A fault detection method, device and storage medium for overhead lines in a wind farm

Abstract

The present invention discloses a fault detection method, device and storage medium for overhead lines in a wind farm, which relates to the technical field of power systems and includes: S1, intelligent data acquisition; S2, extraction and compressed representation of fault signal features; S3, simulation modeling of fault propagation paths; S4, fault mode recognition based on artificial intelligence; S5, adaptive algorithm for fault location; S6, real-time detection of fault arc features; S7, construction of a spatio-temporal dynamic fault warning model; S8, construction of a multi-dimensional fault self-healing control strategy; S9, fault cause analysis based on a knowledge graph. High-frequency signal features are quickly captured in the early stage of fault occurrence, and through multi-dimensional analysis and control strategies, accurate fault location, prediction and real-time response are achieved; at the same time, the combination of fuzzy logic and expert system enables the system to dynamically adjust strategies when facing various fault types, optimize the line operation mode, adjust the load distribution, and start the standby power supply.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and specifically to a fault detection method, device, and storage medium for overhead lines in a wind farm. Background Art

[0002] With the continuous expansion of the scale of wind farms, overhead lines have become an important infrastructure connecting wind farms and power grids. However, due to harsh environmental conditions (such as strong winds, lightning, rain, snow, etc.) and complex factors within the power system (such as voltage fluctuations, short circuits, etc.) during the operation of overhead lines in wind farms, faults are extremely likely to occur. Therefore, how to detect and locate these faults in a timely and accurate manner has become a key challenge for ensuring the stable operation of wind farms.

[0003] Traditional fault detection technologies mostly rely on simple voltage and current monitoring and relay protection devices. These methods usually can only respond passively after a fault occurs through current mutations or changes in line impedance, and it is difficult to quickly and accurately locate the fault. Especially in the face of complex multi-point faults and high-noise environments, traditional technologies appear slow to respond and inefficient. In addition, existing detection technologies lack effective intelligent support in aspects such as fault type identification, fault propagation analysis, and fault self-healing control. The following are the deficiencies of traditional fault detection technologies in each key process:

[0004] Fault signal feature extraction: Existing technologies mostly adopt single signal processing methods, which are difficult to cope with the complexity and diversity of fault signals in overhead lines of wind farms. Multi-scale signal analysis methods are rarely seen in traditional applications. Even if adopted, their feature extraction effects are not ideal in high-noise backgrounds.

[0005] Fault propagation path simulation: Currently, the simulation of fault propagation is mostly based on linear models or simplified electromagnetic field calculations, ignoring the influence of environmental factors such as wind speed and temperature on the fault propagation path, resulting in low prediction accuracy. In addition, most models have poor adaptability to dynamic environments and are difficult to perform accurate fault path analysis under changing natural conditions.

[0006] Fault arc detection: Due to frequency range limitations, traditional detection methods mostly adopt low-frequency or medium-frequency signal processing technologies, making it difficult to capture the ultra-high frequency (UHF) signals generated during arc discharge. This results in the difficulty of identifying arc signals in the early stage, thus delaying the emergency response.

[0007] Multi-dimensional self-healing control strategy: In most existing technologies, the control strategies after a fault occurs mostly adopt preset single response modes, lacking flexibility and being difficult to dynamically adjust according to the severity of the fault and the system state. At the same time, the collaborative optimization of multi-dimensional control (such as load adjustment, standby power startup) in existing systems is insufficient, resulting in a slow self-healing speed and limited effect of the power grid. Summary of the Invention

[0008] Aiming at the deficiencies of the existing technology, the present invention provides a fault detection method, device and storage medium for overhead lines in a wind farm to solve the problems proposed in the above background technology.

[0009] To achieve the above object, the present invention provides the following technical solutions: A fault detection method for overhead lines in a wind farm, comprising the following steps:

[0010] S1. Intelligent data acquisition;

[0011] Based on sensors, conduct all-round monitoring and acquisition of data on the overhead lines of the wind farm, and perform fusion of multi-source data on the acquired data;

[0012] S2. Fault signal feature extraction and compressed representation;

[0013] Using signal processing technologies of wavelet transform and Fourier transform, extract fault feature signals from multi-source data, perform multi-scale analysis on the acquired time series data, separate features in different frequency bands, and compress high-dimensional features into a low-dimensional space through principal component analysis (PCA) or sparse coding methods, retaining the most important fault information;

[0014] S3. Simulation modeling of fault propagation paths;

[0015] Construct a fault propagation model of the overhead lines in the wind farm through electromagnetic field simulation technology, simulate the propagation path of fault signals in the power lines, and calculate the transmission characteristics of fault signals in different line segments based on actual line parameters such as resistance, inductance, capacitance and environmental conditions such as wind speed and temperature using the finite element analysis method;

[0016] S4. Fault mode recognition based on artificial intelligence;

[0017] Use a deep learning model to automatically identify fault modes, perform supervised learning through a large-scale training data set, and automatically identify common fault types such as short circuits, disconnections, and insulation breakdowns;

[0018] S5. Adaptive algorithm for fault location;

[0019] Adopt the impedance method combined with the particle swarm optimization algorithm (PSO) to accurately locate the fault point, calculate the preliminary fault location through the impedance method, and use the PSO algorithm for optimized search to find the fault point coordinates with the smallest error;

[0020] S6. Real-time detection of fault arc characteristics;

[0021] During the detection process, the characteristics of the fault arc are detected in real time using ultra-high frequency (UHF) signals, and the key characteristics of the arc signal are further extracted and amplified through an intelligent filter and a signal enhancement algorithm;

[0022] S7. Construct a spatio-temporal dynamic fault warning model;

[0023] Based on the results of pattern recognition, construct a spatio-temporal dynamic fault warning model. Using geographic information system (GIS) technology, map the fault information onto the physical geographical location of the wind farm in real time. Through spatio-temporal data modeling and analysis, combined with historical data and current real-time monitoring data, predict the possible development trend and impact range of the fault;

[0024] S8. Construct a multi-dimensional fault self-healing control strategy;

[0025] After detecting a fault and completing the preliminary location, automatically activate the multi-dimensional fault self-healing control strategy, including operations such as switching the line operation mode, adjusting the load distribution, and starting the standby power supply. Utilize fuzzy logic control and expert system technology to make the optimal decision instantly to minimize the impact of the fault on the entire power grid;

[0026] S9. Fault cause analysis based on the knowledge graph;

[0027] By establishing the relationships between the data nodes of the wind farm equipment, lines, and environmental factors, use graph algorithms to mine the root causes that may lead to faults. The knowledge graph helps to identify the complex causal relationships hidden in the data, enabling the system to conduct a more accurate root cause analysis of the faults.

[0028] To further optimize this technical solution, in step S1, the sensors include sensor types such as current, voltage, temperature, vibration, and sound wave. Based on multi-source sensors, multiple physical quantities are simultaneously collected to obtain all the data of the overhead lines in the wind farm;

[0029] Through high-precision time synchronization technology, integrate all the data under a unified time reference, and use data fusion algorithms such as Kalman filtering and multi-scale clustering to preliminarily process the data of different sensors to enhance the accuracy and reliability of the fault signal.

[0030] To further optimize this technical solution, when performing multi-scale analysis on the collected time series data in step S2, construct a multi-scale signal decomposition model. By combining the non-linear characteristics and spectral information of the time series signal, conduct in-depth analysis and decomposition of the fault signal. The multi-scale signal decomposition model includes a signal decomposition formula, a kernel function, a multi-scale spectral decomposition formula, and an adaptive scale optimization mechanism.

[0031] To further optimize this technical solution, in the multi-scale signal decomposition model:

[0032] Signal decomposition formula. Assume the original signal is a non-linear time series signal containing multiple frequency components. To decompose this signal, a non-linear kernel function is introduced, and the formula is as follows:

[0033] ;

[0034] where,

[0035] represents the signal component after decomposition at scale ;

[0036] is a scale-related kernel function used to perform convolution operations on the signal at a specific scale;

[0037] represents the multi-scale parameter used to control the width of the kernel function. As increases, the kernel function will capture the low-frequency characteristics of the signal;

[0038] For the kernel function, a non-linear Gaussian-Lorentz composite kernel function is selected to enhance the ability to extract non-linear characteristics of the signal. The kernel function is as follows:

[0039] ;

[0040] where,

[0041] controls the attenuation rate of the Gaussian part and is used to capture the high-frequency part of the signal;

[0042] controls the peak shape of the Lorentz part and is used to extract low-frequency characteristics;

[0043] By combining the advantages of the Gaussian and Lorentz distributions, it can effectively capture the high-frequency and low-frequency characteristics of the signal at different scales and better adapt to the non-linear changes of the fault signal.

[0044] To further optimize this technical solution, in the multi-scale signal decomposition model:

[0045] The multi-scale spectral decomposition formula is used to perform multi-scale transformation on the original signal to obtain a series of decomposed signals at different scales . The signal at each scale is expressed as:

[0046] ;

[0047] In this process, as the scale With the increase of , the kernel function gradually becomes wider, separating the low-frequency components step by step;

[0048] An adaptive scale optimization mechanism, based on which the problem of scale selection in different fault scenarios is solved, and the multi-scale energy ratio of the signal is defined as:

[0049] ;

[0050] By maximizing , the system can automatically select the optimal scale to extract fault features, and the optimal scale satisfies the following conditions:

[0051] ;

[0052] At the selected scale , the energy of the signal is most concentrated and the fault features are most obvious.

[0053] To further optimize this technical solution, when the multi-scale signal decomposition model is used, it includes the following specific processes:

[0054] Initial signal analysis: First, perform a preliminary analysis on the collected original fault signal to judge the overall frequency distribution of the signal, and select a suitable multi-scale parameter range by observing the signal characteristics for subsequent non-linear multi-scale spectral decomposition;

[0055] Gradual execution of multi-scale decomposition: Use the model to decompose the signal step by step, and analyze the signal components at different scales to extract the high-frequency features used to reflect the transient disturbance of the line and the low-frequency features representing the persistent influence of the fault respectively;

[0056] Adaptive selection of the optimal scale: During the decomposition process, calculate the multi-scale energy ratio at each scale in real time, and automatically adjust the scale according to the formula to make the decomposition proceed at the optimal scale , ensuring that no matter how the fault type changes, the model extracts fault features at a suitable scale;

[0057] Compression representation and feature retention: Once the signal is decomposed at the optimal scale, perform compression representation on the decomposed data through principal component analysis (PCA). PCA reduces the high-dimensional multi-scale signal data to low-dimensional while retaining its main fault feature information, which is used to reduce the data dimension and improve the computational efficiency of subsequent fault mode recognition and diagnosis.

[0058] To further optimize this technical solution, in step S3, line parameters, environmental factors, and non-linear losses are introduced into the fault propagation model;

[0059] Line parameters, considering the distributed parameters of overhead lines, such as resistance , inductance , capacitance and admittance produce different responses at different frequencies;

[0060] Environmental factors, such as wind speed , temperature , humidity have a dynamic impact on signal propagation;

[0061] Non-linear losses, the amplitude and frequency characteristics of fault signals are affected by both the electromagnetic characteristics of the line and environmental conditions;

[0062] The fault propagation model is as follows:

[0063] ;

[0064] Among them,

[0065] is the voltage distribution at position and time ;

[0066] is the propagation speed, which is the result of the combined action of medium parameters, electromagnetic constants, and environmental factors;

[0067] is the introduced non-linear loss term, referring to the attenuation of the signal by the line under different environmental conditions; the function expresses the complex relationship between loss and voltage , current and environmental parameters , such as wind speed and temperature.

[0068] To further optimize this technical solution, in step S8, when constructing the multi-dimensional fault self-healing control strategy, it includes the following specific processes:

[0069] I. Construction of a fuzzy logic controller;

[0070] Based on fuzzy logic theory, dealing with the uncertainty and fuzzy information in the input signal, including the following control dimensions:

[0071] Line operation mode switching control, after detecting a fault, switching the operation mode of the line, such as from parallel operation to series operation, to ensure the stability of the system;

[0072] Load distribution adjustment control, which adjusts the load distribution on each line to avoid the overload problem in the faulty section and balance the overall system load;

[0073] Standby power supply startup control: Automatically starts the standby power supply when necessary to maintain the continuity of power supply and reduce the impact of power outage areas;

[0074] Second, use the Mamdani fuzzy inference model for the fuzzy inference process;

[0075] Third, the integration and optimization of the expert system, which uses a pre-set knowledge base and inference engine to provide optimized self-healing strategies in complex fault scenarios;

[0076] Fourth, a multi-dimensional collaborative control strategy;

[0077] Multi-dimensional fault self-healing operations will be executed simultaneously to achieve all-round power grid restoration;

[0078] Fifth, a real-time dynamic adjustment and feedback mechanism;

[0079] Status monitoring and feedback: Real-time monitor the changes in the system status and feedback the monitoring data to the fuzzy controller and the expert system to update the inference parameters;

[0080] Adaptive control optimization: Automatically optimize the fuzzy rules and the expert system strategy according to the feedback data to improve the accuracy and response speed of the self-healing control.

[0081] A fault detection device for an overhead line in a wind farm, the fault detection device includes a memory, a processor, and a fault detection program stored on the memory and executable on the processor. When the fault detection program is executed by the processor, it implements the steps of the fault detection method for the overhead line in the wind farm as described above.

[0082] A computer-readable storage medium, on which a fault detection program is stored. When the fault detection program is executed by a processor, it implements the steps of the fault detection method for the overhead line in the wind farm as described above.

[0083] Compared with the prior art, the present invention provides a fault detection method, device, and storage medium for an overhead line in a wind farm, and has the following beneficial effects:

[0084] The fault detection method, device, and storage medium for the overhead line in the wind farm can quickly capture the high-frequency signal characteristics in the early stage of the fault occurrence, and through multi-dimensional analysis and control strategies, achieve precise fault location, prediction, and real-time response;

[0085] Meanwhile, the combination of fuzzy logic and expert system enables the system to dynamically adjust strategies when facing multiple types of faults, optimize the line operation mode, adjust the load distribution, and start the standby power supply. This not only significantly improves the fault detection and response speed but also enhances the intelligence of fault self-healing and the reliability of power grid operation, effectively reducing the impact of faults on the stability of the entire power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a schematic flow chart of a fault detection method for overhead lines in a wind farm proposed by the present invention;

[0087] Figure 2 It is a schematic diagram of the composition and process of a multi-scale signal decomposition model in a fault detection method for overhead lines in a wind farm proposed by the present invention;

[0088] Figure 3 It is a schematic diagram of the formula for the influence of environmental factors on the propagation speed in the fault propagation model in a fault detection method for overhead lines in a wind farm proposed by the present invention;

[0089] Figure 4 It is a schematic diagram of the non-linear loss model in the fault propagation model in a fault detection method for overhead lines in a wind farm proposed by the present invention;

[0090] Figure 5 It is a schematic flow chart of the fault propagation model in a fault detection method for overhead lines in a wind farm proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0091] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0092] Embodiment 1

[0093] Please refer to Figure 1 , a fault detection method for overhead lines in a wind farm, including the following steps:

[0094] S1. Intelligent data acquisition

[0095] Based on sensors, conduct comprehensive monitoring and acquisition of data on the overhead lines of the wind farm, and perform fusion of multi-source data on the acquired data.

[0096] In this embodiment, in step S1, the sensors include sensor types such as current, voltage, temperature, vibration, and sound wave. Multiple physical quantities are simultaneously collected based on multi-source sensors to obtain all data of the overhead lines in the wind farm.

[0097] Through high-precision time synchronization technology, all data is integrated under a unified time reference. Data fusion algorithms such as Kalman filtering and multi-scale clustering are used to preliminarily process different sensor data to enhance the accuracy and reliability of fault signals.

[0098] S2. Fault signal feature extraction and compressed representation

[0099] Using signal processing techniques of wavelet transform and Fourier transform, fault feature signals are extracted from multi-source data. Multi-scale analysis is performed on the collected time series data to separate features in different frequency bands. The high-dimensional features are compressed into a low-dimensional space through methods such as principal component analysis (PCA) or sparse coding to retain the most important fault information.

[0100] As Figure 2 shown, in this embodiment, in step S2, when performing multi-scale analysis on the collected time series data, a multi-scale signal decomposition model is constructed. By combining the non-linear characteristics and spectral information of the time series signal, in-depth analysis and decomposition of the fault signal are carried out. The multi-scale signal decomposition model includes a signal decomposition formula, a kernel function, a multi-scale spectral decomposition formula, and an adaptive scale optimization mechanism.

[0101] Furthermore, in the multi-scale signal decomposition model:

[0102] Signal decomposition formula. Assume the original signal is a non-linear time series signal containing multiple frequency components. To decompose this signal, a non-linear kernel function is introduced, and the formula is as follows:

[0103] ;

[0104] where

[0105] represents the signal component decomposed at scale ;

[0106] is a scale-related kernel function used to perform convolution operations on the signal at a specific scale;

[0107] represents the multi-scale parameter used to control the width of the kernel function. As increases, the kernel function will capture the low-frequency features of the signal.

[0108] Kernel function. The non - linear Gaussian - Lorentz composite kernel function is selected to enhance the ability to extract non - linear features of the signal. The kernel function is as follows:

[0109] ;

[0110] Among them,

[0111] Controls the attenuation rate of the Gaussian part and is used to capture the high - frequency part of the signal;

[0112] Controls the peak shape of the Lorentz part and is used to extract low - frequency features;

[0113] By combining the advantages of the Gaussian and Lorentz distributions, it can effectively capture the high - frequency and low - frequency characteristics of the signal at different scales and better adapt to the non - linear changes of the fault signal.

[0114] Multi - scale spectral decomposition formula. Perform multi - scale transformation on the original signal to obtain a series of decomposed signals at different scales , analyze the signal at different scales, and the signal at each scale is expressed as:

[0115] ;

[0116] In this process, as the scale increases, the kernel function gradually becomes wider, so that the low - frequency components are gradually separated.

[0117] Adaptive scale optimization mechanism. Based on this mechanism, solve the problem of scale selection under different fault scenarios. Define the multi - scale energy ratio of the signal as:

[0118] ;

[0119] By maximizing , the system can automatically select the optimal scale to extract fault features. The optimal scale satisfies the following conditions:

[0120] ;

[0121] At the selected scale , the energy of the signal is the most concentrated and the fault features are the most obvious.

[0122] When the multi - scale signal decomposition model is used, it includes the following specific processes:

[0123] Initial signal analysis: First, analyze the collected original fault signal Perform a preliminary analysis to determine the overall frequency distribution of the signal. By observing the signal characteristics, select a suitable range of multi-scale parameters for subsequent non-linear multi-scale spectral decomposition;

[0124] Gradual execution of multi-scale decomposition: Use the model to decompose the signal gradually. By analyzing the signal components at different scales of the signal extract the high-frequency characteristics used to reflect the transient disturbances of the line and the low-frequency characteristics representing the persistent effects of faults respectively;

[0125] Adaptive selection of the optimal scale: During the decomposition process, calculate the multi-scale energy ratio at each scale in real time and automatically adjust the scale according to the formula so that the decomposition is carried out at the optimal scale to ensure that regardless of how the fault type changes, the model extracts fault characteristics at an appropriate scale;

[0126] Compression representation and feature retention: Once the signal is decomposed at the optimal scale, perform a compression representation of the decomposed data through principal component analysis (PCA). PCA reduces the high-dimensional multi-scale signal data to a low dimension while retaining its main fault feature information, which is used to reduce the data dimension and improve the computational efficiency of subsequent fault mode recognition and diagnosis.

[0127] By introducing an adaptive kernel function and an energy ratio optimization mechanism, the feature extraction ability of the fault signals of the overhead lines in the wind farm is effectively improved. This model can not only accurately decompose complex fault signals, but also automatically adjust the analysis scale according to different fault scenarios, so that the fault characteristics are fully reflected in the multi-scale space.

[0128] S3. Simulation modeling of the fault propagation path

[0129] Construct a fault propagation model of the overhead lines in the wind farm through electromagnetic field simulation technology to simulate the propagation path of the fault signals in the power lines. Based on the actual line parameters such as resistance, inductance, capacitance and environmental conditions such as wind speed, temperature, use the finite element analysis method to calculate the transmission characteristics of the fault signals in different line segments.

[0130] In this embodiment, in step S3, line parameters, environmental factors and non-linear losses are introduced into the fault propagation model;

[0131] Line parameters, considering the distributed parameters of the overhead lines, such as resistance and inductance and capacitance and admittance produce different responses at different frequencies;

[0132] Environmental factors, such as wind speed , Temperature , Humidity Dynamic influence on signal propagation;

[0133] Nonlinear loss, the amplitude and frequency characteristics of the fault signal are affected by both the electromagnetic characteristics of the line and environmental conditions;

[0134] The fault propagation model is as follows:

[0135] ;

[0136] Wherein,

[0137] is the voltage distribution at position and time ;

[0138] is the propagation speed, which is the result of the combined action of medium parameters, electromagnetic constants and environmental factors;

[0139] is the introduced nonlinear loss term, referring to the attenuation of the signal by the line under different environmental conditions; the function expresses the complex relationship between the loss and the voltage , current and environmental parameters , such as wind speed, temperature.

[0140] Propagation speed is significantly affected by external environmental conditions, and its relationship with wind speed , temperature and humidity is as shown in Figure 3 . The formula describes how the signal propagation speed is dynamically adjusted as the environmental conditions change. For example, a higher wind speed will accelerate the cooling effect, thus affecting the conductivity of the medium; while an increase in temperature may increase the conductivity and slow down the signal propagation.

[0141] To more accurately describe the loss suffered by the fault signal during propagation, a nonlinear loss model is constructed as shown in Figure 4 . The model shows that the loss intensity of the fault signal depends not only on the amplitude and frequency characteristics of the signal itself, but also on the regulation of external environmental factors. For example, higher humidity will reduce the ionization rate and reduce the loss, while an increase in temperature may accelerate the increase in the resistance of the line and increase the loss.

[0142] As shown in Figure 5 , when using this model:

[0143] Initialize the model parameters:

[0144] During the use of the model, first, according to the actual operation data and environmental conditions of the wind farm, the initial values of line parameters (resistance , inductance , capacitance and admittance ) and environmental impact factors (such as wind speed , temperature , humidity ) are initialized. These parameters will be dynamically adjusted in subsequent simulations.

[0145] Fault signal propagation simulation:

[0146] By solving the fault signal propagation equation, the dynamic propagation process of the fault signal in the overhead line is simulated. The finite element analysis (FEA) technology is used to discretize the equation to obtain the voltage and current distribution at discrete points, thereby simulating the propagation characteristics and speed of the fault signal in different line segments.

[0147] Influence of environmental conditions on the propagation path:

[0148] During the simulation process, the influence of environmental factors (such as wind speed, temperature, humidity) on the propagation speed and nonlinear loss is updated in real time. Through this dynamic adjustment, the propagation path of the fault signal under different environments and its changes can be simulated, so as to more accurately predict the location of the fault point.

[0149] Precise location of the fault:

[0150] Using the voltage and current distribution characteristics in the simulation results, the particle swarm optimization algorithm (PSO) can be further combined for the optimal location of the fault point. By adjusting the model parameters, the error between the simulation results and the measurement results of the actual fault signal is minimized, and finally the exact location of the fault point is determined.

[0151] S4. Fault mode recognition based on artificial intelligence

[0152] A deep learning model is used for automatic recognition of fault modes. Through supervised learning with a large-scale training dataset, common fault types such as short circuits, disconnections, and insulation damage are automatically recognized.

[0153] In this embodiment, through an ensemble learning method, such as the Adaboost algorithm, the prediction results of multiple neural network models are fused to improve the accuracy of fault classification. This can not only improve the robustness of detection but also effectively reduce the misjudgment rate caused by environmental interference.

[0154] S5. Adaptive Algorithm for Fault Location

[0155] The impedance method combined with the Particle Swarm Optimization (PSO) algorithm is used to accurately locate the fault point. The preliminary fault location is calculated by the impedance method, and the PSO algorithm is used for optimized search to find the fault point coordinates with the minimum error.

[0156] In this embodiment, the advantage of the particle swarm algorithm is that it can quickly converge to the global optimal solution and maintain high accuracy even under non-ideal conditions (such as noise interference, line non-linearity, etc.). Combining the classical algorithm with the optimization technique significantly improves the speed and accuracy of fault location.

[0157] S6. Real-time Detection of Fault Arc Characteristics

[0158] During the detection process, the characteristics of the fault arc are detected in real time using Ultra-High Frequency (UHF) signals, and the key features of the arc signal are further extracted and amplified through an intelligent filter and a signal enhancement algorithm.

[0159] In this embodiment, an adaptive intelligent filter is designed to dynamically adjust the signal enhancement strategy:

[0160] Adaptive filter structure: The filter is designed using an adaptive algorithm (such as the LMS algorithm or the RLS algorithm), which can dynamically adjust the filtering parameters according to the characteristics of the input signal to achieve the suppression of different frequency noises.

[0161] Nonlinear signal enhancement: Through nonlinear processing techniques (such as amplitude modulation or envelope detection), the key features of the arc signal are further amplified, making the signal more prominent in the background noise.

[0162] S7. Construction of a Spatiotemporal Dynamic Fault Warning Model

[0163] Based on the results of pattern recognition, a spatiotemporal dynamic fault warning model is constructed. Using Geographic Information System (GIS) technology, the fault information is mapped in real time to the physical geographical location of the wind farm. Through spatiotemporal data modeling and analysis, combined with historical data and current real-time monitoring data, the possible development trends and influence ranges of faults are predicted.

[0164] S8. Construction of a Multi-dimensional Fault Self-healing Control Strategy

[0165] After detecting a fault and completing the preliminary location, the multi-dimensional fault self-healing control strategy is automatically activated, including operations such as switching the line operation mode, adjusting the load distribution, and starting the standby power supply. Using fuzzy logic control and expert system technology, the optimal decision is made instantly to minimize the impact of the fault on the entire power grid.

[0166] In this embodiment, in step S8, when constructing the multi-dimensional fault self-healing control strategy, the following specific processes are included:

[0167] I. Construction of a fuzzy logic controller;

[0168] Based on fuzzy logic theory, deal with the uncertainty and fuzzy information in the input signal, including the following control dimensions:

[0169] Line operation mode switching control: After detecting a fault, switch the line operation mode, such as switching from parallel operation to series operation, to ensure the stability of the system;

[0170] Load distribution adjustment control: Adjust the load distribution on each line to avoid overloading of the fault section and balance the overall system load;

[0171] Standby power supply startup control: Automatically start the standby power supply when necessary to maintain the continuity of power supply and reduce the impact of the power outage area.

[0172] Construct a fuzzy rule table based on input parameters (such as fault severity, line voltage fluctuation, load condition). The example is as follows:

[0173] Input variables: Fault severity (slight, medium, severe), voltage fluctuation amplitude (small, medium, large), load level (low, medium, high).

[0174] Output control: Line switching (maintain, partial switching, full switching), load adjustment (no adjustment, partial adjustment, full adjustment), standby power supply status (off, standby, start).

[0175] II. Use the Mamdani fuzzy inference model for the fuzzy inference process;

[0176] III. Integration and optimization of the expert system. Utilize the pre-set knowledge base and inference engine to provide optimized self-healing strategies in complex fault scenarios, including:

[0177] Knowledge base construction: Based on historical fault data, expert experience, and wind farm operation specifications, establish a detailed knowledge base, including various fault types, possible countermeasures, and their priorities.

[0178] Inference engine: When a fault occurs, the inference engine will derive the optimal self-healing strategy according to the current fault information and the rules in the knowledge base. The inference engine adopts a combination of forward reasoning and backward reasoning to draw conclusions at the fastest speed.

[0179] IV. Multi-dimensional collaborative control strategy;

[0180] Multi-dimensional fault self-healing operations will be executed simultaneously to achieve all-round power grid restoration;

[0181] V. Real-time dynamic adjustment and feedback mechanism;

[0182] Status monitoring and feedback: Real-time monitor the changes in the system status, and feed back the monitoring data to the fuzzy controller and the expert system to update the inference parameters;

[0183] Adaptive control optimization: Automatically optimize the fuzzy rules and the expert system strategy according to the feedback data to improve the accuracy and response speed of the self-healing control.

[0184] S9. Fault cause analysis based on knowledge graph

[0185] By establishing the relationships between the data nodes of the wind farm equipment, lines, and environmental factors, and using graph algorithms to mine the root causes that may lead to faults, the knowledge graph helps to identify the complex causal relationships hidden in the data, enabling the system to perform more accurate root cause analysis of faults.

[0186] In this embodiment, a fault detection device for an overhead line of a wind farm, the fault detection device includes a memory, a processor, and a fault detection program stored on the memory and executable on the processor. When the fault detection program is executed by the processor, it implements the steps of the fault detection method for the overhead line of the wind farm as described above.

[0187] In this embodiment, a computer-readable storage medium stores a fault detection program. When the fault detection program is executed by a processor, it implements the steps of the fault detection method for the overhead line of the wind farm as described above.

[0188] The beneficial effects of the present invention are:

[0189] The fault detection method, device, and storage medium for the overhead line of the wind farm can quickly capture the high-frequency signal characteristics in the early stage of the fault, and through multi-dimensional analysis and control strategies, achieve precise fault location, prediction, and real-time response;

[0190] At the same time, the combination of fuzzy logic and the expert system enables the system to dynamically adjust the strategy when facing various fault types, optimize the line operation mode, adjust the load distribution, and start the standby power supply, which not only significantly improves the fault detection and response speed, but also improves the intelligent level of fault self-healing and the reliability of the power grid operation, and can effectively reduce the impact of faults on the stability of the entire power grid.

[0191] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0192] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting faults in overhead lines of a wind farm, characterized in that: The following steps are involved: S1, intelligent data collection; Conduct all-round monitoring and data collection of wind farm overhead lines based on sensors, and integrate multi-source data collected; S2, fault signal feature extraction and compression representation; Using wavelet transform and Fourier transform signal processing technology, fault feature signals are extracted from multi-source data, and multi-scale analysis is performed on the collected time series data to separate features in different frequency bands. High-dimensional features are compressed into low-dimensional space through principal component analysis (PCA) or sparse coding methods to retain the most important fault information. When performing multi-scale analysis on the collected time series data, a multi-scale signal decomposition model is constructed. By combining the nonlinear characteristics and spectrum information of the time series signal, the fault signal is deeply analyzed and decomposed. The multi-scale signal decomposition model includes a signal decomposition formula, a kernel function, a multi-scale spectrum decomposition formula, and an adaptive scale optimization mechanism; S3, simulation modeling of fault propagation path; The fault propagation model of the wind farm overhead line is constructed through electromagnetic field simulation technology to simulate the propagation path of the fault signal in the power line; Line parameters, environmental factors and nonlinear losses are introduced into the fault propagation model; Line parameters, considering the distributed parameters of overhead lines, including resistance ,inductance ,capacitance and admittance Produces different responses at different frequencies; Environmental factors, including wind speed ,temperature ,humidity Dynamic impact on signal propagation; Nonlinear loss: the amplitude and frequency characteristics of the fault signal are affected by both the electromagnetic characteristics of the line and the environmental conditions; The equation of the fault propagation model is as follows: ； in, is in position and time Voltage distribution at is the propagation speed, which is the result of the interaction of medium parameters, electromagnetic constants and environmental factors; is the nonlinear loss term introduced, which refers to the attenuation of the signal under different environmental conditions; function Expressing the loss and voltage , Current And environmental parameters , including the complex relationship between wind speed and temperature; By solving the equations of the fault propagation model, the dynamic propagation process of the fault signal in the overhead line is simulated. The equations are discretized using the finite element analysis (FEA) technology to obtain the voltage and current distribution at discrete points, thereby simulating the propagation characteristics and speed of the fault signal in different line sections. S4, Fault pattern recognition based on artificial intelligence; Use deep learning models to automatically identify fault modes, and use large-scale training data sets for supervised learning to automatically identify common fault types, including short circuits, broken wires, and insulation damage; S5, adaptive algorithm for fault location; The impedance method combined with the particle swarm optimization algorithm (PSO) is used to accurately locate the fault point. The impedance method is used to calculate the preliminary fault location, and the PSO algorithm is used to perform optimization search to find the coordinates of the fault point with the smallest error; S6. Real-time detection of fault arc characteristics; During the detection process, ultra-high frequency (UHF) signals are used to detect the characteristics of the fault arc in real time, and the key features of the arc signal are further extracted and amplified through intelligent filters and signal enhancement algorithms; S7, constructing a spatiotemporal dynamic fault warning model; Based on the results of pattern recognition, a spatiotemporal dynamic fault warning model is constructed. Geographic Information System (GIS) technology is used to map fault information to the physical geographical location of the wind farm in real time. Through spatiotemporal data modeling and analysis, combined with historical data and current real-time monitoring data, the possible development trend and impact range of the fault are predicted; S8. Build a multi-dimensional fault self-healing control strategy; After the fault is detected and initially located, the multi-dimensional fault self-healing control strategy is automatically activated, including switching line operation mode, adjusting load distribution, and starting backup power supply operations. Fuzzy logic control and expert system technology are used to make the best decision in an instant to minimize the impact of the fault on the entire power grid. S9. Fault cause analysis based on knowledge graph; By establishing the relationship between the data nodes of wind farm equipment, lines, and environmental factors, and using graph algorithms to mine the root causes that may lead to failures, the knowledge graph helps identify the complex causal relationships hidden in the data, allowing the system to make a more accurate root cause analysis of the failure.

2. A method for detecting faults in overhead lines of a wind farm according to claim 1, characterized in that: In step S1, the sensors include current, voltage, temperature, vibration, and sound wave sensors, and multiple physical quantities are collected simultaneously based on multi-source sensors to obtain all data of the wind farm overhead lines; Through high-precision time synchronization technology, all data are integrated into a unified time base, and data fusion algorithms, including Kalman filtering and multi-scale clustering, are used to perform preliminary processing on different sensor data to enhance the accuracy and reliability of fault signals.

3. A method for detecting faults in overhead lines of a wind farm according to claim 1, characterized in that: In the multi-scale signal decomposition model: Signal decomposition formula, assuming that the original signal It is a nonlinear time series signal containing multiple frequency components. In order to decompose the signal, a nonlinear kernel function is introduced. The formula is as follows: ； in, Indicated in scale The signal components after decomposition; It is a scale-dependent kernel function, which is used to perform convolution operations on signals at a specific scale; Represents a multi-scale parameter, which is used to control the width of the kernel function. As it increases, the kernel function will capture the low-frequency characteristics of the signal; Kernel function, select the nonlinear Gauss-Lorentz composite kernel function to enhance the nonlinear feature extraction capability of the signal. The kernel function is as follows: ； in, Control the decay speed of the Gaussian part to capture the high-frequency part of the signal; Control the peak shape of the Lorentz part to extract low-frequency features; By combining the advantages of Gaussian and Lorentz distributions, the high-frequency and low-frequency characteristics of the signal can be effectively captured at different scales, and the nonlinear changes of the fault signal can be better adapted.

4. A method for detecting faults in overhead lines of a wind farm according to claim 3, characterized in that: In the multi-scale signal decomposition model: Multi-scale spectral decomposition formula, for the original signal Perform multi-scale transformation to obtain a series of decomposition signals of different scales , the signal is analyzed at different scales, each scale The signal is represented as: ； In this process, as the scale As increases, the kernel function gradually becomes wider, so that the low-frequency components are gradually separated; Adaptive scale optimization mechanism, based on which the scale selection problem in different fault scenarios is solved and the multi-scale energy ratio of the signal is defined for: ； By maximizing , the system can automatically select the optimal scale To extract fault features, the optimal scale The following conditions are met: ； At the selected scale The signal energy is most concentrated and the fault characteristics are most obvious.

5. A method for detecting faults in overhead lines of a wind farm according to claim 4, characterized in that: When the multi-scale signal decomposition model is used, it includes the following specific processes: Initial signal analysis: First, the collected original signal Conduct preliminary analysis to determine the overall frequency distribution of the signal, and select the appropriate multi-scale parameter range by observing the signal characteristics for subsequent nonlinear multi-scale spectral decomposition; Multi-scale decomposition is performed step by step: The signal is decomposed step by step using a multi-scale signal decomposition model. The signal components Through analysis, high-frequency features reflecting transient disturbances of the line and low-frequency features representing the impact of fault persistence are extracted respectively; Adaptively select the optimal scale: During the decomposition process, the multi-scale energy ratio at each scale is calculated in real time , and automatically adjust the scale according to the optimal scale formula so that the decomposition is at the optimal scale This ensures that no matter how the fault type changes, the multi-scale signal decomposition model can extract fault features at an appropriate scale; Compressed representation and feature retention: Once the signal is decomposed at the optimal scale, the decomposed data is compressed and represented through principal component analysis (PCA). PCA reduces the high-dimensional multi-scale signal data to a low dimension while retaining its main fault feature information, which is used to reduce the data dimension and improve the computational efficiency of subsequent fault mode recognition and diagnosis.

6. A method for detecting faults in overhead lines of a wind farm according to claim 1, characterized in that: In step S8, when constructing the multi-dimensional fault self-healing control strategy, the following specific processes are included:

1. Construction of fuzzy logic controller; Based on fuzzy logic theory, it processes uncertainty and fuzzy information in the input signal, including the following control dimensions: Line operation mode switching control, after detecting a fault, switching the line operation mode; Load distribution adjustment control, adjust the load distribution on each line to avoid overload problems in the fault section to balance the overall system load; Backup power supply start-up control: Automatically start the backup power supply when necessary to maintain the continuity of power supply and reduce the impact of power outage areas; Second, the fuzzy reasoning process is carried out using the Mamdani fuzzy reasoning model; 3. Integration and optimization of expert systems, using pre-set knowledge bases and reasoning engines to provide optimized self-healing strategies in complex fault scenarios; 4. Multi-dimensional collaborative control strategy; Multi-dimensional fault self-healing operations will be performed simultaneously to achieve all-round grid restoration; 5. Real-time dynamic adjustment and feedback mechanism; State monitoring and feedback: monitor system state changes in real time, and feed back monitoring data to the fuzzy controller and expert system to update reasoning parameters; Adaptive control optimization: Automatically optimize fuzzy rules and expert system strategies based on feedback data to improve the accuracy and response speed of self-healing control.

7. A fault detection device for overhead lines in a wind farm, characterized in that: The fault detection device includes a memory, a processor, and a fault detection program stored in the memory and executable on the processor. When the fault detection program is executed by the processor, the steps of the wind farm overhead line fault detection method according to any one of claims 1 to 6 are implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a fault detection program, and when the fault detection program is executed by the processor, the steps of the method for detecting a fault of an overhead line in a wind farm are implemented as described in any one of claims 1 to 6.

Images