Wire fault identification method and device of power system, terminal equipment and storage medium

Abstract

The invention discloses a wire fault identification method and device for a power system, terminal equipment and a storage medium. The method comprises the following steps: acquiring a frequency domain signal of a to-be-identified power region; extracting a fault characteristic quantity according to the frequency domain signal; inputting the fault characteristic quantity into a preset wire fault recognition model to obtain a wire fault recognition result; the training of the wire fault recognition model comprises the step of inputting the historical fault characteristic quantity into a to-be-trained wire fault recognition model to obtain an initial smoothing factor; according to the current smoothing factor and a historical lead fault result, repeatedly executing model optimization operation until a preset number of optimization times is reached, and obtaining a trained lead fault identification model; the model optimization operation comprises training according to a current smoothing factor and a historical fault characteristic quantity to obtain a lead fault prediction result; and updating the current smoothing factor according to a loss function calculated according to a historical lead fault result and a predicted lead fault result. According to the invention, the accuracy of wire fault identification is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power transmission line monitoring, and in particular to a method, a device, a terminal device and a storage medium for identifying conductor faults in a power system. Background Art

[0002] As the scale of the power grid continues to expand, the safety and reliability of power grid operation has attracted widespread attention. Transmission lines are the main power components connecting the national power grid. Since transmission lines cross different complex terrains, they are exposed to various environments for a long time and are prone to failures such as splitting. Once a transmission line fails, it will seriously affect the safe and reliable operation of the power grid system. Traditional power grid equipment monitoring operations can only perform video monitoring and recording functions, and cannot perform intelligent active identification and analysis of monitoring targets. It requires manual observation and analysis of images at all times. The human eye is prone to fatigue and manual subjectivity is strong, which affects the accuracy of transmission line fault identification and leads to low accuracy of transmission line status monitoring. Summary of the invention

[0003] The embodiments of the present invention provide a method, an apparatus, a terminal device and a storage medium for identifying conductor faults in an electric power system, which can effectively solve the problem that the prior art cannot perform intelligent and active identification and analysis of monitoring targets, requires manual observation and analysis of images at all times, and the human eye is easily fatigued and has strong subjectivity, which affects the accuracy of transmission line fault identification and leads to low accuracy of transmission line status monitoring.

[0004] An embodiment of the present invention provides a method for identifying a conductor fault in an electric power system, comprising:

[0005] Acquire a frequency domain signal of the power area to be identified according to a preset time interval;

[0006] Performing time-frequency analysis on the frequency domain signal to extract fault feature quantities;

[0007] Inputting the fault characteristic quantity into a preset wire fault identification model to perform fault identification and classification, and obtaining a wire fault identification result of the power area to be identified;

[0008] The training of the wire fault identification model includes:

[0009] Obtain historical fault feature quantities and historical conductor fault results;

[0010] Inputting the historical fault feature quantity into the wire fault identification model to be trained, so as to initialize the parameters of the wire fault identification model and obtain an initial smoothing factor;

[0011] Taking the historical fault feature as input and the predicted conductor fault result as output, the model optimization operation is repeatedly performed according to the current smoothing factor and the historical conductor fault result until a preset number of optimization times is reached to obtain a trained conductor fault identification model; the model optimization operation includes:

[0012] The conductor fault identification model to be trained is trained according to the current smoothing factor and the historical fault feature quantity to obtain the predicted conductor fault result; wherein the current smoothing factor during the first training is the initial smoothing factor;

[0013] According to the historical wire fault results and the predicted wire fault results, a loss function of a wire fault identification model is calculated; and a current smoothing factor is updated according to the loss function.

[0014] Furthermore, the fault characteristic quantities include: voltage, current, temperature and resistance;

[0015] Performing time-frequency analysis on the frequency domain signal to extract fault feature quantities includes:

[0016] Performing Fourier transform on the frequency domain signal to obtain a frequency domain representation after Fourier transform;

[0017] Performing a generalized S transform according to the frequency domain representation, the preset time interval, and the preset number of sampling points to obtain a transient energy sum of the frequency domain signal;

[0018] Feature extraction is performed based on the transient energy and to obtain a fault feature quantity.

[0019] Furthermore, before inputting the fault feature into a preset wire fault identification model, the method further includes:

[0020] According to each fault characteristic quantity, determining the maximum value and the minimum value of each fault characteristic quantity;

[0021] Data normalization is performed according to each fault characteristic quantity, the maximum value and the minimum value of each fault characteristic quantity to obtain the final fault characteristic quantity.

[0022] Furthermore, the historical fault feature quantity is input into the wire fault identification model to be trained to initialize the model parameters and obtain the current smoothing factor, including:

[0023] Inputting the historical fault feature into the wire fault identification model to be trained to initialize the model and obtain the current particle position and the predicted particle position;

[0024] According to the current particle position and the predicted particle position, the current particle fitness value is calculated;

[0025] Compare the current particle fitness value to determine the optimal position of the current particle;

[0026] Compare according to the current particle fitness value until the preset number of iterations is reached to obtain the current smoothing factor;

[0027] According to the current optimal position of the particle, it is converted into a chaotic sequence to obtain a chaotic optimization result; and according to the chaotic optimization result, the current particle position is randomly replaced to update the current particle position.

[0028] Furthermore, it also includes:

[0029] Randomly replace a current particle position according to the chaotic optimization result to obtain the current particle position after replacement;

[0030] According to the current particle fitness value and the preset constraint factor, the position of the current particle after replacement is updated based on the constraint factor;

[0031] The constraint factors include: a first Lagrange multiplier for expressing an inequality constraint, a second Lagrange multiplier for expressing an equality constraint, and a penalty factor.

[0032] Furthermore, according to the historical fault feature quantity and the current smoothing factor, the predicted wire fault result of the wire fault identification model to be trained is obtained, including:

[0033] According to the current smoothing factor, the conductor fault identification model is reconstructed;

[0034] The historical fault feature quantities are input into the reconstructed conductor fault identification model for feature learning, so that the reconstructed conductor fault identification model outputs the predicted conductor fault result.

[0035] Furthermore, the wire fault identification result includes: a fault location identification result and a fault type classification result;

[0036] The fault location identification result includes: out-of-area fault, bus fault, line fault and non-fault location;

[0037] The fault type classification results include: abnormal resistance value, abnormal current and voltage value, abnormal insulation resistance, abnormal temperature and no fault type;

[0038] After obtaining the wire fault identification result of the power area to be identified, it also includes:

[0039] Determine the fault location and fault type of the power area to be identified based on the conductor fault identification result of the power area to be identified;

[0040] Select the corresponding fault protection measures according to the determined fault location and fault type.

[0041] As an improvement of the above solution, another embodiment of the present invention provides a conductor fault identification device for a power system, comprising:

[0042] A signal acquisition module, used to acquire a frequency domain signal of a power area to be identified according to a preset time interval;

[0043] A fault feature extraction module is used to perform time-frequency analysis based on the frequency domain signal to extract fault feature quantities;

[0044] A conductor fault identification module, used for inputting the fault characteristic quantity into a preset conductor fault identification model to perform fault identification and classification, and obtain a conductor fault identification result of the power area to be identified;

[0045] The training of the wire fault identification model includes:

[0046] Obtain historical fault feature quantities and historical conductor fault results;

[0047] Inputting the historical fault feature quantity into the wire fault identification model to be trained, so as to initialize the parameters of the wire fault identification model and obtain an initial smoothing factor;

[0048] Taking the historical fault feature as input and the predicted conductor fault result as output, the model optimization operation is repeatedly performed according to the current smoothing factor and the historical conductor fault result until a preset number of optimization times is reached to obtain a trained conductor fault identification model; the model optimization operation includes:

[0049] The conductor fault identification model to be trained is trained according to the current smoothing factor and the historical fault feature quantity to obtain the predicted conductor fault result; wherein the current smoothing factor during the first training is the initial smoothing factor;

[0050] According to the historical wire fault results and the predicted wire fault results, a loss function of a wire fault identification model is calculated; and a current smoothing factor is updated according to the loss function.

[0051] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, a method for identifying a conductor fault in a power system as described in the above embodiment is implemented.

[0052] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for identifying wire faults in a power system described in the above embodiment.

[0053] By implementing the present invention, at least the following beneficial effects are achieved:

[0054] The present invention provides a method, device, terminal equipment and storage medium for identifying conductor faults in an electric power system. The method can obtain a frequency domain signal of a power area to be identified according to a preset time interval; perform time-frequency analysis according to the frequency domain signal to extract a fault feature; input the fault feature into a preset conductor fault identification model to perform fault identification and classification, and obtain a conductor fault identification result for the power area to be identified; wherein, the training of the conductor fault identification model includes: obtaining a historical fault feature and a historical conductor fault result; inputting the historical fault feature into the conductor fault identification model to be trained so that the conductor fault identification model initializes parameters, and obtains an initial average fault feature. Smoothing factor; taking the historical fault feature quantity as input and the predicted conductor fault result as output, repeating the model optimization operation according to the current smoothing factor and the historical conductor fault result until the preset number of optimizations is reached to obtain the trained conductor fault identification model; the model optimization operation includes: training the conductor fault identification model to be trained according to the current smoothing factor and the historical fault feature quantity to obtain the predicted conductor fault result; wherein the current smoothing factor during the first training is the initial smoothing factor; calculating the loss function of the conductor fault identification model according to the historical conductor fault result and the predicted conductor fault result; and updating the current smoothing factor according to the loss function. The fault feature quantity is extracted through time-frequency analysis, and the preset conductor fault identification model is used for fault identification and classification. It is not necessary to manually observe and analyze the monitoring image of the power system at all times, so as to realize automatic identification of conductor faults and reduce dependence on manual work; at the same time, the current smoothing factor is updated to adapt to different fault conditions and environmental changes, so that the performance of the conductor fault identification model is continuously improved, thereby improving the accuracy of fault identification, being able to more accurately determine the fault location and fault type, and improving the accuracy of transmission line status monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of a method for identifying a conductor fault in an electric power system provided by an embodiment of the present invention;

[0056] Figure 2 It is a schematic diagram of the GRNN model structure provided by an embodiment of the present invention;

[0057] Figure 3 It is another flow chart of a method for identifying a conductor fault in an electric power system provided by one embodiment of the present invention;

[0058] Figure 4 is an operation flow chart of a CQPSO algorithm provided by an embodiment of the present invention;

[0059] Figure 5 It is a structural schematic diagram of a conductor fault identification device for a power system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] See also Figure 1 , is a flow chart of a method for identifying a conductor fault in a power system provided by an embodiment of the present invention, comprising:

[0062] S1. Acquire the frequency domain signal of the power area to be identified according to a preset time interval;

[0063] S2. Perform time-frequency analysis according to the frequency domain signal to extract fault feature quantities;

[0064] S3, inputting the fault feature quantity into a preset wire fault identification model to perform fault identification and classification, and obtaining a wire fault identification result of the power area to be identified;

[0065] The training of the wire fault identification model includes:

[0066] Obtain historical fault feature quantities and historical conductor fault results;

[0067] Inputting the historical fault feature quantity into the wire fault identification model to be trained, so as to initialize the parameters of the wire fault identification model and obtain an initial smoothing factor;

[0068] Taking the historical fault feature as input and the predicted conductor fault result as output, the model optimization operation is repeatedly performed according to the current smoothing factor and the historical conductor fault result until a preset number of optimization times is reached to obtain a trained conductor fault identification model; the model optimization operation includes:

[0069] The conductor fault identification model to be trained is trained according to the current smoothing factor and the historical fault feature quantity to obtain the predicted conductor fault result; wherein the current smoothing factor during the first training is the initial smoothing factor;

[0070] According to the historical wire fault results and the predicted wire fault results, a loss function of a wire fault identification model is calculated; and a current smoothing factor is updated according to the loss function.

[0071] Specifically, in order to improve the classification accuracy of the conductor identification model, it is necessary to extract the fault feature quantity of the time-frequency signal (including frequency domain signals of different lines and different poles). The conductor fault identification model to be trained is trained based on the generalized regression neural network (GRNN). The structure and weight of the conductor fault identification model are determined by the historical fault feature quantity, and the smoothing factor needs to be continuously adjusted to achieve the change of the model prediction effect. The conductor fault identification model constructed according to the smoothing factor obtains the model's predicted conductor fault results and historical conductor fault results, calculates the model's loss function, and then continuously adjusts the smoothing factor according to the loss function until the preset number of iterations is reached to obtain the trained conductor fault identification model.

[0072] In a preferred embodiment of the present invention, the frequency domain signal of the power area to be identified is first obtained according to a preset time interval; then, a time-frequency analysis is performed according to the frequency domain signal to extract the fault feature; then, the fault feature is input into a preset wire fault identification model for fault identification and classification to obtain the wire fault identification result of the power area to be identified. Wherein, the training of the wire fault identification model includes: obtaining historical fault feature and historical wire fault result; inputting the historical fault feature into the wire fault identification model to be trained so as to initialize the parameters of the wire fault identification model and obtain the initial smoothing factor; taking the historical fault feature as input and the predicted wire fault result as output, repeatedly performing the model optimization operation according to the current smoothing factor and the historical wire fault result until the preset optimization times are reached to obtain the trained wire fault identification model; the model optimization operation includes: training the wire fault identification model to be trained according to the current smoothing factor and the historical fault feature to obtain the predicted wire fault result; wherein, the current smoothing factor at the first training is the initial smoothing factor; calculating the loss function of the wire fault identification model according to the historical wire fault result and the predicted wire fault result; and updating the current smoothing factor according to the loss function.

[0073] like Figure 2 As shown, the generalized regression neural network (GRNN) takes RSSI = [RSSI1, RSSI2, ..., RSSIn]T as input and y1, y2 as output, that is, the fault feature quantity is RSSIi, i = 1, 2, ..., n. y1, y2 are the fault location identification results and fault type classification results.

[0074] Preferably, the fault characteristic quantities include: voltage, current, temperature and resistance;

[0075] Performing time-frequency analysis on the frequency domain signal to extract fault feature quantities includes:

[0076] Performing Fourier transform on the frequency domain signal to obtain a frequency domain representation after Fourier transform;

[0077] Performing a generalized S transform according to the frequency domain representation, the preset time interval, and the preset number of sampling points to obtain a transient energy sum of the frequency domain signal;

[0078] Feature extraction is performed based on the transient energy and to obtain a fault feature quantity.

[0079] In a preferred embodiment of the present invention, a generalized S transform method is used to extract fault feature quantities. First, a Fourier transform is performed on the frequency domain signal to obtain a frequency domain representation after the Fourier transform. Wherein, N is the preset number of sampling points, T is the preset time interval, α and β are adjustable factors, and in this embodiment, α=1 and β=2. Then, a generalized S transform is performed according to the frequency domain representation, the preset time interval and the preset number of sampling points to obtain the transient energy sum of the frequency domain signal:

[0080] Among them, S[p, q] is a complex time-frequency matrix, the row vector is the time domain feature at a certain frequency (indicating the time coordinate index in the time-frequency domain), and the column vector is the amplitude-frequency feature at a certain moment (indicating the frequency coordinate index in the time-frequency domain); abs(S[p, q]) is the absolute value of the S[p, q] matrix element. Finally, feature extraction is performed based on the transient energy and to obtain the fault feature quantity, that is, the parameters of voltage, current, temperature, and resistance are extracted. In this embodiment, the preset number of sampling points is 5000, the preset time interval is 0.4s, and the fault voltage, fault current, fault resistance, and fault temperature data are 5000×0.4=2000 dimensional data, thus constituting 2000×4 dimensional feature data as the input of the preset wire fault identification model.

[0081] Preferably, before inputting the fault characteristic quantity into a preset wire fault identification model, the method further includes:

[0082] According to each fault characteristic quantity, determining the maximum value and the minimum value of each fault characteristic quantity;

[0083] Data normalization is performed according to each fault characteristic quantity, the maximum value and the minimum value of each fault characteristic quantity to obtain the final fault characteristic quantity.

[0084] In a preferred embodiment of the present invention, considering the large difference in the value range between the differences in the fault feature quantities and the training data of the model, it is necessary to maximize the expression ability of the fault feature quantities and improve the accuracy of model recognition and classification. Before the fault feature quantities are input into the preset wire fault recognition model, the maximum value and the minimum value of each fault feature quantity are determined according to each fault feature quantity, and then data normalization processing is performed according to each fault feature quantity, the maximum value and the minimum value of each fault feature quantity to obtain the final fault feature quantity, that is, the normalization processing of the fault feature quantity is realized, and the formula is as follows:

[0085] in, is the final fault feature after normalization; E i It is the fault characteristic quantity before inputting the preset conductor fault identification model; min and max are the minimum and maximum value functions.

[0086] Specifically, the historical fault feature quantity is input into the wire fault identification model to be trained to initialize the model parameters, and the current smoothing factor is obtained, including:

[0087] Inputting the historical fault feature into the wire fault identification model to be trained to initialize the model and obtain the current particle position and the predicted particle position;

[0088] According to the current particle position and the predicted particle position, the current particle fitness value is calculated;

[0089] Compare the current particle fitness value to determine the optimal position of the current particle;

[0090] Compare according to the current particle fitness value until the preset number of iterations is reached to obtain the current smoothing factor;

[0091] According to the current optimal position of the particle, it is converted into a chaotic sequence to obtain a chaotic optimization result; and according to the chaotic optimization result, the current particle position is randomly replaced to update the current particle position.

[0092] In a preferred embodiment of the present invention, the CQPSO algorithm is used to optimize the parameters of the wire fault identification model to be trained, and the smoothing factor under the minimum fitness function under each iterative training is found as the current smoothing factor. The historical fault feature quantity is input into the wire fault identification model to be trained to initialize the model, obtain the current particle position and the predicted particle position, and then calculate the current particle fitness value based on the current particle position and the predicted particle position. The fitness function for calculating the particle fitness value is as follows:

[0093]

[0094] in, is the particle fitness value of the ith particle; (x′ i -x i ) is the predicted position coordinate of the ith particle; (x i ,y i ) is the current position coordinate of the ith particle. After calculating the current particle fitness value of each particle in all populations according to the fitness function, the current particle fitness value is taken as the optimal value of the corresponding particle, and then the current particle fitness value is compared to obtain the global optimal value, and then the optimal position of the current particle is determined. The average optimal position of each particle is calculated by the following formula:

[0095]

[0096] in, is the average optimal position of each particle in the j dimension, M is the size of the population, is a local attractor, and the calculation formula is as follows: is the individual optimal position of the particle, is the global optimal position of the particle, is a random number uniformly distributed between 0 and 1.

[0097] According to the current particle fitness value, the quantum particle swarm algorithm is used to perform a global search to find the global optimal value, where the local optimal value is updated by the following formula:

[0098] P i is the optimal position of the ith individual particle; f(X i (t+1)) is the current particle fitness value. Then the particle global optimal value is updated according to the local optimal value: g i (t+1)=min{P1(t+1),P2(t+1),…,P N (t+1)}. The local optimal value and the global optimal value are compared with the previous value respectively. If the current position is better, the current position will replace the previous position as the optimal value, thereby updating the global optimal value.

[0099] Then, a small perturbation is added to the global optimal value corresponding to the current optimal position of the particle, and the optimization is performed twice until the preset number of iterations is reached to obtain the current smoothing factor. The secondary optimization adopts the classical chaotic system mapping model, as shown below:

[0100] Among them, l is the number of chaotic searches; d is the sequence number of the variable, c and μ are control parameters, Represents the position vector of the d-th particle in the l-th search.

[0101] According to the current optimal position of the particle, the chaotic sequence is converted to obtain the chaotic optimization result; and according to the chaotic optimization result, the current particle position is randomly replaced to update the current particle position. The optimal particle corresponding to the optimal position of the current particle is subjected to chaotic optimization, and the optimal position of the current particle is converted to a chaotic sequence, and then according to the chaotic optimization result, the current particle position is randomly replaced to update the current particle position. Then, according to the updated current particle position, the fitness value of the current particle is recalculated until the preset number of iterations is reached to obtain the current smoothing factor.

[0102] Specifically, it also includes:

[0103] Randomly replace a current particle position according to the chaotic optimization result to obtain the current particle position after replacement;

[0104] According to the current particle fitness value and the preset constraint factor, the position of the current particle after replacement is updated based on the constraint factor;

[0105] The constraint factors include: a first Lagrange multiplier for expressing an inequality constraint, a second Lagrange multiplier for expressing an equality constraint, and a penalty factor.

[0106] In a preferred embodiment of the present invention, the position of the particle is updated and constrained by applying the Schrödinger wave equation. First, the current particle position is randomly replaced according to the chaotic optimization result to obtain the current particle position after replacement; then, a preset constraint factor is added to the current particle fitness value to obtain an augmented Lagrange function expression (augmented Lagrange function expression). For the current particle position after replacement, there are both equality constraints and inequality constraints. The augmented Lagrange function is defined as:

[0107]

[0108] Among them, f(x) is the fitness value of the original objective function, λ j is the Lagrange multiplier corresponding to the inequality constraint, that is, the first Lagrange multiplier used to represent the inequality constraint; μ j The Lagrange multiplier corresponding to the equality constraint, i.e., the second Lagrange multiplier used to represent the equality constraint; r j is the penalty factor. The difference between augmented Lagrange and Lagrange function is that the penalty term is added, and the difference between augmented Lagrange and penalty function is that the multiplier term is added. When the Lagrange multiplier method is used to deal with constrained optimization problems, it is also necessary to pay attention to the update strategy of the multiplier. If the multiplier is too small, the algorithm will find it difficult to find a suitable optimal solution; if the multiplier is too large, the algorithm will easily fall into the local optimum and make the calculation more difficult.

[0109] In another embodiment of the present invention, Figure 4 As shown in the figure, the historical fault feature quantity is input into the wire fault identification model to be trained, and the population particle position is initialized; then the particle fitness is calculated to obtain the individual optimal position and the global optimal position in the particle swarm; then the particle position is calculated according to the CQPSO algorithm, and the fitness value is recalculated, and the individual optimal position and the global optimal position are re-updated; the updated particle position is converted into a chaotic sequence, and the particle position is recalculated; until the preset number of iterations or the preset number of optimization searches is met, the final smoothing factor is obtained.

[0110] In schematic form, based on the historical fault feature quantity and the current smoothing factor, the predicted conductor fault result of the conductor fault identification model to be trained is obtained, including:

[0111] According to the current smoothing factor, the conductor fault identification model is reconstructed;

[0112] The historical fault feature quantities are input into the reconstructed conductor fault identification model for feature learning, so that the reconstructed conductor fault identification model outputs the predicted conductor fault result.

[0113] In a preferred embodiment of the present invention, the current smoothing factor σ is used to reconstruct a conductor fault identification model, and historical fault feature quantities are input into the reconstructed conductor fault identification model for feature learning, so that the reconstructed conductor fault identification model outputs a predicted conductor fault result.

[0114] In another preferred embodiment of the present invention, the reconstructed wire fault identification model includes a Softmax classifier, and the classifier can implement fault identification and classification through feature deep learning.

[0115] Specifically, the wire fault identification result includes: a fault location identification result and a fault type classification result;

[0116] The fault location identification result includes: out-of-area fault, bus fault, line fault and non-fault location;

[0117] The fault type classification results include: abnormal resistance value, abnormal current and voltage value, abnormal insulation resistance, abnormal temperature and no fault type;

[0118] After obtaining the wire fault identification result of the power area to be identified, it also includes:

[0119] Determine the fault location and fault type of the power area to be identified based on the conductor fault identification result of the power area to be identified;

[0120] Select the corresponding fault protection measures according to the determined fault location and fault type.

[0121] In a preferred embodiment of the present invention, the fault location and fault type of the power area to be identified can be identified, that is, the conductor fault identification result includes abnormal resistance value in out-of-zone faults, abnormal current and voltage values ​​in out-of-zone faults, abnormal insulation resistance in out-of-zone faults, and abnormal temperature in out-of-zone faults. By analogy, the fault location can include any one of the fault location identification results and their combination. On the basis of the fault location identification results, the specific fault type classification results are further subdivided. Among them, if the fault location identification result is no fault location, the conductor fault identification result is no fault; if the fault type classification result is no fault type, the conductor fault identification result is no fault. Figure 3 As shown, if it is a bus fault, the bus protection action is started; if it is a line fault, the line protection action is started.

[0122] By implementing this embodiment, a frequency domain signal of a power area to be identified is obtained according to a preset time interval; a time-frequency analysis is performed according to the frequency domain signal to extract a fault feature; the fault feature is input into a preset wire fault identification model to identify and classify the fault, and a wire fault identification result of the power area to be identified is obtained; wherein the training of the wire fault identification model includes: obtaining a historical fault feature and a historical wire fault result; inputting the historical fault feature into the wire fault identification model to be trained so as to initialize the parameters of the wire fault identification model and obtain an initial smoothing factor; taking the historical fault feature as input and the predicted wire fault result as output, repeatedly performing a model optimization operation according to the current smoothing factor and the historical wire fault result until a preset number of optimization times is reached to obtain a trained wire fault identification model; the model optimization operation includes: training the wire fault identification model to be trained according to the current smoothing factor and the historical fault feature to obtain a predicted wire fault result; wherein the current smoothing factor at the first training is the initial smoothing factor; calculating the loss function of the wire fault identification model according to the historical wire fault result and the predicted wire fault result; and updating the current smoothing factor according to the loss function. Fault feature quantities are extracted through time-frequency analysis, and fault identification and classification are performed using a preset conductor fault identification model. This eliminates the need for manual observation and analysis of the monitoring images of the power system, and enables automatic identification of conductor faults, reducing reliance on manual labor. At the same time, the current smoothing factor is updated to adapt to different fault conditions and environmental changes, so that the performance of the conductor fault identification model is continuously improved, thereby improving the accuracy of fault identification, being able to more accurately determine the fault location and fault type, and improving the accuracy of transmission line status monitoring.

[0123] See also Figure 5 , is a schematic diagram of a structure of a conductor fault identification device for a power system provided by an embodiment of the present invention, comprising:

[0124] A signal acquisition module, used to acquire a frequency domain signal of a power area to be identified according to a preset time interval;

[0125] A fault feature extraction module is used to perform time-frequency analysis based on the frequency domain signal to extract fault feature quantities;

[0126] A conductor fault identification module, used for inputting the fault characteristic quantity into a preset conductor fault identification model to perform fault identification and classification, and obtain a conductor fault identification result of the power area to be identified;

[0127] The training of the wire fault identification model includes:

[0128] Obtain historical fault feature quantities and historical conductor fault results;

[0129] Inputting the historical fault feature quantity into the wire fault identification model to be trained, so as to initialize the parameters of the wire fault identification model and obtain an initial smoothing factor;

[0130] Taking the historical fault feature as input and the predicted conductor fault result as output, the model optimization operation is repeatedly performed according to the current smoothing factor and the historical conductor fault result until a preset number of optimization times is reached to obtain a trained conductor fault identification model; the model optimization operation includes:

[0131] The conductor fault identification model to be trained is trained according to the current smoothing factor and the historical fault feature quantity to obtain the predicted conductor fault result; wherein the current smoothing factor during the first training is the initial smoothing factor;

[0132] According to the historical wire fault results and the predicted wire fault results, a loss function of a wire fault identification model is calculated; and a current smoothing factor is updated according to the loss function.

[0133] The present invention provides a wire fault identification device for an electric power system. According to a signal acquisition module, a frequency domain signal of a power area to be identified is acquired according to a preset time interval; in a fault feature extraction module, a time-frequency analysis is performed according to the frequency domain signal to extract a fault feature quantity; through a wire fault identification module, the fault feature quantity is input into a preset wire fault identification model to perform fault identification and classification, and a wire fault identification result of the power area to be identified is obtained; wherein the training of the wire fault identification model includes: acquiring a historical fault feature quantity and a historical wire fault result; inputting the historical fault feature quantity into the wire fault identification model to be trained, so that the wire fault identification model initializes parameters, The initial smoothing factor is obtained; the historical fault feature quantity is used as input, the predicted conductor fault result is used as output, and the model optimization operation is repeatedly performed according to the current smoothing factor and the historical conductor fault result until the preset optimization times are reached to obtain the trained conductor fault identification model; the model optimization operation includes: training the conductor fault identification model to be trained according to the current smoothing factor and the historical fault feature quantity to obtain the predicted conductor fault result; wherein, the current smoothing factor during the first training is the initial smoothing factor; according to the historical conductor fault result and the predicted conductor fault result, the loss function of the conductor fault identification model is calculated; and the current smoothing factor is updated according to the loss function. The fault feature quantity is extracted through time-frequency analysis, and the preset conductor fault identification model is used for fault identification and classification. It is not necessary to manually observe and analyze the monitoring image of the power system at all times, so as to realize automatic identification of conductor faults and reduce dependence on manual work; at the same time, the current smoothing factor is updated to adapt to different fault conditions and environmental changes, so that the performance of the conductor fault identification model is continuously improved, thereby improving the accuracy of fault identification, being able to more accurately determine the fault location and fault type, and improving the accuracy of transmission line status monitoring.

[0134] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0135] Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0136] Another embodiment of the present invention further provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements a wire fault identification method of a power system as described in the above embodiment when executing the computer program. The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory.

[0137] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0138] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.

[0139] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for identifying wire faults in a power system described in the above embodiment.

[0140] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium.

[0141] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for identifying conductor faults in a power system, characterized in that: include: Acquire a frequency domain signal of the power area to be identified according to a preset time interval; Performing time-frequency analysis on the frequency domain signal to extract fault feature quantities; Inputting the fault characteristic quantity into a preset wire fault identification model to perform fault identification and classification, and obtaining a wire fault identification result of the power area to be identified; The training of the wire fault identification model includes: Obtain historical fault feature quantities and historical conductor fault results; Inputting the historical fault feature quantity into the wire fault identification model to be trained, so as to initialize the parameters of the wire fault identification model and obtain an initial smoothing factor; Taking the historical fault feature as input and the predicted conductor fault result as output, the model optimization operation is repeatedly performed according to the current smoothing factor and the historical conductor fault result until a preset number of optimization times is reached to obtain a trained conductor fault identification model; the model optimization operation includes: The conductor fault identification model to be trained is trained according to the current smoothing factor and the historical fault feature quantity to obtain the predicted conductor fault result; wherein the current smoothing factor during the first training is the initial smoothing factor; According to the historical wire fault results and the predicted wire fault results, a loss function of a wire fault identification model is calculated; and a current smoothing factor is updated according to the loss function.

2. A method for identifying conductor faults in an electric power system according to claim 1, characterized in that: The fault characteristic quantities include: voltage, current, temperature and resistance; Performing time-frequency analysis on the frequency domain signal to extract fault feature quantities includes: Performing Fourier transform on the frequency domain signal to obtain a frequency domain representation after Fourier transform; Performing a generalized S transform according to the frequency domain representation, the preset time interval, and the preset number of sampling points to obtain a transient energy sum of the frequency domain signal; Feature extraction is performed based on the transient energy and to obtain a fault feature quantity.

3. A method for identifying conductor faults in a power system according to claim 1, characterized in that: Before inputting the fault characteristic quantity into the preset wire fault identification model, the method further includes: According to each fault characteristic quantity, determining the maximum value and the minimum value of each fault characteristic quantity; Data normalization is performed according to each fault characteristic quantity, the maximum value and the minimum value of each fault characteristic quantity to obtain the final fault characteristic quantity.

4. A method for identifying conductor faults in an electric power system according to claim 1, characterized in that: The historical fault feature quantity is input into the conductor fault identification model to be trained to initialize the model parameters and obtain the current smoothing factor, including: Inputting the historical fault feature into the wire fault identification model to be trained to initialize the model and obtain the current particle position and the predicted particle position; According to the current particle position and the predicted particle position, the current particle fitness value is calculated; Compare the current particle fitness value to determine the optimal position of the current particle; Compare according to the current particle fitness value until the preset number of iterations is reached to obtain the current smoothing factor; According to the current optimal position of the particle, it is converted into a chaotic sequence to obtain a chaotic optimization result; and according to the chaotic optimization result, the current particle position is randomly replaced to update the current particle position.

5. A method for identifying conductor faults in an electric power system according to claim 4, characterized in that: Also includes: Randomly replace a current particle position according to the chaotic optimization result to obtain the current particle position after replacement; According to the current particle fitness value and the preset constraint factor, the position of the current particle after replacement is updated based on the constraint factor; The constraint factors include: a first Lagrange multiplier for expressing an inequality constraint, a second Lagrange multiplier for expressing an equality constraint, and a penalty factor.

6. A method for identifying conductor faults in a power system according to claim 1, characterized in that: According to the historical fault feature quantity and the current smoothing factor, the predicted conductor fault result of the conductor fault identification model to be trained is obtained, including: According to the current smoothing factor, the conductor fault identification model is reconstructed; The historical fault feature quantities are input into the reconstructed conductor fault identification model for feature learning, so that the reconstructed conductor fault identification model outputs the predicted conductor fault result.

7. A method for identifying conductor faults in a power system according to claim 1, characterized in that: The wire fault identification result includes: a fault location identification result and a fault type classification result; The fault location identification result includes: out-of-area fault, bus fault, line fault and non-fault location; The fault type classification results include: abnormal resistance value, abnormal current and voltage value, abnormal insulation resistance, abnormal temperature and no fault type; After obtaining the wire fault identification result of the power area to be identified, it also includes: Determine the fault location and fault type of the power area to be identified based on the conductor fault identification result of the power area to be identified; Select the corresponding fault protection measures according to the determined fault location and fault type.

8. A conductor fault identification device for an electric power system, characterized in that: include: A signal acquisition module, used to acquire a frequency domain signal of a power area to be identified according to a preset time interval; A fault feature extraction module is used to perform time-frequency analysis based on the frequency domain signal to extract fault feature quantities; A conductor fault identification module, used for inputting the fault characteristic quantity into a preset conductor fault identification model to perform fault identification and classification, and obtain a conductor fault identification result of the power area to be identified; The training of the wire fault identification model includes: Obtain historical fault feature quantities and historical conductor fault results; Inputting the historical fault feature quantity into the wire fault identification model to be trained, so as to initialize the parameters of the wire fault identification model and obtain an initial smoothing factor; Taking the historical fault feature as input and the predicted conductor fault result as output, the model optimization operation is repeatedly performed according to the current smoothing factor and the historical conductor fault result until a preset number of optimization times is reached to obtain a trained conductor fault identification model; the model optimization operation includes: The conductor fault identification model to be trained is trained according to the current smoothing factor and the historical fault feature quantity to obtain the predicted conductor fault result; wherein the current smoothing factor during the first training is the initial smoothing factor; According to the historical wire fault results and the predicted wire fault results, a loss function of a wire fault identification model is calculated; and a current smoothing factor is updated according to the loss function.

9. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a method for identifying a conductor fault in an electric power system as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for identifying conductor faults in a power system according to any one of claims 1 to 7.

Images