Power distribution network line fault monitoring and early warning method

Abstract

The invention provides a power distribution network line fault monitoring and early warning method, which comprises the following steps of: acquiring line vibration data through a sensor array, analyzing vibration frequency distribution by utilizing fast Fourier transform, extracting dominant frequency, comparing the dominant frequency with an inherent frequency database, judging a resonance risk, and further calculating an amplitude change trend by adopting a sliding window algorithm. And evaluating the fatigue accumulation degree by combining a material fatigue model. A Bayesian probability model is innovatively introduced, the influence of current load and wind action is fused, the resonance risk posterior probability is calculated, a support vector machine algorithm is adopted to classify and evaluate the risk, accurate early warning is achieved, finally, an early warning signal is transmitted through wireless communication, a vibration suppression control signal is generated to adjust a damping device, and the vibration suppression effect is achieved. And stable line operation is realized. According to the invention, accurate identification, real-time early warning and active suppression of the resonance risk of the power transmission line are realized, and the operation safety and reliability of a power grid are improved.

Description

Technical Field

[0001] This invention relates to a method for monitoring and early warning of faults in power distribution network lines. Background Technology

[0002] Fault location in distribution network lines is a critical area for the safe and stable operation of power systems, directly impacting power supply reliability and user experience. With the development of smart grids, accurate and rapid fault location has become essential for improving grid operating efficiency and reducing maintenance costs. However, existing fault location methods often perform poorly in complex environments, especially when facing potential faults caused by line vibrations, exhibiting insufficient detection accuracy and real-time performance, failing to meet the high reliability requirements of modern power grids.

[0003] Currently, most fault location methods rely on traditional sensors or manual inspections, which suffer from slow response times and limited coverage. This is especially true under complex weather conditions or high load operation, making it difficult to capture dynamic changes in the line. For example, vibrations in conductors caused by current and wind, if close to their natural frequency, can trigger resonance, leading to material fatigue and increased fault risk. However, existing technologies struggle to accurately identify such dynamic risks. Furthermore, traditional methods face significant challenges in real-time monitoring without power interruption, limiting the stability and accuracy of data acquisition. The randomness of line resonance is one of the core challenges in fault location. Vibration frequencies are highly uncertain due to environmental factors such as wind speed, temperature, and current load. This randomness makes it difficult for detection systems to accurately distinguish between normal vibrations and dangerous resonances.

[0004] Therefore, real-time monitoring systems need to process complex dynamic data, but existing technologies have bottlenecks in data analysis and processing capabilities, making it difficult to achieve high-precision vibration state identification without power interruption. Furthermore, early warning of resonance risk relies on accurate modeling of vibration characteristics, but current models are not adaptable enough to random factors, resulting in a high rate of false alarms or missed alarms. Summary of the Invention

[0005] This invention proposes a method for monitoring and early warning of faults in power distribution lines. Under uninterrupted power conditions, it uses intelligent means to monitor the vibration status of power distribution lines in real time, accurately identify resonance risks, and improve the accuracy of early warning.

[0006] The technical solution of this invention is implemented as follows:

[0007] A method for monitoring and early warning of faults in power distribution network lines, the method comprising:

[0008] The vibration status data of the line is collected by a sensor array to obtain a time-series signal containing vibration frequency and amplitude. The time-series signal is analyzed by a fast Fourier transform algorithm to obtain the vibration frequency distribution of the line. Based on the line vibration frequency distribution, the dominant vibration frequency is extracted and compared with a preset conductor natural frequency database. If the dominant vibration frequency is within the natural frequency range, then the risk of resonance is determined. To address the risk of resonance, amplitude data is extracted from vibration state data, and a sliding window algorithm is used to calculate the degree of amplitude increase, thus obtaining the amplitude change trend. By comparing the amplitude change trend with a preset amplitude threshold, if the amplitude change trend exceeds the threshold, the material fatigue model is used to calculate the material fatigue accumulation degree based on the number of stress cycles, and the fatigue accumulation degree is obtained. Based on the degree of fatigue accumulation, a Bayesian probability model is adopted. Through probability distribution updates, the probabilistic influences of current load and wind force are integrated. The degree of fatigue accumulation is used as the conditional probability input. Based on the Bayesian formula P(R|F,C,W)=P(F|R)P(C|R)P(W|R)P(R) / P(F,C,W), where R is the resonance risk, F is the degree of fatigue accumulation, C is the current load, W is the wind force, and P is the probability, the posterior probability is calculated to obtain the resonance risk output. By using the resonant risk output, the support vector machine algorithm is adopted. The conditional probability value of the resonant risk output is used as the feature vector input. The Gaussian kernel function is used to process nonlinear data. Random factors are classified by maximizing the classification boundary. The matching degree between the classification result and the preset threshold is calculated to obtain the early warning accuracy assessment. Based on the accuracy assessment of the early warning, if the matching degree of the classification result reaches the preset threshold, the early warning signal is transmitted to the intelligent detection system through the wireless communication module in an uninterrupted power environment to obtain real-time early warning instructions. By using real-time early warning commands, a vibration suppression control signal is generated by a digital signal processor. Based on the dominant vibration frequency, the feedback frequency of the damping device is adjusted and transmitted to the damping device to reduce the line vibration frequency and achieve a stable operating state. Based on the stable operating state, vibration status data is continuously collected through the sensor array. Based on the sampling time interval of the high-frequency vibration data, the above steps are repeated to obtain continuous monitoring results.

[0009] The beneficial effects of this invention are as follows: Vibration data of the transmission line is collected by a sensor array; the vibration frequency distribution is analyzed using Fast Fourier Transform; the dominant frequency is extracted and compared with a natural frequency database to determine the resonance risk; further, a sliding window algorithm is used to calculate the amplitude variation trend; and the degree of fatigue accumulation is assessed using a material fatigue model. A Bayesian probability model is innovatively introduced, integrating the effects of current load and wind force, to calculate the posterior probability of resonance risk. A support vector machine algorithm is used to classify and assess the risk, achieving accurate early warning. Finally, the early warning signal is transmitted wirelessly, and a vibration suppression control signal is generated to adjust the damping device, achieving stable line operation. This invention achieves accurate identification, real-time early warning, and active suppression of transmission line resonance risk, improving the safety and reliability of power grid operation. Detailed Implementation

[0010] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0011] A method for monitoring and early warning of faults in power distribution lines includes:

[0012] Step S101: Collect line vibration state data through sensor array, obtain time-series signals containing vibration frequency and amplitude, and analyze the time-series signals using fast Fourier transform algorithm to obtain the line vibration frequency distribution.

[0013] Vibration time-series signals, including vibration frequency and amplitude, are acquired using a sensor array and stored as raw time-series data. The raw time-series signal is analyzed using a Fast Fourier Transform (FFT) algorithm to obtain the vibration frequency distribution. Based on the vibration frequency distribution, the main frequency components and their corresponding amplitudes are extracted to generate a vibration feature set. If the main frequency components in the vibration feature set exceed a preset threshold, the time-series signal is denoised using a wavelet transform algorithm to obtain a denoised time-series signal. The denoised time-series signal is then re-analyzed using the FFT algorithm to obtain an optimized frequency distribution. Based on the optimized frequency distribution, the relative change trends between frequency components are calculated to generate vibration state features. The vibration state features are then classified using a Support Vector Machine (SVM) algorithm to determine the degree of abnormality in the line vibration state, and the classification results are obtained.

[0014] Step S102: Based on the line vibration frequency distribution, extract the dominant vibration frequency and compare it with a preset conductor natural frequency database. If the dominant vibration frequency is within the natural frequency range, then it is determined that there is a risk of resonance.

[0015] The dominant vibration frequency is extracted from the line vibration frequency distribution. A Fast Fourier Transform (FFT) algorithm is used to process the vibration time-series signal to obtain the dominant vibration frequency. Using a pre-defined conductor natural frequency database, the conductor's natural frequency range is obtained. The Euclidean distance algorithm is used to calculate the matching degree between the dominant vibration frequency and the natural frequency range, yielding the frequency matching result. If the frequency matching result shows that the dominant vibration frequency falls within the natural frequency range, a resonance risk is identified, and a resonance risk indicator is generated. Based on the resonance risk indicator, local features of the vibration time-series signal are obtained. A Short-Time Fourier Transform (SFT) algorithm is used to analyze these local features, resulting in a time-frequency distribution. The frequency change trend is extracted from the time-frequency distribution, and a linear regression algorithm is used to calculate the slope of the frequency change trend, yielding the frequency change characteristics. If the slope of the frequency change characteristics exceeds a pre-defined threshold, the line vibration state is classified as high-risk, generating a high-risk classification result. Based on the high-risk classification result, the energy distribution of the vibration time-series signal is obtained. An entropy calculation method is used to analyze the uniformity of the energy distribution, yielding the vibration stability characteristics.

[0016] Step S103: To address the resonance risk, amplitude data is extracted from the vibration state data, and the amplitude increase is calculated using a sliding window algorithm to obtain the amplitude change trend.

[0017] Amplitude data is obtained from vibration state data, and amplitude data points are separated using signal processing methods to obtain an amplitude time series. For the amplitude time series, a sliding window algorithm is used, with a predefined window size, to calculate the amplitude change rate within each window, resulting in an amplitude change rate series. Change rate features are extracted from the amplitude change rate series; if the change rate exceeds a preset threshold, it is marked as an outlier, resulting in an outlier set. Based on the outlier set, a moving average algorithm is used to smooth the amplitude change rate series, eliminating noise effects, resulting in a smoothed change rate series. Trend features are extracted from the smoothed change rate series, and a trend line is fitted using a linear regression algorithm to determine the amplitude change trend. For the amplitude change trend, if the trend slope exceeds a preset slope threshold, it is judged as a potential resonance risk, resulting in a risk indicator. Based on the risk indicator and the amplitude change trend, a resonance risk assessment result is generated, and the final risk level is determined.

[0018]

[0019] This represents the final risk assessment value. This represents the amplitude weighting coefficient. Indicates the change in amplitude. This represents the risk indicator weighting coefficient. Indicates the risk indicator value, and It is a regulating factor.

[0020] Step S104: Based on the amplitude change trend, a preset amplitude threshold is used for comparison. If the amplitude change trend exceeds the threshold, the material fatigue model is used to calculate the material fatigue accumulation degree based on the number of stress cycles, and the fatigue accumulation degree is obtained.

[0021] Amplitude data is acquired through sensors, and time-series analysis is used to extract the amplitude variation trend. If the amplitude variation trend exceeds a preset amplitude threshold, the stress cycle count is retrieved from the database. Based on the stress cycle count, the Miner linear cumulative damage model, combined with material property parameters, is used to calculate the cumulative fatigue level.

[0022] D represents the cumulative level of fatigue. This represents the number of cycles at the i-th stress level. Let represent the fatigue life at the i-th stress level, and n represent the total number of stress levels. Stress distribution characteristics are retrieved from the material database, and combined with the cumulative fatigue level, it is determined whether the material has reached its fatigue limit. If the cumulative fatigue level exceeds a preset fatigue threshold, fatigue failure is determined, resulting in a fatigue failure judgment. Historical data analysis extracts cycle count records, and a support vector machine algorithm is used to predict future cycle count trends, resulting in a cycle count prediction result. Based on the cycle count prediction result and the cumulative fatigue level, a conditional judgment logic is used to determine whether to adjust the amplitude threshold. If the predicted cycle count exceeds a safe range, the preset amplitude threshold is updated, resulting in an updated amplitude threshold. The fatigue level output module integrates the cumulative fatigue level, fatigue failure judgment, and updated amplitude threshold to generate a fatigue analysis data package, resulting in a fatigue analysis data package.

[0023] Step S105: Based on the degree of fatigue accumulation, a Bayesian probability model is adopted. Through probability distribution updates, the probabilistic influence of current load and wind force is integrated. The degree of fatigue accumulation is used as the conditional probability input. Based on the Bayesian formula P(R|F,C,W)=P(F|R)P(C|R)P(W|R)P(R) / P(F,C,W) (where R is the resonance risk, F is the degree of fatigue accumulation, C is the current load, W is the wind force, and P is the probability), the posterior probability is calculated to obtain the resonance risk output.

[0024] Real-time data on fatigue accumulation, current load, and wind force are acquired and extracted from the device via sensor interfaces to obtain raw datasets. Data preprocessing methods are used to denoise and normalize the raw datasets to obtain standardized datasets. Based on the standardized datasets and a pre-established probability distribution model, the prior probabilities of fatigue accumulation, current load, and wind force are calculated to obtain prior probability values.

[0025]

[0026] This represents the prior probability distribution of fatigue accumulation. This represents the i-th fatigue sample value, and n represents the total number of samples. This represents the mean. Indicates standard deviation.

[0027]

[0028] This represents the prior probability distribution of the current load. This represents the i-th current sample value, m represents the number of sampling points, α represents the weighting coefficient, β represents the attenuation coefficient, and t represents time.

[0029]

[0030] Represents the prior probability distribution of wind force. This represents the wind force value at time t. This represents the standard wind force value, and T represents the observation period. This represents the proportionality coefficient.

[0031] Using a Bayesian model, based on prior probability values ​​and conditional probability formulas P(F|R), P(C|R), and P(W|R), the posterior probability is updated to obtain the probability distribution of resonance risk.

[0032] P(R|F,C,W)=P(F|R)P(C|R)P(W|R)P(R) / P(F,C,W)

[0033] P(R|F, C, W) represents the posterior probability, i.e., the resonance risk probability; P(F|R) represents the conditional probability of fatigue when resonance occurs; P(C|R) represents the conditional probability of current when resonance occurs; P(W|R) represents the conditional probability of wind force when resonance occurs; and P(R) represents the prior probability of resonance. If the probability distribution of resonance risk exceeds a preset threshold, the specific risk value is calculated using the formula P(R|F, C, W), resulting in a resonance risk output. Based on the resonance risk output, a classification algorithm is used to divide the risk level, obtaining risk level labels. Corresponding equipment status data is generated using these risk level labels and stored in the database to obtain status update records.

[0034] Step S106: Through the resonance risk output, the support vector machine algorithm is adopted. The conditional probability value of the resonance risk output is used as the feature vector input. The Gaussian kernel function is used to process the nonlinear data. Random factors are classified by maximizing the classification boundary. The matching degree between the classification result and the preset threshold is calculated to obtain the early warning accuracy assessment.

[0035] The system acquires resonance risk output data, calculates conditional probability values, and generates feature vectors. A Gaussian kernel function is used to transform the nonlinear data in the feature vectors, resulting in a transformed feature vector. A Support Vector Machine (SVM) algorithm is then used to maximize the classification boundary based on the transformed feature vectors, generating a classification model. If the classification boundary converges, the random factor is classified, yielding a classification result; otherwise, the Gaussian kernel function parameters are adjusted, and the classification model is regenerated. The classification result is compared with a preset threshold to calculate the matching degree, generating a matching degree score. Based on the matching degree score, a logistic regression algorithm is used to evaluate the accuracy of the early warning, yielding an accuracy assessment result. If the accuracy assessment result is lower than the preset threshold, the SVM parameters are optimized, and the classification result is recalculated; if it is higher than the preset threshold, the final early warning accuracy assessment result is output.

[0036] Step S107: Based on the accuracy assessment of the early warning, if the matching degree of the classification result reaches the preset threshold, the early warning signal is transmitted to the intelligent detection system through the wireless communication module in an uninterrupted power supply environment to obtain the real-time early warning instruction.

[0037] Raw signal data is acquired from the data acquisition module and preprocessed to generate classification input data. If the classification input data is processed by the support vector machine algorithm to generate classification results, it is determined whether the classification results reach a preset threshold to determine the matching degree. If the matching degree reaches the preset threshold, a warning signal is sent via the wireless communication module under uninterrupted power conditions to obtain the transmission status. Based on the transmission status, the warning signal is encrypted using an encryption algorithm to generate an encrypted signal. Signal features are extracted from the encrypted signal, and the features are analyzed by an intelligent detection system to obtain real-time warning commands. Based on the real-time warning commands, a logistic regression algorithm is used to prioritize the commands and determine the execution order. Based on the execution order, system response data is generated to obtain the final control signal.

[0038] Step S108: Through real-time early warning instructions, a vibration suppression control signal is generated using a digital signal processor. Based on the dominant vibration frequency, the feedback frequency of the damping device is adjusted and transmitted to the damping device to reduce the line vibration frequency and achieve a stable operating state.

[0039] The system acquires real-time early warning commands, collects vibration data from line sensors, and uses a Fast Fourier Transform (FFT) algorithm to determine the dominant vibration frequency. Based on this dominant frequency, a digital signal processor (DSP) generates a vibration suppression signal, adjusts signal parameters, and obtains a control signal. The control signal is used to adjust the feedback frequency of the damping device. If the deviation between the feedback frequency and the dominant vibration frequency exceeds a preset threshold, the signal parameters are recalculated to obtain the adjusted feedback frequency. Using the adjusted feedback frequency, the damping device is driven to perform vibration suppression, reducing the line vibration amplitude, and the suppressed vibration data is obtained. Line vibration characteristics are extracted from the suppressed vibration data. If the vibration amplitude is below a preset threshold, a stable operating state is determined, and the stable operating state is obtained. Based on the stable operating state, the real-time early warning command database is updated, storing the vibration suppression signal and feedback frequency, resulting in an updated command database. The DSP then optimizes the signal generation strategy based on historical data to obtain an optimized vibration suppression signal.

[0040] Step S109: Based on the stable operating state, continuously collect vibration state data through the sensor array. Based on the sampling time interval of the high-frequency vibration data, repeat the above steps in a loop to obtain continuous monitoring results.

[0041] Vibration state data is acquired in real time by a sensor array, generating a high-frequency vibration data stream based on a preset sampling interval. If outliers exist in the high-frequency vibration data stream, the data is frequency-domain transformed using a Fourier transform algorithm to obtain the spectral characteristics of the vibration signal. Based on the spectral characteristics, a k-means clustering algorithm is used to classify the vibration state and determine the distribution pattern of vibration modes. If the distribution pattern of vibration modes deviates from a preset threshold, a sliding window technique is used to perform time-series analysis on the continuous vibration data to obtain the vibration change trend. Based on the vibration change trend, an autoregressive model is used to predict the future vibration state, generating predicted vibration data. By comparing the predicted vibration data with the real-time acquired vibration data, the stability of the vibration state is determined, and a stability assessment result is obtained. Based on the stability assessment result, the sampling interval of the sensor array is updated to optimize the accuracy of subsequent vibration data acquisition.

[0042] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power distribution network line fault monitoring and early warning method, characterized in that, The method comprises: The method comprises:

2. The power distribution network line fault monitoring and early warning method of claim 1, characterized in that, The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: According to the optimized frequency distribution, the relative change trend between frequency components is calculated to generate a vibration state feature, a support vector machine algorithm is used to classify the vibration state feature, the abnormal degree of the line vibration state is judged, and a classification result is obtained.

3. The power distribution network line fault monitoring and early warning method of claim 1, characterized in that, According to the line vibration frequency distribution, the dominant vibration frequency is extracted, and the preset conductor natural frequency database is compared. If the dominant vibration frequency is within the natural frequency range, it is determined that there is a risk of resonance, including: The dominant vibration frequency is extracted from the line vibration frequency distribution, and the fast Fourier transform algorithm is used to process the vibration time sequence signal to obtain the dominant vibration frequency. Through the preset conductor natural frequency database, the conductor natural frequency interval is obtained, and the Euclidean distance algorithm is used to calculate the matching degree of the dominant vibration frequency and the natural frequency interval to obtain a frequency matching result. If the frequency matching result shows that the dominant vibration frequency falls within the natural frequency interval, it is determined that there is a risk of resonance, and a resonance risk flag is generated. According to the resonance risk flag, the local feature of the vibration time sequence signal is obtained, and the short-time Fourier transform algorithm is used to analyze the local feature to obtain a time-frequency distribution. Through the time-frequency distribution, the frequency change trend is extracted, the slope of the frequency change trend is calculated using a linear regression algorithm, and a frequency change feature is obtained. If the slope of the frequency change feature exceeds a preset threshold, it is determined that the line vibration state is high-risk, and a high-risk classification result is generated. According to the high-risk classification result, the energy distribution of the vibration time sequence signal is obtained, and the uniformity of the energy distribution is analyzed using an entropy calculation method to obtain a vibration stability feature.

4. The power distribution network line fault monitoring and early warning method of claim 1, characterized in that, The amplitude data is extracted from the vibration state data, and the sliding window algorithm is used to calculate the amplitude increase degree to obtain the amplitude change trend, including: The amplitude data is obtained from the vibration state data, and the signal processing method is used to separate the amplitude data points to obtain an amplitude time sequence. The sliding window algorithm is used to set a predefined window size, and the amplitude change rate in each window is calculated to obtain an amplitude change rate sequence. The change rate feature is extracted from the amplitude change rate sequence. If the change rate exceeds a preset threshold, it is marked as an abnormal point to obtain an abnormal point set. The moving average algorithm is used to smooth the amplitude change rate sequence to eliminate noise effects to obtain a smoothed change rate sequence. The trend feature is extracted from the smoothed change rate sequence, and the trend line is fitted using a linear regression algorithm to determine the amplitude change trend. For the amplitude change trend, if the trend slope exceeds a preset slope threshold, it is determined to be a potential resonance risk, and a risk flag is obtained. According to the risk flag, the resonance risk assessment result is generated based on the amplitude change trend to determine the final risk level. ， represents a final risk assessment value, represents an amplitude weight coefficient, represents an amplitude change amount, represents a risk flag weight coefficient, represents a risk flag value, and is a regulation factor.

5. The power distribution network line fault monitoring and early warning method of claim 1, wherein, The amplitude threshold is compared with the amplitude change trend. If the amplitude change trend exceeds the threshold, the material fatigue model is used to calculate the material fatigue accumulation degree based on the stress cycle number to obtain a fatigue accumulation degree, including: The amplitude data is obtained through the sensor, and the time sequence analysis method is used to extract the amplitude change trend to obtain the amplitude change trend. If the amplitude change trend exceeds the preset amplitude threshold, the stress cycle number is obtained from the database to obtain the stress cycle number. According to the number of stress cycles, the cumulative fatigue degree is calculated by using the Miner linear cumulative damage model, and the cumulative fatigue degree is obtained, ， D denotes the accumulated fatigue degree, denotes the number of cycles at the i-th stress level, denotes the fatigue life at the i-th stress level, n denotes the total number of stress levels; The stress distribution characteristics are obtained from the material database, and the cumulative fatigue degree is combined to determine whether the material has reached the fatigue limit. If the cumulative fatigue degree exceeds the preset fatigue threshold, the material fatigue failure is determined, and the fatigue failure judgment is obtained; Through historical data analysis, the cycle number record is extracted, and the support vector machine algorithm is used to predict the future cycle number trend, and the cycle number prediction result is obtained. According to the cycle number prediction result, combined with the cumulative fatigue degree, the conditional judgment logic is used to determine whether to adjust the amplitude threshold. If the predicted cycle number exceeds the safe range, the preset amplitude threshold is updated, and the updated amplitude threshold is obtained. Through the fatigue degree output module, the cumulative fatigue degree, the fatigue failure judgment and the updated amplitude threshold are integrated to generate the fatigue analysis data package, and the fatigue analysis data package is obtained.

6. The power distribution network line fault monitoring and early warning method of claim 1, wherein, According to the fatigue accumulation degree, the Bayesian probability model is used to update the probability distribution, and the probability influence of current load and wind action is fused. The fatigue accumulation degree is used as the conditional probability input, and the posterior probability is calculated based on the Bayes formula P(R|F,C,W)=P(F|R)P(C|R)P(W|R)P(R) / P(F,C,W), wherein R is the resonance risk, F is the fatigue accumulation degree, C is the current load, W is the wind action, and P is the probability. The resonance risk output is obtained, including: The real-time collected fatigue accumulation degree, current load and wind action data are obtained through the sensor interface from the equipment, and the original data set is obtained. The data preprocessing method is used to denoise and normalize the original data set, and the standardized data set is obtained; Through the pre-established probability distribution model, the prior probability of fatigue accumulation degree, current load and wind action is calculated based on the standardized data set, and the prior probability value is obtained; ， a prior probability distribution representing fatigue accumulation, represents the ith fatigue sample value, n represents the total number of samples, represents the mean, represents the standard deviation, ， a prior probability distribution representing a current load, represents the ith current sample value, m represents the number of sample points, a represents a weight coefficient, β represents a decay coefficient, and t represents time. ， a prior probability distribution representing the wind effect, a wind value at time t, a standard wind value, T represents an observation period, a proportional coefficient adopts a Bayesian model, and based on a prior probability value and conditional probability formulas P(F|R), P(C|R) and P(W|R), a posterior probability is updated to obtain a probability distribution of the resonance risk, and a resonance risk output is obtained; According to the resonance risk output, the risk level is divided by using the classification algorithm, the risk level label is obtained, the corresponding equipment state data is generated, and the state update record is stored in the database.

7. The power distribution network line fault monitoring and early warning method of claim 1, wherein, Through the resonance risk output, the support vector machine algorithm is used, the conditional probability value of the resonance risk output is used as the feature vector input, the Gaussian kernel function is used to process the nonlinear data, the random factors are classified by maximizing the classification boundary, the matching degree of the classification result and the preset threshold is calculated, and the early warning accuracy evaluation is obtained, including: The resonance risk output data is obtained, the conditional probability value is calculated, and the feature vector is generated. The Gaussian kernel function is used to transform the nonlinear data in the feature vector, and the transformed feature vector is obtained. The support vector machine algorithm is used to generate a classification model based on the transformed feature vector and maximize the classification boundary; If the classification boundary of the classification model converges, the random factors are classified, and the classification result is obtained. If it does not converge, the Gaussian kernel function parameters are adjusted, and the classification model is regenerated. The matching degree is calculated by comparing the classification result with the preset threshold value, a matching degree score is generated, a logistic regression algorithm is used to evaluate the early warning accuracy, and an accuracy evaluation result is obtained; if the accuracy evaluation result is lower than the preset threshold value, the support vector machine parameters are optimized, and the classification result is recalculated; if it is higher than the preset threshold value, the final early warning accuracy evaluation result is output.

8. The power distribution network line fault monitoring and early warning method of claim 1, wherein, According to the early warning accuracy evaluation, if the classification result matching degree reaches the preset threshold value, the wireless communication module is used to transmit the early warning signal to the intelligent detection system in a non-power-off environment to obtain a real-time early warning instruction, including: Obtain the original signal data from the data acquisition module, generate the classification input data through data preprocessing, and generate the classification result after the support vector machine algorithm processing, then determine whether the classification result reaches the preset threshold value to determine the matching degree; if the matching degree reaches the preset threshold value, send the early warning signal through the wireless communication module in a non-power-off environment to obtain the transmission state; According to the transmission state, the early warning signal is encrypted by using an encryption algorithm to generate an encrypted signal, the signal features are extracted from the encrypted signal, the features are analyzed by the intelligent detection system to obtain a real-time early warning instruction, the instruction priority is sorted by using a logistic regression algorithm to determine the execution order, system response data is generated, and a final control signal is obtained.

9. The power distribution network line fault monitoring and early warning method of claim 1, wherein, According to the real-time early warning instruction, a digital signal processor is used to generate a vibration suppression control signal, the feedback frequency of the damping device is adjusted based on the dominant vibration frequency, and the control signal is transmitted to the damping device to reduce the line vibration frequency and obtain a stable operation state, including: Obtain the real-time early warning instruction, collect vibration data from the line sensor, determine the dominant vibration frequency by using a fast Fourier transform algorithm, obtain the dominant vibration frequency, and a digital signal processor generates a vibration suppression signal, adjusts the signal parameters, and obtains a control signal; By using the control signal, the feedback frequency of the damping device is adjusted, if the deviation between the feedback frequency and the dominant vibration frequency exceeds the preset threshold value, the signal parameters are recalculated to obtain the adjusted feedback frequency; the adjusted feedback frequency is used to drive the damping device to execute vibration suppression, reduce the line vibration amplitude, and obtain suppressed vibration data; the line vibration features are extracted from the suppressed vibration data, if the vibration amplitude is lower than the preset threshold value, it is determined as a stable operation state, and a stable operation state is obtained; Through the stable operation state, the real-time early warning instruction database is updated, the vibration suppression signal and the feedback frequency are stored, and an updated instruction database is obtained; the updated instruction database is obtained, the digital signal processor optimizes the signal generation strategy based on historical data to obtain an optimized vibration suppression signal.

10. The power distribution network line fault monitoring and early warning method of claim 1, wherein, According to the stable operation state, the vibration state data is continuously collected by the sensor array, the sampling time interval of the high-frequency vibration data is used to cyclically execute the above steps, and a continuous monitoring result is obtained, including: The vibration state data is collected in real time by the sensor array, and high-frequency vibration data streams are generated based on the preset sampling interval; if there are abnormal values in the high-frequency vibration data streams, the data is converted into the frequency domain by using a Fourier transform algorithm to obtain the frequency spectrum features of the vibration signal; According to the spectrum characteristics, the k-means clustering algorithm is used to classify the vibration state, and the distribution rule of the vibration mode is determined; if the distribution rule of the vibration mode deviates from the preset threshold, the time series analysis is performed on the continuous vibration data through the sliding window technology, and the vibration change trend is obtained; According to the vibration change trend, the autoregressive model is used to predict the future vibration state, and the predicted vibration data is generated; by comparing the predicted vibration data with the real-time collected vibration data, the stability of the vibration state is judged, the stability evaluation result is obtained, and the sampling interval of the sensor array is updated, and the accuracy of the subsequent vibration data collection is optimized.

Images