A cable soft fault distribution parameter inversion method and related equipment
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SOUTH CHINA UNIV OF TECH
- Filing Date
- 2026-03-05
- Publication Date
- 2026-06-02
Smart Images
Figure CN121784461B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the fields of power line communication and cable condition monitoring and fault diagnosis, and in particular to a method and related equipment for inverting cable soft fault distribution parameters. Background Technology
[0002] Power cables are critical infrastructure for power transmission and communication systems. During long-term operation, cables are affected by factors such as electrothermal stress, mechanical stress, and environmental corrosion, causing slow changes in the insulation and conductor properties. This manifests as gradual or segmented anomalies in the distributed parameters per unit length (especially capacitance C and resistance R) in localized sections. These "soft faults," characterized by continuous parameter changes, are difficult to detect effectively in their early stages using traditional impedance-based reflection methods (such as time-domain reflectometry, TDR). However, they continuously degrade signal transmission quality and may eventually evolve into hard faults leading to power outages, posing a potential threat to system safety.
[0003] For cable fault detection, existing technologies mainly include the following types of methods:
[0004] 1) Location method based on reflection principle: This method determines the fault location by analyzing the arrival time and amplitude of the reflected signal at the port. Its basic premise is that the cable parameters are uniform and known, and the fault manifests as a significant impedance change. For soft faults with continuously changing parameters, the reflected signal is weak and mixed with complex background responses, causing this method to fail or its accuracy to drop sharply.
[0005] 2) Inversion methods based on full-wave propagation models: These methods attempt to invert internal parameters by matching the full waveform of the port. They typically require establishing an accurate cable model and using iterative optimization (such as gradient descent or global search) to adjust model parameters to fit the measurement data. However, their gradient calculations often rely on complex "adjoint state method" or "time reversal" techniques, which are mathematically cumbersome, computationally complex, and computationally expensive, limiting their engineering applications in scenarios with long cables and high sampling rates.
[0006] 3) Purely data-driven machine learning methods: Some studies attempt to use neural networks to directly learn the mapping from port waveforms to fault types or locations. These methods heavily rely on large amounts of labeled training data, and the lack of physical constraints within the model leads to poor physical interpretability of the inversion results and insufficient generalization ability when cable types and operating conditions change.
[0007] In recent years, methods combining physical laws with deep learning, such as Physical Information Neural Networks (PINN), have attracted attention. However, mainstream PINNs typically only use the residuals of the physical equations as regularization terms in the loss function, without changing the core structure of the network's forward propagation. For cable transmission line inversion problems with well-defined spatial discrete formats (such as FDTD) and port boundary conditions, designing an end-to-end framework that strictly adheres to physical laws while efficiently utilizing modern automatic differentiation tools for optimization remains a pressing technical challenge. Summary of the Invention
[0008] The main objective of this application is to propose a method, electronic device, storage medium, and program product for inverting cable soft fault distribution parameters. The aim is to achieve stable, efficient, and physically interpretable quantitative inversion of cable distribution parameters (especially capacitance and resistance) using only transient excitation and response data from cable ports, without relying on sensors deployed along the cable line or requiring a large number of fault samples, thereby accurately characterizing the spatial distribution features of soft faults.
[0009] To achieve the above objectives, one aspect of this application proposes a method for inverting cable soft fault distribution parameters, the method comprising:
[0010] Establish a one-dimensional distributed parameter transmission line model of the cable under test. The model parameters include the resistance per unit length distributed along the length of the cable. ,inductance Electrical conductivity and capacitor ;
[0011] The time-domain control equations of the transmission line model are discretized using finite difference time-domain FDTD, and a voltage-current interleaved grid is used to obtain the recursive update relationship of the discrete state variables in adjacent time layers.
[0012] Construct a recurrent neural network (RNN), and explicitly define the recursive update relationship as a single-step state update function of the RNN. Among them, the first The voltage and current fields at each time step are defined as the hidden states of the RNN. Define the port excitation input as and the distribution parameters to be inverted The state update function is embedded as a trainable parameter. The distribution parameters At least including distributed capacitance ;
[0013] The RNN is expanded in the time dimension and forward propagated to obtain the predicted voltage response at each time step port. ;
[0014] Based on the predicted voltage response Measured voltage response at the port Construct the loss function ;
[0015] The loss function is calculated by automatic differentiation. Regarding the distribution parameters The gradient is calculated and iteratively updated using backpropagation. The process continues until the loss function converges, at which point the distributed parameter inversion results corresponding to the cable soft faults are output.
[0016] In some embodiments, the voltage in the voltage-current interleaved grid is defined at discrete nodes. At this point, the current is defined between adjacent nodes. The recursive update relationship includes an alternating progression structure where voltage is updated differentially by adjacent currents and current is updated differentially by adjacent voltages.
[0017] In some embodiments, the single-step state update function Specifically:
[0018]
[0019] Wherein, the function The operational logic is determined by the obtained FDTD recursive update relationship. For RNN in the 1st The hidden state at each time step.
[0020] In some embodiments, the function The computational logic includes updating the port boundary conditions; the calculation of port boundary condition updates and the calculation of cable interior point updates are performed uniformly in the same time-progression step.
[0021] In some embodiments, the port boundary conditions include the impedance boundary at the load end, satisfying the relationship ,in This is the load impedance, used to achieve matched or unmatched loads; This is the load voltage. This is the load current.
[0022] In some embodiments, the port boundary condition includes an equivalent voltage source boundary at the source end, the equivalent voltage source comprising the source internal resistance. This makes the source node voltage update correlated with the current of the adjacent line segment at the port and the excitation input.
[0023] In some embodiments, the predicted voltage response From the hidden state of RNN The port voltage is sampled and then spliced together at all time steps to form the port prediction waveform.
[0024] In some embodiments, the loss function To address the joint loss under multiple excitation conditions and multiple port observations, joint constraints are applied to the predicted and observed responses for different excitation conditions, different ports, and time sampling points.
[0025] In some embodiments, the distribution parameters to be inverted Including distributed capacitance With distributed resistance And during the reverse propagation process, respectively and Perform iterative updates and output the joint inversion result of the two.
[0026] In some embodiments, the loss function The joint loss function is expressed as follows:
[0027]
[0028] in, This is a time-domain waveform mismatch term used to constrain propagation to time and phase characteristics; This is the amplitude attenuation feature mismatch term, used to constrain the amplitude attenuation feature of the waveform envelope; and These are the weighting coefficients.
[0029] In some embodiments, the amplitude attenuation feature mismatch term A logarithmic amplitude difference based on the port waveform envelope is constructed to enhance the constraint on attenuation differences caused by distributed resistance; wherein, the amplitude attenuation characteristic mismatch term It is used as part of the loss function in backpropagation calculation to update the distribution parameters through automatic differentiation.
[0030] In some embodiments, the distribution parameters to be inverted Logarithmic parameterization is used to make updates take place in a relatively variable space to improve gradient scaling and convergence stability.
[0031] In some embodiments, the port response is divided into several segments according to the time dimension, and the distribution parameters are updated using a segmented time backpropagation strategy to reduce the computational complexity and storage overhead under long-term sequence conditions.
[0032] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described above.
[0033] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described above.
[0034] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer program product, including a computer program that, when executed by a processor, implements the method described above.
[0035] The embodiments of this application include at least the following beneficial effects: This application achieves end-to-end inversion of cable distributed parameters under port transient measurement conditions by embedding the finite-difference time-domain modeling process of cable transient propagation into a differentially cyclic computational structure. The inversion process is directly driven by physical propagation errors, exhibiting good physical consistency and interpretability. This application does not rely on sensor deployment along the cable, nor does it require a large number of labeled samples. It can quantitatively reconstruct the distributed capacitance parameters along the cable using only port transient voltage measurement information, avoiding time reversal and explicit adjoint field gradient derivation processes, reducing algorithm complexity, improving the stability of engineering implementation, and effectively characterizing the spatial distribution characteristics of soft faults. By introducing port boundary consistency modeling and a joint inversion strategy involving multiple ports and multiple excitation conditions, this application improves the stability and robustness of distributed parameter inversion and reduces the impact of model mismatch on the inversion results. Furthermore, this application supports a piecewise time backpropagation strategy, effectively reducing the computational complexity of long-term series inversion while ensuring inversion accuracy, making the method feasible for engineering applications. Attached Figure Description
[0036] Figure 1 The overall flowchart of the cable soft fault distribution parameter inversion method provided in the embodiments of this application is shown.
[0037] Figure 2 This is a physical model and discretization diagram of a one-dimensional distributed parameter transmission line for a cable.
[0038] Figure 3 This diagram illustrates the variable definition and update relationships in a voltage-current staggered grid. Voltage is defined at whole-grid points, and current is defined at half-grid points.
[0039] Figure 4 This is a schematic diagram comparing the distributed capacitance result obtained by inverting a cable with local capacitance anomalies with the actual value in Example 1.
[0040] Figure 5 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation
[0041] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.
[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.
[0043] As a crucial component of power and communication systems, power cables are inevitably affected by various factors during long-term operation, including insulation aging, moisture absorption, water treeing, biological corrosion, and thermal and mechanical stress. These effects typically do not immediately lead to open-circuit or short-circuit faults in the cable. Instead, they manifest as slow changes in the cable's dielectric properties and distributed parameters, particularly continuous or segmented abnormal changes in distributed capacitance and resistance parameters within localized areas or specific sections. Faults characterized primarily by gradual parameter changes are often referred to as soft faults. Their early stages are difficult to detect by traditional protection devices, but they gradually alter the signal propagation speed, attenuation characteristics, and reflection characteristics within the cable, potentially evolving into severe hard faults that threaten the safe operation of the system.
[0044] To address the problem of cable fault detection and location, existing technologies widely employ methods based on the principle of electromagnetic wave reflection. This involves injecting a test signal at the cable port and analyzing the arrival time and amplitude of the reflected waveform to determine the fault location. These methods are effective in handling hard faults caused by significant impedance abrupt changes, but they rely on the fundamental assumption that cable parameters are spatially uniform and known. When the fault manifests as a slow change in distributed parameters or a continuous anomaly within a finite spatial scale, the reflected signal is often weak and heavily superimposed on the background response, making it difficult to reliably separate fault features from the port measurement signal. This limits their application in soft fault scenarios.
[0045] To overcome the limitations of single-reflection analysis in complex scenarios, existing research has introduced an analysis approach based on full-wave propagation characteristics. This approach attempts to improve fault identification capabilities by comprehensively modeling and matching the transient response at the port and utilizing multiple scattering and propagation path information. These methods typically require establishing a relatively accurate cable propagation model and iteratively adjusting the model parameters to gradually reduce the difference between the synthesized response and the actual measured signal, thereby indirectly inferring changes in the cable's distributed parameters. However, these methods often rely on explicit gradient derivation or adjoint field calculations, resulting in complex algorithms and high computational costs. This limits their engineering applications when dealing with long cables or high-time-resolution measurement data.
[0046] With the development of computing resources, neural networks and data-driven methods have been gradually introduced into the field of cable condition monitoring. Some studies have attempted to directly learn the mapping relationship between port waveforms and fault types or locations through neural networks. However, such methods usually rely on a large amount of labeled data, and there is a lack of clear constraints between the network model and the actual electromagnetic propagation mechanism of the cable. The physical consistency and interpretability of the inversion results are insufficient, making it difficult to guarantee the generalization ability under different operating conditions and cable parameters.
[0047] In recent years, a modeling approach combining physical models with neural network structures has gained increasing attention. This involves explicitly embedding physical control equations or numerical solution processes into the network structure, ensuring that the network's forward propagation process aligns with the evolution of the physical system. Within this framework, parameter updates no longer rely on purely empirical mappings but are driven by physical propagation errors, thus achieving a balance between physical consistency and data adaptability to some extent. However, existing research largely focuses on fluctuation problems in continuous media. For cable systems with distinct port boundary conditions and distributed parameter transmission line structures, a systematic solution for inverting distributed parameters in soft faults remains lacking.
[0048] In summary, how to stably and interpretably invert the changes in distributed parameters caused by cable soft faults without relying on a large number of prior samples or deploying sensors along the cable line, using only transient measurement information from the cable port, while ensuring physical consistency and computational feasibility, remains a technical problem that urgently needs to be solved in this field.
[0049] To address the problems of strong model dependence, insufficient inversion stability, and difficulty in balancing physical consistency and engineering feasibility in existing technologies for cable soft fault diagnosis, this application aims to propose a cable soft fault distribution parameter inversion method based on a combination of physically consistent finite difference time-domain modeling and a differentiable recurrent neural network. The goal is to quantitatively invert the cable distribution parameters using only transient measurement information from the cable port, without relying on a large number of prior samples or requiring the deployment of sensors along the cable, thereby achieving an effective characterization of the spatial distribution features of soft faults.
[0050] Unlike existing full-wave inversion methods that rely on electromagnetic time reversal or the idea of adjoint fields, this application reconstructs the inversion framework at the computational structure level. The voltage-current recursive update relationship and port boundary update relationship obtained by finite difference time domain discretization are explicitly equivalent to the single-step state update function of a recurrent neural network. The recursive computation graph is expanded in the time dimension to form a differentiable cyclic computation graph, so that the transient propagation process of the cable is consistent with the recurrent neural network in terms of computational structure.
[0051] In the described loop calculation structure, this application embeds the capacitance and resistance parameters distributed along the cable length direction as trainable parameters in the recurrent neural network state update function, so that the distributed parameters directly participate in the physical propagation calculation at each time step and are automatically updated as optimization variables during the back propagation process.
[0052] This application does not require the construction of time-reversed residual waveforms or explicit adjoint fields. Instead, based on the differentiable cyclic computation structure, a loss function is constructed between the port predicted response and the port observed response. The gradient of the loss function with respect to the distribution parameters is directly obtained through automatic differentiation and backpropagation mechanisms, thereby achieving stable inversion and updating of the distribution parameters.
[0053] Furthermore, this application improves the stability and robustness of distributed parameter inversion by introducing multi-excitation operating conditions and port boundary consistency modeling, and through an optional piecewise backpropagation strategy, the proposed method still has engineering application feasibility under long-term series and limited computing resources.
[0054] like Figure 1 As shown, this embodiment provides a method for inverting cable soft fault distribution parameters, including the following steps:
[0055] S1. Establish a one-dimensional distributed parameter transmission line model of the cable under test. The model parameters include the distributed parameter resistance per unit length along the cable length direction. ,inductance Electrical conductivity and capacitor Voltage and current are denoted as follows: and ;
[0056] S2. The transmission line model is discretized using finite-difference time-domain (FDTD) discretization with a voltage-current interleaved grid to obtain the discrete state variables. and And its recursive update relationship in adjacent time layers;
[0057] S3, construct a recurrent neural network (RNN) such that the single-step state update function of the RNN is explicitly defined as an FDTD recursive update relation. Specifically, the RNN is configured to update the state in the first step of the recursive process. The hidden state at each time step is defined as follows: Port excitation at the 1st The input at each time step is defined as And write the RNN state update as , in It is a deterministic update operator composed of discrete recursive update relations of FDTD. The set of distributed parameters to be inverted includes at least the distributed capacitance. Distribution parameters Trainable parameters in an RNN are embedded in the state update function and iteratively updated as optimization variables during backpropagation. Distributed resistance may also be included. ;
[0058] S4. Expand the RNN in the time dimension and perform a forward propagation to obtain the predicted voltage response at each time step port. and the port observed voltage response Constructing the loss function ;
[0059] S5, calculate the loss function with respect to its distribution parameters through backpropagation. The gradient is calculated, and the distribution parameters are iteratively updated to make the loss function converge, and the distribution parameter inversion results corresponding to the cable soft fault are output.
[0060] The solutions of the embodiments of this application will be described in detail and explained below with reference to specific application examples.
[0061] Example 1: Cable Distributed Capacitance Inversion Method Based on Port Transient Response
[0062] This embodiment takes a single cable as the object and provides a detailed description of the cable soft fault distribution parameter inversion method based on physically consistent transient modeling described in this application.
[0063] (I) Establishment of the physical model of the cable
[0064] like Figure 2 As shown, in this embodiment, the cable under test is equivalent to a one-dimensional distributed parameter transmission line model, and a spatial coordinate system is established along the cable length. Cable unit length parameters include resistance ,inductance Electrical conductivity and capacitor Their physical dimensions are Ω / m, H / m, S / m, and F / m, respectively. In this embodiment, soft faults are mainly detected through distributed capacitance. The spatial anomalies are characterized, and these anomalies can be continuous changes or segmented abrupt changes; resistance ,inductance Electrical conductivity It can be given by the cable's nominal parameters or prior measurements, or it can be used in other implementations with... They are included as parameters to be inverted.
[0065] The transient propagation process of a cable satisfies the one-dimensional transmission line time-domain governing equation. (Based on voltage...) With current Indicates the cable's location With time For the transient quantity, we have:
[0066]
[0067]
[0068] The above governing equations form the physical basis for the subsequent finite-difference time-domain discrete modeling and distributed capacitance inversion calculations in this embodiment.
[0069] See Figure 3 To achieve numerical calculation, in one embodiment, space and time are discretized: the cable is discretized into individual... Voltage node, space step size is The time step is A voltage-current interleaved grid discretization method is adopted, where voltage is defined at the nodes and current is defined between adjacent nodes, i.e.:
[0070]
[0071] The staggered grid discretization method satisfies the discrete consistency of local current continuity and voltage loop relationship, which facilitates the construction of a stable time-domain recursive update structure.
[0072] Regarding port measurement and excitation, this embodiment uses the transient voltage at the cable port as a measurable quantity and applies an excitation signal to the cable port. And collect port responses ,in For port location (e.g.) (Location). Given prior information on some distributed parameters, the distributed capacitance is analyzed by the difference between the port transient response and the model predicted response. An inversion update is performed to obtain the spatial distribution of capacitance anomalies corresponding to soft faults.
[0073] (II) Finite Difference Time Domain (FDTD) Discretization and Port Boundary Modeling
[0074] In this embodiment, based on the transmission line time-domain control equations, the finite difference time-domain method is used to discretize the voltage and current in time and space and then advance the time. The time is calculated along the spatial direction in steps of [number missing]. Discretize the cable as Each voltage node, with time steps... Discrete. Employing a voltage-current interleaved grid: voltage Defined in node With integer time layer Above, current Defined between adjacent nodes and at the half-time level, thus forming an interleaved propagation structure for voltage and current updates.
[0075] In one embodiment, the discrete update of the cable interior points can be written in the following recursive form. For interior point nodes... The node voltage is updated as follows:
[0076]
[0077] in , To update the coefficients, the distributed capacitance at the nodes is used. The time-center discretization method can be used to reflect the dielectric loss effect.
[0078] For line segment current The current is updated as follows:
[0079]
[0080] in as well as The conductor loss effect can be reflected by using a time-center discretization method, which is jointly determined. The staggered calculation for each time step can be completed using the above two equations, yielding the transient evolution of voltage and current across the entire cable.
[0081] At the port, this embodiment integrates the excitation and boundary conditions into a unified time-progression framework. For the source port, a Thevenin equivalent voltage source model incorporating the source internal resistance Rg is used to correlate the port voltage update with the applied excitation and adjacent port currents. In one embodiment, the source node voltage can be updated in the following form (taking the left port as an example):
[0082]
[0083] in For the current in the adjacent line segment at the source end, the coefficient is... This reflects the output characteristics of the actual excitation device.
[0084] For non-excited ports, set load boundary conditions; the load can be supplied by impedance. Characterized by the port relationships satisfying:
[0085]
[0086] And solve in the discrete domain in a manner consistent with the above interior point updates, to achieve unified modeling of matched loads, unmatched loads, or other load forms.
[0087] Through the above FDTD discrete update and port boundary modeling, the transient voltage response of the port can be calculated under given distributed parameters, providing a forward calculation basis for the subsequent construction of the loss function and the inversion update of the distributed capacitance parameters.
[0088] (III) Inversion in conjunction with deep learning frameworks
[0089] After completing the FDTD discretization and port boundary modeling described in the second point of Embodiment 1, this embodiment further constructs the time-progression process into a differentiable cyclic computation structure to achieve end-to-end inversion of the distributed capacitance parameters.
[0090] In one implementation, the "voltage update + current update + port boundary update" at each discrete time step is considered as a recurrent unit (RNN unit). This recurrent unit is expanded along the time dimension to form a directed acyclic computation graph consisting of multiple units connected in series, where the... Each unit receives the state from the previous time step and outputs the state for the next time step. Specifically, the voltage field and current field are taken as the system state, denoted as:
[0091]
[0092] The cyclic update of a single time step can then be abstractly represented as:
[0093]
[0094] in Indicates the port excitation input (e.g.) ), This represents the set of distributed parameters to be inverted. In this embodiment, At least includes discrete distributed capacitance Optionally may also include (And other distribution parameters). The above mapping The internal point update formula and the port boundary update relationship given in the second point of Example 1 are used to determine the loop structure. Therefore, the loop structure is physically consistent with the transient propagation process of FDTD.
[0095] In one implementation, by following the state at each time step The port node voltages are extracted as outputs to form the predicted port responses. For example, let the port set be... Then the port prediction output at the nth time step can be expressed as:
[0096]
[0097] The predicted port waveform is obtained by concatenating the port outputs of all time steps according to time. .
[0098] Subsequently, a loss function is constructed based on the predicted port waveform and the observed port waveform. In one embodiment, waveform mismatch in the form of mean square error is used as the objective function. Let there be a total... There are several excitation shots, each of which can apply excitation to one or more ports and acquire responses at one or more ports; let the number of ports be . The number of time sampling points is Then the loss function can be taken as:
[0099]
[0100] in For the first Each working condition is at the port. The Observe the voltage at each sampling point. Predict the voltage for the corresponding model.
[0101] After constructing the loss function, the gradient of the loss function with respect to the distributed capacitance parameter is calculated through backpropagation, and then iteratively updated. In one implementation, gradient descent or adaptive optimization algorithms can be used to update the parameters, which can be exemplarily written as:
[0102]
[0103] in For the number of iterations, To update the step size. Due to the loop structure. Composed of discrete updates of FDTD and for It is differentiable, therefore it can be obtained through an automatic differentiation mechanism. There is no need to explicitly derive complex analytical gradient expressions.
[0104] In a preferred embodiment, to reduce the computational complexity and storage overhead caused by backpropagation of long sequences, the transient response of the port can be divided into several segments according to the time dimension, and the above-mentioned "forward time advancement - loss calculation - gradient update" process is performed on each segment. The piecewise time backpropagation strategy is adopted to realize feasible training under long sequence conditions.
[0105] Through the above steps, this embodiment can achieve quantitative inversion of the distributed capacitance parameters along the cable using only transient measurement information from the cable port, thereby obtaining the spatial distribution results of capacitance anomalies corresponding to soft faults, such as... Figure 4 As shown.
[0106] Example 2: Imaging method for soft fault parameters of resistance and capacitance simultaneously (integrated joint loss)
[0107] This embodiment, based on the cable physical model, FDTD discrete update and port boundary modeling, and cyclic calculation structure construction method described in Embodiment 1, further illustrates the method of jointly inverting the distributed resistance parameters and distributed capacitance parameters along the cable under port transient measurement conditions.
[0108] In this embodiment, the cable under test is equivalent to a one-dimensional distributed parameter transmission line model, whose unit length parameters include resistance. ,inductance Electrical conductivity and capacitor Among them, distributed resistance Distributed capacitance is used to characterize changes in conductor loss caused by conductor degradation or changes in contact conditions. Used to characterize changes in dielectric properties caused by dielectric degradation, etc. Example 2 will... and Simultaneously used as a parameter to be inverted; inductance Electrical conductivity It can be given according to the nominal value of the cable or regarded as a known constant.
[0109] In forward modeling, this embodiment still adopts the staggered mesh FDTD discrete update structure and port boundary condition modeling method described in Embodiment 1, and constructs "voltage update - current update - port boundary update" into a cyclic calculation unit, which is expanded in the time dimension to form a differentiable cyclic calculation structure, so that the port predicted waveform is related to and It also maintains differentiability, thus supporting end-to-end gradient inversion.
[0110] In terms of observation and objective function construction, this embodiment employs multi-excitation conditions and multi-port joint constraints. Let there be a total... There are several excitation conditions (shots), each with an excitation signal injected into one port and transient voltage response acquired at multiple ports; let the number of ports be... The number of time sampling points is In this embodiment, to balance and To investigate the different mechanisms by which port waveforms are affected, a joint loss function is constructed:
[0111]
[0112] in This is a time-domain waveform mismatch term used to constrain propagation to characteristics such as time and phase. This is the amplitude attenuation characteristic mismatch term, used to constrain energy dissipation and attenuation level. and These are the weighting coefficients.
[0113] Among them, the time-domain waveform mismatch term The mean square error form can be used:
[0114]
[0115] For the first Each working condition is at the port. The Observe the voltage at each sampling point. In the current , , and The predicted voltage is obtained by forward modeling using FDTD-RNN.
[0116] Amplitude attenuation characteristic mismatch term The logarithmic amplitude mismatch form of the port waveform envelope is adopted. For each port waveform... Calculate the analytic signal envelope respectively For example, it can be obtained through the Hilbert transform:
[0117]
[0118] And Defined as:
[0119]
[0120] in >0 represents a small constant to prevent numerical instability. By introducing an envelope logarithmic amplitude mismatch term, the interference of phase differences on resistance updates can be reduced, making the loss function more sensitive to attenuation differences caused by distributed resistance, thereby improving the efficiency of the loss function. Inversion stability.
[0121] Regarding parameter updates, this embodiment uses backpropagation calculation. The two types of parameters are then jointly iteratively updated. In one implementation, gradient descent or adaptive optimization algorithms can be used for the update, as exemplified by:
[0122]
[0123] By iteratively executing the process of "forward modeling - loss assessment - gradient calculation - parameter update" until the loss function meets the preset convergence condition, the joint inversion result of the distributed capacitance and distributed resistance along the cable is obtained, realizing the joint imaging of two types of soft faults: insulation degradation and conductor degradation.
[0124] To ensure the physical feasibility of the inversion results, this embodiment can... and Physical constraints are imposed on the range of values (e.g., limited to a preset range), or a parameterization method that guarantees the parameters are positive is adopted; at the same time, a piecewise backpropagation strategy can be used to reduce the computation and storage overhead under long-term sequence conditions, and the observability and robustness of joint inversion can be improved by multi-excitation conditions and multi-port joint constraints.
[0125] In summary, compared with the prior art, the method of this embodiment has at least the following advantages and beneficial effects:
[0126] 1) Strong physical consistency and interpretability: This invention directly constructs the physical core of electromagnetic propagation in cables—the FDTD recursive relationship—as a forward propagation structure of a neural network, ensuring that the entire inversion process is strictly constrained by physical laws and that the inversion results have clear physical meaning.
[0127] 2) High inversion efficiency and simple implementation: By embedding the physical model into a differentiable programming framework and using mature automatic differentiation tools to calculate the gradient, the complex adjoint field derivation or time reversal operation in traditional full-wave inversion is completely avoided, which significantly reduces the difficulty of algorithm implementation and computational cost.
[0128] 3) Strong detection capability for soft faults: Based on the inversion mechanism of full waveform matching, it is sensitive to the continuous change of the distribution parameters characterizing soft faults, and can realize the imaging of parameter gradient regions that are difficult to achieve by traditional reflection methods.
[0129] 4) Good engineering applicability: The proposed strategies such as unified port boundary processing, multi-condition / multi-port joint constraints, and segmented time training effectively improve the stability and robustness of the inversion and reduce memory requirements, making it more suitable for practical engineering scenarios.
[0130] 5) Low data requirements and excellent generalization ability: It does not rely on a large number of fault samples for training. Its "learning" ability comes from the physical model itself, so it has good generalization potential for cables of different models and working conditions.
[0131] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.
[0132] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.
[0133] Please see Figure 5 , Figure 5 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:
[0134] The processor 501 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.
[0135] The memory 502 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 502 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 502 and is called and executed by the processor 501 using the methods described in the embodiments of this application.
[0136] The input / output interface 503 is used to implement information input and output;
[0137] The communication interface 504 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).
[0138] Bus 505 transmits information between various components of the device (e.g., processor 501, memory 502, input / output interface 503, and communication interface 504);
[0139] The processor 501, memory 502, input / output interface 503, and communication interface 504 are connected to each other within the device via bus 505.
[0140] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.
[0141] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.
[0142] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0143] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.
[0144] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented in the embodiments of this program product are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments. The executable computer program code or "code" used to perform the various embodiments can be written in high-level programming languages such as C, C++, Python, Smalltalk, Java, JavaScript, Visual Basic, Structured Query Language (e.g., Transact-SQL), Perl, or in various other programming languages.
[0145] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.
[0146] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.
[0147] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.
[0148] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.
[0149] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0150] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.
[0151] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0152] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0153] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0154] If the integrated unit is implemented as 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 this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0155] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.
Claims
1. A method for inverting distributed parameters of cable soft faults, characterized in that, The method includes the following steps: Establish a one-dimensional distributed parameter transmission line model of the cable under test. The model parameters include the resistance per unit length distributed along the length of the cable. ,inductance Electrical conductivity and capacitor ; The time-domain control equations of the transmission line model are discretized using finite difference time-domain FDTD, and a voltage-current interleaved grid is used to obtain the recursive update relationship of the discrete state variables in adjacent time layers. Construct a recurrent neural network (RNN), and explicitly define the recursive update relationship as a single-step state update function of the RNN. Among them, the first The voltage and current fields at each time step are defined as the hidden states of the RNN. Define the port excitation input as and the distribution parameters to be inverted The state update function is embedded as a trainable parameter. ; The RNN is expanded in the time dimension and forward propagated to obtain the predicted voltage response at each time step port. ; Based on the predicted voltage response Measured voltage response at the port Construct the loss function ; The loss function is calculated by automatic differentiation. Regarding the distribution parameters The gradient is calculated and iteratively updated using backpropagation. Continue until the loss function converges, and output the inversion results of the distributed parameters corresponding to the cable soft fault; In the voltage-current interleaved grid, the voltage is defined at discrete nodes. At this point, the current is defined between adjacent nodes. The recursive update relationship includes an alternating progressive structure in which voltage is updated differentially by adjacent currents and current is updated differentially by adjacent voltages. The function The computational logic includes updating the port boundary conditions; the calculation of port boundary condition updates and the calculation of cable interior point updates are performed uniformly in the same time-progression step.
2. The method according to claim 1, characterized in that, The port boundary conditions include the equivalent voltage source boundary at the source end, and the equivalent voltage source includes the source internal resistance. This makes the source node voltage update correlated with the current of the adjacent line segment at the port and the excitation input.
3. The method according to claim 1, characterized in that, The predicted voltage response From the hidden state of RNN The port voltage is sampled and then spliced together at all time steps to form the port prediction waveform.
4. The method according to claim 1, characterized in that, The loss function To address the joint loss under multiple excitation conditions and multiple port observations, joint constraints are applied to the predicted and observed responses for different excitation conditions, different ports, and time sampling points.
5. The method according to claim 1, characterized in that, The distribution parameters to be inverted Including distributed capacitance With distributed resistance And during the reverse propagation process, respectively and Perform iterative updates and output the joint inversion result of the two.
6. The method according to claim 5, characterized in that, The loss function The joint loss function is expressed as follows: in, This is a time-domain waveform mismatch term used to constrain propagation to time and phase characteristics; This is the amplitude attenuation feature mismatch term, used to constrain the amplitude attenuation feature of the waveform envelope; and These are the weighting coefficients.
7. The method according to claim 1, characterized in that, The port response is divided into several segments according to the time dimension, and the distribution parameters are updated using a segmented time backpropagation strategy to reduce the computational complexity and storage overhead under long-term sequence conditions.
8. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 7.