Intelligent arc detection and power failure early warning method and system

Abstract

The invention relates to the technical field of electric power system fault detection and early warning, in particular to an intelligent electric arc detection and electric power fault early warning method and system, and the method comprises the following steps: S10, collecting historical current data, and extracting four-dimensional features which comprise a flat shoulder width, a peak amplitude, an average value and duration; s20, calculating a mean value and a standard deviation of the four-dimensional features, generating a dynamic threshold interval, and constructing a dynamic threshold model; s30, real-time current data are collected and input into the dynamic threshold value model, and when any one of the four-dimensional features exceeds the corresponding dynamic threshold value interval, it is judged that an arc fault occurs; s40, generating a probability thermodynamic diagram based on adjacent sensor current difference analysis and preset space-time weight calculation, and positioning a fault position; and S50, fault types are classified based on a preset nonlinear scoring model, and a feedback report is generated and pushed to the terminal. The method has the effects of quickly positioning the fault position and classifying the fault type.

Description

Technical Field

[0001] This application relates to the technical field of power system fault detection and early warning, and specifically to an intelligent arc detection and power fault early warning method and system. Background Art

[0002] Arc faults, as common "invisible killers" in modern power systems, their frequency and potential danger cannot be ignored. In order to accurately capture and early warn of this fault, high-frequency electromagnetic wave technology is widely used. However, in actual applications, this technology still faces many challenges, especially the early warning function is still insufficient, and sometimes it is difficult to quickly and timely notify users before an arc fault is about to occur, which may affect the timely handling of faults and the safe operation of the system.

[0003] Therefore, improvement is needed. Summary of the Invention

[0004] To solve the above technical problems, this application provides an intelligent arc detection and power fault early warning method and system.

[0005] The first invention object of this application is achieved through the following technical solutions: An intelligent arc detection and power fault early warning method includes the following steps: S10: Collect historical current data and extract four-dimensional features, where the four-dimensional features include flat shoulder width, peak amplitude, average value, and duration; S20: Calculate the mean and standard deviation of the four-dimensional features and generate a dynamic threshold interval to construct a dynamic threshold model; S30: Collect real-time current data and input it into the dynamic threshold model. When any of the four-dimensional features exceeds the corresponding dynamic threshold interval, it is determined as an arc fault; S40: Generate a probability heat map based on the analysis of the current difference between adjacent sensors and the calculation of preset spatio-temporal weights to locate the fault position; S50: Classify the fault types based on a preset non-linear scoring model. The fault types include continuous faults, short-term faults, and intermittent faults, and generate a feedback report and push it to the terminal.

[0006] In a preferred embodiment, the step of S10: Collect historical current data and extract four-dimensional features, where the four-dimensional features include flat shoulder width, peak amplitude, average value, and duration, includes: S101: Based on a preset historical data collection period, collect historical current data covering different preset load rates and different preset ambient temperature conditions; S102: Based on a preset four-dimensional feature extraction algorithm, extract four-dimensional features from the historical current data, where the four-dimensional features include flat shoulder width and peak amplitude , average value , duration .

[0007] In a preferred embodiment, the step S20: calculating the mean and standard deviation of the four-dimensional features and generating a dynamic threshold interval to construct a dynamic threshold model includes: S201: calculating the mean of the four-dimensional features and the standard deviation , where corresponds to respectively ; S202: dynamically adjusting the threshold interval coefficient based on a preset load rate-temperature correlation rule : , where is a preset initial coefficient, is a preset first environmental correction parameter, is a preset second environmental correction parameter, is a preset third environmental correction parameter; S203: generating a dynamic threshold interval ; S204: constructing a dynamic threshold model based on S201 - S203.

[0008] In a preferred embodiment, the step S30: collecting real-time current data and inputting it into the dynamic threshold model, and when any of the four-dimensional features exceeds the corresponding dynamic threshold interval, determining it as an arc fault includes: S301: collecting real-time current data based on a preset sampling frequency and inputting it into the dynamic threshold model; S302: synchronously calculating the real-time four-dimensional feature values in the dynamic threshold model ; S303: when any of the four-dimensional feature values exceeds the corresponding dynamic threshold interval , that is, satisfying: , determining it as an arc fault and executing step S40; S304: when not satisfying S303, returning to execute step S30.

[0009] In a preferred embodiment, the step S40: generating a probability heat map based on the analysis of the current difference between adjacent sensors and the calculation of preset spatio-temporal weights to locate the fault position includes: S401: collecting the current sequences of sensors in a ring circuit ; S402: calculating the current difference between adjacent sensors , where ; S403: Screen the peak interval that continuously exceeds the dynamic threshold ; S404: Calculate the fault probability based on the preset spatio-temporal weight , where is the preset time decay factor; S405: Based on the preset formula , normalize the probability value ; S406: Based on the preset heatmap rules, , generate a probability heatmap and compare the probability values in the probability heatmap; S407: When the probability value is the largest, the corresponding sensor interval is the fault location.

[0010] In a preferred embodiment, the step S50: Classify the fault types based on the preset non-linear scoring model, where the fault types include continuous faults, short-term faults, and intermittent faults, generate a feedback report and push it to the terminal, includes: SA1: Construct a preset historical fault data set, and the preset historical fault data set includes groups of historical fault data; SA2: The label of the preset historical fault data set is the fault type, and the fault types include continuous faults, short-term faults, and intermittent faults; SA3: Construct a preset neural network model, where the input layer includes four-dimensional features extracted based on the preset historical fault data set ; SA4: The hidden layer includes 8 neurons and an activation function, and the activation function includes ReLU; SA5: The output layer includes a Softmax function and the corresponding probability of the fault type output based on the Softmax function ; SA6: Minimize the cross-entropy loss function based on the preset Adam optimizer: , output the trained preset weights , where is the preset L2 regularization coefficient, is the preset model parameter; SA7: Store the trained preset weights in the preset early warning system database.

[0011] In a preferred embodiment, the step S50: Classify the fault types based on the preset non-linear scoring model, where the fault types include continuous faults, short-term faults, and intermittent faults, generate a feedback report and push it to the terminal, further includes: S501: Calculate the non - linear weighted score based on a preset non - linear scoring model and the trained preset weights , and calculate the non - linear weighted score: ; S502: Classify the fault type based on a preset classification threshold: When is the preset first fault threshold, it is a continuous fault; S503: When is the preset second fault threshold, it is a short - term fault; S504: When , it is an intermittent fault; S505: Generate a feedback report and push it to the terminal. The feedback report includes the fault location, fault type, and corresponding repair suggestions.

[0012] The second inventive object of the present application is achieved through the following technical solutions: The first module includes: Collect historical current data and extract four - dimensional features. The four - dimensional features include the flat shoulder width, peak amplitude, average value, and duration; The first model module: Calculate the mean and standard deviation of the four - dimensional features and generate a dynamic threshold interval, and construct a dynamic threshold model; The first input module: Collect real - time current data and input it into the dynamic threshold model. When any of the four - dimensional features exceeds the corresponding dynamic threshold interval, it is determined as an arc fault; The first positioning module: Generate a probability heat map based on the analysis of the current difference between adjacent sensors and the calculation of preset spatio - temporal weights, and locate the fault position; The first generation module: Classify the fault type based on a preset non - linear scoring model. The fault types include continuous faults, short - term faults, and intermittent faults, generate a feedback report and push it to the terminal.

[0013] The third inventive object of the present application is achieved through the following technical solutions: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above - mentioned intelligent arc detection and power fault warning method.

[0014] The fourth inventive object of the present application is achieved through the following technical solutions: A computer - readable storage medium stores a computer program. When the computer program is executed by a processor, it implements the steps of the above - mentioned intelligent arc detection and power fault warning method.

[0015] In summary, the present application includes at least one of the following beneficial technical effects: First, the system collects historical current data, which is the basis for analyzing the operating state of the power system. By extracting four-dimensional features - flat shoulder width, peak amplitude, average value, and duration, the shape and change trend of the current signal can be comprehensively captured. These features represent the smoothness, mutation degree, overall level, and duration of the current waveform respectively, providing a key information basis for subsequent fault detection. Next, the system calculates the mean and standard deviation of these four-dimensional features, and then generates a dynamic threshold interval to construct a dynamic threshold model. The mean reflects the average level of the features, while the standard deviation describes the fluctuation range of the features. Through the dynamic threshold model, the threshold can be adaptively adjusted according to historical data to adapt to different operating environments and conditions, improving the accuracy and flexibility of fault detection. The real-time current data is input into the dynamic threshold model, and the system will monitor the comparison between these data and the dynamic threshold interval in real time. Once any of the four-dimensional features exceeds its corresponding dynamic threshold interval, the system immediately determines it as an arc fault. This real-time monitoring and dynamic threshold comparison mechanism can quickly and accurately identify arc faults, providing a guarantee for taking timely measures. After determining an arc fault, the system further generates a probability heat map based on the analysis of the current difference between adjacent sensors and the calculation of preset spatio-temporal weights. The current difference analysis can reveal the propagation and influence of the fault between different sensors, while the spatio-temporal weights consider the distribution characteristics of the fault in time and space. Through the probability heat map, the fault location can be visually located, providing a clear direction for fault troubleshooting and repair. S50: Finally, the system classifies the fault types based on a preset non-linear scoring model, including continuous faults, transient faults, and intermittent faults. This classification helps us better understand the nature and impact of the faults, providing a basis for formulating corresponding treatment strategies. The generated feedback report will detail the fault information and analysis results, and notify relevant personnel in a timely manner by pushing it to the terminal, realizing rapid response and handling of faults. Description of the Drawings

[0016] Figure 1 is a flowchart of an implementation of an embodiment of an intelligent arc detection and power fault warning method of the present application; Figure 2 is a flowchart of an implementation of step S10 in an embodiment of an intelligent arc detection and power fault warning method of the present application; Figure 3 is a flowchart of an implementation of step S20 in an embodiment of an intelligent arc detection and power fault warning method of the present application; Figure 4 is a flowchart of an implementation of step S30 in an embodiment of an intelligent arc detection and power fault warning method of the present application; Figure 5 is a flowchart of an implementation of step S40 in an embodiment of an intelligent arc detection and power fault warning method of the present application; Figure 6 is a schematic block diagram of a computer device according to the present application. Specific embodiments

[0017] The following will further elaborate on the present application Figure 1-6 in conjunction with the accompanying drawings.

[0018] In one embodiment, as Figure 1 shown, the present application discloses an intelligent arc detection and power fault warning method, which specifically includes the following steps: S10: Collect historical current data and extract four-dimensional features, where the four-dimensional features include flat shoulder width, peak amplitude, average value, and duration; S20: Calculate the mean and standard deviation of the four-dimensional features and generate a dynamic threshold interval to construct a dynamic threshold model; S30: Collect real-time current data and input it into the dynamic threshold model. When any one of the four-dimensional features exceeds the corresponding dynamic threshold interval, it is determined as an arc fault; S40: Generate a probability heat map based on the analysis of the current difference between adjacent sensors and the calculation of preset spatio-temporal weights to locate the fault position; S50: Classify the fault types based on a preset non-linear scoring model. The fault types include continuous faults, short-term faults, and intermittent faults, and generate a feedback report and push it to the terminal.

[0019] In this embodiment, S10: First, the system collects historical current data, which is the basis for analyzing the operating state of the power system. By extracting four-dimensional features - flat shoulder width, peak amplitude, average value, and duration, the shape and change trend of the current signal can be comprehensively captured. These features respectively represent the smoothness, mutation degree, overall level, and duration of the current waveform, providing a key information basis for subsequent fault detection. S20: Next, the system calculates the mean and standard deviation of these four-dimensional features, and then generates a dynamic threshold interval to construct a dynamic threshold model. The mean reflects the average level of the features, while the standard deviation describes the fluctuation range of the features. Through the dynamic threshold model, the threshold can be adaptively adjusted according to historical data to adapt to different operating environments and conditions, improving the accuracy and flexibility of fault detection. S30: The real-time current data is input into the dynamic threshold model, and the system will monitor the comparison between these data and the dynamic threshold interval in real time. Once any of the four-dimensional features exceeds its corresponding dynamic threshold interval, the system immediately determines it as an arc fault. This real-time monitoring and dynamic threshold comparison mechanism can quickly and accurately identify arc faults, providing a guarantee for taking timely measures. S40: After determining an arc fault, the system further generates a probability heat map based on the analysis of the current difference between adjacent sensors and the calculation of preset spatio-temporal weights. The current difference analysis can reveal the propagation and influence of faults between different sensors, while the spatio-temporal weights consider the distribution characteristics of faults in time and space. Through the probability heat map, the fault location can be intuitively located, providing a clear direction for fault troubleshooting and repair. S50: Finally, the system classifies the fault types based on a preset non-linear scoring model, including continuous faults, transient faults, and intermittent faults. This classification helps us better understand the nature and impact of faults, providing a basis for formulating corresponding treatment strategies. The generated feedback report will detail the fault information and analysis results, and notify relevant personnel in a timely manner by pushing it to the terminal, realizing rapid response and handling of faults.

[0020] Figure 2 , step S10 includes: S101: Based on a preset historical data collection period, collect historical current data covering different preset load rates and different preset environmental temperature conditions; S102: Based on a preset four-dimensional feature extraction algorithm, extract four-dimensional features from the historical current data, and the four-dimensional features include flat shoulder width , peak amplitude , average value , and duration .

[0021] In this embodiment, S101: The system collects historical current data covering different load rates and ambient temperature conditions according to a preset historical data collection period. The principle of this step is that the operating state of the power system is greatly affected by factors such as load rate and ambient temperature. By collecting data under different working conditions, the comprehensiveness and adaptability of the subsequent analysis model can be ensured. S102: After obtaining rich historical current data, the system uses a preset four-dimensional feature extraction algorithm to extract the flat shoulder width , peak amplitude , average value and duration of these four key features. These features respectively represent specific attributes of the current waveform, such as the smoothness of the waveform, the severity of the mutation, the overall current level, and the duration of the anomaly, providing an important data basis for subsequent fault detection and early warning. Through step S101, the system collects historical current data covering a variety of working conditions, ensuring that the input data of the analysis model is representative and comprehensive, thereby improving the accuracy of fault detection. The four-dimensional features extracted in step S102 can accurately describe the abnormal conditions of the current waveform, providing effective input for the subsequent dynamic threshold model and making the fault detection more accurate and sensitive. These two steps provide a solid data and feature basis for subsequent fault detection, location, and classification, and are an important prerequisite for the entire intelligent arc detection and power fault early warning method.

[0022] Figure 3 , step S20 includes: S201: Calculate the mean and standard deviation of the four-dimensional features, where respectively correspond to ; S202: Dynamically adjust the threshold interval coefficient based on a preset load rate-temperature correlation rule: , where is a preset initial coefficient, is a preset first environmental correction parameter, is a preset second environmental correction parameter, is a preset third environmental correction parameter; S203: Generate a dynamic threshold interval ; S204: Based on S201 - S203, construct a dynamic threshold model.

[0023] In this embodiment, S201: The system first extracts the four-dimensional features (flat shoulder width , peak amplitude , average value and duration ), perform statistical analysis and calculate its mean value and standard deviation . The mean value reflects the average level of the feature, while the standard deviation describes the fluctuation range of the feature. These two provide the basic data for the subsequent setting of the dynamic threshold. S202: Considering that the operating state of the power system is affected by the load rate and environmental temperature, the system dynamically adjusts the threshold interval coefficient based on the preset load rate-temperature correlation rule. This correlation rule is obtained based on historical data and experience summary, which can ensure that the threshold interval is adaptively adjusted with the change of working conditions, improving the accuracy of fault detection. S203: After obtaining the adjusted threshold interval coefficient , the system generates a dynamic threshold interval . These threshold intervals are not fixed, but dynamically change according to the real-time working conditions, and can more flexibly adapt to the complex operating environment of the power system. S204: Based on the above steps, the system constructs a dynamic threshold model. This model combines the statistical characteristics of historical data and the dynamic adjustment of real-time working conditions, and can more accurately identify arc faults. By dynamically adjusting the threshold interval coefficient, the system can adapt to the working conditions under different load rates and environmental temperatures, ensuring the accuracy and reliability of fault detection. Based on the statistical characteristics of historical data and the dynamic adjustment of real-time working conditions, the dynamic threshold model can more accurately identify arc faults, reducing the situations of false alarms and missed alarms. The dynamic threshold model is not fixed, but dynamically changes according to the real-time working conditions, and can more flexibly adapt to the complex operating environment of the power system. The constructed dynamic threshold model provides an important tool and basis for the subsequent real-time current data analysis and fault determination.

[0024] Figure 4 , Step S30 includes: S301: Collect real-time current data based on the preset sampling frequency and input it into the dynamic threshold model; S302: In the dynamic threshold model, synchronously calculate the real-time four-dimensional eigenvalue ; S303: When any one of the four-dimensional eigenvalues exceeds the corresponding dynamic threshold interval , that is, it satisfies: , it is determined as an arc fault, and execute step S40; S304: When S303 is not satisfied, return to execute step S30.

[0025] In this embodiment, S301: The system collects current data in real time based on a preset sampling frequency. These data are an important basis for analyzing the operating state of the power system. The collected data is input into the dynamic threshold model to provide real-time and accurate data support for subsequent eigenvalue calculation and fault determination. S302: In the dynamic threshold model, the system synchronously calculates real-time four-dimensional eigenvalues , which include the flat shoulder width, peak amplitude, average value, and duration, and can comprehensively and accurately reflect the instantaneous changes and abnormal conditions of the current signal. S303: The system compares the calculated real-time four-dimensional eigenvalues with a preset dynamic threshold range. When any eigenvalue exceeds the corresponding threshold range , it is determined as an arc fault. This dynamic threshold setting can be automatically adjusted according to the actual operating conditions of the power system, improving the accuracy and flexibility of fault determination. S304: If the real-time four-dimensional eigenvalues do not exceed the corresponding dynamic threshold range, the system determines that no arc fault has occurred at present, and continues to return to step S30 to maintain continuous monitoring and real-time analysis of the power system. Through the implementation steps of S301 - S304, the system realizes precise acquisition, feature extraction, and fault determination of real-time current data of the power system. This method based on the dynamic threshold model and real-time feature calculation can effectively capture the early signals of arc faults, issue early warnings in a timely manner, and avoid potential safety hazards. At the same time, the continuous monitoring and loop determination mechanism of the system ensure the stable operation of the power system, improving the reliability and safety of power supply. In addition, the application of the dynamic threshold enables the system to adapt to different operating environments and working conditions, improving the accuracy and flexibility of fault determination.

[0026] Figure 5 , Step S40 includes: S401: Collect the current sequences of sensors in the loop circuit ; S402: Calculate the current differences between adjacent sensors , where ; S403: Screen the peak intervals that continuously exceed the dynamic threshold ; S404: Calculate the fault probability based on a preset spatio-temporal weight , where is a preset time decay factor; S405: Based on a preset formula , normalize the probability value ; S406: Based on a preset heat map rule, , generate a probability heat map and compare the probability values in the probability heat map; S407: When the probability value is the largest, the corresponding sensor interval is the fault location.

[0027] In this embodiment, S401: The system first collects the current sequences of each sensor in the ring circuit , and these sequence data reflect the current changes at each point in the circuit, which is an important basis for locating the fault location. S402: By calculating the current differences between adjacent sensors , abnormal fluctuations of the current can be captured, and these fluctuations may be caused by faults. The magnitude and change trend of the current differences provide key bases for subsequent fault determination. S403: The system filters out the peak intervals that continuously exceed the dynamic threshold , and these intervals represent significant abnormalities of the current and are very likely to be the areas where faults occur. S404: Based on the preset spatio-temporal weights, the system calculates the fault probability . The spatio-temporal weights consider the temporal and spatial distributions of current abnormalities, making the calculation of the fault probability more accurate. S405: Through a preset formula to normalize the probability value , making the probability values between different sensors comparable. S406: The system generates a probability heat map according to the preset heat map rules, intuitively showing the fault probabilities of each sensor interval. By comparing the probability values in the heat map, the areas where faults may occur can be preliminarily judged. S407: When the probability value is the largest, the corresponding sensor interval is determined as the fault location. This method utilizes the principle of probability statistics, improving the accuracy and reliability of fault location. Through the implementation steps of S401 - S407, the system realizes the precise location of faults in the power system. First, by collecting and analyzing the current sequence data, the system can timely detect abnormal changes in the current. Secondly, by using the dynamic threshold and spatio-temporal weights to calculate the fault probability, the accuracy and flexibility of fault determination are improved. Finally, by generating the probability heat map and comparing the probability values, the system can intuitively and accurately locate the fault location. The implementation effect of this method is remarkable, not only improving the efficiency of fault troubleshooting, but also reducing the economic losses and potential safety hazards caused by faults, and enhancing the stability and reliability of the power system.

[0028] The S50 steps include: SA1: Construct a preset historical fault data set, and the preset historical fault data set includes groups of historical fault data; SA2: The label of the preset historical fault data set is the fault type, and the fault type includes continuous faults, short-term faults, and intermittent faults; SA3: Construct a preset neural network model, where the input layer includes four-dimensional features extracted based on the preset historical fault data set ; SA4: The hidden layer includes 8 neurons and an activation function, and the activation function includes ReLU; SA5: The output layer includes a Softmax function and the corresponding probabilities of the fault types output based on the Softmax function ; SA6: Minimize the cross-entropy loss function based on a preset Adam optimizer: , and output the preset weights after training , where is a preset L2 regularization coefficient, is a preset model parameter; SA7: Store the preset weights after training into a preset warning system database.

[0029] In this embodiment, SA1: Constructing a preset historical fault data set is to provide sufficient training data so that the neural network model can learn the characteristics of different fault types. The historical fault data includes parameters such as current and voltage in various fault situations, providing real and diverse inputs for model training. SA2: Labeling the fault types for the data set, including continuous faults, short-term faults, and intermittent faults, is to enable the model to clearly distinguish these fault types during training, so that accurate classification can be achieved during prediction. SA3: Constructing a preset neural network model, where the input layer extracts four-dimensional features from the historical fault data set, and these features are parameters that have been screened and verified and have an important impact on fault type determination. SA4: The design of the hidden layer, including 8 neurons and a ReLU activation function, aims to extract the non-linear relationships in the data and enhance the representation ability of the model. SA5: The output layer uses a Softmax function, which can output the corresponding probabilities of each fault type , facilitating subsequent fault type determination. SA6: Minimize the cross-entropy loss function based on the Adam optimizer , which is to continuously adjust the model parameters during training to make the fault type probabilities predicted by the model as close as possible to the true labels. The settings of the preset L2 regularization coefficient and model parameters are to prevent the model from overfitting and improve the generalization ability of the model. SA7: Store the preset weights Stored in the early warning system database, it is for quickly loading the model during real-time fault detection to predict the fault type. Through the implementation steps of SA1 - SA7, the system successfully constructed a non-linear scoring model based on neural network, which can accurately classify continuous faults, transient faults and intermittent faults. The specific effects are as follows: The model learned the characteristics of different fault types through training, and can classify real-time fault data in a short time, improving the efficiency of fault handling. The model adopted an advanced neural network structure and optimization algorithm, with strong non-linear relationship extraction ability and generalization ability, and can accurately predict the fault type. The feedback report generated by the model details the fault type and related information, providing valuable reference for operation and maintenance personnel and helping to quickly locate and solve faults. The model architecture and parameter settings have a certain degree of flexibility and can be adjusted and optimized according to actual needs to adapt to different application scenarios.

[0030] Step S50 also includes: S501: Based on the preset non-linear scoring model and the trained preset weights , calculate the non-linear weighted score: ; S502: Classify the fault type based on the preset classification threshold: When is the preset first fault threshold, it is a continuous fault; S503: When is the preset second fault threshold, it is a transient fault; S504: When is the case, it is an intermittent fault; S505: Generate a feedback report and push it to the terminal. The feedback report includes the fault location, fault type and corresponding repair suggestions.

[0031] In this embodiment, S501: Use the preset non-linear scoring model and the trained preset weights , non - linear weighted scoring calculation is performed on the real - time collected fault data. This step deeply learns the complex features of the input data through a model to obtain a comprehensive score reflecting the severity and type of the fault. S502 - S504: According to the preset classification thresholds, the calculated non - linear weighted score is compared with these thresholds to classify the fault types. Different thresholds correspond to different fault types, such as continuous faults, transient faults, and intermittent faults. This threshold classification method is simple and effective and can quickly determine the fault type. S505: Generate a feedback report based on the fault location, fault type, and corresponding repair suggestions, and push it to the terminal. The generation of the feedback report is based on the results of the previous fault detection and classification, providing detailed fault information and repair suggestions for the operation and maintenance personnel, facilitating their quick response and handling. Through the non - linear scoring model and the threshold classification method, the system can accurately distinguish different types of faults, providing clear guidance for subsequent fault handling. The feedback report generated by the system is pushed to the terminal in real - time to ensure that the operation and maintenance personnel can obtain the fault information in a timely manner, shortening the fault response time. The repair suggestions included in the feedback report are based on historical data and expert experience, providing effective guidance for the operation and maintenance personnel to handle faults and improving the fault repair efficiency. Quick and accurate fault classification and real - time feedback mechanism contribute to the timely handling of faults in the power system, reducing the safety risks and economic losses caused by faults.

[0032] It should be understood that the magnitudes of the sequence numbers of the steps in the above - mentioned embodiments do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0033] In one embodiment, an intelligent arc detection and power fault warning system is provided. This intelligent arc detection and power fault warning system corresponds to the intelligent arc detection and power fault warning method in the above - mentioned embodiment. This intelligent arc detection and power fault warning system includes: The first including module: Collect historical current data and extract four - dimensional features, where the four - dimensional features include flat shoulder width, peak amplitude, average value, and duration; The first model module: Calculate the mean and standard deviation of the four - dimensional features and generate a dynamic threshold interval, and construct a dynamic threshold model; The first input module: Collect real - time current data and input it into the dynamic threshold model. When any one of the four - dimensional features exceeds the corresponding dynamic threshold interval, it is determined as an arc fault; The first positioning module: Based on the analysis of the current difference between adjacent sensors and the calculation of preset spatio - temporal weights, generate a probability heat map to locate the fault position; The first generation module: Classify the fault types based on a preset non - linear scoring model. The fault types include continuous faults, transient faults, and intermittent faults, generate a feedback report and push it to the terminal.

[0034] Optionally, it further includes: The first acquisition module: Based on a preset historical data acquisition period, acquire historical current data covering different preset load rates and different preset environmental temperature conditions; The second inclusion module: Based on a preset four-dimensional feature extraction algorithm, extract four-dimensional features from the historical current data, and the four-dimensional features include the flat shoulder width , peak amplitude , average value , and duration .

[0035] Optionally, it further includes: The first calculation module: Calculate the mean value and standard deviation of the four-dimensional features, where corresponds to respectively; The first adjustment module: Dynamically adjust the threshold interval coefficient : , where is a preset initial coefficient, is a preset first environmental correction parameter, is a preset second environmental correction parameter, is a preset third environmental correction parameter; The second generation module: Generate a dynamic threshold interval ; The second model module: Based on S201 - S203, construct a dynamic threshold model.

[0036] Optionally, it further includes: The input module: Acquire real-time current data based on a preset sampling frequency and input it into the dynamic threshold model; The third calculation module: In the dynamic threshold model, synchronously calculate the real-time four-dimensional feature value ; The second determination module: When any one of the four-dimensional feature values exceeds the corresponding dynamic threshold interval , that is, it satisfies: , it is determined as an arc fault, and step S40 is executed; The first return module: When S303 is not satisfied, return to execute step S30.

[0037] Optionally, it further includes: The second acquisition module: Acquire the current sequences of sensors in the loop circuit ; Fourth calculation module: Calculate the current difference between adjacent sensors , where ; First screening module: Screen the peak intervals that continuously exceed the dynamic threshold ; Fifth calculation module: Calculate the fault probability based on the preset spatio-temporal weight , where is the preset time decay factor; Normalization module: Based on the preset formula , normalize the probability value ; First comparison module: Based on the preset heat map rule, , generate a probability heat map and compare the probability values in the probability heat map; Location module: When the probability value is the largest, the corresponding sensor interval is the fault location.

[0038] Optionally, it further includes: Third inclusion module: Construct a preset historical fault dataset, and the preset historical fault dataset includes groups of historical fault data; Fourth inclusion module: The label of the preset historical fault dataset is the fault type, and the fault type includes continuous fault, short-term fault, and intermittent fault; Input layer module: Construct a preset neural network model, where the input layer includes four-dimensional features extracted based on the preset historical fault dataset ; First layer module: The hidden layer includes 8 neurons and an activation function, and the activation function includes ReLU; Second layer module: The output layer includes a Softmax function and the corresponding probability of the fault type output based on the Softmax function ; First output module: Minimize the cross-entropy loss function based on the preset Adam optimizer: , and output the trained preset weight , where is the preset L2 regularization coefficient, is the preset model parameter; First storage module: Store the trained preset weight in the preset early warning system database.

[0039] Optionally, it further includes: Sixth calculation module: Calculate the non-linear weighted score based on the preset non-linear scoring model and the trained preset weight : ; Classification module: Classify the fault type based on a preset classification threshold: When is the preset first fault threshold, it is a continuous fault; First classification module: When is the preset second fault threshold, it is a transient fault; Second classification module: When it is an intermittent fault; Sixth module: Generate a feedback report and push it to the terminal. The feedback report includes the fault location, fault type, and corresponding repair suggestions.

[0040] For the specific limitations of an intelligent arc detection and power fault warning system, reference can be made to the limitations of an intelligent arc detection and power fault warning method in the above text, which will not be elaborated here. Each module in the above intelligent arc detection and power fault warning system can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0041] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 6 shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store fault types. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements an intelligent arc detection and power fault warning method.

[0042] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, an intelligent arc detection and power fault warning method is implemented.

[0043] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, an intelligent arc detection and power fault warning method is implemented.

[0044] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0045] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

Claims

1. An intelligent arc detection and power fault early warning method, characterized in that: The following steps are involved: S10: Collect historical current data and extract four-dimensional features, where the four-dimensional features include flat shoulder width, peak amplitude, average value, and duration; S20: Calculate the mean and standard deviation of the four-dimensional features and generate a dynamic threshold interval to construct a dynamic threshold model; S30: collecting real-time current data and inputting it into a dynamic threshold model, and determining it as an arc fault when any of the four-dimensional features exceeds the corresponding dynamic threshold interval; S40: Based on the current difference analysis of adjacent sensors and the preset time-space weight calculation, a probability heat map is generated to locate the fault position; S50: Classify the fault type based on a preset nonlinear scoring model, where the fault type includes a continuous fault, a short-term fault, and an intermittent fault, generate a feedback report, and push it to the terminal.

2. The intelligent arc detection and power fault early warning method according to claim 1 is characterized in that: The step of S10: collecting historical current data and extracting four-dimensional features, wherein the four-dimensional features include flat shoulder width, peak amplitude, average value, and duration, includes: S101: Based on a preset historical data collection cycle, historical current data covering different preset load rates and different preset ambient temperature conditions are collected; S102: Based on a preset four-dimensional feature extraction algorithm, extract four-dimensional features from historical current data, wherein the four-dimensional features include flat shoulder width , peak amplitude ,average value , Duration .

3. The intelligent arc detection and power fault early warning method according to claim 1 is characterized in that: The step S20: calculating the mean and standard deviation of the four-dimensional features and generating a dynamic threshold interval, and constructing a dynamic threshold model comprises: S201: Calculate the mean of the four-dimensional features and standard deviation ,in Corresponding to ; S202: Dynamically adjust the threshold interval coefficient based on the preset load rate-temperature association rule : ,in To preset the initial coefficient, To preset the first environment correction parameter, To preset the second environment correction parameters, To preset the third environment correction parameters; S203: Generate dynamic threshold interval ; S204: Based on S201-S203, a dynamic threshold model is constructed.

4. The intelligent arc detection and power fault early warning method according to claim 1, characterized in that: The step S30: collecting real-time current data and inputting it into a dynamic threshold model, and determining that it is an arc fault when any of the four-dimensional features exceeds the corresponding dynamic threshold interval, includes: S301: Collecting real-time current data based on a preset sampling frequency and inputting it into a dynamic threshold model; S302: In the dynamic threshold model, synchronously calculate the real-time four-dimensional eigenvalue ; S303: When any of the four-dimensional eigenvalues Exceeds the corresponding dynamic threshold interval When , it satisfies: , it is determined to be an arc fault, and step S40 is executed; S304: When S303 is not satisfied, return to step S30.

5. The intelligent arc detection and power fault early warning method according to claim 1, characterized in that: The step S40: generating a probability heat map based on the current difference analysis of adjacent sensors and the preset time-space weight calculation to locate the fault position includes: S401: Acquisition Ring Circuit The current sequence of the sensors ; S402: Calculate the current difference between adjacent sensors ,in ; S403: Screening for continuous exceeding of dynamic threshold The peak interval of S404: Calculating the failure probability based on the preset spatiotemporal weights ,in is the preset time decay factor; S405: Based on preset formula , normalized probability value ; S406: Based on the preset heat map rules, , generate a probability heat map and compare the probability values ​​in the probability heat map; S407: When the probability value is the largest, the corresponding sensor section is the fault location.

6. The intelligent arc detection and power fault early warning method according to claim 1, characterized in that: The step S50: classifying the fault type based on the preset nonlinear scoring model, wherein the fault type includes a continuous fault, a short-term fault, and an intermittent fault, and generating a feedback report and pushing it to the terminal comprises: SA1: Construct a preset historical fault data set, which includes Group historical fault data; SA2: The label of the preset historical fault data set is the fault type, and the fault type includes continuous fault, short-term fault, and intermittent fault; SA3: Construct a preset neural network model, where the input layer includes four-dimensional features extracted based on the preset historical fault data set ; SA4: The hidden layer includes 8 neurons and an activation function, wherein the activation function includes ReLU; SA5: The output layer includes the Softmax function and the corresponding probability of the fault type based on the output of the Softmax function ; SA6: Minimize the cross entropy loss function based on the preset Adam optimizer: , output the preset weight after training ,in, is the preset L2 regularization coefficient, is the preset model parameters; SA7: Set the preset weights after training Stored in the preset warning system database.

7. The intelligent arc detection and power fault early warning method according to claim 1, characterized in that: The step S50: classifying the fault type based on a preset nonlinear scoring model, wherein the fault type includes a continuous fault, a short-term fault, and an intermittent fault, generating a feedback report and pushing it to the terminal, further includes: S501: Based on the preset nonlinear scoring model and the trained preset weights , calculate the nonlinear weighted score: ; S502: Classify the fault type based on the preset classification threshold: When the first fault threshold is preset, it is a continuous fault; S503: When When the second fault threshold is preset, it is a short fault; S504: When When , it is an intermittent fault; S505: Generate a feedback report and push it to the terminal, where the feedback report includes the fault location, fault type, and corresponding repair suggestions.

8. An intelligent arc detection and power fault warning system, characterized in that: include: The first module includes: collecting historical current data and extracting four-dimensional features, wherein the four-dimensional features include flat shoulder width, peak amplitude, average value, and duration; The first model module: calculates the mean and standard deviation of the four-dimensional features and generates a dynamic threshold interval to construct a dynamic threshold model; The first input module collects real-time current data and inputs it into the dynamic threshold model. When any of the four-dimensional features exceeds the corresponding dynamic threshold interval, it is determined to be an arc fault. The first positioning module: generates a probability heat map based on the current difference analysis of adjacent sensors and the preset time and space weight calculation to locate the fault position; The first generation module: classifies the fault type based on a preset nonlinear scoring model, wherein the fault type includes a continuous fault, a short-term fault, and an intermittent fault, generates a feedback report, and pushes it to the terminal.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of an intelligent arc detection and power fault warning method as claimed in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the intelligent arc detection and power fault warning method according to any one of claims 1 to 7.

Images