Power line fault positioning and early warning system

Abstract

The invention discloses an electric power line fault positioning and early warning system, relates to the technical field of electric power, and solves the problem of insufficient early warning capability of existing processing methods. The signal acquisition unit captures a plurality of broadband transient signals and full-waveform backscattering signals on a power line, and the broadband transient signals and the full-waveform backscattering signals are converted into a digital signal sequence through the data acquisition and processing unit. The chaos fingerprint database module stores reference known boundary features; the threatening early warning module detects a weak threatening signal from the digital signal sequence in a normal monitoring state, and outputs a threatening early warning signal when the weak threatening signal is detected; when the self-adaptive positioning module monitors that a hard fault occurs, chaos features are extracted from the full-waveform backscattering signals, a dynamic transformation operator is solved based on reference known boundary features, the dynamic transformation operator is used for correcting the off-line chaos fingerprint reference library to generate a corrected fingerprint library, and the corrected fingerprint library is used for positioning the full-waveform backscattering signals. And the chaos features are matched in the corrected fingerprint database, so that the fault position can be solved. The method is used for fault positioning and early warning.

Description

Technical Field

[0001] This invention relates to the field of power technology, specifically to a power line fault location and early warning system. Background Technology

[0002] Ensuring the safe and stable operation of the power grid is the core task of the power system. As the channel for power transmission, the health status of power lines directly affects the reliability of power supply. Therefore, achieving rapid and accurate location of power line faults and early warning of potential fault hazards has always been a key focus and challenge in the field of power technology.

[0003] In fault location technology, fingerprint matching based on traveling waves and transient signals has attracted attention due to its theoretically high accuracy. This requires pre-constructing a signal feature database containing signals corresponding to different fault locations and types through offline simulation or field experiments. However, power lines are long-distance facilities exposed outdoors, and their electrical parameters are not constant but are affected by a combination of time-varying factors such as ambient temperature, air humidity, line icing, material aging, and real-time load fluctuations. These time-varying factors cause the transient response characteristics of the line to change dynamically during actual operation. This leads to a mismatch between the pre-established signal feature database based on ideal or baseline conditions and the real-time physical state of the power line. When a real fault occurs, the collected signal features are morphologically inconsistent with the baseline features stored in the database. Comparing with this mismatched signal feature database directly results in deviations and incorrect location, making it difficult to meet the high-precision location requirements of actual operation and maintenance.

[0004] In terms of fault early warning, before catastrophic hard faults such as short circuits and open circuits occur, early defects such as micro-cracks in power line insulators and partial discharges at equipment joints generate some weak, high-frequency transient signals, i.e., precursor signals. However, in actual operating power systems, there is a large amount of electromagnetic interference and background noise far exceeding the intensity of these precursor signals. The amplitude of these weak precursor signals is far below the background noise floor, resulting in an extremely low signal-to-noise ratio at the signal acquisition end. Existing detection systems and traditional signal processing methods ignore these early potential hazards as random noise when faced with such a strong noise background, leading to insufficient system early warning capabilities. This prevents timely alarms before faults escalate into serious accidents, thus missing the best opportunity for preventative maintenance. Summary of the Invention

[0005] The purpose of this invention is to address the problem that existing detection systems and traditional signal processing methods often ignore early warning signals as random noise in the face of strong noise backgrounds, resulting in insufficient early warning capabilities. This invention proposes a power line fault location and early warning system.

[0006] A power line fault location and early warning system, the system comprising multiple signal acquisition units, data acquisition and processing units, a chaotic fingerprint database module, a precursor warning module and an adaptive positioning module;

[0007] Multiple signal acquisition units are used to capture multiple broadband transient signals and full-waveform backscattered signals on the power line and transmit them to the data acquisition and processing unit.

[0008] The data acquisition and processing unit is electrically connected to multiple signal acquisition units and is used to convert multiple broadband transient signals and full-waveform backscattered signals into digital signal sequences, which are then transmitted to the early warning module and the adaptive positioning module, respectively.

[0009] The chaotic fingerprint library module is used to store an offline chaotic fingerprint benchmark library, which contains known boundary features of the benchmark.

[0010] The early warning module communicates with the data acquisition and processing unit and the chaotic fingerprint database module. It is used to detect weak early warning signals from broadband transient digital signal sequences under normal monitoring conditions, based on the nonlinear dynamic system model and the principle of stochastic resonance, and output early warning signals when weak early warning signals are detected.

[0011] An adaptive positioning module is used to extract chaotic features from a full-waveform backscattered digital signal sequence when a hard fault is detected, and to solve a dynamic transformation operator based on the known boundary features of the benchmark stored in the chaotic fingerprint library module. The dynamic transformation operator is used to correct the offline chaotic fingerprint benchmark library to generate a corrected fingerprint library, and the fault location is calculated by matching the chaotic features in the corrected fingerprint library.

[0012] Preferably, the nonlinear dynamic system model is a bistable system, and the dynamic behavior of the nonlinear dynamic system model is described by the Langevin equation.

[0013] Preferably, before extracting the chaotic features, the adaptive positioning module is further used to deconstruct the real-time full-waveform backscatter signal to separate the real-time known boundary full-waveform backscatter signal and the real-time unknown fault full-waveform backscatter signal.

[0014] The chaotic features include real-time known boundary features and real-time unknown features extracted from real-time known boundary full waveform backscattered signals and real-time unknown fault full waveform backscattered signals, respectively.

[0015] Preferably, chaotic features are extracted, and the specific process is as follows:

[0016] Perform phase space reconstruction on the time series to construct chaotic attractors;

[0017] Furthermore, a quantitative calculation of chaotic features is performed on the chaotic attractor to generate a chaotic feature descriptor including the correlation dimension and the maximum Lyapunov exponent, which serves as the chaotic feature.

[0018] Preferably, before performing the phase space reconstruction, the adaptive positioning module is further used to determine the time delay using the average mutual information method and to determine the embedding dimension using the pseudo-nearest neighbor method.

[0019] Preferably, the dynamic transformation operator is modeled as an affine transformation, which includes a transformation matrix and a translation vector;

[0020] The adaptive positioning module is used to solve for the transformation matrix and the translation vector by minimizing the Euclidean distance between the real-time known boundary features and the reference known boundary features after the affine transformation.

[0021] Preferably, the adaptive positioning module is configured to match the chaotic features in the corrected fingerprint database to calculate the fault location. The specific process is as follows:

[0022] Calculate the Euclidean distance between the real-time unknown feature and the corrected baseline feature corresponding to each entry in the corrected fingerprint database;

[0023] In addition, the minimum value of the Euclidean distance is found to determine the location of the fault.

[0024] Preferably, the offline chaotic fingerprint benchmark library and the benchmark known boundary features are generated by constructing an electromagnetic transient model of a power line and performing offline simulation;

[0025] The offline simulation generation includes: simulating and acquiring the full waveform backscattered signal of the reference fault and the full waveform backscattered signal of the reference known boundary;

[0026] Chaotic feature extraction is performed on the backscattered signal of the full waveform of the baseline fault and the backscattered signal of the full waveform of the baseline known boundary obtained from the simulation to obtain the baseline chaotic features and the baseline known boundary features.

[0027] Preferably, the signal acquisition unit is one or more combinations of a high-frequency current transformer, a capacitive coupler, or an optical sensor;

[0028] The data acquisition and processing unit includes a local processor, which is either a digital signal processor or a field-programmable gate array.

[0029] Preferably, the system further includes a main control and output unit.

[0030] The main control and output unit is used to suspend or stop the operation of the early warning module and start the adaptive positioning module when a hard fault trigger signal is detected.

[0031] The main control and output unit is a central server or an industrial control computer.

[0032] The beneficial effects of this invention are:

[0033] This invention, by setting up a precursor warning module and utilizing the principle of random resonance, treats the inherent background noise of the line as an auxiliary energy source, nonlinearly amplifies the weak precursor features submerged under the noise background, so that the weak precursor features can be clearly detected and identified, realizing early warning before hard faults occur, providing valuable time for taking preventive maintenance measures, and improving the operational safety of the power system.

[0034] This invention sets up an adaptive positioning module and uses the deviation between real-time known boundary features and reference features to solve a dynamic transformation operator. The adaptive positioning module uses the dynamic transformation operator to perform dynamic correction on the entire chaotic fingerprint reference library, generating a corrected fingerprint library that is aligned with the current physical state. This ensures that the matching of fault features is carried out in the same real-time feature space, thereby improving the accuracy of fault location and environmental adaptability.

[0035] This invention uses a dynamic transformation operator to correct the benchmark library, avoiding the complex work of having to perform large-scale, time-consuming offline simulations to adapt to changes in line status. The adaptive correction process involves only one mapping operation, with low computational load and fast response speed. While ensuring high positioning accuracy, it reduces computational overhead and data maintenance costs, and improves real-time response capability and engineering practicality. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of a power line fault location and early warning system.

[0037] Figure 2 Flowchart for building a chaotic fingerprint benchmark library;

[0038] Figure 3 Flowchart for implementing early warning signals;

[0039] Figure 4 Flowchart for adaptive fault localization. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0042] Example:

[0043] Reference Figure 1 An embodiment provides a power line fault location and early warning system, the system including multiple signal acquisition units 1, data acquisition and processing units 2, chaotic fingerprint database module 3, early warning module 4 and adaptive positioning module 5;

[0044] Multiple signal acquisition units 1 are used to capture multiple broadband transient signals and full-waveform backscattered signals on the power line and transmit them to the data acquisition and processing unit 2;

[0045] The data acquisition and processing unit 2 is electrically connected to multiple signal acquisition units 1, and is used to convert multiple broadband transient signals and full waveform backscattered signals into digital signal sequences, which are then transmitted to the early warning module 4 and the adaptive positioning module 5, respectively.

[0046] Chaotic fingerprint library module 3 is used to store an offline chaotic fingerprint benchmark library, which contains benchmark known boundary features;

[0047] The early warning module 4 is connected to the data acquisition and processing unit 2 and the chaotic fingerprint library module 3. It is used to detect weak early warning signals from broadband transient digital signal sequences under normal monitoring conditions, based on the nonlinear dynamic system model and the principle of random resonance, and output early warning signals when weak early warning signals are detected.

[0048] The adaptive positioning module 5 is used to extract chaotic features from the digital signal sequence of full waveform backscattering when a hard fault is detected, and solve the dynamic transformation operator based on the benchmark known boundary features stored in the chaotic fingerprint library module 3. The dynamic transformation operator is used to correct the offline chaotic fingerprint benchmark library to generate a corrected fingerprint library, and the fault location is calculated by matching the chaotic features in the corrected fingerprint library.

[0049] Specifically, multiple signal acquisition units 1 are deployed at monitoring points of the power line, such as the main end, the end, or key branch nodes of the line.

[0050] Multiple signal acquisition units 1 are one or more combinations of high-frequency current transformers, capacitive couplers, or optical sensors;

[0051] The data acquisition and processing unit 2 includes an analog-to-digital conversion module and a local processor. The analog-to-digital conversion module is used to convert the analog signals captured by the signal acquisition unit into digital signal sequences. The sampling rate meets the requirements for capturing high-frequency transient signals. The local processor is either a digital signal processor or a field-programmable gate array.

[0052] Furthermore, the nonlinear dynamic system model is a bistable system, and the dynamic behavior of the nonlinear dynamic system model is described by the Langevin equation.

[0053] Furthermore, before extracting the chaotic features, the adaptive positioning module 5 is also used to deconstruct the real-time full-waveform backscatter signal to separate the real-time known boundary full-waveform backscatter signal and the real-time unknown fault full-waveform backscatter signal.

[0054] The chaotic features include real-time known boundary features and real-time unknown features extracted from real-time known boundary full waveform backscattered signals and real-time unknown fault full waveform backscattered signals, respectively.

[0055] Further refinement, extraction of chaotic features, specific process:

[0056] Perform phase space reconstruction on the time series to construct chaotic attractors;

[0057] Furthermore, a quantitative calculation of chaotic features is performed on the chaotic attractor to generate a chaotic feature descriptor including the correlation dimension and the maximum Lyapunov exponent, which serves as the chaotic feature.

[0058] Furthermore, before performing the phase space reconstruction, the adaptive positioning module 5 is also used to determine the time delay using the average mutual information method and to determine the embedding dimension using the pseudo-nearest neighbor method.

[0059] Further specifying, the dynamic transformation operator is modeled as an affine transformation, which includes a transformation matrix and a translation vector;

[0060] The adaptive positioning module 5 is used to solve for the transformation matrix and the translation vector by minimizing the Euclidean distance between the real-time known boundary features and the reference known boundary features after the affine transformation.

[0061] Further specifying, the adaptive positioning module 5 is configured to match the chaotic features in the corrected fingerprint database to calculate the fault location, specifically as follows:

[0062] Calculate the Euclidean distance between the real-time unknown feature and the corrected baseline feature corresponding to each entry in the corrected fingerprint database;

[0063] In addition, the minimum value of the Euclidean distance is found to determine the location of the fault.

[0064] Furthermore, the offline chaotic fingerprint benchmark library and the benchmark known boundary features are generated by constructing an electromagnetic transient model of a power line and performing offline simulation;

[0065] The offline simulation generation includes: simulating and acquiring the full waveform backscattered signal of the reference fault and the full waveform backscattered signal of the reference known boundary;

[0066] Chaotic feature extraction is performed on the backscattered signal of the full waveform of the baseline fault and the backscattered signal of the full waveform of the baseline known boundary obtained from the simulation to obtain the baseline chaotic features and the baseline known boundary features.

[0067] Furthermore, the system also includes a main control and output unit.

[0068] The main control and output unit (6) is used to suspend or stop the operation of the early warning module 4 and start the adaptive positioning module 5 when a hard fault trigger signal is detected;

[0069] The main control and output unit (6) is a central server or an industrial control computer.

[0070] The following is a detailed description of the content of this embodiment:

[0071] The early warning module is configured to receive and analyze in real time the broadband signal sequence captured by the signal acquisition unit and digitized by the data acquisition and processing unit. Based on the principle of nonlinear random resonance, the early warning module uses the background noise in the signal to nonlinearly amplify the weak early warning signal submerged in the noise. When a resonance feature that meets the conditions is detected, the early warning module outputs an early warning signal.

[0072] An adaptive positioning module is configured to activate upon detection of a hard fault. The adaptive positioning module captures the full-waveform backscattered signal after the fault and extracts the chaotic features of the full-waveform backscattered signal. The adaptive positioning module is also configured to perform adaptive correction. The adaptive positioning module separates and analyzes the reflected wave features from known boundaries and uses them as a reference benchmark to dynamically correct the benchmark fingerprint stored in the chaotic fingerprint library module. By matching the reflection features of the unknown fault point with the corrected fingerprint library, the fault location is calculated.

[0073] The chaotic fingerprint library module is configured to persistently store the offline chaotic fingerprint benchmark library and the benchmark known boundary features in the memory of the main control and output units. The offline chaotic fingerprint benchmark library stores the mapping relationship between the fault location and the benchmark chaotic features, and the benchmark known boundary features are the reference for subsequent correction. The chaotic fingerprint library module provides data query and write operations for the adaptive positioning module.

[0074] The early warning module is configured to receive and analyze broadband signal sequences digitized by the data acquisition and processing unit in real time. Broadband signal sequence It consists of two parts, namely:

[0075] ;

[0076] in, It is a broadband signal sequence. The target is a weak precursor signal with an amplitude below the noise floor. This is the inherent broadband background noise of the system.

[0077] The early warning module is implemented by transmitting a complete broadband signal sequence. As input, it is injected into a pre-defined nonlinear dynamic system model. The nonlinear dynamic system model is a bistable system, and its dynamic behavior is described by the Langevin equations:

[0078] ;

[0079] in, It is the output time series of this nonlinear system. and The system potential well parameters are preset, and , The early warning module utilizes The energy that makes Stochastic resonance occurs with nonlinear systems, thus affecting the output. In China The corresponding nonlinear amplification of characteristic frequency components is further configured in the early warning module for calculation. power spectral density ,when exist The peak value within the characteristic frequency band exceeds the preset warning threshold. At that time, the early warning module outputs an early warning signal to the main control and output unit.

[0080] The adaptive positioning module is configured to activate after the system detects a hard fault. The adaptive positioning module acquires the real-time full-waveform backscattered signal (i.e., the real-time full-waveform backscattered signal) from the data acquisition and processing unit after the fault, using a high sampling rate. The adaptive positioning module is further configured to perform real-time full-waveform backscatter signal deconstruction. Separate into two components:

[0081] Real-time known boundary full waveform backscattered signal This component is mainly caused by the reflected waveform resulting from the impedance mismatch at the known boundaries of the line;

[0082] Real-time unknown fault full waveform backscattered signal This component is mainly composed of unknown fault points. The reflected and scattered waveforms generated at that location.

[0083] The core function of the adaptive positioning module is to perform adaptive correction, which first... Phase space reconstruction and chaotic feature extraction are performed to obtain real-time known boundary features. Subsequently, the adaptive localization module retrieves an offline-stored baseline known boundary feature from the chaotic fingerprint database module. .

[0084] The adaptive positioning module is configured to be used for comparison and To solve a dynamic transformation operator for the deviation between them. Operator The distortion of the chaotic feature space caused by real-time changes in the line state is characterized by the use of operators in the adaptive positioning module. Offline chaotic fingerprint benchmark library in the chaotic fingerprint library module Perform item-by-item transformations to generate a corrected fingerprint database. .

[0085] Finally, the adaptive positioning module... Extracting real-time unknown features and in the corrected fingerprint database The search and matching are performed to solve for the unique match. Best matching fault location And output it.

[0086] The chaotic fingerprint database module is configured to persistently store two core datasets in the memory of the main control and output units:

[0087] Offline Chaotic Fingerprint Benchmark Library Stored Discrete fault locations ( (and the baseline fault full waveform backscattered signal obtained through offline simulation calculation:)

[0088] ;

[0089] Benchmark known boundary features The backscattered signal of the known boundary full waveform is obtained through offline simulation benchmark. It is obtained by extracting chaotic features and is the dynamic transformation operator for adaptive positioning module calculation. The chaotic fingerprint library module provides a reference benchmark for the adaptive positioning module. and The data reading interface supports... Real-time generation, storage, and querying.

[0090] See attached document Figure 2 In the offline preparation stage of this embodiment, the data generation process in the chaotic fingerprint database module primarily involves constructing a high-precision electromagnetic transient model of the power line.

[0091] The electromagnetic transient model is implemented on a dedicated simulation platform. This model is used to generate transient waveforms caused by hard faults and precursor signals in subsequent simulations.

[0092] In electromagnetic transient models, the power line itself is characterized using a high-frequency distributed parameter model, such as a frequency-dependent model or the J. Marti model, in order to accurately reproduce the attenuation and dispersion effects exhibited by the high-frequency transient traveling wave during propagation in subsequent simulations.

[0093] The electromagnetic transient model also includes the line topology, such as branch lines or T-junctions. Impedance mismatch points in these topologies are important sources of full-waveform backscattered signals and must be characterized in the electromagnetic transient model. In addition, the model must include the boundary conditions at both ends of the line. These boundary conditions are modeled as equivalent impedance networks to characterize the comprehensive impedance characteristics of substation buses, circuit breakers, voltage or current transformers, and other connected equipment at high frequencies. Accurate modeling of these boundary conditions is the basis for subsequent extraction of full-waveform backscattered signals from known baseline boundaries and the implementation of adaptive correction.

[0094] See attached document Figure 3 After completing the electromagnetic transient model construction, the simulation and extraction of the benchmark full-waveform backscatter signal are carried out, including the extraction of the benchmark fault full-waveform backscatter signal and the extraction of the benchmark known boundary full-waveform backscatter signal.

[0095] The backscattered signal of the full waveform of the reference fault is extracted on an electromagnetic transient model.

[0096] First, along the length of the line model, a preset high-resolution spatial interval is set. Discrete simulated fault points ,in:

[0097] ;

[0098] Then, for each fault point The model simulates a hard fault event, such as a metallic short circuit or a low-resistance ground fault. In each fault simulation, at a preset monitoring point ( Record a complete high-frequency transient time series from the moment the fault occurs; this series captures the state of the fault point. The initial traveling wave, and all subsequent reflections and scattering waveforms generated as it propagates between the line topology and the two end boundaries, are defined in a complete time series corresponding to the location. The reference fault full waveform backscattered signal is denoted as .

[0099] The purpose of extracting the full waveform backscattered signal from the known boundary of the benchmark is to obtain the reference anchor point required for subsequent adaptive correction. This anchor point characterizes the electromagnetic reflection characteristics of the monitoring point itself. A preset high-frequency test pulse signal is injected into the circuit model, and simultaneously, at the same monitoring point ( Record the reflected echo signal caused by the test pulse. Since the test pulse first encounters... The boundary is such that the first part of the recorded echo signal that arrives at the boundary and has the most distinctive characteristics is the reflection feature of the known boundary.

[0100] Using time windows or other signal separation methods, the reflected waveform generated solely by the known boundary is extracted from the echo signal. This waveform is defined as the full-waveback scattered signal of the reference known boundary, denoted as... ,Should Independent of any fault location It is not a fingerprint of the monitoring point itself in the electromagnetic transient model.

[0101] See attached document Figure 2 After simulating and extracting the baseline full-waveform backscattered signal, phase space reconstruction and chaotic feature extraction are then performed. The aim is to transform the acquired one-dimensional, seemingly disordered time series into a high-dimensional state-space attractor that reveals its intrinsic dynamic characteristics. This attractor is then quantitatively described and applied sequentially to each generated time series. For clarity, general notation is used below. Represents an input time series (i.e.) It can be any one or ).

[0102] First, performing phase space reconstruction requires determining two key parameters: time delay. and embedding dimension .

[0103] Time delay The determination of time delay can be achieved using the average mutual information method to calculate different time delays. Below, the original signal With delayed signal Mutual information between ,choose The first local minimum that appears As the optimal time delay, it ensures that the reconstructed phase space coordinates are statistically independent.

[0104] Embedding dimension The determination of the embedding dimension can be achieved using the pseudo-nearest neighbor method, when the embedding dimension increases from... Increase to At that time, the percentage of points in phase space that are mistakenly grouped together due to their low dimension. , choose to The initial drop below a preset, sufficiently small threshold As the optimal embedding dimension, it ensures that the attractor of the original dynamical system is in It is fully expanded in 3D space, eliminating the self-intersection of trajectories.

[0105] Determining the optimal parameters and Then, for the time series Reconstruction is performed to generate a series of phase space vectors. :

[0106] ;

[0107] in , For the total number of phase space vectors generated, all The resulting set represents the chaotic attractors corresponding to the time series of the full-waveform backscattered signal. Then, chaotic feature extraction involves quantitatively calculating the reconstructed chaotic attractors to generate a multi-dimensional chaotic feature descriptor. ,Should This is the fingerprint used for subsequent matching. A vector may include one or more of the following components:

[0108] Correlation dimension This is used to quantify the geometric complexity or fractal properties of attractors by calculating correlation integrals. To determine, Defined as:

[0109] ;

[0110] in, It is the radius. It is the Herveside step function. It is the Euclidean distance between two phase space vectors. pass and The scaling relationship between them is determined as follows:

[0111] ;

[0112] Maximum Lyapunov index The predictability of system dynamics is used to quantify the predictability of attractor trajectories; it is the average exponential divergence rate of the attractor trajectory, a positive value. It is a clear indicator of chaotic behavior. This can be achieved by tracking the initial neighbor vector pairs in phase space. and Over time Average divergence distance of evolution To calculate, the divergent behavior approximately satisfies:

[0113] ;

[0114] in It is the sampling time interval, through the sampling... By fitting the logarithmic slope of the evolution curve, the solution can be obtained. .

[0115] By calculating one or more of the above descriptors, for each Generate a unique chaotic feature vector:

[0116] ;

[0117] Should The vector is that Chaotic fingerprints.

[0118] See attached document Figure 3 After extracting chaotic features from all benchmark full-waveform backscattered signals, the core of generating the benchmark fingerprint database lies in establishing a persistent mapping relationship data structure, namely, an offline chaotic fingerprint benchmark database. And store it in the chaotic fingerprint database module.

[0119] The physical carrier of the chaotic fingerprint library module is the memory of the main control and output unit. This offline chaotic fingerprint benchmark library... A data structure organized as a lookup table or a collection of key-value pairs, in which the keys are defined. Location of each discrete simulation fault point ,in The value is related to that The unique corresponding baseline chaotic feature vector obtained in the previous calculation The benchmark chaotic eigenvector Based on correlation dimension Maximum Lyapunov index It consists of one or more chaotic descriptor components; therefore, the complete structure of this benchmark library can be represented as:

[0120] ;

[0121] this When the system is running online, the benchmark dictionary used for matching by the adaptive positioning module provides a benchmark mapping relationship from a chaotic feature to a physical location.

[0122] In addition to generating and storing In addition, the system must also store separately the full waveform backscattered signal with a known baseline boundary. The calculated chaotic feature is defined as the baseline known boundary feature. ,Should It is also persistently stored in the chaotic fingerprint database module, but not used as... Members of a set

[0123] Its function is to serve as a reference anchor point during the online adaptive correction phase. The adaptive positioning module will use this point during online operation. With real-time captured boundary reflection features The comparison is performed, and the result of the comparison is the solution of the dynamic transformation operator. The basis for this.

[0124] See attached document Figure 3 The first function in the online operation phase of this embodiment is executed by the early warning module.

[0125] The operating principle of the early warning module differs from traditional filtering and noise reduction techniques based on linear systems. Traditional filtering techniques convert broadband signal sequences... Background noise in The technical concept of the early warning module in this embodiment is to treat the background noise as interference and filter it out. Treating it as an auxiliary energy source, this is achieved by constructing a pre-defined nonlinear dynamic system. Full injection, utilizing The energy works in conjunction with this nonlinear system to nonlinearly transfer and concentrate the energy to The weak precursor signals contained within are to be detected. characteristic frequency Up, thus achieving The features are magnified.

[0126] The nonlinear dynamic system used in this early warning module is a bistable system, and its dynamic behavior is described by a Langevin equation:

[0127] ;

[0128] in, It is the output time series of this nonlinear system. It is a complete broadband signal sequence obtained from the data acquisition and processing unit. This is the potential function of the bistable system. In this example, it is set as a quartic polynomial with two stable equilibrium points and one unstable equilibrium point, i.e., a double potential well:

[0129] ;

[0130] in, and These are pre-set system parameters that are greater than zero. and Together, they determine the depth of the two potential wells and the height of the potential barrier between them. , the potential function Substituting the above dynamic equations, we obtain the specific operational model of this early warning module:

[0131] ;

[0132] in, For the system's output state variables, and Here are the potential well parameters. For inclusion and The system input, the resonant amplification mechanism of this operating model is: when only noise is input... When used as input, the system state Random, uncontrolled transitions occur between two potential wells, with the transition rate determined by the noise intensity. Decide.

[0133] When a weak, quasi-periodic precursor signal With noise Together as When inputting, this Will to The symmetry of the element applies a tiny, periodic modulation.

[0134] when cycle With noise When the induced average potential well transition times match, the system undergoes stochastic resonance. At this point, the system state... The transition behavior will change from random to quasi-synchronous with The periodic modulation, this quasi-synchronous behavior, results in the output signal The energy spectrum redistribution, the noise that was originally distributed over a wide frequency band The energy is nonlinearly concentrated into characteristic frequency In terms of its harmonics, this is specifically manifested in the system output. exist peak power spectral density at Significantly higher than the original weak signal Its own spectral peaks.

[0135] See attached document Figure 4 The early warning process is executed by the early warning module, which acquires broadband signal sequences in real time. The early warning module then acquires the digitized broadband signal sequences in real time via a data acquisition and processing unit. ,Should It is used as the sole input to the nonlinear dynamic system.

[0136] Solving for the output of the nonlinear system, the early warning module will acquire... As a driving force, we substitute it into the bistable system model:

[0137] ;

[0138] in, and For the pre-defined system parameters used to define the potential well shape, due to Includes random noise The above equation is a stochastic differential equation. This early warning module uses a numerical integration method with discrete time steps to achieve this. The equation is solved iteratively, and this solution process generates an expression that is similar to the input. Correspondingly, a completely new output time series .

[0139] The power spectrum of the output signal is calculated by the early warning module for the generated output time series. Perform spectrum analysis, by analyzing... The power spectral density estimate is calculated using a fast Fourier transform algorithm on a time window of data. .

[0140] The precursor warning module performs resonance detection and threshold judgment, and calculates the power spectrum. In this process, one or more preset precursor characteristic frequency bands are monitored. These precursor characteristic frequency bands are based on the weak precursor signals to be detected. The physical properties are preset.

[0141] The early warning module searches for spectral peaks within the aforementioned characteristic frequency bands and extracts the maximum peak value. And it is compared with a preset, fixed warning threshold. Compare them.

[0142] Output a warning signal if a spectral peak is detected. satisfy If the conditions are met, the system determines that random resonance has occurred, indicating the presence of a weak precursor signal. In this situation, the early warning module immediately generates an early warning signal and sends it to the main control and output unit, which then performs subsequent alarm or recording operations. If the conditions are not met... If the conditions are met, the system determines that there are no warning signals and returns to continue executing the next cycle of real-time monitoring.

[0143] See attached document Figure 4 In the online operation phase of this embodiment, when the system detects a sudden change in power frequency current or voltage and determines that a hard fault has occurred, the system function switches to the adaptive positioning module and immediately executes the fault adaptive positioning process.

[0144] The primary step in the adaptive fault localization process is the real-time acquisition and deconstruction of the full-waveform backscatter signal. During this process, the adaptive localization module first captures a high-sampling-rate digital signal sequence at the instant the fault occurs and for a period thereafter through the data acquisition and processing unit. This digital signal sequence is the real-time full-waveform backscatter signal monitored online, denoted as... ,Should Includes unknown fault points The generated initial wave, and the complete waveform information of all reflections, scattering, attenuation and dispersion of the initial wave and subsequent waveforms between the line topology, towers and known boundaries at both ends of the line.

[0145] In capture Then, the adaptive positioning module immediately performs waveform decomposition on it. The purpose of this waveform decomposition operation is to... The signal is separated into two core components for subsequent processing: the real-time known boundary full-waveform backscattered signal. and real-time unknown fault full waveform backscatter signal .

[0146] There are two methods for waveform deconstruction. One is based on the time window method, which utilizes the arrival time difference of different reflected waves. Since the known boundary is the impedance mismatch point closest to the signal acquisition unit, the reflected waves generated by this known boundary... It must be The first feature component to arrive, therefore, the adaptive localization module in The initial part of the extracted signal, The length is set to be less than the shortest propagation time of the reflected wave generated by any fault point. In this way, the signal within this time window is separated and identified as... ,exist After being extracted, the adaptive positioning module is in A slightly later time window Internal signal extraction, this Corresponding to the main by The arrival time of the generated reflected wave, within which the signal is separated and identified as... .

[0147] Another deconstruction method is based on template matching. The adaptive positioning module retrieves the backscattered full-waveform of the known boundary from the chaotic fingerprint database module. This adaptive positioning module will As a waveform template, in the captured Perform cross-correlation or matched filtering operations in the middle. In the waveform template The signal segment that produces the maximum correlation peak or matched response is identified and extracted as... .

[0148] exist After being identified, Zhongyu The strongly correlated subsequent full-waveform backscattered signal portion was separated and extracted as These two separated time series and It is output as input for adaptive correction.

[0149] See attached document Figure 4 After completing the real-time full-waveform backscatter signal capture and deconstruction, the adaptive positioning module immediately performs adaptive correction of the chaotic fingerprint database.

[0150] The principle lies in the offline chaotic fingerprint benchmark library stored in the chaotic fingerprint database module. It is generated based on a fixed set of offline electromagnetic transient model parameters. However, the real-time high-frequency transient characteristics of the physical circuit will change with environmental factors. This change leads to a mismatch between the offline model and physical reality. This mismatch causes chaotic features corresponding to a real-time captured full-waveform backscattered signal. This will systematically deviate from its corresponding baseline features in the offline library. The goal is to quantify this mismatch in real time using a known boundary as a common reference point, and to analyze the entire... Perform dynamic corrections to align it with the current physical reality.

[0151] First, real-time feature calculation is performed, and the adaptive localization module processes the two separated time series. and The phase space reconstruction (PSR) and chaotic feature extraction methods are executed separately. This computation process generates two real-time chaotic feature vectors with known boundary features in real time. and real-time unknown features Next, the dynamic transformation operator is executed. The adaptive localization module retrieves the stored baseline known boundary features from the chaotic fingerprint database module for generation. The adaptive positioning module compares and The deviation between them is used to solve a dynamic transformation operator. The dynamic transformation operator It is used to describe the mapping distortion from the baseline feature space to the real-time feature space caused by model mismatch.

[0152] Assuming chaotic characteristics It is 3D vectors, dynamic transformation operators It is modeled as an affine transformation:

[0153] ;

[0154] in, It is The transformation matrix, It is The translation vector of the operator This is determined by solving a minimization problem, which aims to minimize... and Find the distance between them.

[0155] ;

[0156] in Let Euclidean distance (norm 12) be the factor. By minimizing this factor, the system obtains a unique transformation that characterizes the difference between the current line state and the model. Then, fingerprint database correction is performed, and the adaptive positioning module uses the dynamic transformation operator obtained in the previous step. It is applied to the offline chaotic fingerprint benchmark library stored in the chaotic fingerprint library module. All entries within, that is, for Each of the baseline chaotic features ( The adaptive positioning module calculates its corrected features. :

[0157] ;

[0158] All calculations obtained Its corresponding position Combined, they form a completely new, calibrated fingerprint database. :

[0159] ;

[0160] The corrected fingerprint database It is a temporary database, valid only for the current fault, whose chaotic feature space has been aligned according to the current physical state of the line. Connecting synchronous real-time unknown features They are then output together to the chaotic fingerprint matching system to perform the final localization calculation.

[0161] See attached document Figure 4 After completing the adaptive correction of the chaotic fingerprint database, the execution process of the adaptive localization module enters the final step: fault localization, which receives two core input data:

[0162] The previously calculated real-time unknown features ;

[0163] And the previously generated corrected fingerprint database:

[0164] ;

[0165] By performing chaotic feature matching, in Search and The closest entry, used to pinpoint the final location of the fault, includes:

[0166] First, chaotic feature matching is performed, and the adaptive localization module traverses the corrected fingerprint database. All of them Each entry, for each The adaptive positioning module calculates real-time unknown features. Corrected reference features corresponding to this entry Distance between ,Should A vector is dimensional vector, distance This is achieved by calculating the Euclidean distance between the two vectors:

[0167] ;

[0168] in This represents the Euclidean distance between vectors.

[0169] Then, the positioning results are output, after all calculations are completed. Dist Then, the adaptive localization module performs a minimization search operation across all Dist. middle Find the index corresponding to the minimum value of this distance. The search process can be represented as:

[0170] ;

[0171] The index The corresponding corrected fingerprint database The key in The physical location identified as the best match for the backscattered signal characteristics of the real-time fault full waveform is then used by the adaptive positioning module. Defined as the final fault location result ,Right now The adaptive positioning module will use the positioning result The data is sent to the main control and output unit, which then displays, stores, or triggers corresponding protection actions.

[0172] See attached document Figure 1 The main control and output unit is the central processing and coordination core of the power line fault location and early warning system in this embodiment.

[0173] In terms of physical implementation, the main control and output unit can be a high-performance industrial control computer, a digital signal processor platform, or an embedded system based on a field-programmable gate array and a multi-core processor. The main control and output unit communicates and exchanges data bidirectionally with the data acquisition and processing unit, the early warning module, the adaptive positioning module, and the chaotic fingerprint database module through a high-speed data bus.

[0174] One of the core functions of the main control and output unit is the coordination and management of system status.

[0175] When the system is in normal monitoring mode, the main control and output units run the early warning module by default and continuously. The main control and output units also monitor external hard fault trigger signals in real time. .

[0176] The second core function of the main control and output unit is the triggering and switching of functional modules.

[0177] Warning: When the precursor warning module detects a resonance spectral peak When an early warning signal is output, the main control and output unit receives the signal and immediately executes the early warning output operation.

[0178] Location: When the main control and output unit detects an external hard fault trigger signal When a hard fault occurs, it immediately determines that a hard fault has occurred in the line. At this time, it will immediately suspend or stop the operation of the early warning module and immediately start the adaptive positioning module to execute the complete fault location process from the capture of the full waveform backscatter signal.

[0179] The third core function of this main control and output unit is data interface and storage management.

[0180] The main control and output unit acts as the database manager for the chaotic fingerprint library module:

[0181] Read: When the adaptive positioning module is triggered, the main control and output unit is responsible for high-speed retrieving the offline chaotic fingerprint benchmark library from the non-volatile memory of the chaotic fingerprint library module. and benchmark known boundary features They are then loaded into a cache for high-speed calculation and matching by the adaptive positioning module.

[0182] Write: If the system has online learning or model self-optimization capabilities, the main control and output units can safely write the updated model parameters or fingerprint features back to the chaotic fingerprint library module.

[0183] The fourth core function of this main control and output unit is result output and human-computer interaction:

[0184] Warning output: When a warning signal is received from the early warning module, the main control and output unit converts it into specific alarm information, including driving the local audible and visual alarm, displaying a prompt that the early warning signal exceeds the limit on the local HMI, and sending a warning status code to the host computer or SCADA system via industrial Ethernet.

[0185] Location output: When the final fault location result calculated by the adaptive positioning module is received. At that time, the main control and output unit is responsible for persistently storing and displaying the result. This main control and output unit can also be used on the local HMI to... The fault is clearly displayed at the corresponding location on the line topology map, along with a timestamp of the fault occurrence. Simultaneously, this precise location result is formatted and sent to the remote control system via a communication protocol.

[0186] Event logging: The main control and output unit is also responsible for maintaining a detailed event log with high-precision timestamps to record all early warning events, hard fault triggering events, and location results, providing key data support for subsequent fault analysis and diagnosis.

[0187] This system acquires data at different locations along the line by constructing a high-precision electromagnetic transient simulation model. The reference fault full waveform backscatter signal and the full waveform backscattered signal of the reference known boundary .

[0188] Subsequently, the system applies nonlinear dynamics theory to reconstruct the phase space of these waveforms and extract the correlation dimension. and the maximum Lyapunov index Chaotic eigenvectors Ultimately, a storage location is generated. Compared with the benchmark features Offline chaotic fingerprint benchmark library for mapping relationships and store separately As a correction anchor point.

[0189] During online operation, the system operates in dual-mode. Under normal conditions, the early warning module utilizes the stochastic resonance effect of the bistable system to amplify and detect weak early warning signals against a noisy background. To enable early warning, once a hard fault is detected. The system immediately switches to the adaptive positioning module, which compares the real-time known boundary features. Known boundary features of the benchmark To determine the deviation, a dynamic transformation operator is solved in real time. To compensate for model mismatch caused by environmental changes.

[0190] Finally, the adaptive positioning module uses this operator Applied to the entire benchmark library After obtaining the corrected fingerprint database and through in Matching unknown features in real time This enables high-precision adaptive positioning.

[0191] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A power line fault location and early warning system, characterized in that, The system includes multiple signal acquisition units (1), data acquisition and processing units (2), a chaotic fingerprint database module (3), an early warning module (4), and an adaptive positioning module (5). Multiple signal acquisition units (1) are used to capture multiple broadband transient signals and full-waveform backscattered signals on the power line and transmit them to the data acquisition and processing unit (2). The data acquisition and processing unit (2) is electrically connected to multiple signal acquisition units (1) to convert multiple broadband transient signals and full waveform backscattered signals into digital signal sequences, which are then transmitted to the early warning module (4) and the adaptive positioning module (5), respectively. The chaotic fingerprint library module (3) is used to store an offline chaotic fingerprint benchmark library, which contains known boundary features of the benchmark. The precursor warning module (4) is connected to the data acquisition and processing unit (2) and the chaotic fingerprint library module (3) to detect weak precursor signals from broadband transient digital signal sequences based on the nonlinear dynamic system model and the principle of random resonance under normal monitoring conditions, and output a precursor warning signal when a weak precursor signal is detected. The adaptive positioning module (5) is used to extract chaotic features from the digital signal sequence of full waveform backscattering when a hard fault is detected, and solve the dynamic transformation operator based on the known boundary features of the benchmark stored in the chaotic fingerprint library module (3). The dynamic transformation operator is used to correct the offline chaotic fingerprint benchmark library to generate a corrected fingerprint library, and the fault location is calculated by matching the chaotic features in the corrected fingerprint library.

2. The power line fault location and early warning system according to claim 1, characterized in that, The nonlinear dynamic system model is a bistable system, and its dynamic behavior is described by the Langevin equation.

3. The power line fault location and early warning system according to claim 1, characterized in that, Before extracting the chaotic features, the adaptive positioning module (5) is also used to deconstruct the real-time full-waveform backscatter signal to separate the real-time known boundary full-waveform backscatter signal and the real-time unknown fault full-waveform backscatter signal. The chaotic features include real-time known boundary features and real-time unknown features extracted from real-time known boundary full waveform backscattered signals and real-time unknown fault full waveform backscattered signals, respectively.

4. The power line fault location and early warning system according to claim 3, characterized in that, Extracting chaotic features, the specific process is as follows: Perform phase space reconstruction on the time series to construct chaotic attractors; Furthermore, a quantitative calculation of chaotic features is performed on the chaotic attractor to generate a chaotic feature descriptor including the correlation dimension and the maximum Lyapunov exponent, which serves as the chaotic feature.

5. A power line fault location and early warning system according to claim 4, characterized in that, Before performing the phase space reconstruction, the adaptive positioning module (5) is also used to determine the time delay using the average mutual information method and to determine the embedding dimension using the pseudo-nearest neighbor method.

6. The power line fault location and early warning system according to claim 1, characterized in that, The dynamic transformation operator is modeled as an affine transformation, which includes a transformation matrix and a translation vector. The adaptive positioning module (5) is used to solve for the transformation matrix and the translation vector by minimizing the Euclidean distance between the real-time known boundary features and the reference known boundary features after the affine transformation.

7. A power line fault location and early warning system according to claim 1, characterized in that, The adaptive positioning module (5) is configured to match the chaotic features in the corrected fingerprint database and calculate the fault location. The specific process is as follows: Calculate the Euclidean distance between the real-time unknown feature and the corrected baseline feature corresponding to each entry in the corrected fingerprint database; In addition, the minimum value of the Euclidean distance is found to determine the location of the fault.

8. A power line fault location and early warning system according to claim 1, characterized in that, The offline chaotic fingerprint benchmark library and the benchmark known boundary features are generated by constructing an electromagnetic transient model of a power line and performing offline simulation. The offline simulation generation includes: simulating and acquiring the full waveform backscattered signal of the reference fault and the full waveform backscattered signal of the reference known boundary; Chaotic feature extraction is performed on the backscattered signal of the full waveform of the baseline fault and the backscattered signal of the full waveform of the baseline known boundary obtained from the simulation to obtain the baseline chaotic features and the baseline known boundary features.

9. A power line fault location and early warning system according to claim 1, characterized in that, The signal acquisition unit (1) is one or more of a combination of a high-frequency current transformer, a capacitive coupler, or an optical sensor. The data acquisition and processing unit (2) includes a local processor, which is either a digital signal processor or a field-programmable gate array.

10. A power line fault location and early warning system according to claim 1, characterized in that, The system also includes a main control and output unit. The main control and output unit (6) is used to suspend or stop the operation of the early warning module (4) and start the adaptive positioning module (5) when a hard fault trigger signal is detected. The main control and output unit (6) is a central server or an industrial control computer.

Images