A metro platform door power supply fault detection method and system

Abstract

This invention discloses a fault detection method and system for power supply systems in subway platforms. The method includes collecting voltage amplitude, current waveform, spectral distribution parameters, node coordinates, and timestamps of power supply nodes to form power supply monitoring data; identifying abnormal fluctuation patterns in the power supply monitoring data, determining fault characteristic data, analyzing fault timing characteristics, and conducting risk assessment to obtain fault warning data; inputting the fault warning data into a fault classification model to output fault mode labels; performing correlation analysis and analyzing the impact of node spacing on signal propagation based on the fault mode labels, fault warning data, and node coordinates to determine the signal strength contour distribution; combining the power supply monitoring data, performing weight allocation and weighted processing, and mapping it to the power equipment coordinate system to obtain fault location information; converting the fault mode labels and fault location information into electronic map annotations and generating a comprehensive fault report. This method achieves accurate detection and location of power supply system faults.

Description

Technical Field

[0001] This invention relates to the field of fault detection technology, and in particular to a fault detection method and system for the power supply of subway platform doors. Background Technology

[0002] Currently, the stable operation of the subway platform screen door power system is a core foundation for ensuring the safety and efficiency of rail transit. Its failure can lead to platform screen door malfunction and subsequent safety accidents. However, in the complex, high-frequency operation environment, the complex dynamic changes in parameters such as voltage, current, and spectrum of the power equipment contain crucial information. Efficiently capturing these subtle changes and identifying fault signs is a core challenge. Even more difficult is accurately associating these abstract electrical signal anomalies with the spatial location of specific power equipment or line segments to achieve real-time precise positioning. This requires overcoming two core technological bottlenecks: first, deeply analyzing the complex fault characteristics hidden in the operating parameters and capturing anomalies in real time; and second, establishing a deterministic mapping relationship between the extracted electrical signal features and the spatial coordinates of the equipment. Therefore, there is an urgent need for intelligent algorithms capable of simultaneously processing high-dimensional time-series data feature extraction and spatial location information fusion to achieve a direct conversion from complex operating states to precise fault location, thereby improving safety early warning and response capabilities.

[0003] In one existing technology, a distributed sensor network is deployed to monitor each power node. Voltage and current sensors are installed in the power distribution cabinet at the platform gate, and spectrum analyzers are added to key nodes. Fixed threshold parameters are set (e.g., an alarm is triggered when voltage fluctuation exceeds ±10% or harmonic distortion rate is greater than 5%). Data is collected in real time and transmitted to the central processor via a bus. A preset rule base is used for matching and judgment. When a current drop exceeds 30%, it is marked as a short circuit anomaly. When the system triggers an alarm, it only outputs a warning signal and the associated power distribution cabinet number. Finally, operators carry portable testing instruments to the designated area and determine the specific equipment location of the fault through point-by-point manual measurement.

[0004] Existing technologies rely on manual secondary inspection and fixed threshold rules, resulting in vague fault location and an inability to adapt to nonlinear characteristics under varying operating conditions. Therefore, accurate detection and location of faults in the power supply system of subway platform screen doors cannot be achieved. Summary of the Invention

[0005] This invention provides a method and system for fault detection of subway platform door power supply, so as to achieve accurate detection and location of faults in the subway platform door power supply system.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a fault detection method for the power supply of subway platform doors, comprising: The voltage amplitude, current waveform, spectral distribution parameters, node coordinates, and data acquisition timestamps of the power nodes are collected to form power monitoring data. Based on the power monitoring data, abnormal fluctuation patterns in the power signal are identified, and abnormal fluctuation features are extracted to determine fault characteristic data. Based on the fault characteristic data, the fault timing characteristics are analyzed, and a risk assessment is performed to obtain fault early warning data; The fault warning data is input into a pre-built fault classification model, and fault mode labels are output. Based on the fault mode label, the fault warning data, and the node coordinates, an association matrix analysis is performed, and the influence of node spacing on signal propagation is analyzed to determine the signal intensity contour distribution within the radiation radius of the fault source. Based on the signal strength contour distribution and the power monitoring data, weight allocation and weighting are performed, and the data is mapped to a preset power equipment coordinate system to obtain fault location information. The fault mode labels and fault location information are converted into electronic map annotations and a comprehensive fault report is generated.

[0007] In one optional implementation, the step of identifying abnormal fluctuation patterns in the power signal based on the power monitoring data, extracting abnormal fluctuation features, and determining fault characteristic data includes: Based on the power monitoring data, the time-domain signal is converted into a frequency-domain signal using a fast Fourier transform to obtain the amplitude and phase of the spectral components and generate frequency-domain feature data. When the amplitude of the spectral component in the frequency domain feature data exceeds the preset amplitude threshold, the corresponding power monitoring data is taken as abnormal fluctuation data and time-frequency analysis is performed to extract the time-frequency distribution features as abnormal fluctuation features. Based on the abnormal fluctuation characteristics, the abnormality types are classified according to the preset threshold classification rules, and abnormality classification labels are output. Based on the anomaly classification label, the anomaly classification label is bound to the power device identifier and sensor node identifier through database association query to generate fault feature data containing abnormal fluctuation data, abnormal fluctuation characteristics, anomaly type, device identifier and timestamp.

[0008] In one optional implementation, the step of analyzing the fault timing characteristics based on the fault characteristic data and performing a risk assessment to obtain fault early warning data includes: Based on the fault characteristic data, the time series change analysis is performed using the Long Short-Term Memory Network algorithm to obtain the fault time series characteristics and expected fault time. Based on the fault timing characteristics and the expected fault time, a potential risk assessment is performed using the random forest algorithm to obtain the fault risk level. The fault characteristic data, the fault timing characteristics, the expected fault time, and the fault risk level are used as fault early warning data.

[0009] In one optional implementation, the process of constructing the fault classification model includes: Obtain historical fault warning data and corresponding historical fault modes; An initial fault classification model is constructed using the support vector machine algorithm, and the classification boundary is initialized using the RBF kernel function, with a penalty coefficient set. Based on the historical fault warning data, the curvature adjustment parameter of the classification boundary is calculated using the kernel density estimation method, and the curvature of the classification boundary is dynamically adjusted through the RBF kernel function. Based on the historical failure modes, the penalty coefficient is optimized using a grid search method to determine the error tolerance threshold; Cross-validate the adjusted classification boundaries and calculate the set of classification accuracy and error distribution after boundary adjustment; When the error distribution set exceeds the error tolerance threshold, the gradient descent method is used to re-optimize the curvature parameter of the classification boundary, and the classification boundary is iteratively updated until the error distribution set does not exceed the error tolerance threshold. By using RBF kernel function mapping, the optimized classification boundary is matched with the fault mode in a multi-dimensional space to determine the hyperplane equation corresponding to each defect data. By combining the time series characteristics of the historical failure modes, the sliding window method is used to extract dynamic trend parameters. The dynamic trend parameters are then correlated with the hyperplane equation to construct the final failure classification model.

[0010] In one optional implementation, the step of performing correlation matrix analysis based on the fault mode label, the fault early warning data, and the node coordinates, and analyzing the impact of node spacing on signal propagation to determine the signal intensity contour distribution within the fault source radiation radius includes: Based on the fault warning data, the fault feature categories are obtained by classifying the fault feature data using the DBSCAN clustering algorithm. Based on the fault feature category and the node coordinates, a correlation analysis is performed using the Pearson correlation coefficient to obtain the feature correlation matrix; Based on the feature correlation matrix, the signal attenuation gradient coefficient is calculated using the gradient descent algorithm to generate fault spatial distribution data. The node spacing is calculated based on the node coordinates, and combined with the fault spatial distribution data, spatial interpolation and attenuation law analysis are performed using the Kriging interpolation algorithm to obtain the node spacing influence factor. Based on the node spacing influence factor and the signal attenuation gradient coefficient, the signal intensity contour distribution within the radiation radius of the fault source is calculated using the Kriging contour generation algorithm.

[0011] In one optional implementation, the step of weighting and processing the signal strength contour distribution and the power monitoring data, and mapping them to a preset power equipment coordinate system to obtain fault location information, includes: Based on the signal strength contour distribution and the power monitoring data, the weight factor of each power node is calculated using the spatial kernel density estimation algorithm to obtain the spatial weight factor. Based on the spatial weighting factor and the signal intensity contour distribution, a weighted average algorithm is used to perform data fusion processing to obtain weighted distribution data. The weighted distribution data is mapped to a preset power equipment coordinate system through a three-dimensional affine transformation, and fault location information including fault location coordinates and equipment number is output.

[0012] In one optional implementation, the step of converting the fault mode label and the fault location information into electronic map annotations and generating a comprehensive fault report includes: Based on the fault mode label and the fault location information, map projection and labeling are performed using the Mercator projection transformation algorithm to obtain an electronic fault map. Integrate the fault electronic map and the fault early warning data to output a comprehensive fault report.

[0013] Secondly, the present invention provides a fault detection system for the power supply of subway platform doors, comprising: The data acquisition module is used to collect the voltage amplitude, current waveform, spectral distribution parameters, node coordinates, and data acquisition timestamps of the power nodes, forming power monitoring data. The fault feature analysis module is used to identify abnormal fluctuation patterns in the power signal based on the power monitoring data, extract abnormal fluctuation features, and determine fault feature data. The fault early warning analysis module is used to analyze the fault timing characteristics based on the fault characteristic data, perform risk assessment, and obtain fault early warning data. The fault mode analysis module is used to input the fault warning data into a pre-built fault classification model and output fault mode labels. The signal strength analysis module is used to perform correlation matrix analysis based on the fault mode label, the fault warning data and the node coordinates, and to analyze the influence of node spacing on signal propagation, and determine the signal strength contour distribution within the radiation radius of the fault source. The fault location module is used to perform weight allocation and weighted processing based on the signal strength contour distribution and the power monitoring data, and map them to a preset power equipment coordinate system to obtain fault location information; The fault report generation module is used to convert the fault mode labels and fault location information into electronic map annotations and generate a comprehensive fault report.

[0014] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the fault detection method for the power supply of subway platform doors as described in any one of the above.

[0015] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the fault detection method for the power supply of subway platform doors as described above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) Frequency domain amplitude and phase features are extracted by fast Fourier transform (FFT), and transient abnormal fluctuations are captured by wavelet transform time-frequency analysis. Dynamic amplitude threshold is used to replace fixed threshold to solve the problem of missed detection of nonlinear features caused by changes in operating conditions in traditional methods, and improve the detection sensitivity of complex fault modes such as voltage / current fluctuations and harmonic distortion.

[0017] (2) Based on the LSTM network model, the temporal evolution law of fault characteristics is modeled, and the spatial correlation of faults is analyzed by integrating DBSCAN clustering and Pearson correlation coefficient. The influence of node spacing on signal attenuation is quantified by Kriging interpolation algorithm, and a fault radiation model is constructed by combining the distribution of signal intensity contour lines. This solves the blind spot of traditional manual point-by-point investigation and realizes accurate spatial mapping within the radius of the fault source.

[0018] (3) A fault classifier is constructed using RBF kernel support vector machine. The curvature of the classification boundary is dynamically adjusted by kernel density estimation. The penalty coefficient is iteratively optimized by combining grid search and gradient descent method to solve the misclassification problem under noise interference. A sliding window is introduced to extract dynamic trend parameters and associate hyperplane equations to enhance the identification of transient fault modes (such as voltage drop).

[0019] (4) Calculate the power node weight factor based on spatial kernel density estimation, and use the weighted average algorithm to fuse the signal strength contour distribution and real-time monitoring data; map to the equipment coordinate system through three-dimensional affine transformation, establish a deterministic association between abstract electrical signal characteristics (such as spectrum anomalies) and physical equipment location (distribution cabinet number / cable coordinates), and eliminate the problem of separation between fault labels and location information in traditional methods. Attached Figure Description

[0020] Figure 1 This is a schematic flowchart of the subway platform door power supply fault detection method provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the fault detection system for subway platform door power supply provided in the second embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Reference Figure 1 The first embodiment of the present invention provides a fault detection method for the power supply of subway platform doors, comprising the following steps: S11 collects the voltage amplitude, current waveform, spectral distribution parameters, node coordinates, and data acquisition timestamp of the power node to form power monitoring data; S12, Identify abnormal fluctuation patterns in the power signal based on the power monitoring data, extract abnormal fluctuation features, and determine fault feature data. S13, Based on the fault characteristic data, analyze the fault timing characteristics and conduct a risk assessment to obtain fault early warning data; S14, input the fault warning data into the pre-built fault classification model and output the fault mode label; S15, perform correlation matrix analysis based on the fault mode label, the fault early warning data and the node coordinates, and analyze the influence of node spacing on signal propagation to determine the signal intensity contour distribution within the radiation radius of the fault source; S16. Based on the signal strength contour distribution and the power monitoring data, perform weight allocation and weighting processing, and map to the preset power equipment coordinate system to obtain fault location information; S17, convert the fault mode label and the fault location information into electronic map annotations and generate a comprehensive fault report.

[0023] In step S11, the voltage amplitude, current waveform, spectral distribution parameters, node coordinates, and data acquisition timestamps of the power supply nodes are collected to form power supply monitoring data.

[0024] Specifically, voltage sensors deployed in the power distribution cabinet at the platform gate capture real-time voltage amplitude information of power nodes, while current sensors record current waveform characteristics, including instantaneous changes and phase relationships, through high-frequency sampling. A spectrum analyzer performs spectral feature analysis on the power signal to generate spectral distribution parameters, including fundamental and harmonic component characteristics. Node coordinate data is sourced from a pre-set equipment spatial location database and stored in a three-dimensional coordinate system format. Timestamps are generated by a high-precision time synchronization module to mark the acquisition time. All data is transmitted to the preprocessing module via a real-time bus. Voltage signals undergo analog-to-digital conversion, current waveforms are truncated using window functions, and downsampled spectral parameters maintain their original set. Node coordinates and timestamps are bound together to form a spatiotemporal identifier. The preprocessing module performs spatiotemporal alignment of multi-source data, aligning the voltage amplitude and current waveform spectral distribution parameters by timestamp and encapsulating them into a structured data package containing timestamp fields, spatial coordinate fields, voltage feature fields, current waveform arrays, and a set of spectral parameters. The final generated power monitoring data stream is transmitted to the central analysis system. This process, through the fusion and binding of electrical parameters and spatial location, overcomes the limitations of separating features and location in traditional monitoring, forming the foundational input for subsequent fault spatial location.

[0025] In step S12, abnormal fluctuation patterns in the power signal are identified based on the power monitoring data, and abnormal fluctuation features are extracted to determine fault feature data.

[0026] In one specific implementation, the step of identifying abnormal fluctuation patterns in the power signal based on the power monitoring data, extracting abnormal fluctuation features, and determining fault characteristic data includes: Based on the power monitoring data, the time-domain signal is converted into a frequency-domain signal using a fast Fourier transform to obtain the amplitude and phase of the spectral components and generate frequency-domain feature data. When the amplitude of the spectral component in the frequency domain feature data exceeds the preset amplitude threshold, the corresponding power monitoring data is taken as abnormal fluctuation data and time-frequency analysis is performed to extract the time-frequency distribution features as abnormal fluctuation features. Based on the abnormal fluctuation characteristics, the abnormality types are classified according to the preset threshold classification rules, and abnormality classification labels are output. Based on the anomaly classification label, the anomaly classification label is bound to the power device identifier and sensor node identifier through database association query to generate fault feature data containing abnormal fluctuation data, abnormal fluctuation characteristics, anomaly type, device identifier and timestamp.

[0027] Specifically, the input data, which is the power monitoring data generated in step S11, includes a voltage amplitude field, a current waveform array field, and a set of spectral distribution parameters. First, a fast Fourier transform is performed on the current waveform array field to convert the time-domain signal into a frequency-domain signal. The amplitude and phase characteristics of the spectral components are then calculated to form the frequency-domain feature data.

[0028] When the amplitude of a specific spectral component in the frequency domain feature data exceeds a preset amplitude threshold (this preset threshold is determined through statistical distribution of historical normal operation data, for example, taking the mean amplitude under normal operating conditions plus three times the standard deviation as the threshold), the power monitoring data for the corresponding time period is marked as abnormal fluctuation data. For abnormal fluctuation data, a wavelet transform algorithm is used for time-frequency analysis, decomposing the current waveform into sub-frequency bands of different scales, calculating the energy distribution characteristics and time-domain abrupt change points at each scale, and generating time-frequency distribution features containing the scale energy distribution matrix and time abrupt change coordinates, which serve as the abnormal fluctuation features.

[0029] Based on the time-frequency distribution characteristics, anomaly types are classified according to preset threshold classification rules. This rule base includes multi-level judgment conditions (for example, when the high-frequency band energy surge coefficient is greater than a specific threshold and the density of abrupt change points exceeds a benchmark value, it is judged as a surge anomaly. Here, the surge coefficient is a quantitative indicator obtained by calculating the relative change rate of the current high-frequency band energy relative to the historical normal operation benchmark value. The benchmark value is determined based on the statistical analysis of historical normal operation data (such as taking the mean plus three times the standard deviation). This coefficient reflects the intensity of the energy surge and is used to capture transient abnormal events (such as voltage surges)). Anomaly classification labels are output (such as code A representing voltage surge, code B representing harmonic distortion). Finally, through database association query operations, the anomaly classification labels are bound to the device identification field and sensor node identification field in the power monitoring data to generate structured fault feature data, which includes five core fields: abnormal fluctuation data field (stores the original abnormal current waveform), abnormal fluctuation feature field (stores the time-frequency distribution matrix), anomaly type field (stores the classification label), device identification field (stores the power distribution equipment number), and timestamp field (records the time of fault occurrence).

[0030] This step overcomes the limitations of traditional fixed-threshold detection by combining dynamic threshold determination with wavelet time-frequency analysis: the dynamic threshold is adaptively adjusted based on historical operating data, avoiding misjudgments caused by environmental changes; wavelet transform jointly captures transient anomaly features in the time and frequency domains, improving the detection sensitivity for complex fault modes such as voltage drops and current harmonics. The resulting fault feature data provides high-quality input with clear equipment location identification and complete anomaly features for subsequent time-series analysis, laying the foundation for accurate location decisions.

[0031] In step S13, the fault timing characteristics are analyzed based on the fault characteristic data, and a risk assessment is performed to obtain fault warning data.

[0032] In one specific implementation, the step of analyzing the fault timing characteristics based on the fault characteristic data and performing a risk assessment to obtain fault early warning data includes: Based on the fault characteristic data, the time series change analysis is performed using the Long Short-Term Memory Network algorithm to obtain the fault time series characteristics and expected fault time. Based on the fault timing characteristics and the expected fault time, a potential risk assessment is performed using the random forest algorithm to obtain the fault risk level. The fault characteristic data, the fault timing characteristics, the expected fault time, and the fault risk level are used as fault early warning data.

[0033] Specifically, the input data is the fault feature data generated in step S12, which includes an abnormal fluctuation feature field, an equipment identification field, and a timestamp field. First, the fault feature data is processed by a Long Short-Term Memory (LSTM) network algorithm: the timestamp field is divided into continuous sequences according to a fixed time window (e.g., one hour is an analysis unit). The abnormal fluctuation feature is used as the input vector. The network structure includes a forget gate to control the memory strength of historical features, an input gate to filter the weights of current features, and an output gate to generate fault evolution features. After multiple time step iterations, the fault time sequence features (including feature change trend vectors) and the expected fault time are output.

[0034] Based on the fault timing characteristics and expected fault time, the random forest algorithm is used to assess the risk: an ensemble model containing multiple decision trees is constructed, the input feature change trend vector, equipment type code and expected time distance value are input, each tree is split by the Gini coefficient (for example, when the change rate of feature A is greater than a certain value, the left subtree is split), and the output probabilities of all trees are summarized to calculate the fault risk level (e.g., level 1 to 4 represent low risk to emergency risk respectively, and the risk level threshold is determined by cluster analysis of historical fault data).

[0035] Finally, the original fault feature data, fault time-series feature vectors, expected fault timestamps, and fault risk level codes are integrated to generate structured fault early warning data. This process breaks through the limitations of traditional static threshold early warning: Long Short-Term Memory (LSTM) networks capture the nonlinear evolution of fault features over time (such as the gradual deterioration pattern of voltage fluctuations), and random forests fuse features from multiple device types to quantify the probability of potential risks, providing decision inputs including time dimensions and risk levels for subsequent fault classification.

[0036] In step S14, the fault warning data is input into a pre-built fault classification model, and fault mode labels are output.

[0037] In one specific implementation, the process of constructing the fault classification model includes: Obtain historical fault warning data and corresponding historical fault modes; An initial fault classification model is constructed using the support vector machine algorithm, and the classification boundary is initialized using the RBF kernel function, with a penalty coefficient set. Based on the historical fault warning data, the curvature adjustment parameter of the classification boundary is calculated using the kernel density estimation method, and the curvature of the classification boundary is dynamically adjusted through the RBF kernel function. Based on the historical failure modes, the penalty coefficient is optimized using a grid search method to determine the error tolerance threshold; Cross-validate the adjusted classification boundaries and calculate the set of classification accuracy and error distribution after boundary adjustment; When the error distribution set exceeds the error tolerance threshold, the gradient descent method is used to re-optimize the curvature parameter of the classification boundary, and the classification boundary is iteratively updated until the error distribution set does not exceed the error tolerance threshold. By using RBF kernel function mapping, the optimized classification boundary is matched with the fault mode in a multi-dimensional space to determine the hyperplane equation corresponding to each defect data. By combining the time series characteristics of the historical failure modes, the sliding window method is used to extract dynamic trend parameters. The dynamic trend parameters are then correlated with the hyperplane equation to construct the final failure classification model.

[0038] Specifically, historical fault warning data (including abnormal fluctuation feature vectors, equipment type codes, and fault risk level fields) and corresponding historical fault modes (such as short circuit, overload, and other mode labels) are acquired as the basic data for model training. The initial model is constructed using the support vector machine algorithm, and the classification boundary in the multidimensional feature space is initialized using the radial basis kernel function to set the initial penalty coefficient (based on a preset baseline value from the distribution range of historical data).

[0039] Based on historical fault warning data, the kernel density estimation method is used to calculate the data point distribution density as a parameter for adjusting the curvature of the classification boundary (increasing the curvature of the boundary in high-density areas and decreasing the curvature in low-density areas). The shape of the classification boundary is dynamically adjusted through the radial basis kernel function (e.g., forming a curved interface in the overload fault data cluster area). According to the historical fault mode labels, different combinations of penalty coefficients are traversed through the grid search method (e.g., the coefficient range is set to an exponential multiple of the benchmark value). The penalty coefficients are optimized according to the principle of maximizing the cross-validation accuracy, and finally the error tolerance threshold is determined (e.g., the classification error rate must be lower than a certain percentage).

[0040] Cross-validation is performed on the curvature-adjusted classification boundary to calculate the adjusted classification accuracy and the error distribution set (including the type and number of misclassified samples). When the error distribution set exceeds the tolerance threshold (e.g., the misclassification rate of a certain type of fault exceeds the upper limit), gradient descent is used to update the curvature parameters along the negative gradient direction of the loss function (with each iteration reducing the step size), and the boundary is repeatedly adjusted until the error distribution meets the tolerance threshold requirement. The optimized classification boundary is matched with the fault mode through radial basis function mapping to determine the separating hyperplane equations of various fault data in the feature space (e.g., overload and short-circuit faults correspond to independent hyperplanes).

[0041] By combining the time series characteristics of historical failure modes (such as overload failures often accompanied by characteristic fluctuations lasting several minutes), the sliding window method is used to extract the feature change rate within the window as a dynamic trend parameter (such as the rate of increase of risk level per unit time). This parameter is then correlated with the hyperplane equation (for example, data points with high change rates need to be far away from the classification boundary) to construct the final failure classification model.

[0042] The fault warning data is input into the constructed fault classification model, and the fault mode label is output.

[0043] In summary, the radial basis kernel function dynamically adjusts the boundary to adapt to nonlinear distributions, gradient descent optimization improves robustness to noise interference, and sliding window association of time-varying features enhances the model's ability to identify transient faults (such as voltage sags), providing technical support for accurate fault classification.

[0044] In step S15, an association matrix analysis is performed based on the fault mode label, the fault warning data, and the node coordinates, and the influence of node spacing on signal propagation is analyzed to determine the signal intensity contour distribution within the radiation radius of the fault source.

[0045] In one specific implementation, the step of performing correlation matrix analysis based on the fault mode label, the fault early warning data, and the node coordinates, and analyzing the impact of node spacing on signal propagation to determine the signal intensity contour distribution within the fault source radiation radius includes: Based on the fault warning data, the fault feature categories are obtained by classifying the fault feature data using the DBSCAN clustering algorithm. Based on the fault feature category and the node coordinates, a correlation analysis is performed using the Pearson correlation coefficient to obtain the feature correlation matrix; Based on the feature correlation matrix, the signal attenuation gradient coefficient is calculated using the gradient descent algorithm to generate fault spatial distribution data. The node spacing is calculated based on the node coordinates, and combined with the fault spatial distribution data, spatial interpolation and attenuation law analysis are performed using the Kriging interpolation algorithm to obtain the node spacing influence factor. Based on the node spacing influence factor and the signal attenuation gradient coefficient, the signal intensity contour distribution within the radiation radius of the fault source is calculated using the Kriging contour generation algorithm.

[0046] Specifically, the input data includes the fault mode labels (such as overload type codes) output in step S14, the fault warning data (including abnormal fluctuation feature vectors and fault risk levels) generated in step S13, and node coordinates (three-dimensional spatial location data). First, based on the multi-dimensional feature vectors (such as risk level values ​​and abnormal fluctuation intensity values) in the fault warning data, classification is performed using the DBSCAN clustering algorithm: setting the neighborhood radius parameter (determined by the standard deviation of the spatial distribution of historical data), the minimum sample number parameter (set according to the equipment density value), calculating the density connection region of each data point, and outputting fault feature category labels (such as category C representing high-density fault clusters and category D representing discrete abnormal points).

[0047] Based on fault feature category labels and node coordinates, Pearson correlation coefficients are calculated: Node coordinates with the same category label are grouped into spatial locations, and the abnormal fluctuation feature vectors of each node within the group are extracted. The ratio of covariance to standard deviation of the feature vectors between groups is calculated to generate a feature correlation matrix (rows and columns are device node numbers, and matrix element values ​​are correlation coefficients ranging from -1 to 1). Based on this matrix, the signal attenuation gradient coefficient is calculated using a gradient descent algorithm: Random coefficient values ​​are initialized, and a signal strength prediction function is constructed (node ​​A signal strength = node B signal strength × attenuation coefficient × reciprocal of distance, where node B is preferentially selected as the core node (fault source node) with the highest abnormal fluctuation intensity in the DBSCAN clustering results; if the fault source is not clear, the node with the closest spatial distance to node A is selected. The mean square error between the predicted value and the actual monitored value is used as the loss function, and the coefficients are iteratively updated along the negative gradient direction until convergence. The signal attenuation gradient coefficient is output, and fault spatial distribution data (including the theoretical signal strength values ​​of each node) is generated.

[0048] The pairwise distances between nodes are calculated using the Euclidean distance formula based on node coordinates. Combined with fault spatial distribution data, the attenuation pattern is analyzed using a Kriging interpolation algorithm: with the device spacing as the independent variable and the deviation between actual and theoretical signal strength as the dependent variable, a semi-variogram model (such as an exponential model) is fitted, outputting the node spacing influence factor (signal attenuation per unit distance). Finally, combining the signal attenuation gradient coefficient and the spacing influence factor, a Kriging contour line generation algorithm is used: with the fault source location as the center, grid points are divided according to the radiation radius; the signal strength interpolation of each grid point is calculated based on the spatial covariance matrix; and points of equal intensity are connected to form a closed contour line distribution map.

[0049] DBSCAN clustering identifies spatial clustering patterns of faults (such as the abnormal concentration of equipment clusters on the south side of the distribution cabinet), Pearson coefficient quantifies the cross-equipment propagation correlation of fault characteristics (a correlation coefficient greater than 0.8 indicates a strong correlation), and Kriging algorithm integrates distance attenuation law with spatial correlation (the correlation coefficient of the measured attenuation curve fitted by the exponential model reaches 0.92). The final generated contour distribution achieves millimeter-level spatial mapping of the fault radiation range, providing key spatial topological basis for accurate positioning.

[0050] In step S16, based on the signal strength contour distribution and the power monitoring data, weight allocation and weighting are performed, and the data is mapped to a preset power equipment coordinate system to obtain fault location information.

[0051] In one specific implementation, the step of weighting and processing the signal strength contour distribution and the power monitoring data, and mapping them to a preset power equipment coordinate system to obtain fault location information, includes: Based on the signal strength contour distribution and the power monitoring data, the weight factor of each power node is calculated using the spatial kernel density estimation algorithm to obtain the spatial weight factor. Based on the spatial weighting factor and the signal intensity contour distribution, a weighted average algorithm is used to perform data fusion processing to obtain weighted distribution data. The weighted distribution data is mapped to a preset power equipment coordinate system through a three-dimensional affine transformation, and fault location information including fault location coordinates and equipment number is output.

[0052] Specifically, the input data consists of the signal strength contour distribution output in step S15 (including the signal strength values ​​of each spatial point within the radiation radius) and the power monitoring data collected in step S11 (including node coordinate fields and real-time spectrum parameters). First, a spatial kernel density estimation algorithm is used to calculate the weighting factor: Centered on the three-dimensional coordinates of each power node, a kernel function is constructed to calculate the spatial influence weight based on the contour distribution density around the node (e.g., the number of contour lines per unit area) and the anomaly degree of real-time spectrum parameters (e.g., the degree to which harmonic distortion rate deviates from the baseline value). The spatial weighting factor for each node is then output (the value ranges from 0 to 1; a larger value indicates a higher probability of the node being associated with a fault). The kernel function uses a Gaussian attenuation model, and its weight calculation is based on the spatial distance attenuation characteristics between the target node and surrounding nodes, and the real-time spectrum anomaly degree. The bandwidth value of the key parameter is dynamically adjusted according to the average distance between nodes within the radiation radius of the fault source. Finally, by superimposing the distance contribution of surrounding nodes and integrating the real-time spectrum anomaly data, a normalized spatial weighting factor is output.

[0053] Based on the spatial weighting factors and the contour distribution of signal intensity, weighted average data fusion is performed: the signal intensity value of each spatial point in the contour distribution is multiplied by the weighting factor of the corresponding node, the weighted values ​​of all nodes are summed, and then divided by the sum of the weighting factors to generate weighted distribution data of fused spatial features and real-time monitoring data (the fused intensity value of each spatial point is stored in matrix form). Each continuous spatial point in the contour distribution is associated with its nearest neighbor power node in its local area, and the weighting factor of that node is used as the fusion coefficient of the spatial point. When the spatial point is located in the multi-node overlapping influence area, the weighting factor ratio of each node is allocated according to the inverse distance principle to ensure that the weight of discrete nodes and the continuous spatial signal intensity are dynamically mapped. Finally, the weighted distributed data is mapped to a preset power equipment coordinate system through a three-dimensional affine transformation: First, the reference transformation parameters of the equipment coordinate system are obtained—including the coordinate offset with the main distribution cabinet as the origin, the rotation angle of the equipment layout (X / Y / Z axes), and the preset scale factor; based on these parameters, an affine transformation matrix is ​​constructed: this matrix transforms the spatial point coordinates in the global coordinate system into precise position coordinates in the equipment topology coordinate system by combining translation operations (coordinate origin offset), rotation operations (angle correction), and scaling operations (scale adaptation) in three-dimensional space. The spatial point coordinates in the weighted distributed data are then converted into absolute coordinates in the equipment coordinate system (e.g., mapping spatial points X / Y / Z to distribution cabinet number + cable tray location code), and structured fault location information is output (including the absolute coordinate field of the fault point, the associated equipment number field, and the cable identification field).

[0054] This process achieves precise conversion from abstract signals to physical locations: spatial kernel density estimation integrates contour distribution density and real-time spectral anomalies (such as increasing the weight of high-frequency harmonic concentration areas to above 0.9), solving the problem of traditional methods neglecting the spatial attenuation characteristics of electrical signals; the weighted average algorithm balances historical spatial distribution and real-time monitoring data (such as enhancing the weight of transient current fluctuation areas), eliminating the positioning bias of a single data source; three-dimensional affine transformation maps the weighted data to operable equipment locations (such as locating the 5th terminal of the A-line trough in distribution cabinet No. 3) through a preset equipment coordinate system (such as establishing a topological coordinate system with the main distribution cabinet as the origin), solving the problem of the separation between electrical signal characteristics and physical location, and realizing the positioning from "fault area" to "faulty equipment".

[0055] In step S17, the fault mode label and the fault location information are converted into electronic map annotations and a comprehensive fault report is generated.

[0056] In one specific implementation, based on the fault mode label and the fault location information, a map is projected and labeled using the Mercator projection transformation algorithm to obtain a fault electronic map. Integrate the fault electronic map and the fault early warning data to output a comprehensive fault report.

[0057] Specifically, in step S17, the specific implementation process for generating a comprehensive fault report is as follows: First, based on the fault mode label (such as overload type code) and fault location information (including the absolute coordinates of the fault point in the equipment coordinate system and the associated equipment number), spatial coordinate transformation is performed using the Mercator projection transformation algorithm: the preset station electronic map reference coordinate system parameters (such as the central meridian value and latitude range) are read, and the three-dimensional coordinates (X equipment, Y equipment, Z equipment) in the equipment coordinate system are converted into planar projection coordinates (X map, Y map). The conversion process includes coordinate translation (matching the map origin offset), scaling (adapting to the map resolution), and angle correction (eliminating terrain distortion). Then, according to the fault mode label setting labeling rules (e.g., overload type uses a red icon, short circuit type uses a triangle symbol), the equipment number text (such as "distribution cabinet A-line trough 3") is superimposed at the projection coordinate position to generate a fault electronic map.

[0058] Integrate the fault electronic map with the fault warning data from step S13 (including fault risk level and expected fault time fields) to construct a structured report: the map annotation layer and the warning data table are automatically linked through the timestamp field. For example, when a local map annotation point is triggered, the real-time risk level (e.g., "high risk") and expected fault time (e.g., "remaining maintenance time") of the corresponding equipment are displayed in conjunction. The final output is a comprehensive fault report, which includes a layered rendered electronic map (fault location annotation layer, equipment topology layer), a warning data table (equipment number, risk level, expected time), and a correlation analysis summary (e.g., "3 overload anomalies in the south power distribution area, priority maintenance is recommended").

[0059] This process achieves spatial and structural fusion of fault information: Mercator projection solves the high-fidelity conversion from equipment coordinates to map coordinates, the annotation rule base establishes a strong correlation between fault types and visual symbols, and the spatiotemporal linkage mechanism breaks through the limitations of traditional static report display, enabling maintenance personnel to intuitively locate the faulty equipment and simultaneously obtain risk status.

[0060] Reference Figure 2 The second embodiment of the present invention provides a fault detection system for the power supply of subway platform doors, comprising: The data acquisition module is used to collect the voltage amplitude, current waveform, spectral distribution parameters, node coordinates, and data acquisition timestamps of the power nodes, forming power monitoring data. The fault feature analysis module is used to identify abnormal fluctuation patterns in the power signal based on the power monitoring data, extract abnormal fluctuation features, and determine fault feature data. The fault early warning analysis module is used to analyze the fault timing characteristics based on the fault characteristic data, perform risk assessment, and obtain fault early warning data. The fault mode analysis module is used to input the fault warning data into a pre-built fault classification model and output fault mode labels. The signal strength analysis module is used to perform correlation matrix analysis based on the fault mode label, the fault warning data and the node coordinates, and to analyze the influence of node spacing on signal propagation, and determine the signal strength contour distribution within the radiation radius of the fault source. The fault location module is used to perform weight allocation and weighted processing based on the signal strength contour distribution and the power monitoring data, and map them to a preset power equipment coordinate system to obtain fault location information; The fault report generation module is used to convert the fault mode labels and fault location information into electronic map annotations and generate a comprehensive fault report.

[0061] It should be noted that the fault detection device for the power supply of a subway platform door provided in this embodiment of the invention is used to execute all the process steps of the fault detection method for the power supply of a subway platform door in the above embodiment. The working principle and beneficial effect of the two are one-to-one, so they will not be described again.

[0062] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a fault detection program for the power supply of subway platform doors. When the processor executes the computer program, it implements the steps in the aforementioned embodiments of the fault detection methods for the power supply of subway platform doors, for example... Figure 1 Step S11 is shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the fault detection module for the power supply of subway platform doors.

[0063] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0064] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0065] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0066] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0067] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0068] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0069] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for detecting a failure of a power supply of a subway platform gate, characterized by, include: The voltage amplitude, current waveform, spectral distribution parameters, node coordinates, and data acquisition timestamps of the power nodes are collected to form power monitoring data. Based on the power monitoring data, abnormal fluctuation patterns in the power signal are identified, and abnormal fluctuation features are extracted to determine fault characteristic data. Based on the fault characteristic data, the fault timing characteristics are analyzed, and a risk assessment is performed to obtain fault early warning data; The fault warning data is input into a pre-built fault classification model, and fault mode labels are output. Based on the fault mode label, the fault warning data, and the node coordinates, an association matrix analysis is performed, and the influence of node spacing on signal propagation is analyzed to determine the signal intensity contour distribution within the radiation radius of the fault source. Based on the signal strength contour distribution and the power monitoring data, weight allocation and weighting are performed, and the data is mapped to a preset power equipment coordinate system to obtain fault location information. The fault mode labels and fault location information are converted into electronic map annotations and a comprehensive fault report is generated.

2. The method for detecting a failure of a power supply of a subway platform gate according to claim 1, characterized by, The step of identifying abnormal fluctuation patterns in the power signal based on the power monitoring data, extracting abnormal fluctuation features, and determining fault characteristic data includes: Based on the power monitoring data, the time-domain signal is converted into a frequency-domain signal using a fast Fourier transform to obtain the amplitude and phase of the spectral components and generate frequency-domain feature data. When the amplitude of the spectral component in the frequency domain feature data exceeds the preset amplitude threshold, the corresponding power monitoring data is taken as abnormal fluctuation data and time-frequency analysis is performed to extract the time-frequency distribution features as abnormal fluctuation features. Based on the abnormal fluctuation characteristics, the abnormality types are classified according to the preset threshold classification rules, and abnormality classification labels are output. Based on the anomaly classification label, the anomaly classification label is bound to the power device identifier and sensor node identifier through database association query to generate fault feature data containing abnormal fluctuation data, abnormal fluctuation characteristics, anomaly type, device identifier and timestamp.

3. The method for detecting a failure of a power supply of a subway platform gate according to claim 1, characterized by, The step of analyzing the fault timing characteristics based on the fault characteristic data and performing risk assessment to obtain fault early warning data includes: Based on the fault characteristic data, a time-series change analysis is performed using a long short-term memory network algorithm to obtain the fault time-series characteristics and expected fault time. Based on the fault timing characteristics and the expected fault time, a potential risk assessment is performed using the random forest algorithm to obtain the fault risk level. The fault characteristic data, the fault timing characteristics, the expected fault time, and the fault risk level are used as fault early warning data.

4. The method for detecting a failure of a power supply of a subway platform gate according to claim 1, characterized by, The process of constructing the fault classification model includes: Obtain historical fault warning data and corresponding historical fault modes; An initial fault classification model is constructed using the support vector machine algorithm, and the classification boundary is initialized using the RBF kernel function, with a penalty coefficient set. Based on the historical fault warning data, the curvature adjustment parameter of the classification boundary is calculated using the kernel density estimation method, and the curvature of the classification boundary is dynamically adjusted through the RBF kernel function. Based on the historical failure modes, the penalty coefficient is optimized using a grid search method to determine the error tolerance threshold; Cross-validate the adjusted classification boundaries and calculate the set of classification accuracy and error distribution after boundary adjustment; When the error distribution set exceeds the error tolerance threshold, the gradient descent method is used to re-optimize the curvature parameter of the classification boundary, and the classification boundary is iteratively updated until the error distribution set does not exceed the error tolerance threshold. By using RBF kernel function mapping, the optimized classification boundary is matched with the fault mode in a multi-dimensional space to determine the hyperplane equation corresponding to each defect data. By combining the time series characteristics of the historical failure modes, the sliding window method is used to extract dynamic trend parameters. The dynamic trend parameters are then correlated with the hyperplane equation to construct the final failure classification model.

5. The method for detecting a failure of a power supply of a subway platform gate according to claim 1, characterized by, The step of performing correlation matrix analysis based on the fault mode label, the fault early warning data, and the node coordinates, and analyzing the impact of node spacing on signal propagation to determine the signal intensity contour distribution within the fault source radiation radius includes: Based on the fault warning data, the fault feature categories are obtained by classifying the fault feature data using the DBSCAN clustering algorithm. Based on the fault feature categories and the node coordinates, a feature correlation matrix is ​​obtained by performing correlation analysis using the Pearson correlation coefficient. Based on the feature correlation matrix, the signal attenuation gradient coefficient is calculated using the gradient descent algorithm to generate fault spatial distribution data. The node spacing is calculated based on the node coordinates, and combined with the fault spatial distribution data, spatial interpolation and attenuation law analysis are performed using the Kriging interpolation algorithm to obtain the node spacing influence factor. Based on the node spacing influence factor and the signal attenuation gradient coefficient, the signal intensity contour distribution within the radiation radius of the fault source is calculated using the Kriging contour generation algorithm.

6. The method for detecting a failure of a power supply of a subway platform gate according to claim 1, characterized by, The step of weighting and processing the signal strength contour distribution and power monitoring data, and mapping them to a preset power equipment coordinate system to obtain fault location information, includes: Based on the signal strength contour distribution and the power monitoring data, the weight factor of each power node is calculated using the spatial kernel density estimation algorithm to obtain the spatial weight factor. Based on the spatial weighting factor and the signal intensity contour distribution, a weighted average algorithm is used to perform data fusion processing to obtain weighted distribution data. The weighted distribution data is mapped to a preset power equipment coordinate system through a three-dimensional affine transformation, and fault location information including fault location coordinates and equipment number is output.

7. The method of claim 1, wherein The step of converting the fault mode labels and fault location information into electronic map annotations and generating a comprehensive fault report includes: Based on the fault mode label and the fault location information, map projection and labeling are performed using the Mercator projection transformation algorithm to obtain an electronic fault map. Integrate the fault electronic map and the fault early warning data to output a comprehensive fault report.

8. A subway platform gate power supply failure detection system characterized by comprising: include: The data acquisition module is used to collect the voltage amplitude, current waveform, spectral distribution parameters, node coordinates, and data acquisition timestamps of the power nodes, forming power monitoring data. The fault feature analysis module is used to identify abnormal fluctuation patterns in the power signal based on the power monitoring data, extract abnormal fluctuation features, and determine fault feature data. The fault early warning analysis module is used to analyze the fault timing characteristics based on the fault characteristic data, perform risk assessment, and obtain fault early warning data. The fault mode analysis module is used to input the fault warning data into a pre-built fault classification model and output fault mode labels. The signal strength analysis module is used to perform correlation matrix analysis based on the fault mode label, the fault warning data and the node coordinates, and to analyze the influence of node spacing on signal propagation, and determine the signal strength contour distribution within the radiation radius of the fault source. The fault location module is used to perform weight allocation and weighted processing based on the signal strength contour distribution and the power monitoring data, and map them to a preset power equipment coordinate system to obtain fault location information; The fault report generation module is used to convert the fault mode labels and fault location information into electronic map annotations and generate a comprehensive fault report.

9. An electronic device, comprising: The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the fault detection method for the power supply of subway platform doors as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the fault detection method for the power supply of subway platform doors as described in any one of claims 1 to 7.

Images