Model-data driven power transmission line forest fire fault detection method and system

Abstract

The invention discloses a model-data driven power transmission line forest fire fault detection method and system, and relates to the technical field of power transmission line forest fire fault detection, and the method comprises the steps: obtaining voltage and current recording data at two ends, achieving the signal synchronization, and constructing an equivalent mathematical model suitable for different line lengths; a fault branch parameter identification model is established in combination with the model and recording data, the arc length and the transition resistance are acquired by adopting a feature extraction method, and the identification precision is optimized through voltage inversion and joint loss; and finally, evaluating the insulation strength based on an identification result, and judging the occurrence and development trend of a forest fire fault in combination with a resistance change trend and environmental factors. Through the combination of physical constraint and data driving, the interpretability and robustness of the model are enhanced, the noise interference and sampling frequency requirements are reduced, and accurate and sensitive fault detection can be realized under the condition of limited samples.

Description

Technical Field

[0001] The present invention relates to the technical field of transmission line wildfire fault detection, and in particular to a model-data driven transmission line wildfire fault detection method and system. Background Art

[0002] Transmission lines are a core component of the power system, and their safe and stable operation is directly related to the reliability of power supply and socioeconomic development. However, transmission lines often traverse complex environments such as forests and mountainous areas, making them extremely vulnerable to wildfires. Wildfires not only cause transmission line faults such as short circuits and tripping, but can also trigger large-scale power outages, even resulting in equipment damage and casualties. Therefore, the detection and early warning of transmission line wildfire faults are of great significance and are key to ensuring the safe operation of power systems.

[0003] Traditional methods for detecting wildfires along power transmission lines rely primarily on satellite remote sensing, ground-based sensor networks, and manual inspections. While these methods are somewhat effective in monitoring wildfires, they suffer from significant limitations in real-time performance, accuracy, and coverage. Particularly in complex terrain and adverse weather conditions, these traditional methods struggle to meet the demands for efficient and accurate detection.

[0004] In recent years, with the rapid development of artificial intelligence technology, wildfire detection methods based on neural networks have gradually become a research hotspot due to their powerful data processing and nonlinear modeling capabilities. However, existing neural network-based methods mainly rely on data-driven methods and have the following limitations:

[0005] High dependence on data quality and quantity: Data-driven methods require a large amount of high-quality training data. However, in practical applications, wildfire fault samples are scarce and unevenly distributed, resulting in insufficient model training and affecting detection accuracy.

[0006] Lack of understanding of physical mechanisms: Existing methods mainly rely on data-driven methods and fail to fully integrate the physical mechanisms of wildfire failures, resulting in false positives or missed positives in complex environments or unknown failure scenarios.

[0007] Insufficient explainability: The decision-making process of data-driven methods lacks transparency, making it difficult to provide in-depth analysis of the causes of failures, limiting their reliability and verifiability in practical applications. Summary of the Invention

[0008] In view of the above-mentioned problems, the present invention is proposed.

[0009] Therefore, the technical problem solved by the present invention is that the existing wildfire fault detection methods have the problems of strong data dependence, insufficient physical understanding, and low interpretability, making it difficult to solve the technical problem of how to achieve high robustness and accuracy in complex scenarios.

[0010] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0011] In the first aspect, the present invention provides a model-data driven method for detecting wildfire faults in power transmission lines, comprising obtaining three-phase voltage and current recording data at both ends of the transmission line and extracting line parameters. In the obtained data, the time consistency of multi-source recording signals is achieved through signal synchronization technology, an equivalent mathematical model of the transmission line wildfire fault is constructed, and a fault branch parameter identification model is constructed based on the equivalent mathematical model and the recording signal. In the constructed identification model, key parameters are extracted from the fault signal through first feature extraction, the identified arc length and transition resistance are substituted into the fault model for voltage inversion estimation, and a joint loss function including physical residuals and observation errors is constructed. The insulation strength is calculated based on the identification results and compared with the withstand voltage threshold corrected for the flame environment to determine whether a wildfire fault has occurred, and the wildfire development trend is evaluated in combination with the resistance change trend.

[0012] As a preferred solution of the model-data-driven transmission line wildfire fault detection method described in the present invention, the signal synchronization technology includes reconstructing the time base of the voltage signal and the current signal respectively, reconstructing the time deviation by the voltage zero crossing point and jointly determining the time offset with the cross-correlation analysis, and using quadratic interpolation to improve the accuracy of signal alignment in the recorded data of the measuring point with large offset.

[0013] As a preferred solution of the model-data driven transmission line wildfire fault detection method described in the present invention, the equivalent mathematical model includes a first part, a short line lumped parameter model composed of series resistance, inductance and parallel capacitance, and a second part, a long line distributed parameter model constructed based on telegraph equations.

[0014] This design improves the adaptability and accuracy of equivalent mathematical models by dividing the modeling approach based on the actual length of the transmission line. Short lines use a simplified RLC lumped model, which is computationally efficient and suitable for short-distance transmission scenarios; long lines use a distributed parameter model, which accurately describes the fluctuations and delay effects of long-distance transmission. This differentiation scheme ensures that the model accurately reflects the dynamic electrical changes caused by wildfire faults under various operating conditions and is a key step in building a physically constrained foundation model.

[0015] As a preferred solution of the model-data-driven transmission line wildfire fault detection method described in the present invention, the fault branch parameter identification model includes constructing a multi-channel input structure including continuous wavelet transform, combining wavelet coefficients under different scale parameters to form a multi-dimensional tensor input, and after the convolutional neural network extracts the high-order features of each frequency band, a multi-task learning structure is used to output the dynamic arc length and transition resistance respectively to achieve parallel feature mapping.

[0016] This solution combines the time-frequency resolution of wavelet analysis with the deep feature extraction capabilities of convolutional neural networks, effectively capturing multi-scale transient features in voltage and current waveforms. By simultaneously outputting multiple target parameters through a multi-task learning architecture, it achieves parallel optimization of the model structure and improves computational efficiency. This approach enhances the model's ability to characterize complex wildfire-induced signal morphologies and facilitates the accurate extraction of dynamic arc characteristics in non-standardized fault scenarios.

[0017] As a preferred embodiment of the model-data driven transmission line wildfire fault detection method of the present invention, the step of substituting the identified arc length and transition resistance into the fault model for voltage inversion estimation includes comparing the fault voltage inversion equation constructed in the model with the recorded data to generate a physically constrained residual as a first loss function;

[0018] A second loss function is generated based on the mean deviation between the observed samples and the prior samples, and the neural network parameters are iteratively updated through a dual-loss coupled optimization method.

[0019] By comparing the inversion results of the equivalent mathematical model with the actual recorded data, a physical residual is generated, which serves as a supervisory signal for the neural network, guiding it to follow the physical laws of the electrical system. Furthermore, a data compensation term is constructed by statistically analyzing the mean deviation between the observed samples and the predicted results to address potential deviations from purely physical modeling. The dual-loss function collaborative optimization mechanism improves the model's interpretability and convergence stability, making the recognition results more realistic and reliable in engineering deployment.

[0020] As a preferred solution of the model-data driven transmission line wildfire fault detection method described in the present invention, the calculation of insulation strength based on the identification results includes establishing a correction function of flame temperature and conductivity based on the air thermal breakdown theory, introducing ambient humidity and arc oscillation frequency as correction factors, dynamically correcting the breakdown voltage threshold in a multi-parameter coupling manner, and performing fault judgment based on the instantaneous withstand voltage model constructed based on the dynamic arc length.

[0021] As a preferred embodiment of the model-data-driven transmission line wildfire fault detection method described in the present invention, the comparison with the flame environment-corrected withstand voltage threshold includes constructing a first-order slope curve based on the derivative change of the dynamic transition resistance in multiple time periods, and introducing meteorological data such as wind speed, temperature and humidity as external variables to establish a resistance-environmental factor joint regression model to infer the direction and speed of wildfire spread.

[0022] In a second aspect, the present invention provides a model-data driven transmission line wildfire fault detection system, comprising:

[0023] In a third aspect, the present invention provides an electronic device, comprising:

[0024] memory and processor;

[0025] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the high-temperature pipeline health online monitoring method are implemented.

[0026] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the high-temperature pipeline health online monitoring method.

[0027] Compared with the prior art, the present invention has the following beneficial effects: the model-data driven transmission line wildfire fault detection method provided by the present invention enhances the interpretability and robustness of the model through the combination of physical constraints and data-driven, and reduces noise interference and sampling frequency requirements, and can achieve accurate and sensitive fault detection under limited sample conditions. In addition, the method has strong noise resistance and scalability, is applicable to a variety of transmission line fault scenarios, and provides reliable technical support for the safe and stable operation of the power system. The distinction between long and short models ensures that the model can accurately reflect the dynamic electrical changes caused by wildfire faults under various working conditions. The fault branch parameter identification model outputs multiple target parameters simultaneously through a multi-task learning structure, realizes parallel optimization of the model structure and improves computational efficiency, and enhances the model's ability to characterize complex wildfire-induced signal morphologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 This is an overall flow chart of a model-data driven transmission line wildfire fault detection method provided by the first embodiment of the present invention.

[0030] Figure 2 A schematic structural diagram of a wildfire fault equivalent mathematical model of a model-data driven transmission line wildfire fault detection method provided in the first embodiment of the present invention.

[0031] Figure 3 A schematic diagram of a physical model-inspired neural network for a model-data driven method for detecting wildfire faults in power transmission lines provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0032] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0033] Example 1, reference Figure 1 , as one embodiment of the present invention, provides a model-data driven method for detecting wildfire faults in power transmission lines, comprising:

[0034] S1: Obtain three-phase voltage and current recording data at both ends of the transmission line and extract line parameters. In the acquired data, achieve time consistency of multi-source recording signals through signal synchronization technology, construct an equivalent mathematical model of the transmission line wildfire fault, and construct a fault branch parameter identification model based on the equivalent mathematical model and the recording signals.

[0035] Furthermore, the system acquires transmission line data, including recordings of the three-phase voltage and current signals on the secondary side at both ends of the line. Time alignment is used to ensure the synchronization of these signals, while also extracting parameters such as the line's resistance, inductance, and capacitance. Specifically, the system first calculates the time deviation Δt using the timestamp or zero-crossing point of the power frequency voltage. Then, linear interpolation is performed on the data from the earlier side. After alignment, phase consistency is verified using Fourier transforms or correlation analysis, and alignment accuracy is assessed by calculating the mean square error.

[0036] Signal synchronization technology includes reconstructing the time base of voltage and current signals separately, reconstructing the time deviation by voltage zero-crossing point and jointly determining the time offset by cross-correlation analysis. In the waveform data of measuring points with large offset, quadratic interpolation is used to improve the accuracy of signal alignment.

[0037] It should be noted that the equivalent mathematical model includes the first part, a short line lumped parameter model consisting of series resistance, inductance and parallel capacitance, and the second part, a long line distributed parameter model constructed based on the telegraph equation.

[0038] In the embodiment of the present invention, as a preferred solution, a lumped parameter model is used for short lines, and an equivalent mathematical model of a distributed parameter model based on the telegraph equation is used for long lines, specifically:

[0039] Construct an equivalent mathematical model of transmission line wildfire fault, including: transmission line equivalent model and fault branch equivalent model.

[0040] Preferably, when the transmission line is a short line with a length of less than 100 km, the transmission line equivalent model is preferably a π-type equivalent model, which treats the transmission line as equivalent to a series resistor and inductor, with capacitors connected in parallel at both ends. The resistor and inductor are used to simulate the impedance characteristics of the line, reflecting the energy loss and magnetic field effect of the line. The parallel capacitor is used to simulate the distributed capacitance characteristics of the line, reflecting the electric field effect of the line, such as Figure 2 When the transmission line is long (more than 100 km), the transmission line equivalent model preferably uses the distributed parameter equivalent model, which is based on the telegraph equation and considers the resistance, inductance, capacitance, and conductance of the transmission line as parameters distributed along the length of the line.

[0041] Preferably, the fault branch equivalent model employs a controllable arc model, which uses arc length as a key variable to describe the dynamic relationship between arc voltage, resistance, and length. The arc voltage is proportional to arc length, while arc resistance dynamically adjusts with changes in arc length and current. Using the energy balance equation, the model effectively describes the arc's energy dissipation process and, combined with the arc length change rate, comprehensively reflects the arc's dynamic characteristics.

[0042] When the transmission line is a short line, in the process of constructing the equivalent mathematical model of the transmission line wildfire fault, the fault voltage u and current i satisfy formula (1) and formula (2).

[0043]

[0044] Where x is the fault distance. r is the resistance per unit length of the transmission line, in Ω / km. l is the resistance per unit length of the transmission line, in H / km; c is the resistance per unit length of the transmission line, in F / km; u F is the voltage at the fault point. F =i+i′ is the fault branch current, and i′ is the opposite side current. is the rate of change of arc conductivity; is the steady-state conductance. V C The steady-state field strength in the arc column is approximately constant, and the empirical value V can be taken C =15V / cm. L C is the arc length, in cm. τ C is the arc time constant.

[0045] In another feasible embodiment, the equivalent mathematical model may also uniformly adopt a multi-segment distributed parameter model and perform modeling through equivalent segment discretization. Specifically:

[0046] The entire transmission line is uniformly divided into a finite number of equally spaced subsegments, regardless of length. Within each subsegment, constant distributed parameters such as resistance, inductance, capacitance, and conductance are set as the local characteristic parameters of that segment. The subsegments are coupled via boundary voltages and currents, forming a cascaded network structure for the entire line. By establishing differential voltage-current equations for each segment in the time or frequency domain, the overall approach approximates a continuous distributed parameter model. This method retains the ability to describe the distributed characteristics of long lines while offering high flexibility and numerical stability, adapting to lines of arbitrary lengths, including short, medium-long, and multi-scenario transformer-sectioned lines. Furthermore, this modeling approach facilitates coupled solution with the fault point parameters output by the neural network, enhancing the model's adaptability to complex real-world environments (such as terrain changes, arcing effects, and transient characteristics), improving fault simulation accuracy and the engineering deployability of the detection model.

[0047] It should also be noted that a preferred solution for the first feature extraction in the present invention is to perform time-frequency feature extraction and mapping based on continuous wavelet transform + convolutional neural network (CWT+CNN), specifically:

[0048] A multi-channel input structure including continuous wavelet transform is constructed, and the wavelet coefficients under different scale parameters are combined to form a multi-dimensional tensor input. After the convolutional neural network extracts the high-order features of each frequency band, a multi-task learning structure is used to output the dynamic arc length and transition resistance respectively, realizing parallel feature mapping.

[0049] The fault branch parameter identification model based on neural network consists of a continuous wavelet transform layer (CWTL), a convolutional layer (CL), a pooling layer (PL) and a fully connected layer (FCL). The continuous wavelet transform layer is used to extract the time-frequency characteristics of the fault signal and capture the multi-scale characteristics of the transient voltage and current signals. The convolution layer extracts high-order features through local perception and weight sharing. The pooling layer reduces the dimension of the features and enhances the robustness of the model. The fully connected layer maps the extracted features to parameters such as the dynamic arc length and transition resistance of the fault branch, such as Figure 3 shown.

[0050] In another feasible embodiment, the first feature extraction can be performed based on empirical mode decomposition (EMD) + machine learning regression model, specifically:

[0051] First, EMD processing is performed on the collected transmission line voltage and current recording signals, decomposing the original non-stationary signals into several intrinsic mode functions (IMFs) with local time-domain characteristics. Each IMF represents an oscillation mode in a specific frequency band. Subsequently, a series of statistical features are extracted from these IMF signals, including time-frequency statistical indicators such as frequency-domain energy distribution, instantaneous frequency, envelope amplitude, kurtosis, and entropy, to reflect the nonlinear characteristics of arc behavior in the signals. These multidimensional features constitute the feature vector input dataset of the fault signal. Combined with the dynamic arc length and transition resistance annotated from field or simulation data as supervisory labels, a support vector regression (SVR) model or a random forest (RF) regression model is trained. This method fully utilizes the adaptive decomposition capabilities of EMD and the nonlinear mapping capabilities of machine learning models to achieve accurate prediction of arc parameters, with strong noise immunity and adaptability to small sample data.

[0052] S2: In the constructed identification model, key parameters are extracted from the fault signal through the first feature extraction, the identified arc length and transition resistance are substituted into the fault model for voltage inversion estimation, and a joint loss function including physical residuals and observation errors is constructed.

[0053] Furthermore, the first feature extraction is based on continuous wavelet transform + convolutional neural network (CWT + CNN) to extract and map time-frequency features. The continuous wavelet transform layer converts the input fault voltage u and current i into time-frequency signals according to formula (1), as shown in formula (3), which is expressed as:

[0054]

[0055] Among them, a is the scale parameter, b is the translation parameter, ψ(t) is the wavelet basis function, and ψ* is its complex conjugate. Then, using the convolutional neural layer, the time-frequency signal feature Z is extracted, as shown in formula (4)

[0056] Z=f(W*Y+b) (4)

[0057] Among them, W is the convolution kernel, * represents the convolution operation, b is the bias term, and f is the activation function.

[0058] Then, the pooling layer is used to reduce the dimension of feature Z, as shown in formula (5)

[0059] P=Pooling(Z) (5)

[0060] Among them, Pooling is the maximum pooling operation.

[0061] Finally, the extracted features are mapped to the dynamic arc length and transition resistance of the fault branch through the fully connected layer, as shown in formula (6)

[0062]

[0063] Among them, W f is the weight matrix of the fully connected layer, b f is the bias term.

[0064] Substituting the identified arc length and transition resistance into the fault model for voltage inversion estimation involves comparing the fault voltage inversion equation constructed in the model with the recorded data to generate a physical constraint residual as the first loss function. Specifically:

[0065] The physical constraint model is constructed, and the equivalent mathematical model of the transmission line wildfire fault is converted into a residual format, and the residual is used as the loss of the neural network. Specifically, the identified And the measured i, i′ are substituted into (1)(2) to estimate the voltage And, solve the loss Λ1

[0066]

[0067] The second loss function is generated based on the mean deviation between the observed sample and the prior sample, and the neural network parameters are iteratively updated through a dual-loss coupled optimization method. The second loss function is specifically:

[0068] Construct a data-based compensation model and use observation data to compensate for known physical information. Specifically, the observation data is first converted into loss Λ2

[0069]

[0070] Then, the physical constraint model is compensated by weighted differencing.

[0071] It should be noted that the identification results are verified by comparing the predicted values ​​of the equivalent mathematical model with the actual observed values ​​and calibrating the model output to ensure the accuracy and reliability of the results.

[0072] The neural network parameters are dynamically adjusted based on the recognition results. The Adam optimizer combined with L2 regularization is used to update network hyperparameters to optimize the model's convergence speed and generalization performance. The Adam optimizer accelerates model training through adaptive learning rate adjustment. L2 regularization penalizes large weights to prevent model overfitting and improve its robustness in complex fault scenarios.

[0073] S3: Calculate the insulation strength based on the identification results and compare it with the voltage withstand threshold corrected for the flame environment to determine whether a wildfire fault has occurred. The wildfire development trend is then assessed based on the resistance change trend.

[0074] Furthermore, the insulation strength is calculated based on the identification results, including the air thermal breakdown theory, by establishing a correction function of flame temperature and conductivity, introducing ambient humidity and arc oscillation frequency as correction factors, dynamically correcting the breakdown voltage threshold in a multi-parameter coupling manner, and performing fault judgment based on the instantaneous withstand voltage model constructed based on the dynamic arc length.

[0075] Use the pre-built fault branch parameter identification model to identify the recorded data of voltage and current signals, and output parameters such as the dynamic arc length and dynamic transition resistance of the fault branch, including:

[0076] The collected voltage and current signals are preprocessed, including signal denoising, signal synchronization and other operations.

[0077] The time-frequency characteristics of the signal are obtained through the continuous wavelet transform layer to capture the multi-scale characteristics of the transient process.

[0078] The extracted time-frequency features are input into the convolution layer to extract high-order features. Then, the features are reduced in dimension using the pooling layer. Finally, the features are mapped into parameters such as dynamic arc length and dynamic transition resistance through the fully connected layer.

[0079] In the above embodiment, based on the arc length of the identified fault branch, the insulation strength of the transmission line is evaluated and wildfire fault detection is achieved. The specific steps are as follows:

[0080] Based on the theory of thermal breakdown in air, the withstand voltage per unit length of the gap under flame, θ, is calculated as the dielectric strength threshold. Specifically, when the flame completely envelops the conductor, it forms a high-temperature, high-conductivity region, causing significant changes in the electric field distribution and breakdown characteristics of the gap. Based on the theory of thermal breakdown in air, the breakdown voltage of the gap under flame can be expressed as:

[0081] θ=E th ·f(T,σ) (9)

[0082] Where: E th is the breakdown field strength in air at room temperature (typically around 30 kV / cm). f(T,σ) is a correction function of flame temperature and conductivity, representing the effect of the flame on the breakdown field strength. T is the flame temperature (K) and σ is the flame conductivity (S / m).

[0083] The correction function f(T,σ) of flame temperature and conductivity can be further expressed as:

[0084]

[0085] Where α is an empirical coefficient related to flame characteristics.

[0086] It should be noted that the comparison with the voltage withstand threshold after correction for the flame environment includes constructing a first-order slope curve based on the derivative changes of the dynamic transition resistance in multiple time periods, and introducing meteorological data such as wind speed, temperature and humidity as external variables to establish a resistance-environmental factor joint regression model to infer the direction and speed of wildfire spread.

[0087] Estimate the gap withstand voltage per unit length based on the identified fault branch arc length As the insulation strength, as shown in formula (11), it is compared with the preset threshold. If the insulation strength is lower than or equal to the threshold, it is determined that the transmission line fault is caused by a wildfire.

[0088]

[0089] Among them, U f is the voltage at the moment of fault. d is the air gap distance

[0090] The present invention combines the physical model of transmission line faults to construct a physical-data-driven deep learning neural network model. The model input is constructed with transient voltage and current, and parameters such as dynamic arc length and dynamic transition resistance are output. Ultimately, a transmission line wildfire fault detector is formed, which is deployed and applied to both ends of the transmission line, that is, the outgoing end of the substation, for transmission line wildfire fault detection.

[0091] Example 2 is the second embodiment of the present invention, which is different from the previous embodiment in that:

[0092] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0093] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0094] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0095] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0096] Example 3 is the third embodiment of the present invention. This embodiment provides a system for a model-data driven transmission line wildfire fault detection method, including a recording signal modeling module, a physical constraint optimization module, and a wildfire trend determination module.

[0097] The recording signal modeling module is used to obtain the three-phase voltage and current recording data at both ends of the transmission line and extract the line parameters. In the acquired data, the time consistency of multi-source recording signals is achieved through signal synchronization technology, and an equivalent mathematical model of the transmission line wildfire fault is constructed. Based on the equivalent mathematical model and the recording signal, a fault branch parameter identification model is constructed.

[0098] The physical constraint optimization module is used to extract key parameters from the fault signal through the first feature extraction in the constructed identification model, substitute the identified arc length and transition resistance into the fault model for voltage inversion estimation, and construct a joint loss function that includes physical residuals and observation errors.

[0099] The wildfire trend determination module is used to calculate the insulation strength based on the identification results and compare it with the voltage withstand threshold corrected for the flame environment to determine whether a wildfire fault has occurred, and to evaluate the wildfire development trend based on the resistance change trend.

[0100] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A model-data driven method for detecting wildfire faults in power transmission lines, characterized in that: include: Acquire three-phase voltage and current recording data at both ends of the transmission line and extract line parameters. Use signal synchronization technology to achieve temporal consistency between multi-source recording signals within the acquired data. Construct an equivalent mathematical model for transmission line wildfire faults. Based on this equivalent mathematical model and the recording signals, construct a fault branch parameter identification model. In the constructed identification model, key parameters are extracted from the fault signal through the first feature extraction. The identified arc length and transition resistance are substituted into the fault model for voltage inversion estimation, and a joint loss function including physical residual and observation error is constructed. The insulation strength is calculated based on the identification results and compared with the voltage withstand threshold corrected for the flame environment to determine whether a wildfire fault has occurred. The development trend of the wildfire is then evaluated based on the resistance change trend.

2. The model-data driven transmission line wildfire fault detection method according to claim 1, characterized in that: The signal synchronization technology includes reconstructing the time base of the voltage signal and the current signal respectively, reconstructing the time deviation by the voltage zero crossing point and jointly determining the time offset with the cross-correlation analysis, and using quadratic interpolation to improve the accuracy of signal alignment in the waveform data of the measuring point with large offset.

3. The model-data driven transmission line wildfire fault detection method according to claim 2, characterized in that: The equivalent mathematical model includes a first part, which is a short line lumped parameter model consisting of series resistance, inductance and parallel capacitance, and a second part, which is a long line distributed parameter model constructed based on telegraph equations.

4. The model-data driven transmission line wildfire fault detection method according to claim 3, characterized in that: The fault branch parameter identification model includes constructing a multi-channel input structure including continuous wavelet transform, combining wavelet coefficients under different scale parameters to form a multi-dimensional tensor input, extracting high-order features of each frequency band through a convolutional neural network, and adopting a multi-task learning structure to output dynamic arc length and transition resistance respectively, realizing parallel feature mapping.

5. The model-data driven transmission line wildfire fault detection method according to claim 4, characterized in that: Substituting the identified arc length and transition resistance into the fault model for voltage inversion estimation includes comparing the fault voltage inversion equation constructed in the model with the recorded data to generate a physical constraint residual as a first loss function; A second loss function is generated based on the mean deviation between the observed samples and the prior samples, and the neural network parameters are iteratively updated through a dual-loss coupled optimization method.

6. The model-data driven transmission line wildfire fault detection method according to claim 5, characterized in that: The calculation of insulation strength based on the identification results includes establishing a correction function of flame temperature and conductivity based on air thermal breakdown theory, introducing ambient humidity and arc oscillation frequency as correction factors, dynamically correcting the breakdown voltage threshold in a multi-parameter coupling manner, and performing fault judgment based on an instantaneous withstand voltage model constructed based on the dynamic arc length.

7. The model-data driven transmission line wildfire fault detection method according to claim 6, characterized in that: The comparison with the voltage withstand threshold after correction for the flame environment includes constructing a first-order slope curve based on the derivative changes of the dynamic transition resistance in multiple time periods, introducing meteorological data such as wind speed, temperature and humidity as external variables, and establishing a resistance-environmental factor joint regression model to infer the direction and speed of wildfire spread.

8. A system using the model-data driven transmission line wildfire fault detection method according to any one of claims 1 to 7, characterized in that: It includes waveform signal modeling module, physical constraint optimization module, and wildfire trend determination module; The recorded signal modeling module is used to obtain three-phase voltage and current recorded data at both ends of the transmission line and extract line parameters. In the obtained data, the time consistency of multi-source recorded signals is achieved through signal synchronization technology, and an equivalent mathematical model of the transmission line wildfire fault is constructed. Based on the equivalent mathematical model and the recorded signal, a fault branch parameter identification model is constructed; The physical constraint optimization module is used to extract key parameters from the fault signal through the first feature extraction in the constructed identification model, substitute the identified arc length and transition resistance into the fault model to perform voltage inversion estimation, and construct a joint loss function including physical residual and observation error; The wildfire trend determination module is used to calculate the insulation strength based on the identification result, and compare it with the withstand voltage threshold corrected by the flame environment to determine whether a wildfire fault has occurred, and evaluate the wildfire development trend in combination with the resistance change trend.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the model-data driven power transmission line wildfire fault detection method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the model-data driven transmission line wildfire fault detection method according to any one of claims 1 to 7 are implemented.

Images