Fault diagnosis and positioning lightning protection system for power plant line

Abstract

The invention relates to the field of power system monitoring, and discloses a fault diagnosis and positioning lightning protection system for a power plant line, and the system comprises distributed optical fiber sensors which are disposed along the power line and are used for monitoring the temperature, strain and sound data of the line in real time; an optical time domain reflectometer for collecting backscattered signals along the fiber optic line to determine the location of loss and reflection events; the data processing unit is provided with a digital signal processor and a machine learning model, and is used for analyzing the data collected from the optical fiber sensor and identifying a fault mode; and the fault positioning module is used for fault positioning based on time difference. By integrating an advanced optical fiber sensor and an artificial intelligence algorithm, the physical state of the power line can be monitored in real time, and the system not only can identify a current fault, but also can predict possible problems in the future. The high-precision and quick-response fault detection capability greatly reduces the loss and risk caused by fault response delay.

Description

Technical Field

[0001] The invention relates to the technical field of power system monitoring, and in particular to a fault diagnosis and positioning lightning protection system for power plant lines. Background Art

[0002] With the continuous development and expansion of power systems, the operation, maintenance and management of power plant lines have become increasingly complex. Power lines are often affected by various external factors, especially under extreme weather conditions such as lightning, and line failures occur frequently. Rapid and accurate diagnosis and location of faults are essential to ensure the stable operation of power systems and power supply safety.

[0003] Existing power line fault diagnosis technologies mainly rely on traditional sensor technology and primary fault analysis methods. These methods can monitor basic electrical parameters such as current and voltage, and diagnose whether there is a fault in the line by analyzing the changes in these parameters. However, these technologies often cannot provide accurate information about the specific location of the fault, and it is difficult to predict possible problems in the future.

[0004] The main problem with existing technologies is that they are not fast enough and accurate enough to diagnose faults. In traditional systems, fault location often takes a long time and the misjudgment rate is relatively high. This not only prolongs the time to restore normal power supply, but also may lead to unnecessary maintenance costs and waste of resources. In addition, traditional methods are often unable to cope with complex faults caused by extreme weather, and are unable to effectively prevent and manage the resulting chain reactions, increasing system instability and operational risks. Summary of the invention

[0005] In view of the deficiencies of the prior art, the present invention provides a fault diagnosis and positioning lightning protection system for power plant lines, which solves the problem that fault positioning in traditional systems often takes a long time and has a relatively high misjudgment rate.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a fault diagnosis and positioning lightning protection system for power plant lines, comprising: Distributed fiber optic sensors are deployed along power lines to monitor line temperature, strain, and sound data in real time; Optical time domain reflectometry, which collects backscattered signals along a fiber line to determine the location of loss and reflection events; a data processing unit configured with a digital signal processor and a machine learning model for analyzing data collected from the fiber optic sensor and identifying failure modes; The fault location module applies fault location based on time difference.

[0007] Preferably, the fault location module formula is: in is The data measured by the sensor position, At a known fault location Data under the conditions The probability density function of is used to calculate the fault location.

[0008] Preferably, the distributed optical fiber sensor uses a technology based on Rayman scattering and Brillouin scattering to detect temperature and strain at different wavelengths, thereby monitoring the physical state of the power line in real time.

[0009] Preferably, the optical time domain reflectometer has a high-precision time measurement capability, which is used to calculate the time delay and loss of the reflected signal, and the formula is: in is the input power, is the attenuation coefficient, is the distance along the fiber, is in position This data is used to analyze and locate potential faults.

[0010] Preferably, the data processing unit is further configured with a bandpass filter for signal denoising, and adopts a discrete wavelet transform processing module, which is used to extract key time-frequency features from complex signals for subsequent analysis.

[0011] Preferably, the discrete wavelet transform is implemented by the following formula: in is the wavelet basis function, Represents different scales and displacements for multi-scale analysis of power line data.

[0012] Preferably, the machine learning model is a convolutional neural network designed to automatically identify abnormal patterns in signals processed from discrete wavelet transforms and classify the health status of the line.

[0013] Preferably, the convolutional neural network includes at least three convolutional layers and at least two pooling layers, each convolutional layer is followed by a ReLU activation function to enhance nonlinear learning ability, and the pooling layer is used to reduce feature dimensions and optimize the computational efficiency and performance of the model.

[0014] Preferably, the training process of the machine learning model includes: using a historical data set for supervised learning, the historical data set consists of labeled normal and fault state data, and using a back-propagation algorithm to optimize network weights.

[0015] Preferably, after the model is trained, a fault prediction step is implemented, including real-time input of data collected from the optical fiber sensor and the optical time domain reflectometer, and output of the fault type and location prediction results after model processing.

[0016] Working principle: The system densely deploys distributed fiber optic sensors based on Rayman scattering and Brillouin scattering along the power lines. These sensors can detect the temperature and strain of the line at different wavelengths, thereby monitoring the physical state of the power line in real time. The data of the fiber optic sensor is collected by an optical time domain reflectometer. The optical time domain reflectometer can measure the optical signal loss and reflection events along the fiber path, providing accurate information about the fiber status and potential faults. The collected data is first processed by a digital signal processor, including denoising and signal enhancement. The bandpass filter in the system is used to remove noise of irrelevant frequencies and retain the signal part related to the change of the power line state. Next, a multi-scale signal analysis is performed using a discrete wavelet transform to extract key time-frequency features from the complex power line data, and the extracted features are input into a deep convolutional neural network for further analysis. The deep convolutional neural network learns the deep features of the signal through multiple layers of convolution and pooling layers, and classifies them to distinguish between normal, abnormal and faulty states of the line. The training process of the deep convolutional neural network includes supervised learning using a large amount of labeled historical data, optimizing the model's weights and parameters to improve the accuracy of the prediction. The fault prediction step includes real-time input of the latest sensor and optical time domain reflectometer data into the trained model, and the model outputs the prediction results of the fault type and location after processing. These real-time data enable the system to respond quickly to potential fault events. The system also includes a performance evaluation and feedback module for collecting operating data and maintenance results to optimize system configuration and AI model parameters. This module ensures that the system continues to evolve and improve over time and changes in the power grid environment. Through this detailed description of the working principle, the system can provide an efficient and accurate power plant line fault diagnosis and positioning lightning protection solution, significantly improving the reliability and safety of the power system.

[0017] The present invention provides a fault diagnosis and positioning lightning protection system for power plant lines. It has the following beneficial effects: 1. By integrating advanced fiber optic sensors and artificial intelligence algorithms, the present invention can monitor the physical state of power lines, such as temperature and strain, in real time, and quickly and accurately identify faults or abnormal conditions. Using deep learning models, the system can not only identify current faults, but also predict possible problems in the future. This high-precision and fast-response fault detection capability greatly reduces the losses and risks caused by delayed fault response.

[0018] 2. The present invention can automatically evaluate the impact scope after detecting a fault, and intelligently adjust the grid load in the surrounding area according to the preset risk management strategy to prevent the fault from spreading. At the same time, by implementing intelligent fault isolation, the system can locally isolate the fault point to prevent the entire network from being affected, thereby maintaining the stable operation of the grid. This dynamic load management and fault isolation strategy greatly improves the operating efficiency and reliability of the grid.

[0019] 3. The integrated intelligent response and automated fault handling functions of the present invention can automatically plan maintenance work, including dispatching the nearest maintenance team and optimizing the maintenance path. This automated maintenance scheduling not only speeds up fault recovery, but also optimizes resource allocation, thereby reducing maintenance costs. The system can also provide maintenance recommendations based on historical data and current operating data to achieve predictive maintenance, further reducing the incidence of unexpected failures and related costs.

[0020] 4. The present invention significantly improves the overall safety and stability of the power grid through real-time monitoring and immediate response to faults, combined with intelligent fault isolation and automatic recovery systems. In extreme weather conditions, such as lightning, the system can effectively prevent and manage power line faults caused by them, reduce the chain reaction caused by faults, and ensure the continuity and reliability of power supply. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is the main framework diagram of the present invention; Figure 2 It is a flow chart of the distributed optical fiber sensor of the present invention; Figure 3 is a flow chart of the machine learning model of the present invention; Figure 4 It is a flow chart of the data processing unit of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] Please refer to the attached Figure 1 - Attachment Figure 4 The embodiment of the present invention provides a fault diagnosis and positioning lightning protection system for power plant lines, including: Distributed fiber optic sensors are deployed along power lines to monitor line temperature, strain, and sound data in real time; Optical time domain reflectometry, which collects backscattered signals along a fiber line to determine the location of loss and reflection events; a data processing unit configured with a digital signal processor and a machine learning model for analyzing data collected from the fiber optic sensor and identifying failure modes; The fault location module applies fault location based on time difference.

[0024] The fault location module formula is: in is The data measured by the sensor position, At a known fault location Data under the conditions The probability density function of is used to calculate the fault bit.

[0025] Distributed fiber optic sensors use technologies based on Rayman scattering and Brillouin scattering to detect temperature and strain at different wavelengths, thereby monitoring the physical status of power lines in real time.

[0026] The optical time domain reflectometer has high-precision time measurement capability and is used to calculate the time delay and loss of the reflected signal. The formula is: in is the input power, is the attenuation coefficient, is the distance along the fiber, is in position This data is used to analyze and locate potential faults.

[0027] The data processing unit is further configured with a bandpass filter for signal denoising and employs a discrete wavelet transform processing module, which is used to extract key time-frequency features from complex signals for subsequent analysis.

[0028] The discrete wavelet transform is implemented by the following formula: in is the wavelet basis function, Represents different scales and displacements for multi-scale analysis of power line data.

[0029] The machine learning model is a convolutional neural network designed to automatically identify abnormal patterns in the signal processed from discrete wavelet transform and classify the health status of the line.

[0030] The convolutional neural network consists of at least three convolutional layers and at least two pooling layers. Each convolutional layer is followed by a ReLU activation function to enhance nonlinear learning capabilities. The pooling layer is used to reduce feature dimensions and optimize the computational efficiency and performance of the model.

[0031] The training process of the machine learning model includes: supervised learning using historical data sets, which consist of labeled normal and fault state data, and optimizing network weights using the back-propagation algorithm.

[0032] After the model is trained, the fault prediction step is implemented, including real-time input of data collected from fiber optic sensors and optical time domain reflectometers, and output of fault type and location prediction results after model processing.

[0033] In one embodiment, the distributed optical fiber sensors of Rayman scattering and Brillouin scattering are selected to continuously monitor the temperature and strain changes along the entire length of the optical fiber. These parameters are important indicators for judging the health status of the power line. The selection of sensor interval is based on the balance between monitoring accuracy and cost-effectiveness to ensure that small physical changes can be captured, so as to detect problems in time at the early stage of faults; The use of optical time domain reflectometer for data acquisition is another core component of this system. The optical time domain reflectometer can detect faults and damage in optical fibers by analyzing the reflected light signal. Its principle is based on measuring the intensity and time of the light signal reflected from the fault point to determine the loss and location. Through the formula: Not only can the fault point be located, but also the fault severity can be estimated. is the input optical power, is the absorption coefficient of the optical fiber, It is the reflectivity, and these data are crucial for subsequent analysis.

[0034] In one embodiment, after data is collected from the fiber optic sensor and the optical time domain reflectometer, the data processing unit begins its work. First, the high-frequency and low-frequency noise in the signal is removed by a bandpass filter, retaining the frequency components related to the fault. Next, a discrete wavelet transform is performed, which is a powerful signal processing tool that can analyze signals at different frequency and time scales, so that even complex or non-stationary signals can be effectively processed. The discrete wavelet transform is expressed by the formula Implementation, where It is the mother wavelet function, which is used to extract useful time-frequency features from the signal; Using features extracted from discrete wavelet transform, deep learning models are trained to identify the health status of power lines. The deep learning model processes data through multiple convolutional layers and pooling layers, each followed by an activation function to extract nonlinear features in the data and reduce the spatial dimension of the features, optimizing the model's operational efficiency. This training process not only relies on historical labeled data, but also updates and optimizes model parameters by continuously learning new fault modes.

[0035] In one embodiment, once the system detects abnormal behavior, the fault location module is immediately activated. The fault point is accurately calculated using the formula mentioned above: This formula estimates the fault location by maximizing the posterior probability, where Indicates that from The data obtained from the sensors, Indicates a given fault location This method not only improves the positioning accuracy, but also speeds up the fault response speed.

[0036] In a preferred embodiment, in order to further enhance the performance of the power plant line fault diagnosis and positioning lightning protection system, the combination of the data preprocessing step and the aforementioned technical solution can be described in detail, especially how to use these preprocessed data in a hybrid neural network model, that is, an optimization model combining a convolutional neural network (CNN) and a long short-term memory network (LSTM). The following is a detailed technical solution description and an integration of the formula part.

[0037] Combining data preprocessing with hybrid neural network models Data preprocessing steps In this embodiment, data preprocessing is a key step for the successful implementation of the machine learning model. Its main goal is to convert the raw time series data collected from the fiber optic sensor into a format suitable for neural network processing.

[0038] Signal denoising and normalization: Denoising: Preliminary data processing using a bandpass filter to remove irrelevant frequency components caused by electromagnetic interference or noise from the device itself.

[0039] Normalization: The Min-Max normalization method is used to scale the data between 0 and 1. The formula is: in is the original data, and are the minimum and maximum values ​​in the data respectively.

[0040] Feature extraction: The discrete wavelet transform (DWT) is used to further extract the time-frequency features of the signal and provide input to the CNN model. This step decomposes the signal into different frequency bands, helping the model to better understand the frequency variation patterns in the data.

[0041] Hybrid model combining CNN and LSTM The hybrid neural network model combines the spatial feature extraction capability of CNN and the time series data processing capability of LSTM, providing a powerful tool for real-time monitoring and fault prediction of power lines.

[0042] CNN Layers: The CNN layer mainly processes the data preprocessed by DWT, and uses multiple convolutional layers and pooling layers to extract key spatial features. Each convolutional layer is followed by a ReLU activation function to increase the nonlinear ability of the model, and the pooling layer is used to reduce the data dimension and the amount of calculation.

[0043] LSTM Layer: The LSTM layer is connected after the CNN layer to handle the long-term dependency problem of time series. LSTM processes time series data through the following update formula: in is the sigmoid activation function, is the hyperbolic tangent function, , , , are the activation states of the forget gate, input gate, and output gate, respectively. is the cell state, is the output state, and is a learning parameter.

[0044] In a preferred embodiment, 1. Fault detection and intelligent response In this preferred embodiment, the system first uses the previously described fiber optic sensing technology and deep learning models (CNN and LSTM) for fault detection. These models can analyze the status of the power line in real time and identify potential fault points in a timely manner. Once the system detects an abnormality, the fault response procedure will be immediately initiated.

[0045] Fault Response Algorithm Using the time series prediction results obtained from the LSTM model, the system can predict the trend of fault development and the possible impact range. Specific fault response strategies include: here, is the sensor data, At the fault location The probability distribution of the data under the condition .

[0046] Dynamic load management: Dynamically adjust the grid load in the surrounding area based on the fault location and impact to reduce the impact of the fault on the entire system.

[0047] Intelligent isolation decision algorithm: Based on the severity of the fault and the predicted future trend, the system automatically decides whether to initiate local isolation. The probability threshold output by the machine learning model is used to determine the isolation condition: in, is the predicted probability of failure occurring, is the risk threshold set 2. Automated Failure Recovery The system not only intelligently isolates the faulted area, but also automatically initiates the recovery process after isolation. This includes using optimization algorithms to plan maintenance work and using automated switchgear to restore normal operations, including: Fault repair priority algorithm: Automatically arrange the priority and resources of maintenance tasks based on the severity of the fault impact and the urgency of recovery.

[0048] Automatic restoration procedure: After the necessary repairs are completed, the system will automatically assess the integrity of the line and gradually restore power supply.

[0049] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A lightning protection system for fault diagnosis and positioning of power plant lines, characterized in that: include: Distributed fiber optic sensors are deployed along power lines to monitor line temperature, strain, and sound data in real time; Optical time domain reflectometry, which collects backscattered signals along a fiber line to determine the location of loss and reflection events; a data processing unit configured with a digital signal processor and a machine learning model for analyzing data collected from the fiber optic sensor and identifying failure modes; The fault location module applies fault location based on time difference.

2. The fault diagnosis and positioning lightning protection system for power plant lines according to claim 1 is characterized in that: The fault location module formula is: in is The data measured by the sensor position, At a known fault location Data under the conditions The probability density function of is used to calculate the fault location.

3. The fault diagnosis and positioning lightning protection system for power plant lines according to claim 1 is characterized in that: The distributed optical fiber sensor uses technology based on Rayman scattering and Brillouin scattering to detect temperature and strain at different wavelengths, thereby monitoring the physical state of the power line in real time.

4. The fault diagnosis and positioning lightning protection system for power plant lines according to claim 1, characterized in that: The optical time domain reflectometer has high-precision time measurement capability and is used to calculate the time delay and loss of the reflected signal. The formula is: in is the input power, is the attenuation coefficient, is the distance along the fiber, is in position This data is used to analyze and locate potential faults.

5. The fault diagnosis and positioning lightning protection system for power plant lines according to claim 1, characterized in that: The data processing unit is further configured with a bandpass filter for signal denoising and adopts a discrete wavelet transform processing module, which is used to extract key time-frequency features from complex signals for subsequent analysis.

6. The fault diagnosis and positioning lightning protection system for power plant lines according to claim 5, characterized in that: The discrete wavelet transform is implemented by the following formula: in is the wavelet basis function, Represents different scales and displacements for multi-scale analysis of power line data.

7. The fault diagnosis and positioning lightning protection system for power plant lines according to claim 1, characterized in that: The machine learning model is a convolutional neural network designed to automatically identify abnormal patterns in signals processed from discrete wavelet transforms and classify the health status of the line.

8. The fault diagnosis and positioning lightning protection system for power plant lines according to claim 7, characterized in that: The convolutional neural network includes at least three convolutional layers and at least two pooling layers. Each convolutional layer is followed by a ReLU activation function to enhance nonlinear learning capabilities. The pooling layer is used to reduce feature dimensions and optimize the computational efficiency and performance of the model.

9. The fault diagnosis and positioning lightning protection system for power plant lines according to claim 1, characterized in that: The training process of the machine learning model includes: using historical data sets for supervised learning, the historical data sets are composed of labeled normal and fault state data, and using a back-propagation algorithm to optimize network weights.

10. The fault diagnosis and positioning lightning protection system for power plant lines according to claim 1, characterized in that: After the model is trained, a fault prediction step is implemented, including real-time input of data collected from the optical fiber sensor and the optical time domain reflectometer, and output of the fault type and location prediction results after model processing.

Images