A street lamp line ground fault intelligent diagnosis system and method
By optimizing the combination of variational mode decomposition and kernel extreme learning machine classifier using the parrot optimization algorithm, the problems of low accuracy and large influence of human factors in the diagnosis of grounding faults in street light lines are solved, and efficient and reliable fault identification is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHENGZHOU XUEFU ELECTRONICS ENG TECH CO LTD
- Filing Date
- 2026-02-02
- Publication Date
- 2026-06-05
AI Technical Summary
The existing street light line grounding fault diagnosis system has low accuracy, is greatly affected by human factors, and the existing methods cannot effectively identify grounding faults, resulting in frequent safety hazards.
The Parrot optimization algorithm is used to optimize the variational mode decomposition method to preprocess the residual current data. A multidimensional feature vector is constructed by combining the Pearson correlation coefficient and the sample entropy value. A kernel extreme learning machine classifier is used for fault diagnosis to establish the mapping relationship between the feature vector and the fault state.
It enables rapid and accurate diagnosis of grounding faults in street light lines, improving the accuracy of diagnosis and anti-interference capabilities, and is suitable for safety monitoring and fault identification of urban lighting systems.
Smart Images
Figure CN122153560A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power equipment operation monitoring technology, and more specifically to an intelligent diagnostic system and method for grounding faults in street light lines. Background Technology
[0002] Grounding faults are a major cause of leakage current hazards in street light circuits, leading to energized street light poles. When the leakage current exceeds the safety threshold for electric shock, contact with the human body poses a fatal risk, especially during the rainy season when street light electric shock accidents are frequent. According to the requirements of "Installation and Operation of Residual Current Operated Protective Devices," outdoor electrical installations must be equipped with residual current operated protective devices. However, due to the long length of street light circuits, the grounding fault current is relatively small, and the distributed capacitance current to the ground is relatively large during normal operation, the total residual current easily exceeds the typical operating threshold of the residual current operated protective device, causing frequent malfunctions. In actual engineering projects, residual current operated protective devices are often removed, resulting in a lack of grounding protection and electric shock protection, posing a serious safety hazard.
[0003] Currently, intelligent diagnosis of grounding faults in street light lines is insufficient, and the accuracy and stability of diagnostic methods need to be improved, failing to meet the application needs in practical engineering. In terms of data processing, variational mode decomposition is an effective method for non-stationary data analysis, but the combination of key parameters before decomposition generally needs to be set manually, which has the disadvantage of being greatly affected by human factors. Inappropriate parameter settings can lead to mode aliasing, over-decomposition, or under-decomposition, seriously affecting the extraction of fault features and the accuracy of subsequent diagnosis.
[0004] Therefore, overcoming the limitations of human experience, determining the optimal data decomposition parameters, and achieving accurate and reliable diagnosis of grounding faults in street light lines have become the focus of efforts for those skilled in the art. Summary of the Invention
[0005] The purpose of this invention is to address the problems of unreliability of existing fault protection systems, significant influence of human factors, and low accuracy of existing street light grounding fault diagnosis systems and methods in fault diagnosis. This invention provides an intelligent diagnosis system and method for street light line grounding faults.
[0006] To achieve the above objectives, the present invention specifically adopts the following technical solution: A smart diagnostic system for grounding faults in street light lines includes a data acquisition and preprocessing unit, a feature extraction unit, and a fault diagnosis unit. The data acquisition and preprocessing unit acquires raw residual current data of the street light lines using a residual current transformer and an oscilloscope. It then preprocesses the residual current data using a variational mode decomposition method optimized by the Parrot optimization algorithm. The processed data is input into the feature extraction unit, which selects components with high correlation to fault characteristics based on the Pearson correlation coefficient. Subsequently, it calculates the sample entropy value to construct a multidimensional feature vector set representing the line state, which is then input into the fault diagnosis unit. The fault diagnosis unit uses historical data to train a kernel extreme learning machine classifier, establishing a mapping relationship between feature vectors and fault states, thereby achieving rapid and accurate classification of real-time data features and fault diagnosis of the lines.
[0007] Furthermore, the data acquisition and preprocessing unit includes a data acquisition module and a data preprocessing module, which acquire raw residual current data and decompose the raw data to better extract fault information from the data.
[0008] Furthermore, the fault diagnosis unit should issue a fault alarm signal so that relevant personnel can handle the target fault in a timely manner.
[0009] Furthermore, the data preprocessing module employs one or more of the frequency domain method and the spatial domain method.
[0010] To achieve the above objectives, the present invention provides the following technical solution: an intelligent diagnostic method for grounding faults in street light lines, comprising the following steps: S1: Collect residual current time-domain data of the line under normal and various fault conditions; S2: The Parrot optimization algorithm is used to optimize the parameters of variational mode decomposition. The population size, number of iterations and parameter search range of the Parrot optimization algorithm are set. The minimum envelope entropy of the decomposed data is used as the fitness function. The optimal parameter combination is found through iteration. Using this optimal parameter, variational mode decomposition is performed on all the collected residual current data to obtain several intrinsic mode components. S3: Calculate the Pearson correlation coefficient between each intrinsic mode component and the corresponding original data, select the effective components with a correlation coefficient greater than the threshold, calculate the sample entropy of these effective components respectively, combine the sample entropy values of all data to form a feature vector sample set, and normalize and label it. S4: Divide the feature vector sample set into a training set and a test set. Use the training set to train the classifier and optimize its parameters. Finally, use the trained model for real-time or offline diagnosis. Input the feature vector of new data to output the diagnostic results.
[0011] Furthermore, the preprocessing of the raw residual current data in S2 includes the following steps: Step S21: Optimize the parameters of variational mode decomposition using the parrot optimization algorithm; Step S22: Decompose the original residual current data using the improved variational mode decomposition.
[0012] Furthermore, the feature extraction of the data in S3 includes the following steps: Step S31: Screen intrinsic mode components using Pearson correlation coefficient; Step S32: Calculate the sample entropy of different intrinsic mode components and construct a multidimensional feature vector set.
[0013] Furthermore, the specific steps in S4 for training the classifier using the kernel extreme learning machine model are as follows: Step S41: Divide the constructed multidimensional feature vector set into a training set and a test set, and label the sample categories in the training set to form a label file; Step S42: Train the kernel extreme learning machine classification model using the training set; Step S43: Test the test machine using the trained kernel extreme learning machine classification model to obtain the classification results.
[0014] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention uses the parrot optimization algorithm to adaptively determine the optimal decomposition parameters of VMD, which overcomes the limitations of manually setting parameters, ensures the quality of data decomposition, lays the foundation for extracting high-quality fault features, and solves the problems of lack of reliability of existing fault protection, large influence of human factors, and low accuracy of existing street light grounding fault diagnosis systems and methods in existing fault diagnosis. It also uses the calculation of its sample entropy value to construct a multi-dimensional feature vector set representing the line state, and inputs it into the fault diagnosis unit. The fault diagnosis unit uses the feature vector set to train the kernel extreme learning machine classifier to realize the rapid and accurate classification of real-time data features and line fault diagnosis. It can effectively identify the normal and grounding fault states of street light lines, and has the advantages of high diagnostic accuracy, strong anti-interference ability, and good real-time performance. It is suitable for safety monitoring and fault identification of urban lighting systems.
[0015] (2) The present invention uses sample entropy as the core feature quantity, which can effectively capture the change in the complexity of residual current data caused by grounding fault. The feature discrimination is significantly higher than that of the effective value or harmonic features of residual current in traditional street light systems.
[0016] (3) The present invention uses a kernel extreme learning machine as a classifier, which combines the advantages of fast learning speed and strong generalization ability of kernel method, and reduces the impact of environmental factors on the identification of fault target features to a certain extent.
[0017] (4) This invention provides a complete system and method from data acquisition and intelligent processing to fault diagnosis. It does not rely on fixed thresholds, has strong adaptability, and can effectively distinguish between normal leakage current and fault current, providing an effective technical solution for solving the problem of safety protection of street light circuits. Attached Figure Description
[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a system framework diagram as described in Embodiment 1 of the present invention; Figure 2 This is a structural diagram of the data acquisition and preprocessing unit provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the data acquisition circuit provided in Embodiment 2 of the present invention; Figure 4 This is a flowchart of the variational mode decomposition optimal parameter solution provided in Embodiment 4 of the present invention; Figure 5 This is a structural diagram of the feature extraction unit provided in Embodiment 1 of the present invention; Figure 6 This is a structural diagram of the intelligent diagnostic unit provided in Embodiment 1 of the present invention. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention. Example 1
[0021] like Figure 1As shown, an intelligent diagnostic system for grounding faults in street light lines includes a data acquisition and preprocessing unit, a feature extraction unit, and a fault diagnosis unit. The data acquisition and preprocessing unit acquires raw residual current data of the street light lines using a residual current transformer and an oscilloscope. It then preprocesses the residual current data using a variational mode decomposition method optimized by the Parrot optimization algorithm. The processed data is then input into the feature extraction unit, which selects components with high correlation to fault features based on the Pearson correlation coefficient. Subsequently, it calculates the sample entropy value to construct a multidimensional feature vector set representing the line state, which is then input into the fault diagnosis unit. The fault diagnosis unit uses historical data to train a kernel extreme learning machine classifier, establishing a mapping relationship between feature vectors and fault states, thereby achieving rapid and accurate classification of real-time data features and line fault diagnosis.
[0022] After acquiring and preprocessing the raw residual current data, the data acquisition and preprocessing unit inputs the processed data into the feature extraction unit; for example... Figure 5 As shown, the feature extraction unit, based on the Pearson correlation coefficient and sample entropy algorithm, selects effective components and calculates their feature values, constructs a multi-dimensional feature vector, and inputs it into the intelligent diagnostic unit; as shown... Figure 6 As shown, the intelligent diagnostic unit first receives the feature vector, and then trains and predicts the classification model constructed by the kernel extreme learning machine. The trained classifier completes the diagnosis and early warning of grounding faults.
[0023] In this example, the Parrot Optimization Algorithm is used to progressively optimize the target parameters. This algorithm has a fast convergence speed, its random structure effectively avoids getting trapped in local optima, and it has a strong global optimization capability.
[0024] The parrot optimization algorithm updates the population location primarily through foraging behavior, maintenance behavior, communication behavior, and avoidance of strangers behavior. For example, foraging behavior mainly estimates the approximate location of food by referring to the location of the entire population or considering the owner's location, and then the parrots fly to their respective locations. The formula for location change is: (1) (2) In the formula, This represents the individual's current location. The updated position; This is the optimal location currently being sought; for Distribution reflects the parrots' flight patterns; D represents the individual dimension. To observe the location of the entire population; This represents the maximum number of iterations. This represents the average position of the entire population at present. is the number of iterations; N is the population size.
[0025] The Parrot optimization algorithm selects the minimum envelope entropy as the fitness function. Envelope entropy serves as a standard for evaluating signal sparsity; a smaller local entropy value indicates richer fault features in the component signals. When the envelope entropy value is minimized, the optimal combination of the number of decomposition levels and the penalty factor in variational mode decomposition is achieved. The formula for calculating the minimum envelope entropy is: (3) In the formula, n is the number of signal samples; Pi is the envelope amplitude of the signal under the Hilbert transform.
[0026] This example uses the Pearson correlation coefficient to screen the K intrinsic mode components with high correlation, i.e., the components with richer fault characteristics. The formula for calculating the Pearson correlation coefficient is: (4) In the formula, , These are the original signal and the decomposed sub-signals, respectively. , These are the means of the corresponding variables.
[0027] In this example, the sample entropy values of each intrinsic mode component after screening are calculated and used as the fault feature vector. Sample entropy is a feature quantity that characterizes the complexity of a time series. Sample entropy makes up for the shortcomings of traditional approximate entropy. Compared with other entropy calculation methods, the sample entropy calculation process requires less data and is more robust. The formula for calculating sample entropy is: (5) In the formula, The dimension that constitutes the vector space for the signal; U is the tolerance; U is the length of the time series; for The proportion of similar points between two sequences in a given dimension; for The proportion of similar points between two sequences in a given dimension. Example 2
[0028] like Figure 2 As shown, based on Embodiment 1, the data acquisition and preprocessing unit includes a data acquisition module and a data preprocessing module, such as... Figure 3 As shown, the data acquisition module acquires the raw residual current data of the street light circuit through a residual current transformer and an oscilloscope. The data preprocessing module uses a variational mode decomposition method based on the parrot optimization algorithm to preprocess the residual current data. The data preprocessing module uses one or more of the frequency domain method and the spatial domain method.
[0029] Preferably, the data preprocessing module uses the parrot optimization algorithm to optimize the parameters of variational mode decomposition, making the fault information of the residual current data after decomposition more concentrated and reducing the influence of manually determined parameters. Example 3
[0030] This embodiment of the intelligent diagnosis method for grounding faults in street light lines is based on the intelligent diagnosis system for grounding faults in street light lines in Embodiment 1. The intelligent diagnosis method for grounding faults in street light lines includes the following steps: S1: Collect residual current time-domain data of the line under normal and various fault conditions; S2: The Parrot optimization algorithm is used to optimize the parameters of variational mode decomposition. The population size, number of iterations and parameter search range of the Parrot optimization algorithm are set. The minimum envelope entropy of the decomposed data is used as the fitness function. The optimal parameter combination is found through iteration. Using this optimal parameter, variational mode decomposition is performed on all the collected residual current data to obtain several intrinsic mode components. S3: Calculate the Pearson correlation coefficient between each intrinsic mode component and the corresponding original data, select the effective components with a correlation coefficient greater than the threshold, calculate the sample entropy of these effective components respectively, combine the sample entropy values of all data to form a feature vector sample set, and normalize and label it. S4: Divide the feature vector sample set into a training set and a test set. Use the training set to train the classifier and optimize its parameters. Finally, use the trained model for real-time or offline diagnosis. Input the feature vector of new data to output the diagnostic results. Example 4
[0031] Based on Example 3, the preprocessing of the raw residual current data in S2 includes the following steps: Step S21: As Figure 4 As shown, the parameters of variational mode decomposition are optimized using the parrot optimization algorithm; Step S22: Decompose the original residual current data using the improved variational mode decomposition.
[0032] Variational mode decomposition is used to preprocess the original residual current signal. This process can effectively separate background noise, power frequency fundamental wave and interference components in the fault process, significantly improve the time and frequency resolution of the signal, and highlight the weak fault characteristics in the original waveform. The parrot optimization algorithm is used to determine the optimal parameters of variational mode decomposition, which overcomes the blindness and uncertainty of traditional manual parameter setting, and lays a precise and reliable foundation for subsequent feature extraction and intelligent diagnosis. Example 5
[0033] Based on Example 3, the feature extraction of data in S3 includes the following steps: Step S31: Screen intrinsic mode components using Pearson correlation coefficient; Step S32: Calculate the sample entropy of different intrinsic mode components and construct a multidimensional feature vector set.
[0034] The processed residual current data is input into the feature extraction unit. The feature extraction unit selects the intrinsic mode components that are highly correlated with the fault characteristics based on the Pearson correlation coefficient, effectively eliminating the interference of noise and irrelevant components. The feature vector constructed from the sample entropy values not only overcomes the shortcomings of insufficient discrimination of traditional single effective values or harmonic features, but also significantly enhances the characterization ability of street light circuit fault status through multi-dimensional entropy feature fusion. The operating environment of street light circuits is complex, and the extracted residual current data has a lot of interference, which further increases the difficulty of fault diagnosis. Therefore, effective extraction of feature information is a necessary prerequisite for judging the operating status of street light circuits. Example 6
[0035] Based on Example 3, the specific steps in S4 for training the classifier using the kernel extreme learning machine model are as follows: Step S41: Divide the constructed multidimensional feature vector set into a training set and a test set, and label the sample categories in the training set to form a label file; Step S42: Train the kernel extreme learning machine classification model using the training set; Step S43: Test the test machine using the trained kernel extreme learning machine classification model to obtain the classification results.
[0036] The fault diagnosis unit first receives a multi-dimensional feature vector sample set composed of different sample entropy values. The normalized feature vector sample set is divided into a training set and a test set according to the proportion. One-Hot encoding is used to label "normal state" and "grounding fault state". Then, the kernel extreme learning machine classification model is supervisedly trained using the training set to optimize its regularization coefficient. The trained model is finally used to perform online classification diagnosis on the feature vectors collected and processed in real time, and output a clear "normal" or "fault" judgment result to achieve the purpose of diagnosis and early warning.
[0037] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.
Claims
1. A smart diagnostic system for grounding faults in street light lines, comprising a data acquisition and preprocessing unit, a feature extraction unit, and a fault diagnosis unit, characterized in that: The data acquisition and preprocessing unit acquires the original residual current data of the street light circuit through a residual current transformer and an oscilloscope, preprocesses the residual current data using a variational mode decomposition method based on the parrot optimization algorithm, and inputs the processed data into the feature extraction unit. The feature extraction unit selects components with high correlation to fault characteristics based on the Pearson correlation coefficient. The sample entropy value is calculated to construct a multi-dimensional feature vector set representing the line state. This set is then input into the fault diagnosis unit, which uses historical data to train a kernel extreme learning machine classifier to establish a mapping relationship between the feature vector and the fault state.
2. The intelligent diagnostic system for grounding faults in street light lines according to claim 1, characterized in that: The data acquisition and preprocessing unit includes a data acquisition module and a data preprocessing module. It acquires raw residual current data and decomposes the raw data to extract fault information from the data.
3. The intelligent diagnostic system for grounding faults in street light lines according to claim 2, characterized in that: The fault diagnosis unit should issue a fault alarm signal, allowing relevant personnel to promptly address the target fault.
4. The intelligent diagnostic system for grounding faults in street light lines according to claim 3, characterized in that: The data preprocessing module employs one or more of the frequency domain method and the spatial domain method.
5. A method for intelligent diagnosis of grounding faults in street light lines, characterized in that, Includes the following steps: S1: Collect residual current time-domain data of the line under normal and various fault conditions; S2: The Parrot optimization algorithm is used to optimize the parameters of variational mode decomposition. The population size, number of iterations and parameter search range of the Parrot optimization algorithm are set. The minimum envelope entropy of the decomposed data is used as the fitness function. The optimal parameter combination is found through iteration. Using this optimal parameter, variational mode decomposition is performed on all the collected residual current data to obtain several intrinsic mode components. S3: Calculate the Pearson correlation coefficient between each intrinsic mode component and the corresponding original data, select the effective components with a correlation coefficient greater than the threshold, calculate the sample entropy of these effective components respectively, combine the sample entropy values of all data to form a feature vector sample set, and normalize and label it. S4: Divide the feature vector sample set into a training set and a test set. Use the training set to train the classifier and optimize its parameters. Finally, use the trained model for real-time or offline diagnosis. Input the feature vector of new data to output the diagnostic results.
6. The intelligent diagnostic method for grounding faults in street light lines according to claim 5, characterized in that, The preprocessing of the raw residual current data in step S2 includes the following steps: Step S21: Find the minimum value of the envelope entropy using the parrot optimization algorithm to optimize the parameters of the variational mode decomposition; Step S22: Decompose the original residual current data using the improved variational mode decomposition.
7. The intelligent diagnostic method for grounding faults in street light lines according to claim 5, characterized in that, The feature extraction of data in step S3 includes the following steps: Step S31: Screen intrinsic mode components using Pearson correlation coefficient; Step S32: Calculate the sample entropy of different intrinsic mode components and construct a multidimensional feature vector set.
8. The intelligent diagnostic method for grounding faults in street light lines according to claim 5, characterized in that, The specific steps for training the classifier using the kernel extreme learning machine model in S4 are as follows: Step S41: Divide the constructed multidimensional feature vector set into a training set and a test set, and label the sample categories in the training set to form a label file; Step S42: Train the kernel extreme learning machine classification model using the training set; Step S43: Test the test machine using the trained kernel extreme learning machine classification model to obtain the classification results.
9. The intelligent diagnostic method for grounding faults in street light lines according to claim 6, characterized in that, In step S21, the Parrot optimization algorithm is used to determine the optimal parameters for variational mode decomposition of street light line data.