fault analysis method has low equipment requirements and small investment, and has strong engineering practicality, so it has been widely studied. However, in actual situations, the line may have inaccurate or unknown parameters due to reasons such as long commissioning time or incomplete parameter records, which leads to large ranging errors in the ranging method based on the fault analysis method, and the accuracy of line modeling will also affect the accuracy of fault location.
[0004] Therefore, it is of practical significance to propose a transmission line fault location method that is not affected by line modeling and parameter errors. With the widespread application of artificial intelligence technology in power systems, artificial intelligence algorithms also provide a solution to the fault location problem of power lines without the need for line models and line parameters. Summary of the invention
[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a double-terminal hybrid line asynchronous fault ranging method based on BP neural network, which does not rely on accurate line parameters, has high ranging accuracy, and is not affected by factors such as fault location, fault type, transition resistance, fault initial phase angle, load current and asynchronous data.
[0006] The present invention solves the technical problem by the following technical solutions:
[0007] A method for non-synchronous fault location of a double-terminal hybrid line based on BP neural network, characterized in that the steps of the method are:
[0008] Construct hybrid line fault location function;
[0009] Obtaining electrical quantities based on a hybrid line fault location function;
[0010] The electrical quantity data are preprocessed using the least square method;
[0011] Calculate the preprocessed electrical quantity data, design and train the fault location model based on BP neural network;
[0012] When a line fault occurs, the trained BP neural network fault location model is used to perform fault location measurement.
[0013] Moreover, the hybrid line fault location function is:
[0014]
[0015] Where: Z c1 and γ1 are the positive sequence wave impedance and positive sequence propagation constant of the cable line respectively;
[0016] Z c2 and γ2 are the positive sequence wave impedance and positive sequence propagation constant of the overhead line respectively;
[0017] l1 is the length of the cable line;
[0018] l2 is the length of the overhead line;
[0019] L is the total length of the hybrid line;
[0020] x is the distance between the fault point f and the M end.
[0021] Moreover, the for:
[0022]
[0023] in: and They are the positive sequence voltage and current at the M terminal and the positive sequence voltage and current at the N terminal after the line fault respectively;
[0024] δ is the asynchronous angle of the data collected by the M end relative to the data collected by the N end, with the time when the data is collected by the N end as a reference.
[0025] Furthermore, A2 and A1 are:
[0026]
[0027] Moreover, the A2 and A1 relationships include that when the two-section cable-overhead line hybrid line operates normally, the positive sequence voltage and positive sequence current at both ends of the line have the following relationship:
[0028]
[0029] in: and They are respectively the positive-sequence voltage and current at the M terminal and the positive-sequence voltage and current at the N terminal when the line is operating normally.
[0030] Moreover, the method of obtaining electrical quantities based on the hybrid line fault distance measurement function is specifically as follows: when a short circuit fault occurs in the transmission line, the sampling data of the voltage and current at both ends of the line before and after the fault are obtained, and the fundamental components of the voltage and current at both ends are calculated respectively using the full-wave Fourier algorithm, and then the positive sequence components of the voltage and current at both ends are calculated using the symmetrical component method.
[0031] Moreover, the use of the least square method to pre-process the above electrical quantity data is specifically as follows:
[0032] The positive sequence voltage at the M terminal before the line fault and positive sequence current As the independent variable of the linear function, the positive sequence voltage at the N terminal before the line fault and positive sequence current As the dependent variable of the linear function, the linear function in equation (4) is fitted using the least squares method to obtain the linear function coefficient matrix A2A1e jδ The values of .
[0033] Furthermore, the calculation of the preprocessed electrical quantity data and the design and training of the fault location model based on the BP neural network are specifically as follows:
[0034] The positive sequence voltage at terminal M after the line fault and positive sequence current Substituting into the fitted linear function, the results are respectively consistent with the positive sequence voltage at the N terminal after the line fault and positive sequence current Perform subtraction and then convert the result and Perform division operation, The real part k1 and imaginary part k2 of are the final line electrical characteristic quantities;
[0035] The fault distance measurement model based on BP neural network comprises three layers, namely, input layer, hidden layer and output layer. The number of neurons in the input layer and the number of neurons in the output layer are 2 and 1 respectively, the number of neurons in the hidden layer is selected to be 10, and the activation function is the tansig function.
[0036] After data preprocessing, the line electrical characteristics k1 and k2 are used as the input of the BP neural network, and the corresponding fault location is used as the output of the BP neural network. The network is trained to fit the distance measurement function in equation (1).
[0037] Moreover, when a line fault occurs, the fault distance measurement is performed using the trained BP neural network fault distance measurement model. Specifically, when a line fault occurs, the fault data of the line is collected for preprocessing, and then the obtained line electrical characteristic quantities k1 and k2 are input into the trained BP neural network to output the corresponding fault location.
[0038] The advantages and beneficial effects of the present invention are:
[0039] The method of the present invention realizes the feature transformation and feature aggregation of the original voltage and current by performing data preprocessing on the original voltage and current at both ends of the line, so that the preprocessed data can fully reflect the characteristics of the electrical quantities at both ends of the line under different fault types and different fault distances. The preprocessed data is input into the BP neural network for training, which can achieve the effect of reducing the amount of fault data samples required for network training, thereby matching the actual situation of less historical fault data in power grid operation. Based on the trained BP neural network, accurate fault distance measurement can be achieved by using the positive sequence voltage and positive sequence current sampled asynchronously at both ends after the transmission line fault. The distance measurement accuracy is not affected by factors such as fault location, fault type, transition resistance, fault initial phase angle, load current and asynchronous data, and has a very high distance measurement accuracy. In addition, the method of the present invention does not rely on line parameters when performing fault distance measurement, so it is not affected by line parameter errors. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] Figure 1 A schematic diagram of a system for a two-section cable-overhead line hybrid line applied to the present invention when the fault point is located in the cable line;
[0041] Figure 2 A schematic diagram of a system for a two-section cable-overhead line hybrid line applied to the present invention when the fault point is located in the overhead line;
[0042] Figure 3 It is a schematic diagram of the BP neural network structure used for double-terminal hybrid line fault location in the present invention;
[0043] Figure 4 It is a schematic diagram of the overall process of the double-terminal hybrid line fault location method in the present invention;
[0044] Figure 5 This is a diagram of the distance measurement results when different types of short circuit faults occur at different locations of the present invention;
[0045] Figure 6It is a distance measurement result diagram of the power transmission line of the present invention when different types of short-circuit faults occur at different transition resistances at different times;
[0046] Figure 7 This is a diagram showing the distance measurement results when different types of short-circuit faults occur in the power transmission line of the present invention under different load current conditions and through different transition resistances. DETAILED DESCRIPTION
[0047] The present invention is further described in detail below through specific examples. The following examples are only illustrative and not restrictive, and the protection scope of the present invention cannot be limited thereto.
[0048] A method for non-synchronous fault location of a double-terminal hybrid line based on BP neural network, the innovation of which lies in: the steps of the method are:
[0049] Construct hybrid line fault location function,
[0050]
[0051] Where: Z c1 and γ1 are the positive sequence wave impedance and positive sequence propagation constant of the cable line respectively;
[0052] Z c2 and γ2 are the positive sequence wave impedance and positive sequence propagation constant of the overhead line respectively;
[0053] l1 is the length of the cable line;
[0054] l2 is the length of the overhead line;
[0055] L is the total length of the hybrid line;
[0056] x is the distance between the fault point f and the M end.
[0057] Said for:
[0058]
[0059] in: and They are the positive sequence voltage and current at the M terminal and the positive sequence voltage and current at the N terminal after the line fault respectively;
[0060] δ is the asynchronous angle of the data collected by the M end relative to the data collected by the N end, with the time when the data is collected by the N end as a reference.
[0061] The A2 and A1 are:
[0062]
[0063] The A2 and A1 relationships include the following relationship between the positive sequence voltage and the positive sequence current at both ends of the line when the two-section cable-overhead line hybrid line operates normally:
[0064]
[0065] in: and They are respectively the positive-sequence voltage and current at the M terminal and the positive-sequence voltage and current at the N terminal when the line is operating normally.
[0066] The electrical quantity is obtained according to the hybrid line fault location function. When a short circuit fault occurs in the transmission line, the sampling data of the voltage and current at both ends of the line before and after the fault are taken, and the fundamental components of the voltage and current at both ends are calculated respectively using the full-wave Fourier algorithm, and then the positive sequence components of the voltage and current at both ends are obtained using the symmetrical component method.
[0067] The above electrical quantity data are preprocessed using the least square method, and the positive sequence voltage at the M terminal before the line fault is and positive sequence current As the independent variable of the linear function, the positive sequence voltage at the N terminal before the line fault and positive sequence current As the dependent variable of the linear function, the linear function in equation (4) is fitted using the least squares method to obtain the linear function coefficient matrix A2A1e jδ The values of .
[0068] The pre-processed electrical quantity data is calculated, and a fault location model based on BP neural network is designed and trained to calculate the positive sequence voltage at the M terminal after the line fault. and positive sequence current Substituting into the fitted linear function, the results are respectively consistent with the positive sequence voltage at the N terminal after the line fault and positive sequence current Perform subtraction and then convert the result and Perform division operation, The real part k1 and imaginary part k2 of are the final line electrical characteristic quantities;
[0069] The fault distance measurement model based on BP neural network comprises three layers, namely, input layer, hidden layer and output layer. The number of neurons in the input layer and the output layer are 2 and 1 respectively, the number of neurons in the hidden layer is 10, and the activation function is the tansig function; its expression is:
[0070]
[0071] After data preprocessing, the line electrical characteristics k1 and k2 are used as the input of the BP neural network, and the corresponding fault location is used as the output of the BP neural network. The network is trained to fit the distance measurement function in equation (1).
[0072] When a line fault occurs, the fault distance measurement is performed using the trained BP neural network fault distance measurement model. The line fault data is collected for preprocessing, and then the obtained line electrical characteristics k1 and k2 are input into the trained BP neural network to output the corresponding fault location.
[0073] Figure 1 , 2 FIG. 1 is a fault diagram of a two-section cable-overhead line hybrid line applied to the present invention, wherein the distance between the fault point f and the M end is x, the length of the cable line is l1, the length of the overhead line is l2, the total length of the hybrid line is L, and J is the connection point between the cable line and the overhead line. The voltage and current at each point of the line are as follows: Figure 1 , 2 The present invention discloses a method for non-synchronous fault location of a double-terminal hybrid line based on a BP neural network, which mainly includes two parts: analysis of the principle of hybrid line fault location and realization of a fault location algorithm.
[0074] 1. Analysis of hybrid line fault location principle
[0075] Based on the distributed parameter model, the distance measurement principle of the two-stage cable-overhead line hybrid line is analyzed. The present invention takes into account the asynchrony of the data sampling at both ends. Taking the time of data collection at the N end as a reference, the asynchrony angle of the data collected at the M end relative to the data collected at the N end is δ. That is, when considering the asynchrony of the data collected at both ends, if the positive sequence voltage and positive sequence current at the N end are respectively and Then the positive sequence voltage and positive sequence current at terminal M are and
[0076] (1) Cable line failure
[0077] like Figure 1 As shown in the figure, when a fault occurs at point f, according to the uniform transmission line equation, the positive sequence voltage at the M end is and positive sequence current Calculate the positive sequence voltage at the fault point and the positive sequence current flowing into the fault point from the M side for:
[0078]
[0079] In the formula, Z c1 and γ1 are the positive-sequence wave impedance and positive-sequence propagation constant of the cable line respectively.
[0080] The positive sequence voltage at the fault point and the positive sequence current flowing from the fault point to the N side Calculate the positive sequence voltage at the N terminal and positive sequence current for:
[0081]
[0082] In the formula, Z c2 and γ2 are the positive-sequence wave impedance and positive-sequence propagation constant of the overhead line, respectively.
[0083] according to in is the positive sequence current flowing into the fault point, and it can be obtained from formula (7):
[0084]
[0085] Substituting formula (6) into formula (8), we can obtain:
[0086]
[0087] Where:
[0088]
[0089] make
[0090]
[0091] Simplifying formula (9) yields:
[0092]
[0093] Divide both sides of the equal sign in equation (12) to eliminate the unknown quantity. Then, after equation transformation, the distance measurement equation can be obtained as follows:
[0094]
[0095] (2) Overhead line failure
[0096] like Figure 2 As shown in the figure, when a fault occurs at point f, according to the uniform transmission line equation, the positive sequence voltage at the M end is and positive sequence current Calculate the positive sequence voltage at the fault point and the positive sequence current flowing into the fault point from the M side for:
[0097]
[0098] The positive sequence voltage at the fault point and the positive sequence current flowing from the fault point to the N side Calculate the positive sequence voltage at the N terminal and positive sequence current for:
[0099]
[0100] according to Formula (15) can be transformed into:
[0101]
[0102] Substituting formula (14) into formula (16), we can obtain:
[0103]
[0104] Simplifying formula (17), we can get:
[0105]
[0106] Divide both sides of the equal sign in equation (18) to eliminate the unknown quantity. Then, after equation transformation, the distance measurement equation can be obtained as follows:
[0107]
[0108] (3) Two-section cable-overhead line hybrid line fault
[0109] When the two-section cable-overhead line hybrid line operates normally, the positive sequence voltage and positive sequence current at both ends of the line have the following relationship:
[0110]
[0111] Where: and They are respectively the positive-sequence voltage and current at the M terminal and the positive-sequence voltage and current at the N terminal when the line is operating normally.
[0112] make
[0113]
[0114] Where: and They are respectively the positive sequence voltage and current at the M terminal and the positive sequence voltage and current at the N terminal after the line fault.
[0115] Combined with the above analysis, the distance measurement function of the two-section cable-overhead line hybrid line is:
[0116]
[0117] Where: f(x) is an expression including line parameters and fault location.
[0118] 2. Implementation of fault distance measurement algorithm
[0119] The above analysis of the two-section cable-overhead line hybrid line shows that the positive sequence voltage and positive sequence current at both ends of the line before and after the fault can be used to obtain the distance function From formula (22), we can see that the ranging function of the hybrid line at both ends is a piecewise function, and the parameters Z of each section of the line need to be given in advance. ci , γ i , l i In fact, the parameters of each section of a two-terminal hybrid line may be unknown or inaccurate, so it is difficult to directly construct and solve the ranging function. The present invention uses the least squares method and the function fitting ability of the BP neural network, and proposes a data-mechanism-integrated two-terminal hybrid line asynchronous fault ranging algorithm based on the positive sequence voltage and positive sequence current data at both ends before and after the transmission line fault.
[0120] It can be seen from formula (20) that when the transmission line operates normally, the positive sequence voltage and current at the M terminal are linearly related to the positive sequence voltage and current at the N terminal, where each coefficient depends on the asynchronous angle δ and the line parameters (Z ci , γ i , l i ), which is independent of factors such as fault type and fault location. Since the asynchronous angle and line parameters before and after the transmission line fault are the same, the data at both ends of the transmission line after the fault and the coefficients are substituted into the corresponding formula (21), the distance function can be obtained. Then, the fault location function in equation (22) is fitted by BP neural network to establish the fault location model of double-terminal hybrid line. The structure of BP neural network is as follows: Figure 3 As shown, the overall process of the double-terminal hybrid line fault location algorithm of the present invention is as follows Figure 4 As shown, the specific steps are as follows:
[0121] (1) Data preprocessing based on least squares method:
[0122] When a short circuit occurs in a transmission line, the voltage and current sampling data at both ends of the line before and after the fault are taken, and the fundamental components of the voltage and current at both ends are calculated using the full-wave Fourier algorithm, and then the positive sequence components of the voltage and current at both ends are calculated using the symmetrical component method. and positive sequence current As the independent variable of the linear function, the positive sequence voltage at the N terminal before the line fault and positive sequence current As the dependent variable of the linear function, the least squares method is used to fit the linear function in equation (20). Finally, the positive sequence voltage at the M terminal after the line fault is and positive sequence current Substituting into the fitted linear function, the results are respectively consistent with the positive sequence voltage at the N terminal after the line fault and positive sequence current Perform subtraction and then convert the result and Perform division operation, The real part k1 and imaginary part k2 are the final line electrical characteristics.
[0123] (2) Design and training of fault location model based on BP neural network:
[0124] The BP neural network model includes three layers, namely, an input layer, a hidden layer and an output layer. The number of neurons in the input layer and the number of neurons in the output layer are 2 and 1 respectively. The present invention comprehensively considers the output error and overfitting phenomenon, selects the number of neurons in the hidden layer as 10, and the activation function is the tansig function. After data preprocessing, the obtained line electrical characteristics k1 and k2 are used as the input of the BP neural network, and the corresponding fault position is used as the output of the BP neural network to train the network. The weights and thresholds from the output layer to the hidden layer and from the hidden layer to the input layer are adjusted layer by layer by using the reverse propagation of the error until the BP neural network reaches the predetermined error. At this time, the functional relationship between the line electrical characteristics and the fault position is stored in the BP neural network in the form of connection weights and thresholds between neurons.
[0125] (3) Fault distance measurement
[0126] When a line fault occurs, the fault data of the line is collected for preprocessing, and then the obtained line electrical characteristics k1 and k2 are input into the trained BP neural network to output the corresponding fault location.
[0127] After analyzing and summarizing the distance measurement principle of the double-terminal hybrid line, the method of the present invention realizes the feature transformation and feature aggregation of the original voltage and current by preprocessing the data of the original voltage and current at both ends of the line, so that the preprocessed data can fully reflect the characteristics of the electrical quantities at both ends of the line under different fault types and different fault distances. The preprocessed data is input into the BP neural network for training, which can achieve the effect of reducing the amount of fault data samples required for network training, thereby matching the actual situation of less historical fault data in power grid operation. The method of the present invention takes into account the sampling asynchronous angle of the data at both ends of the line, does not require line parameters when performing fault distance measurement, is not affected by parameter errors and asynchronous data, and the ranging accuracy is not affected by factors such as fault type, fault location, transition resistance, fault initial phase angle and load current.
[0128] The present invention builds a 500kV cable-overhead line two-stage transmission line simulation model in the Matlab / Simulink environment to verify the above algorithm. Figure 1 As shown. The system frequency is 50Hz, the total line length L is 100km, the length of the cable line l1 is 30km, the length of the overhead line l2 is 70km, and the power supply parameters on the M and N sides are 500∠65°kV and 500∠0°kV respectively. The equivalent impedance of the system on both sides is: Z M =(2.4578+j76.3962)Ω; Z N =(2.4515+j76.0822)Ω. The cable line parameters are: [9] : Positive sequence impedance Z d1 =(0.0240+j0.1485)Ω / km; zero sequence impedance Z d0 =(0.4120+j0.4819)Ω / km; positive sequence capacitance C d1 =0.2811μF / km; zero sequence capacitance C d0 =0.1529μF / km. Overhead line parameters are [9] : Positive sequence impedance Z j1 =(0.0346+j0.4216)Ω / km; zero sequence impedance Z j0 =(0.0300+j1.1400)Ω / km; positive sequence capacitance C j1 =0.00864μF / km; zero sequence capacitance C j0 =0.006175μF / km. In the simulation, the data sampling frequency at both ends is 10kHz, and the full-wave Fourier algorithm is used to extract the fundamental phasor for distance measurement. The fault distance is based on the M end as the starting point, and the calculation formula for the relative distance measurement error is defined as
[13] :
[0129]
[0130] This simulation uses Matlab editor to write a program to change the corresponding parameters of the transmission line simulation model to realize the acquisition of fault data. The simulation collected 304 sets of fault data to train the BP neural network. The set fault conditions include 4 types of faults (three-phase short circuit, single-phase ground short circuit, two-phase phase-to-phase short circuit, two-phase ground short circuit) and 4 types of transition resistance (0Ω, 30Ω, 100Ω, 300Ω). For each fault condition, 19 sets of fault data at different fault locations were collected.
[0131] In order to evaluate the performance of the fault location algorithm proposed in the present invention, the trained BP neural network is used to perform fault location measurement for different fault types at different fault locations, different transition resistances, different fault initial phase angles and different load currents.
[0132] (1) Influence of transition resistance and fault type
[0133] The transmission line passes through 0Ω, 30Ω, 100Ω, and 300Ω transition resistances respectively. The distance measurement results when different types of short-circuit faults occur at different locations are as follows: Figure 5 As shown. Figure 5 It can be seen that when a single-phase ground short circuit and a three-phase short circuit occur, the relative ranging error of the proposed algorithm is relatively high, but it is still maintained within 0.25% overall. It can be seen that the proposed algorithm has a high ranging accuracy when various short circuit faults occur through different transition resistances.
[0134] (2) Influence of initial fault phase angle
[0135] The distance measurement results when different types of short-circuit faults occur in the transmission line at different times through different transition resistances are as follows: Figure 6 shown. Figure 6 a~d are the situations of single-phase grounding fault, two-phase grounding short-circuit fault, three-phase short-circuit fault and two-phase interphase short-circuit fault respectively. Figure 6 It can be seen that the relative ranging error of the proposed algorithm under different fault initial phase angles and different transition resistances does not exceed 0.4%, and it has high ranging accuracy.
[0136] (3) Impact of load current
[0137] The distance measurement results of transmission lines under different load current conditions and different types of short-circuit faults occurring through different transition resistances are as follows: Figure 7 As shown, Figure 7 a~d are the cases of single-phase grounding fault, two-phase grounding short circuit fault, three-phase short circuit fault and two-phase interphase short circuit fault respectively. The load current is represented by the phase angle difference of the power supply at both ends. Figure 7 It can be seen that the relative ranging error of the proposed algorithm under different load currents is kept within 0.4%, and it has high ranging accuracy.
[0138] In summary, the effectiveness of the algorithm proposed in the present invention is not affected by factors such as fault location, fault type, transition resistance, fault initial phase angle and load current, and its relative ranging error is always kept within 0.4%, with high ranging accuracy.
[0139] Although the embodiments and drawings of the present invention are disclosed for illustrative purposes, those skilled in the art will appreciate 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 asynchronous fault location of a double-terminal hybrid line based on BP neural network, characterized in that: The steps of the method are: Construct hybrid line fault location function; Obtaining electrical quantities based on a hybrid line fault location function; The electrical quantity data are preprocessed using the least square method; Calculate the preprocessed electrical quantity data, design and train the fault location model based on BP neural network; When a line fault occurs, the trained BP neural network fault location model is used to perform fault location measurement.
2. The method for asynchronous fault location of a double-terminal hybrid line based on a BP neural network according to claim 1 is characterized in that: The hybrid line fault location function is: Where: Z c1 and γ1 are the positive sequence wave impedance and positive sequence propagation constant of the cable line respectively; Z c2 and γ2 are the positive sequence wave impedance and positive sequence propagation constant of the overhead line respectively; l1 is the length of the cable line; l2 is the length of the overhead line; L is the total length of the hybrid line; x is the distance between the fault point f and the M end.
3. The method for asynchronous fault location of a double-terminal hybrid line based on a BP neural network according to claim 2 is characterized in that: Said for: in: and They are the positive sequence voltage and current at the M terminal and the positive sequence voltage and current at the N terminal after the line fault respectively; δ is the asynchronous angle of the data collected by the M end relative to the data collected by the N end, with the time when the data is collected by the N end as a reference.
4. The method for asynchronous fault location of a double-terminal hybrid line based on a BP neural network according to claim 3 is characterized in that: The A2 and A1 are:
5. The method for asynchronous fault location of a double-terminal hybrid line based on a BP neural network according to claim 4 is characterized in that: The A2 and A1 relationships include the following relationship between the positive sequence voltage and the positive sequence current at both ends of the line when the two-section cable-overhead line hybrid line operates normally: in: and They are respectively the positive-sequence voltage and current at the M terminal and the positive-sequence voltage and current at the N terminal when the line is operating normally.
6. The method for asynchronous fault location of a double-terminal hybrid line based on a BP neural network according to claim 1, characterized in that: The method of obtaining electrical quantities according to the hybrid line fault distance measurement function is as follows: when a short circuit fault occurs in the transmission line, sampling data of the voltage and current at both ends of the line before and after the fault are obtained, the fundamental components of the voltage and current at both ends are calculated respectively using the full-wave Fourier algorithm, and then the positive sequence components of the voltage and current at both ends are obtained using the symmetrical component method.
7. The method for asynchronous fault location of a double-terminal hybrid line based on a BP neural network according to claim 1, characterized in that: The use of the least square method to pre-process the electrical quantity data is specifically as follows: The positive sequence voltage at the M terminal before the line fault and positive sequence current As the independent variable of the linear function, the positive sequence voltage at the N terminal before the line fault and positive sequence current As the dependent variable of the linear function, the linear function in equation (4) is fitted using the least squares method to obtain the linear function coefficient matrix A2A1e jδ The values of .
8. The method for asynchronous fault location of a double-terminal hybrid line based on a BP neural network according to claim 1, characterized in that: The calculation of the preprocessed electrical quantity data and the design and training of the fault distance measurement model based on the BP neural network are specifically as follows: The positive sequence voltage at terminal M after the line fault and positive sequence current Substituting into the fitted linear function, the results are respectively consistent with the positive sequence voltage at the N terminal after the line fault and positive sequence current Perform subtraction and then convert the result and Perform division operation, The real part k1 and imaginary part k2 of are the final line electrical characteristic quantities; The fault distance measurement model based on BP neural network comprises three layers, namely, input layer, hidden layer and output layer. The number of neurons in the input layer and the number of neurons in the output layer are 2 and 1 respectively, the number of neurons in the hidden layer is selected to be 10, and the activation function is the tansig function. After data preprocessing, the line electrical characteristics k1 and k2 are used as the input of the BP neural network, and the corresponding fault location is used as the output of the BP neural network. The network is trained to fit the distance measurement function in equation (1).
9. The method for asynchronous fault location of a double-terminal hybrid line based on a BP neural network according to claim 1, characterized in that: When a line fault occurs, the fault distance measurement is performed using the trained BP neural network fault distance measurement model. Specifically, when a line fault occurs, the fault data of the line is collected for preprocessing, and then the obtained line electrical characteristics k1 and k2 are input into the trained BP neural network to output the corresponding fault location.
