Power cable comprehensive on-line monitoring system

Abstract

The invention provides a power cable comprehensive online monitoring system, and relates to the technical field of data processing, and the system comprises the steps: extracting a partial discharge distribution point set in a multi-dimensional feature fusion data set, and calculating the weight factor of each distribution point; based on the weight factors, a weighted point set concave packet generation algorithm is adopted to construct a partial discharge source probability distribution boundary, and a boundary feature point set is obtained; inputting the boundary feature point set and the multi-dimensional feature fusion data set into a convolution-long and short-term memory hybrid neural network to generate a diagnosis result; and a third-level early warning signal is generated based on a diagnosis result, early warning information is pushed to a remote operation and maintenance terminal and a mobile APP through an MQTT protocol, a digital work order service is automatically triggered to generate a maintenance task, and positioning verification is executed. According to the invention, the accuracy of monitoring and fault diagnosis is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an integrated online monitoring system for power cables. Background Art

[0002] As the core carrier of urban and industrial power distribution systems, the operating status of power cables is directly related to the safety and stability of the power system. As grid load density increases and cable service life increases, the risk of abnormal ground currents, joint overheating, and partial discharge failures increases. Traditional single-parameter monitoring and manual diagnosis methods are unable to meet reliability requirements.

[0003] In addition, the cable status is affected by environmental temperature, humidity, and contamination factors. Some existing systems store environmental parameters as independent data and do not deeply integrate them into the state feature extraction process, resulting in insufficient model generalization capabilities; some traditional systems rely on a single communication method, which is prone to data packet loss in complex electromagnetic environments or network congestion scenarios, affecting the timeliness of monitoring. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a comprehensive online monitoring system for power cables, which improves the accuracy of monitoring and fault diagnosis.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows: In a first aspect, a comprehensive online monitoring system for power cables comprises: The data acquisition module is used to collect the time domain waveform of the power cable grounding current, the joint temperature field distribution, the partial discharge pulse sequence and the environmental parameters to obtain a multi-type raw data set; The data processing module is used to input multiple types of raw data sets into the edge computing node, perform data preprocessing, and generate standardized status data packets; The data fusion module is used to transmit standardized status data packets to the cloud platform via a dual-channel redundant network of TCP / IP and power carrier. It uses a dynamic time warping algorithm to align the temporal and spatial dimensions of current, temperature, and partial discharge data, and integrates environmental parameters to generate a multi-dimensional feature fusion data set. The fault diagnosis module is used to extract the set of partial discharge distribution points from the multi-dimensional feature fusion data set and calculate the weight factor of each distribution point. Based on the weight factor, the weighted point set concave hull generation algorithm is used to construct the probability distribution boundary of the partial discharge source to obtain a set of boundary feature points. The boundary feature point set and the multi-dimensional feature fusion data set are input into the convolution-long short-term memory hybrid neural network to generate the diagnosis result. The intelligent response module is used to generate three-level warning signals based on diagnostic results, push warning information to remote operation and maintenance terminals and mobile apps through the MQTT protocol, and automatically trigger the digital work order service to generate maintenance tasks and perform positioning verification.

[0006] Furthermore, multiple types of raw data sets are input into the edge computing node for data preprocessing to generate standardized status data packets, including: The time-domain waveform data of the power cable grounding current in the multi-type original data set is processed to generate the current waveform data after the power frequency is eliminated; the temperature field data is generated using the Kriging interpolation method based on the discrete point data of the power cable joint temperature field distribution in the multi-type original data set; the temperature field data is calculated and identified to obtain the temperature field analysis result data; Processing the partial discharge pulse sequence data in the multi-type original data set to obtain a partial discharge pulse feature data set; and processing the environmental parameters in the multi-type original data set to generate effective environmental parameter data; The current waveform data after power frequency removal, temperature field analysis result data, partial discharge pulse characteristic data set and effective environmental parameter data are collected and processed to generate a multi-source state data set; the multi-source state data set is normalized to obtain a standardized state data packet.

[0007] Furthermore, the standardized status data packets are transmitted to the cloud platform via a dual-channel redundant network of TCP / IP and power carrier. A dynamic time warping algorithm is used to align the temporal and spatial dimensions of the current, temperature, and partial discharge data. Environmental parameters are integrated to generate a multi-dimensional feature fusion data set, including: The standardized status data packet is transmitted to the cloud platform in parallel through the TCP / IP network channel and the power carrier communication channel; the data packets received by the dual channels are verified to be correct through redundancy verification and then a cloud-side reception verification data packet is generated; The verification data packets received on the cloud are processed to obtain the current time series, temperature time series, and partial discharge pulse time series. Dynamic time warping algorithm and nonlinear time axis transformation are then used to generate the current series, temperature series, and partial discharge series aligned in time and space. According to the current sequence, temperature alignment sequence and partial discharge alignment sequence aligned in the time and space dimensions, the normalized environmental parameter data is extracted from the verification data packet received through the cloud, and the current alignment sequence, temperature alignment sequence, partial discharge alignment sequence and environmental parameter data are integrated according to the time point to generate a multi-dimensional feature fusion dataset.

[0008] Furthermore, the PD distribution point set in the multi-dimensional feature fusion data set is extracted, and the weight factor of each distribution point is calculated. Based on the weight factor, the weighted point set concave hull generation algorithm is used to construct the probability distribution boundary of the PD source, and a set of boundary feature points is obtained. The boundary feature point set and the multi-dimensional feature fusion data set are input into the convolution-long short-term memory hybrid neural network to generate the diagnosis results, including: The spatiotemporal distribution point set of partial discharge pulses is extracted from the multi-dimensional feature fusion data set. Based on the spatiotemporal distribution point set of partial discharge pulses, the time aggregation weight factor of each point's timestamp and the relative intensity weight factor of each point's amplitude are calculated, and the two are fused to generate a joint weight factor to obtain a weighted partial discharge site set. The PD distribution point set and the corresponding weight factors of each point are generated by using the weighted point set concave hull generation algorithm to calculate a set of convex boundary points that can surround all PD distribution points, construct the probability distribution boundary of the high-weight area, and obtain the boundary feature point set. The boundary feature point set is mapped to generate a spatial probability matrix and fused with the time series data. The matrix is ​​input into the convolution-LSTM hybrid neural network to extract spatiotemporal features and generate diagnostic results.

[0009] Furthermore, a three-level warning signal is generated based on the diagnosis results. The warning information is pushed to the remote operation and maintenance terminal and mobile app via the MQTT protocol, automatically triggering the digital work order system to generate maintenance tasks and perform location verification, including: Analyze the cable insulation status risk level value in the diagnosis result; match the risk level value with the preset three-level warning threshold range to obtain the warning signal level; extract the spatial location coordinates and timestamp data of the partial discharge source in the diagnosis result; encapsulate the warning signal level, location coordinates, and timestamp into a structured MQTT protocol message; Push warning information in parallel via the cloud platform to remote operation and maintenance terminals and mobile apps using structured MQTT protocol messages. Generate maintenance priority identifiers based on warning signal levels. Bind location coordinates, warning signal levels, and priority identifiers to generate structured maintenance instructions. Automatically trigger the digital work order service to generate electronic maintenance work orders. The spatial positioning service of the on-site verification terminal is driven by the positioning coordinates in the maintenance work order to execute local discharge source positioning and receive feedback on the verification results.

[0010] Furthermore, the standardized status data packet is transmitted to the cloud platform in parallel through the TCP / IP network channel and the power carrier communication channel; the data packets received by the dual channels are verified to be correct through redundancy verification to generate a cloud-side reception verification data packet, including: The standardized status data packet is copied into two copies at the same time and transmitted to the cloud platform in parallel via the TCP / IP network channel and the power carrier communication channel respectively. The cloud platform receives the data packet copies of the two channels respectively, which are recorded as copy A and copy B, and performs preliminary data integrity checks on each. Based on the status of replicas A and B after preliminary verification, a preset time window is started for waiting; if both replicas are valid within the window period, valid replicas A and B are obtained; otherwise, an exception is triggered and the process is terminated; Apply the same preset verification algorithm to valid replicas A and B, respectively, to generate verification identifiers A and B, and compare the two. If they match, generate replicas A and B with the same content. Otherwise, trigger data inconsistency exception processing and terminate this process. Based on the confirmed content of replicas A and B, one of the replicas is selected according to the preset selection strategy, and a cloud-received verification data packet is generated and output.

[0011] Furthermore, the verification data packets received on the cloud are processed to obtain the current time series, temperature time series, and partial discharge pulse time series. Dynamic time warping algorithm and nonlinear time axis transformation are used to generate the current series, temperature alignment series, and partial discharge alignment series aligned in time and space dimensions, including: Based on the cloud, it receives verification data packets, performs predefined protocol parsing operations, and generates original current time series, temperature time series, and partial discharge pulse time series; The temperature time series is designated as the reference series, and a dynamic time warping algorithm is applied between the current time series and the temperature time series. The nonlinear time warping path between the two series is calculated to generate a current-temperature warping mapping relationship. Apply a dynamic time warping algorithm between the PD pulse time series and the temperature time series; calculate the tortuous path and generate a PD-temperature warping mapping relationship; Based on the extracted original time series and mapping relationship, nonlinear time axis transformation is performed to generate current alignment sequence and partial discharge alignment sequence that are completely aligned with the time axis of the temperature series, and the original temperature series is used as the temperature alignment sequence.

[0012] Furthermore, the PD distribution point set and the corresponding weight factors of each point are used to generate the weighted point set concave hull algorithm. A set of convex boundary points that can surround all PD distribution points is calculated to construct the probability distribution boundary of the high-weight area, and the boundary feature point set is obtained, including: Based on the weighted partial discharge site set, the initial concave hull radius parameter setting value and radius adjustment step size are configured; Based on the weighted partial discharge site set and the initial radius setting value, a concave hull generation algorithm with a joint weight factor as the spatial density weight is used to calculate and generate a set of candidate boundary points that enclose all points. Calculate the sum of the joint weight factors of all points in the area enclosed by the concave hull formed by the candidate boundary point set to obtain the enclosed weight sum; calculate the sum of the joint weight factors of all points in the weighted partial discharge site set to obtain the total weight sum; Calculate the wrapping rate of high-weight areas, dynamically adjust the radius parameter based on the comparison between the wrapping rate and the target threshold, and perform candidate boundary calculation and evaluation cyclically until the wrapping rate meets the requirements, and finally obtain a set of boundary feature points.

[0013] In a second aspect, a computing device includes: one or more processors; The storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the system.

[0014] According to a third aspect, a computer-readable storage medium stores a program, which implements the system when executed by a processor.

[0015] The above solution of the present invention includes at least the following beneficial effects: By using edge computing nodes to preprocess data, invalid data transmission is reduced and processing efficiency is improved. A dual-channel redundant network design using TCP / IP and power carrier effectively mitigates data loss caused by single transmission link failures, ensuring the real-time and continuity of monitoring data. A weighted point set concave hull algorithm is used to construct the probability distribution boundary of partial discharge sources. Combined with a convolutional-long short-term memory hybrid neural network, this fully exploits the spatial characteristics and temporal associations of the data, improving the recognition and early warning capabilities of partial discharge faults and reducing false alarms and missed alerts. In addition, through full-time online monitoring, precise fault location and rapid maintenance response, potential cable defects can be discovered in a timely manner and intervention can be made in advance, effectively improving power outages caused by fault expansion, improving the reliability and safety of power system operation, and reducing economic losses caused by faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a schematic diagram of a comprehensive online monitoring system for power cables provided by an embodiment of the present invention.

[0017] Figure 2 This is a flowchart of an integrated online monitoring system for power cables provided by an embodiment of the present invention, which transmits standardized status data packets to a cloud platform through TCP / IP and a power carrier dual-channel redundant network, uses a dynamic time warping algorithm to align the temporal and spatial dimensions of current, temperature, and partial discharge data, and integrates environmental parameters to generate a multi-dimensional feature fusion data set. DETAILED DESCRIPTION

[0018] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0019] like Figure 1As shown, an embodiment of the present invention provides a comprehensive online monitoring system for power cables, comprising: The data acquisition module is used to collect the time domain waveform of the power cable grounding current, the joint temperature field distribution, the partial discharge pulse sequence and the environmental parameters to obtain a multi-type raw data set; The data processing module is used to input multiple types of raw data sets into the edge computing node, perform data preprocessing, and generate standardized status data packets; The data fusion module is used to transmit standardized status data packets to the cloud platform via a dual-channel redundant network of TCP / IP and power carrier. It uses a dynamic time warping algorithm to align the temporal and spatial dimensions of current, temperature, and partial discharge data, and integrates environmental parameters to generate a multi-dimensional feature fusion data set. The fault diagnosis module is used to extract the set of partial discharge distribution points from the multi-dimensional feature fusion data set and calculate the weight factor of each distribution point. Based on the weight factor, the weighted point set concave hull generation algorithm is used to construct the probability distribution boundary of the partial discharge source to obtain a set of boundary feature points. The boundary feature point set and the multi-dimensional feature fusion data set are input into the convolution-long short-term memory hybrid neural network to generate the diagnosis result. The intelligent response module is used to generate three-level warning signals based on diagnostic results, push warning information to remote operation and maintenance terminals and mobile apps through the MQTT protocol, and automatically trigger the digital work order service to generate maintenance tasks and perform positioning verification.

[0020] In an embodiment of the present invention, edge computing nodes are used to pre-process data, reducing the amount of invalid data transmission and improving processing efficiency. A dual-channel redundant network design of TCP / IP and power carrier effectively mitigates data loss caused by failures in a single transmission link, ensuring the real-time and continuity of monitoring data. The weighted point set concave hull algorithm is used to construct the probability distribution boundary of the partial discharge source. Combined with a convolutional-long short-term memory hybrid neural network, the spatial characteristics and temporal associations of the data are fully exploited, improving the recognition and early warning capabilities of partial discharge faults and reducing false alarms and missed alarms. In addition, through full-time online monitoring, precise fault location and rapid maintenance response, potential cable defects can be discovered in a timely manner and intervention can be made in advance, effectively improving power outages caused by fault expansion, improving the reliability and safety of power system operation, and reducing economic losses caused by faults. In a preferred embodiment of the present invention, a multi-type raw data set is input into an edge computing node, data preprocessing is performed, and a standardized status data packet is generated, including: The time-domain waveform data of the power cable grounding current in the multi-type original data set is processed to generate the current waveform data after the power frequency is eliminated; the temperature field data is generated using the Kriging interpolation method based on the discrete point data of the power cable joint temperature field distribution in the multi-type original data set; the temperature field data is calculated and identified to obtain the temperature field analysis result data; Processing the partial discharge pulse sequence data in the multi-type original data set to obtain a partial discharge pulse feature data set; and processing the environmental parameters in the multi-type original data set to generate effective environmental parameter data; The current waveform data after power frequency removal, temperature field analysis result data, partial discharge pulse characteristic data set and effective environmental parameter data are collected and processed to generate a multi-source state data set; the multi-source state data set is normalized to obtain a standardized state data packet.

[0021] In an embodiment of the present invention, a sliding average method is used to perform baseline correction on the original current time domain waveform (sampling frequency 1kHz-10kHz); the window size (5-20 sampling points) is dynamically adjusted according to the sampling frequency. For example, when sampling at 1kHz, the window takes 20 points (corresponding to 20ms), and when sampling at 10kHz, the window takes 5 points (corresponding to 0.5ms). The sliding average is calculated to eliminate the slow trend caused by sensor drift and retain the high-frequency fluctuation component.

[0022] Based on the 50Hz power frequency of my country's power system, the center frequency of the band-stop filter is set to 50Hz and the bandwidth is 2Hz (covering the range of 49-51Hz) to ensure accurate targeting of the power frequency interference band; to ensure the interference suppression effect, the stopband attenuation is set to ≥40dB (which can reduce the power frequency signal amplitude to less than 1% of the original), and the passband ripple is set to ≤1dB (reducing the attenuation of useful signals). Considering the computing power limitations of edge computing nodes, a 4-8th order Butterworth band-stop filter is selected (the 4th order can meet basic attenuation requirements, and the 8th order is suitable for strong interference scenarios. The higher the order, the steeper the filtering curve, but the computational complexity increases).

[0023] Based on the above parameters, a filter is processed in the edge computing node to determine the order coefficients. The preprocessed ground current time-domain waveform data (with trend removed) is input into the filter, and signal filtering is achieved through iterative calculations. The output value at each moment is the weighted sum of the input values ​​at the current moment and the previous 3-7 moments, depending on the order. The ratio of the residual amplitude of the 50Hz frequency band signal to the original amplitude is calculated to ensure an attenuation rate of ≥90%. The amplitude loss rate in the useful signal frequency band (such as the 1-10kHz frequency band related to partial discharge) is also checked to be ≤5% to avoid useful signal distortion. The filter value range is 4-8, with a stopband attenuation of 40-60dB and a passband ripple of 0.5-1dB.

[0024] Standard deviation (σ): The value is 0.5-2 sampling intervals, and is adjusted according to the sampling frequency: when sampling at 1kHz (sampling interval 1ms), σ is 1-2ms (i.e., 1-2 sampling intervals); when sampling at 10kHz (sampling interval 0.1ms), σ is 0.05-0.1ms (i.e., 0.5-1 sampling interval), balancing noise removal and feature retention. The window length is 6-8 times σ (covering more than 99% of the energy of the Gaussian distribution). For example, when σ = 1ms, the window size is 6-8ms (corresponding to 6-8 sampling points); when σ = 0.1ms, the window size is 0.6-0.8ms (corresponding to 6-8 sampling points). The weight at the center of the window is the highest (close to 1), and gradually decreases towards the ends (edge ​​point weight ≤ 0.01), and the total weight is normalized to 1 (to avoid overall changes in the signal amplitude after smoothing). Perform sliding window processing on the current waveform data: the smoothed value of each sampling point is equal to the sum of the products of the original values ​​of all points in the window and the corresponding weights. Boundary processing: For areas where the beginning and end of the waveform are less than the length of a window, the mirror filling method is used (copying adjacent data with the boundary point as the symmetry axis) to avoid distortion of the boundary points. Value range: σ = 0.5-2 sampling intervals, window size = 6-8 times σ, weight value 0.01-1 (after normalization).

[0025] The original discrete temperature points come from distributed optical fiber or infrared sensors (sampling point spacing is 0.05-0.2 meters), and outliers are eliminated. The temperature validity range is set to -20°C to 90°C (covering the normal operating temperature of the cable and extreme environments). Data outside this range is considered a sensor failure and deleted. For missing points (≤5%), the average of the three adjacent valid points is used to fill in the gaps to ensure data continuity.

[0026] Construction, training and implementation of the variation function model (taking the spherical model as an example): The spherical model is selected to quantify the spatial correlation of the temperature field, and three core parameters are defined: nugget value (C0): reflects the measurement error, with a value of 0.1-2℃ 2 (According to the sensor accuracy, ±0.5℃ sensor corresponds to C0≤0.25℃ 2 ); Sill value (C0+C): reflects the total spatial variation, with a value of 5-20℃ 2 (corresponding to a maximum actual temperature difference of 3-5°C); Range (a): reflects the spatial correlation distance, with a value of 0.1-0.5 meters (needs to be less than the cable connector length of 0.5-2 meters).

[0027] The model logic is: when the distance is less than or equal to the range, the semivariance approaches the sill value as the distance increases; when the distance is greater than the range, the semivariance stabilizes at the sill value.

[0028] Model training: The spatial distance of temperature point pairs (0.05-1 m) was divided into 10-15 lag groups (each group width 0.05-0.1 m), each group containing ≥30 point pairs, and the semi-variance of each group (the average of the square of the temperature difference) was calculated; parameter initialization: C0 took 10%-20% of the semi-variance of the minimum lag group, the base value took 90%-100% of the semi-variance of the maximum lag group, and the range took 1 / 3-1 / 2 of the total lag range.

[0029] Iterative optimization: By adjusting the parameters (range ± 0.05 m, C0 ± 0.1 ° C 2 , base value ±1℃ 2 ), minimize the root mean square error (RMSE) between the model prediction value and the experimental value, and iterate until the RMSE change is ≤ 0.05℃ 2 Or until 30 times; constraints: C0 ≤ base value × 10%, range ≤ 1 / 2 of the joint length.

[0030] Model implementation and verification: Visual verification: Plot the experimental values ​​against the fitting curve, requiring that more than 80% of the points fall within ±0.5°C of the curve. 2 Cross-validation: use 80% of the data for training and 20% for validation to ensure that the RMSE difference between the validation set and the training set is less than 10%; if it does not meet the standard, readjust the lag distance grouping or change the model (such as the exponential model).

[0031] The cable joint area (length 0.5-2 meters) is divided into three-dimensional grids of 0.01-0.05 meters. The temperature value of each grid point is calculated based on the trained variation function model to generate continuous temperature field distribution data.

[0032] Calculate the maximum temperature, minimum temperature, average temperature, and maximum temperature gradient of the temperature field (the ratio of the temperature difference between adjacent grids to the distance, reflecting the heat diffusion rate).

[0033] Hotspot identification: Set the temperature threshold to 70-80°C (based on the long-term allowable operating temperature of 90°C for cross-linked polyethylene cables), mark the area exceeding the threshold as a hotspot, and record its spatial coordinates and temperature value.

[0034] The standard deviation of the temperature field is calculated. Under normal operation, it should be ≤5°C. Exceeding this threshold indicates uneven temperature distribution. The amplitude threshold is set to 3-5 times the background noise mean (to distinguish valid pulses from noise), and the width threshold is set to 0.1-10μs (to filter out nanosecond electromagnetic interference and millisecond mechanical noise). Pulse signals that meet the characteristics of partial discharge are retained. For each valid pulse, the peak value (0.1-100mV), rise time (10-100ns), fall time (50-500ns), and pulse width (100ns-1μs) are extracted. Spectral analysis is used to extract the three main frequency components (1-100MHz) and their corresponding amplitudes. The pulse repetition rate (1-1000 times / minute) and the mean and variance of the amplitude distribution are calculated within one minute.

[0035] Through analysis, redundant features with a correlation degree of ≤0.3 with partial discharge types (such as corona and surface discharge) are eliminated, and 8-15 core features are retained to form a feature data set.

[0036] Environmental parameters (temperature, humidity, air pressure, and rainfall) were filtered for outliers, with valid ranges set as: -40°C to 50°C ambient temperature, 0-100% humidity, 80-110kPa air pressure, and 0-50mm / h rainfall. Missing values ​​≤ 5 minutes were filled using linear interpolation.

[0037] Determine the missing value type: Scan the original environmental parameter sequence (sampling interval is 1 minute), mark the missing value position, and classify it into short missing values ​​based on the missing value duration: continuous missing values ​​≤ 5 data points (corresponding to ≤ 5 minutes, meeting the preprocessing requirements); long missing values: continuous missing values ​​> 5 data points (this type of missing value is not suitable for linear interpolation and needs to be marked as an invalid segment and processed separately).

[0038] For short missing intervals, determine the valid data points before and after them (recorded as the front point A and the back point B). It is required that points A and B are normal data that have not been eliminated (such as the ambient temperature is within the range of -40℃-50℃), and there are no other valid points between A and B (to avoid multiple interpolation).

[0039] Assume that the timestamp of the front point A is t1 (e.g. 10:00), and the corresponding parameter value is v1 (e.g. 25°C); the timestamp of the back point B is t2 (e.g. 10:03), and the corresponding parameter value is v2 (e.g. 27°C); calculate the time interval Δt = t2 - t1 (unit: minutes) between A and B, with a value range of 2-6 minutes (because the missing value is ≤5 minutes, add 1 valid point before and after, and the minimum total interval is 2 minutes).

[0040] Number of missing points: Assume there are n missing points between A and B (n=1-5), and the timestamp of each missing point is t 1+1 , t 1+2 ,...,t 1+n (1 minute interval).

[0041] Slope calculation logic: Determine the parameter change rate (slope) per unit time based on the parameter difference and time interval between points A and B. For example, if point A is 25°C (10:00) and point B is 27°C (10:03), Δt = 3 minutes, and the total parameter change Δv = 2°C, then the rate of change per minute is 2°C / 3 ≈ 0.67°C / minute.

[0042] Fill point by point: The first missing point (10:01): value = v1 + 1 × rate of change (25 + 0.67 ≈ 25.67°C); the second missing point (10:02): value = v1 + 2 × rate of change (25 + 1.33 ≈ 26.33°C); and so on, until all missing points are filled. The calculated value of the last missing point must be consistent with v2 at point B (error ≤ 0.1°C to ensure linear continuity).

[0043] Error range control: Calculate whether the interpolation result is consistent with the changing trend of the adjacent valid points (for example, when the temperature rises as a whole, the interpolation result must increase monotonically), and the deviation of a single interpolation point from the value calculated according to the linear trend is ≤0.5℃ (humidity ≤2%, air pressure ≤0.5kPa).

[0044] Anomaly correction: If the interpolation result exceeds the reasonable range of the parameter (such as the temperature after interpolation is greater than 50℃), the linear interpolation will be abandoned and the historical average value of the parameter for the same period (such as the average value of the same period in the past 7 days) will be used to fill the gap to ensure that the physical meaning of the data is reasonable.

[0045] Through correlation analysis, the correlation between each parameter and the cable joint temperature is calculated. Parameters with a correlation degree ≥ 0.6 (usually ambient temperature and humidity) are retained, and weakly correlated parameters such as air pressure are eliminated. The retained environmental parameters are converted into dimensionless indices (such as humidity index = actual humidity ÷ 100, temperature index = (actual temperature + 40) ÷ 90) to facilitate cross-parameter fusion analysis.

[0046] The four types of processed data (current waveform after power frequency removal, temperature field analysis results, partial discharge pulse characteristics, and effective environmental parameters) are aligned by timestamp, with a unified sampling interval of 1 minute to form a multi-source state data set.

[0047] The min-max normalization method is used to map the data to the interval [0, 1]. The minimum and maximum values ​​of each parameter are taken as the extreme values ​​of their historical normal operating range (such as ground current 0-10A, temperature field maximum temperature 20-90℃), and finally a standardized status data packet is generated.

[0048] Power frequency rejection and outlier processing effectively remove noise and interference while retaining key features. Normalization and time alignment address the heterogeneity of multi-source data, ensuring that data of different types and magnitudes can be integrated for analysis. Steps such as temperature field hotspot identification and partial discharge pulse feature extraction specifically mine key information related to cable faults, improving the input validity of subsequent diagnostic models. Edge node preprocessing reduces invalid data transmission, conserving computing power for in-depth analysis on the cloud platform and improving system real-time performance.

[0049] In a preferred embodiment of the present invention, the standardized status data packet is transmitted to the cloud platform via a TCP / IP and power carrier dual-channel redundant network. A dynamic time warping algorithm is used to align the temporal and spatial dimensions of current, temperature, and partial discharge data, and environmental parameters are integrated to generate a multi-dimensional feature fusion data set, including: The standardized status data packet is transmitted to the cloud platform in parallel through the TCP / IP network channel and the power carrier communication channel; the data packets received by the dual channels are verified to be correct through redundancy verification and then a cloud-side reception verification data packet is generated; The verification data packets received on the cloud are processed to obtain the current time series, temperature time series, and partial discharge pulse time series. Dynamic time warping algorithm and nonlinear time axis transformation are then used to generate the current series, temperature series, and partial discharge series aligned in time and space. According to the current sequence, temperature alignment sequence and partial discharge alignment sequence aligned in the time and space dimensions, the normalized environmental parameter data is extracted from the verification data packet received through the cloud, and the current alignment sequence, temperature alignment sequence, partial discharge alignment sequence and environmental parameter data are integrated according to the time point to generate a multi-dimensional feature fusion dataset.

[0050] In the embodiment of the present invention, the TCP / IP channel configuration adopts the industrial Ethernet protocol, the transmission rate is 100Mbps-1Gbps, industrial dedicated ports (such as 502, 8080) are selected, the data packet format is JSON (including timestamp, device ID, data type, payload field), the sending interval is consistent with the standardized data packet generation frequency (1 minute / time), and the timeout retransmission threshold is set to 3 times (if exceeded, switching to the backup path).

[0051] Power line carrier channel configuration: low-voltage power line carrier communication (PLC) is used, the operating frequency is 3-500kHz (avoiding power frequency interference), the modulation method is orthogonal frequency division multiplexing (OFDM), the transmission rate is 500kbps-2Mbps, and the data packet is additionally coded with RS code channel (error correction capability 4-8 bytes), which is adaptable to the power line noise environment.

[0052] The edge node sends the same standardized status data packet to both channels simultaneously, marked with the channel identifier ("CH1" for TCP / IP and "CH2" for power carrier), ensuring that the data packets are transmitted independently at the physical link layer.

[0053] The cloud platform sorts dual-channel data packets by timestamp (error is allowed to be ±100ms), and data packets with the same timestamp are considered to be homologous data.

[0054] Calculate the CRC16 checksum for each data packet (polynomial 0x8005, initial value 0xFFFF) and compare it with the checksum field of the data packet. If the checksum passes, it is marked as a valid packet. If both channels are valid, select the TCP / IP channel data (with lower transmission delay) as the baseline and the power line carrier data as the backup.

[0055] If verification fails on a single channel, valid data from the other channel is used directly. If both channels fail, a data retransmission request is triggered (limited to packets within the last 5 minutes). If the number of retransmissions exceeds the limit, the data is marked as missing (the percentage must be ≤ 0.1%). Finally, a cloud-based verification packet is generated (including a complete timestamp and multiple data types).

[0056] The verification data packet received from the cloud is split into the current time series: the sampling frequency is 1kHz-10kHz, and each data point contains a timestamp (accurate to milliseconds) and a normalized current value (0-1); the temperature time series: the sampling frequency is 1 time / minute, and contains a timestamp (accurate to minutes) and temperature field statistical characteristics (maximum temperature, average temperature, etc., all normalized); the partial discharge pulse time series: non-uniformly spaced sampling (recorded when the pulse is triggered), including the pulse occurrence time (accurate to microseconds) and characteristic vectors (peak value, frequency, etc., normalized); duplicate timestamp data in the time series is removed (the latest value is retained) to ensure that the timestamp within a single sequence is unique.

[0057] Dynamic Time Warping (DTW) algorithm construction, training, and implementation: The core goal is to align sequences with different sampling frequencies and start times (e.g., current sequences with high frequency and high density, and temperature sequences with low frequency and low density). Three key elements are defined: Distance metric: Euclidean distance is used to measure the similarity between two data points (applicable to normalized data, with a value range of 0-√2); Path constraint: Sakoe-Chiba window constraint is used (the window width is 5%-10% of the sequence length, for example, the window width for a 1000-point sequence is 50-100 points) to limit the search range of the alignment path and avoid physically unreasonable matches (such as aligning a sudden current change with a stable temperature segment); Cumulative distance: The sum of the cumulative distances from the start point to the end point of the sequence is used to evaluate the overall alignment effect (the smaller the value, the higher the alignment).

[0058] Training data: 30 days of historical normal operation data (including current, temperature, and partial discharge sequences) were selected to ensure coverage of different operating conditions (such as peak and off-peak loads). Window width optimization: The physical consistency of the aligned sequences (such as the matching degree between the current peak and the temperature rise trend) was calculated for different window widths (3%-15% of the sequence length). The minimum window with a matching degree ≥ 85% was selected (balancing accuracy and computational complexity), and the final window width was determined to be 5%-10%. Distance threshold setting: The cumulative distance after normal data alignment was calculated, and the 95th percentile value was used as the threshold (usually 0.5-2.0, after normalization). If the threshold was exceeded, it was determined to be an alignment anomaly (requiring manual review).

[0059] Sequence preprocessing: The current sequence is downsampled to 1 time / minute (the same frequency as the temperature sequence), retaining the peak and average values ​​within each minute as features. Distance matrix construction: The Euclidean distance between each point pair in the current feature sequence (length N) and the temperature sequence (length M) is calculated to form an N×M matrix. Path search: Starting from the upper left corner of the matrix (the starting point), the path with the smallest cumulative distance is selected within the constraint window, allowing one-to-one, one-to-many, or many-to-one matching (for example, matching one current peak point to three temperature rise segments). Nonlinear time axis transformation: The timestamp of the temperature sequence is adjusted according to the adapted path, so that each temperature point is aligned with the corresponding current feature point (error ≤ 10 seconds). Partial discharge sequence alignment: Using the same logic, the partial discharge pulse time is mapped to the aligned time axis (based on the current sequence time), ensuring that the pulse occurrence time matches the changing trends of current and temperature.

[0060] Finally, a current alignment sequence, temperature alignment sequence, and partial discharge alignment sequence are generated that are aligned in time and space dimensions (time stamps are uniformly accurate to seconds and have consistent lengths).

[0061] Normalized environmental parameter data (ambient temperature index, humidity index, range 0-1) is extracted from the verification data packet received from the cloud. Its timestamp is consistent with the temperature sequence (1 time / minute). Through timestamp matching, the environmental parameters are associated with the aligned current, temperature, and partial discharge sequences (four data items with the same timestamp constitute a feature sample).

[0062] Based on the aligned time axis (maximum sequence length, typically 24 × 60 = 1440 points / day), check whether each time point contains four data items: if environmental parameters are missing at a time point, linear interpolation is used to fill in the gaps (error ≤ 0.05); if partial discharge data are missing, it is marked as "no pulse" (eigenvectors are all zeros).

[0063] After integration, each sample contains: current alignment features (3-5 statistics), temperature alignment features (4-6 statistics), partial discharge alignment features (8-15 core features), and environmental parameters (2 indices), forming a structured data record.

[0064] 10% of the samples were randomly selected to check the timestamp consistency (error ≤ 1 second) and data integrity (missing rate ≤ 0.5%) to ensure that the multi-dimensional features corresponded one-to-one in time, and finally a multi-dimensional feature fusion dataset was generated.

[0065] The dual-channel redundant design of TCP / IP and power carrier ensures a data transmission success rate of ≥99.9%. Automatic switching in the event of a single channel failure prevents data loss. The integrated dataset includes multi-dimensional features such as current, temperature, partial discharge, and environmental conditions, with consistent temporal and spatial dimensions, improving model recognition accuracy. Algorithm parameters (such as window width and distance threshold) are optimized based on historical data training, balancing real-time performance with computing power requirements.

[0066] In a preferred embodiment of the present invention, a set of partial discharge distribution points is extracted from a multi-dimensional feature fusion data set, and a weight factor of each distribution point is calculated. Based on the weight factor, a weighted point set concave hull generation algorithm is used to construct a partial discharge source probability distribution boundary to obtain a set of boundary feature points. The set of boundary feature points and the multi-dimensional feature fusion data set are input into a convolutional-long short-term memory hybrid neural network to generate a diagnosis result, including: The spatiotemporal distribution point set of partial discharge pulses is extracted from the multi-dimensional feature fusion data set. Based on the spatiotemporal distribution point set of partial discharge pulses, the time aggregation weight factor of each point's timestamp and the relative intensity weight factor of each point's amplitude are calculated, and the two are fused to generate a joint weight factor to obtain a weighted partial discharge site set. The PD distribution point set and the corresponding weight factors of each point are generated by using the weighted point set concave hull generation algorithm to calculate a set of convex boundary points that can surround all PD distribution points, construct the probability distribution boundary of the high-weight area, and obtain the boundary feature point set. The boundary feature point set is mapped to generate a spatial probability matrix and fused with the time series data. The matrix is ​​input into the convolution-LSTM hybrid neural network to extract spatiotemporal features and generate diagnostic results.

[0067] In an embodiment of the present invention, partial discharge pulse data is screened from a multi-dimensional feature fusion data set, and the spatiotemporal attributes of each pulse are extracted to form a spatiotemporal distribution point set; each point contains: a timestamp (accurate to milliseconds, consistent with the aligned time axis); spatial coordinates (based on the relative position of the three-dimensional model of the cable connector, such as axial distance 0-2 meters and radial distance 0-0.5 meters); and pulse amplitude (normalized to 0-1, corresponding to the original amplitude 0.1-100mV).

[0068] The point set needs to remove duplicate points (those with the same space-time coordinates and pulses with the largest amplitude within ±1ms of the timestamp) to ensure that there is no redundant data in a single point set.

[0069] A sliding time window is used to calculate the clustering degree of statistical points, and the window size is set to 5-10 minutes (covering the typical burst period of partial discharge pulses). For each point, the total number of pulses in its window is calculated (denoted as N). The time clustering weight factor = N ÷ the maximum number of pulses in the window (value range 0-1). For example, if there are a maximum of 50 pulses in a window and a window containing 30 pulses at a certain point has a time weight of 0.6. The window sliding step is set to 1-2 minutes to ensure that the time resolution matches the overall monitoring frequency (1 minute / time).

[0070] Taking the maximum amplitude of all pulses in the point set as the benchmark (denoted as Vmax), the relative intensity weight factor of each point = the amplitude of the point ÷ Vmax (value range 0-1); to avoid small-amplitude pulses being ignored, the minimum intensity threshold is set to 0.1 (the weight of points with amplitudes lower than Vmax×10% is 0.1, not 0) to ensure the rationality of the weight of weak signals.

[0071] Combined weight factor = 0.4 × time aggregation weight + 0.6 × relative intensity weight (the time weight accounts for 40%-60%, which can be dynamically adjusted according to the type of partial discharge. For example, the time weight ratio is increased for periodic partial discharge).

[0072] The combined weight after fusion ranges from 0.1 to 1 (because the lowest intensity weight is 0.1). The higher the value, the greater the contribution of the point to fault diagnosis. Finally, a weighted partial discharge site set is generated (each point is attached with a combined weight).

[0073] Weight threshold determination: The weight threshold is set based on the noise level in the multi-dimensional feature fusion data, with a value range of 0.3-0.7; if the monitoring environment noise is high (such as a complex cable laying environment), the threshold is set at 0.5-0.7 to filter out low-weight noise points; if the noise is low (such as a laboratory or a new cable system), the threshold is set at 0.3-0.5 to retain more potential effective signals. High-weight key extraction: Points with a joint weight factor higher than the above threshold are screened out to form a high-weight key subset; the number of points in this subset usually accounts for 20%-50% of the total number of original distribution points (if the proportion is too low, effective information may be lost, and if it is too high, redundant points will be retained). The detailed process of constructing the probability distribution boundary of partial discharge sources using the weighted point set concave hull generation algorithm: From the weighted partial discharge site set, low-weight points with a joint weight factor of less than 0.1 (considered as noise or interference signals) are eliminated, and the valid point set (denoted as P, containing N points, N is usually 50-500, depending on the monitoring duration and discharge activity) is retained; the spatial coordinates of the valid points (such as the distance value along the cable length) are mapped to a unified coordinate system (for example, with the cable starting point as the origin, the coordinate range is normalized to a relative value of 0-100) to avoid the boundary calculation being affected by the absolute distance difference; ensure that the joint weight factor of all retained points is within the range of 0.1-1.0 (if the original calculation exceeds this range, it is adjusted to this interval through linear scaling), providing a unified benchmark for subsequent weighted calculations.

[0074] Initial convex hull generation (basis of boundary calculation): Calculate the minimum convex hull (that is, the smallest convex polygon that can completely surround all points) for the valid point set P: first find the point with the smallest y coordinate in the point set (if there are multiple points, choose the one with the smallest x coordinate) as the starting point, and then sort the remaining points in ascending order according to the polar angle with the starting point. Then use the stack structure to determine the direction of the points in turn (left turns are retained, right turns are eliminated), and finally form the vertex sequence of the convex hull (denoted as H, containing M vertices, M is usually 5-30, less than the number of points in the point set P).

[0075] Connect the convex hull vertices in order to form M continuous edge segments (denoted as L1, L2, ..., L m ), each line segment is determined by two adjacent vertices, and the line segment length range is usually 5-50 (in normalized coordinates, corresponding to the actual cable length of 0.5-5m).

[0076] Convex hull edge evaluation and candidate concavity point identification: For each convex hull edge L i (Connect vertices A and B), traverse all valid points in P that are not in L i Points on, calculate the distance from each point to line segment L i The vertical distance (denoted as d) is selected to select the point with d>0 (i.e., the point located at L i The inner point may become the concavity reference point).

[0077] Weighted distance sorting: For the selected inner points, calculate the "weighted distance" = d × the joint weight factor of the points (the higher the weight, the greater the influence of the point on edge adjustment). Then sort them from large to small by weighted distance, and take the top 3-5 points as candidate concave points for the edge (these points are the key points most likely to "pull" the edge inward).

[0078] Weight-based edge concavity adjustment: For each edge L i and its candidate concave points, to determine whether the concave conditions are met: the joint weight factor of the candidate points is ≥ 0.5 (to ensure that the point has sufficient reference value), and the point to L iThe distance should be ≥ 10% of the edge length (to avoid meaningless fine-tuning due to being too close).

[0079] If there is a candidate point that meets the conditions (denoted as C), the original edge L i (AB) is split into two segments: AC and CB. At this time, the "convex corner" of the original convex hull is replaced by a "concave corner" (point C becomes the new boundary vertex).

[0080] After splitting, the total length of the new line segments AC and CB is equal to the original line segment L. i The length difference is ≤30% (to prevent boundary distortion caused by excessive concavity); if the weight of candidate point C is greater than 0.8 (high weight), the concavity is retained first; if the weight is between 0.5-0.8, it is necessary to check whether there are other high-weight points (≥0.7) around that support the concavity. If so, retain it; otherwise, discard it.

[0081] Multi-round iterative optimization: Repeat the above concavity operation for all edges for 2-3 rounds (avoid boundary fragmentation caused by excessive iterations). After each round, recalculate the convex hull vertex sequence and update the boundary segments.

[0082] After concave adjustment, the boundary may contain too many vertices (for example, the distance between some adjacent vertices is less than 2 normalized units, corresponding to an actual distance of less than 0.2m). The Douglas-Peucker algorithm is used to simplify the boundary, retain key vertices (the error threshold is set to 1-2 normalized units), and reduce the number of feature points (ultimately retaining 10-50 to balance accuracy and computational efficiency).

[0083] Douglas-Peucker algorithm steps: Assume that the set of boundary points to be simplified is an ordered sequence S=[Q0, Q1, ..., Q n ] (arrange in clockwise / counterclockwise order, connect end to end to form a closed boundary), set the distance threshold ε (set according to the accuracy requirements, such as 1-2 units in normalized coordinates, corresponding to 0.1-0.2m of the actual cable length); the smaller the threshold, the more points retained after simplification, and the closer the shape is to the original boundary; the larger the threshold, the higher the degree of simplification, and the fewer the number of points.

[0084] Take the first point Q0 and the last point Q in the point set n , connected to form the initial line segment M=Q0-Q n .

[0085] Calculate all the intermediate points (Q1 to Q n-1 ) to the perpendicular distance of line segment M, find the point Q with the largest distance k , and the distance is d k .

[0086] Judgment and recursive processing: If d k >ε:description of Q k It is the key feature point that affects the boundary shape and must be retained; split the point set S into two sub-point sets: S1=[Q0, Q1, ..., Q k ] and S2=[Q k , Q k+1 ,...,Q n ], repeat the recursive simplification for the two sub-point sets respectively; if d k ≤ε: This means that the distances from all intermediate points to the line segment M are less than the threshold. These points can be considered redundant points and are directly removed. The simplified line segment is M, and only Q0 and Q n ; After the recursion is completed, all retained key feature points are spliced ​​in the original order to form a simplified boundary point set.

[0087] Example description: Assume that the original boundary has 100 vertices and the threshold ε = 1.5 (normalized coordinates): First round: take the first and last points Q0 and Q 99 , calculate the distance between the middle 98 points and the line segment Q0-Q 99 distance, find the farthest point Q 50 (distance d=3>1.5), retain Q 50 , split into [Q0-Q 50 ] and [Q 50 -Q 99 ].

[0088] Second round: [Q0-Q 50] , find the farthest point Q 20 (d=2>1.5), retain and split; for [Q 50 -Q 99 ], find the farthest point Q 70 (d=1.2≤1.5), remove the middle point and keep only Q 50 and Q 99 .

[0089] By repeating the recursion, 100 points may eventually be reduced to 15-20 key vertices, which not only retains the concave and convex features of the boundary but also greatly reduces the amount of data.

[0090] Ensure that the streamlined vertex sequence is connected end to end to form a closed concave polygonal boundary. The boundary must completely surround all points with weights ≥ 0.3 in the valid point set (low-weight points can be excluded from the boundary and regarded as minor interference). Arrange the vertices of the final boundary in clockwise or counterclockwise order to form a set of boundary feature points (each point contains standardized coordinates and corresponding weights), which will serve as one of the input features of the subsequent neural network.

[0091] Network structure construction: Input layer: Receives the spatial probability matrix (size 32×32 or 64×64, where the matrix value is the joint weight of the corresponding position, reflecting the spatial distribution) mapped to the boundary feature point set; and the time series of the multi-dimensional feature fusion dataset (including current, temperature, and environmental parameters, with a length of 10-30 minutes).

[0092] Convolutional layer: Set up 2-3 convolutional layers. The first layer uses 32 3×3 filters (to extract local spatial features), and the second layer uses 64 3×3 filters (to extract combined features). Each layer is followed by 2×2 maximum pooling (to reduce the dimension). The activation function uses ReLU (to avoid gradient disappearance).

[0093] LSTM layer: Set up 1-2 layers of LSTM units (64-128 units / layer), receive the spliced ​​data of the spatial features and time series output of the convolution layer, and learn temporal associations (such as the lag relationship between partial discharge pulses and temperature rise).

[0094] Output layer: Set the number of nodes according to the diagnostic target (e.g., 3 types of faults + 1 type of normal, a total of 4 nodes), and output the probability of each category (range 0-1).

[0095] Network training process: 70% of the historical monitoring data was selected as the training set (containing normal and faulty labels), 20% as the validation set, and 10% as the test set, ensuring that faulty samples accounted for 30%-40% (balanced category distribution). The optimizer used was Adam (learning rate 0.001-0.01, decaying by 10% every 5 epochs). Parameters were adjusted through iterative optimization for 30-50 iterations (until the validation set accuracy no longer improved). A dropout layer (ratio 0.2-0.3) and L2 regularization (coefficient 0.0001) were added to prevent overfitting (the difference in accuracy between the training set and validation set must be less than 5%).

[0096] Input data preprocessing: The spatial probability matrix is ​​normalized to [0, 1], and the time series is spliced ​​into continuous segments at the minute level.

[0097] The convolutional layer outputs a spatial feature map (reflecting the spatial morphology of high-weight areas), the LSTM layer outputs a temporal feature vector (reflecting the trend of parameter changes), and after fusion by the fully connected layer, the output is a probability distribution.

[0098] The category with the highest probability is taken as the final result. If the maximum probability is <0.7 (insufficient credibility), it is marked as "pending review" and the manual verification process is triggered.

[0099] Value range: 32-64 convolution filters, 64-128 LSTM units, learning rate 0.001-0.01, 30-50 training iterations, spatial matrix size 32×32-64×64, time series segment length 10-30 minutes.

[0100] Weighting factors enhance high-value partial discharge signals, reduce noise interference, and focus the model on fault-related features, improving diagnostic sensitivity. Weighted concave hull boundaries accurately enclose high-weight regions, transforming three-dimensional spatial features into structured boundary points. This resolves ambiguity in partial discharge source location and reduces spatial positioning errors. The convolution-LSTM network simultaneously captures both the spatial morphology and temporal evolution of partial discharges (such as changes in pulse frequency and amplitude during fault development), significantly improving diagnostic accuracy compared to single models (pure convolution or pure LSTM).

[0101] In a preferred embodiment of the present invention, a three-level warning signal is generated based on the diagnosis results, and the warning information is pushed to the remote operation and maintenance terminal and mobile app via the MQTT protocol, automatically triggering the digital work order system to generate maintenance tasks and perform location verification, including: Analyze the cable insulation status risk level value in the diagnosis result; match the risk level value with the preset three-level warning threshold range to obtain the warning signal level; extract the spatial location coordinates and timestamp data of the partial discharge source in the diagnosis result; encapsulate the warning signal level, location coordinates, and timestamp into a structured MQTT protocol message; Push warning information in parallel via the cloud platform to remote operation and maintenance terminals and mobile apps using structured MQTT protocol messages. Generate maintenance priority identifiers based on warning signal levels. Bind location coordinates, warning signal levels, and priority identifiers to generate structured maintenance instructions. Automatically trigger the digital work order service to generate electronic maintenance work orders. The spatial positioning service of the on-site verification terminal is driven by the positioning coordinates in the maintenance work order to execute local discharge source positioning and receive feedback on the verification results.

[0102] The cable insulation status risk level value is extracted from the diagnosis results (normalized to a range of 0-100, where 0 indicates no risk and 100 indicates extremely high risk).

[0103] Preset three-level warning threshold range: Level 1 warning (the highest level): risk level 70-100 (corresponding to severe insulation degradation and severe partial discharge); Level 2 warning: risk level 30-70 (corresponding to mild insulation aging and moderate partial discharge); Level 3 warning: risk level 0-30 (corresponding to slight insulation abnormality and weak or occasional partial discharge).

[0104] Match the analyzed risk level value with the threshold range to determine the warning signal level (e.g., risk level 85 matches level 1 warning).

[0105] Extract the spatial coordinates of the PD source from the diagnostic results: This is based on the three-dimensional coordinate system of the cable laying path, including the axial distance (0-5 km, accurate to ±1 m), radial depth (0-1 m, accurate to ±0.1 m), and elevation (±50 m, referenced to ground level). Extract timestamp data: accurate to the second (in the format of YYYY-MM-DDHH:MM:SS, consistent with the time the diagnostic results were generated, with an error of ≤10 seconds).

[0106] Set the message subject to "PowerCable / Warning / [Device ID]" (the device ID is a unique 10-20-digit code, such as "CBL-20230723-001").

[0107] The payload field contains structured data: warning level ("Level 1", "Level 2", "Level 3"); positioning coordinates (axial, radial, and elevation values ​​and units); timestamp (in the format string described above); and risk level value (retaining one decimal place, such as 85.2).

[0108] The message QoS (Quality of Service) is set to 1 (ensuring at least one delivery), and the message retention is set to "on" (new access terminals can obtain historical warnings).

[0109] The cloud platform pushes messages to the remote operation and maintenance terminal (industrial tablet, supporting 4G / 5G) and mobile app (iOS / Android system) simultaneously through MQTT broker.

[0110] Remote terminal: After receiving the alarm, it triggers an audible and visual alarm (volume ≥ 80dB, light flashing frequency 2 times / second), and displays the complete warning information on the screen (font size is enlarged to 24, and the positioning coordinates are highlighted).

[0111] Mobile app: Push notification message (title includes "Warning Level + Device ID", content summary shows risk level and time), click to jump to the details page (including map location entrance).

[0112] After the first push fails, retry three times with intervals of 10s, 30s, and 60s. If it still fails, it will be recorded in the log and trigger a manual intervention reminder.

[0113] Set priorities based on alert levels: Level 1 warning: priority is "urgent" (marked as "P0"), requiring a response within 1 hour and arrival at the scene within 4 hours; Level 2 warning: priority is "high" (marked as "P1"), requiring a response within 2 hours and arrival at the scene within 12 hours; Level 3 warning: priority is "general" (marked as "P2"), requiring a response within 24 hours and completion of verification within 72 hours.

[0114] The maintenance instructions include: equipment ID, warning level, priority identification, location coordinates (format after latitude and longitude conversion, error ±5m), and diagnosis result summary (such as "partial discharge source is located at connector A3, amplitude 120mV").

[0115] Work order numbering format: "WO-last two digits of the year-month-day-serial number" (e.g., "WO-25-07-23-005"); includes the following fields: work order status (initially "pending"), responsible team (automatically assigned based on equipment ownership), estimated completion time (calculated based on priority), and associated warning ID (bound to the unique identifier of the MQTT message).

[0116] The work order will be pushed to the responsible team terminal through the system within 5 minutes after it is generated, and will be updated to the operation and maintenance management platform dashboard simultaneously.

[0117] After maintenance personnel arrive at the approximate area with an on-site verification terminal (supporting GPS and Beidou dual-mode positioning, with a positioning accuracy of ±3m), the terminal automatically activates the spatial positioning service based on the positioning coordinates in the work order: it starts the partial discharge detector (detection frequency band 30kHz-30MHz) and narrows the detection range based on the coordinate information (initial radius 10m, dynamically adjusted with signal strength: reduced to 3m for strong signals and expanded to 20m for weak signals).

[0118] The terminal screen displays the real-time positioning deviation (the straight-line distance between the current position and the target coordinates, in meters). When the deviation is ≤5 meters, a prompt "Enter the verification area" will be displayed.

[0119] On-site personnel enter the verification results through the terminal, with the result type being: "Confirmed fault", "Suspected fault", or "No fault". Additional information includes: text description (e.g., "The temperature at connector A3 is 38°C, and a partial discharge signal is detected"), on-site photos (≥2 photos, including equipment identification and measuring instrument readings), and positioning correction coordinates (if the actual position deviates from the work order coordinates by >5m, manual adjustment is required).

[0120] Feedback information is uploaded to the cloud platform via 4G / 5G, triggering an update of the work order status (such as "verified") and synchronously linked to the early warning record to form a closed loop.

[0121] Three-level threshold intervals are tied to priorities, ensuring rapid response to high-risk faults, orderly handling of low-risk issues, and optimized resource allocation. Automated processes, from early warning location to work order generation and on-site verification, shorten troubleshooting time. The MQTT protocol, combined with a retry mechanism and multi-terminal push notifications, improves the delivery rate of early warning information. Dual-mode positioning and dynamic detection range adjustment enhance verification success rates.

[0122] In a preferred embodiment of the present invention, the standardized status data packet is transmitted to the cloud platform in parallel via the TCP / IP network channel and the power carrier communication channel; the data packets received by the dual channels are verified to be correct through redundancy verification, and then a cloud-side reception verification data packet is generated, including: The standardized status data packet is copied into two copies at the same time and transmitted to the cloud platform in parallel via the TCP / IP network channel and the power carrier communication channel respectively. The cloud platform receives the data packet copies of the two channels respectively, which are recorded as copy A and copy B, and performs preliminary data integrity checks on each. Based on the status of replicas A and B after preliminary verification, a preset time window is started for waiting; if both replicas are valid within the window period, valid replicas A and B are obtained; otherwise, an exception is triggered and the process is terminated; Apply the same preset verification algorithm to valid replicas A and B, respectively, to generate verification identifiers A and B, and compare the two. If they match, generate replicas A and B with the same content. Otherwise, trigger data inconsistency exception processing and terminate this process. Based on the confirmed content of replicas A and B, one of the replicas is selected according to the preset selection strategy, and a cloud-received verification data packet is generated and output.

[0123] In an embodiment of the present invention, one standardized status data packet is completely replicated to generate two copies (copy A and copy B) with completely identical content. Copy A is transmitted via a TCP / IP network channel, and copy B is transmitted via a power carrier communication channel, with both channels starting transmission simultaneously.

[0124] Transmission rate range: TCP / IP channels are typically 10Mbps-10Gbps (limited by network bandwidth); power line carrier channels are typically 10kbps-200Mbps (affected by power line noise and transmission distance).

[0125] The cloud platform receives copies of the two channels respectively and performs preliminary checks on each copy, including packet length verification: checking whether the length of the received packet is within a preset range (e.g., 128 bytes to 2048 bytes, as defined by the data content). If it exceeds the range, it is deemed invalid.

[0126] Identifier integrity check: Checks whether the start identifier (such as the fixed 2-byte "0xAA55") at the header of the data packet and the end identifier (such as "0x55AA") at the tail are complete. If they are missing, the data packet is considered invalid.

[0127] Simple checksum check: Calculate the cumulative sum of all data bytes in the data packet and compare it with the checksum field carried in the data packet. The two must be completely consistent (the error range allowed is 0), otherwise it is considered invalid. If the first valid copy (such as copy A) arrives, a preset time window is started to wait for the second copy (copy B).

[0128] Time window value range: usually 50ms-500ms (set according to the maximum transmission delay difference between the two channels. For example, the window can be appropriately enlarged when the power carrier channel delay is high).

[0129] Result determination: If both copies in the window pass the preliminary check (both are valid), the process continues; otherwise, exception handling is triggered (such as re-requesting data) and the process terminates.

[0130] Verification identifier generation and comparison: Apply the same preset verification algorithm (such as CRC32 or SHA-1) to valid copies A and B respectively to generate verification identifiers. Taking CRC32 as an example: Calculate the 32-bit CRC value for the complete data content of the copy (excluding the check field itself) to obtain verification identifiers A and B. The value range is 0-0xFFFFFFFF (32-bit unsigned integer).

[0131] Comparison rules: Verification identifier A and verification identifier B must be completely consistent (the numerical difference is 0). Otherwise, the data is considered inconsistent, triggering exception processing and terminating.

[0132] Based on the preset strategy, one of the replicas A and B with consistent content is selected as the cloud-side receiving verification data packet output: Selection strategy examples: first-come first-served (select the replica that arrives first); channel priority (such as the preset TCP / IP channel priority); random selection.

[0133] Dual-channel parallel transmission and redundant verification prevent data loss caused by single-channel failures, improving transmission success rates. Multi-level verification (initial integrity + verification identifier comparison) ensures received data is consistent with the original data, reducing the risk of transmission errors. Exception handling mechanisms promptly address channel delays, data corruption, and other issues, ensuring process stability.

[0134] An embodiment of the present invention further provides a computing device comprising: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, executes the system described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0135] The embodiment of the present invention further provides a computer-readable storage medium storing instructions, which, when executed on a computer, causes the computer to execute the system described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0136] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A comprehensive online monitoring system for power cables, characterized in that: include: The data acquisition module is used to collect the time domain waveform of the power cable grounding current, the joint temperature field distribution, the partial discharge pulse sequence and the environmental parameters to obtain a multi-type raw data set; The data processing module is used to input multiple types of raw data sets into the edge computing node, perform data preprocessing, and generate standardized status data packets; The data fusion module is used to transmit standardized status data packets to the cloud platform via a dual-channel redundant network of TCP / IP and power carrier. It uses a dynamic time warping algorithm to align the temporal and spatial dimensions of current, temperature, and partial discharge data, and integrates environmental parameters to generate a multi-dimensional feature fusion data set. The fault diagnosis module is used to extract the set of partial discharge distribution points from the multi-dimensional feature fusion data set and calculate the weight factor of each distribution point. Based on the weight factor, the weighted point set concave hull generation algorithm is used to construct the probability distribution boundary of the partial discharge source to obtain a set of boundary feature points. The boundary feature point set and the multi-dimensional feature fusion data set are input into the convolution-long short-term memory hybrid neural network to generate the diagnosis result. The intelligent response module is used to generate three-level warning signals based on diagnostic results, push warning information to remote operation and maintenance terminals and mobile apps through the MQTT protocol, and automatically trigger the digital work order service to generate maintenance tasks and perform positioning verification.

2. The comprehensive online monitoring system for power cables according to claim 1, characterized in that: Input multiple types of raw data sets into the edge computing node, perform data preprocessing, and generate standardized status data packets, including: The time-domain waveform data of the power cable grounding current in the multi-type original data set is processed to generate the current waveform data after the power frequency is eliminated; the temperature field data is generated using the Kriging interpolation method based on the discrete point data of the power cable joint temperature field distribution in the multi-type original data set; the temperature field data is calculated and identified to obtain the temperature field analysis result data; Processing the partial discharge pulse sequence data in the multi-type original data set to obtain a partial discharge pulse feature data set; and processing the environmental parameters in the multi-type original data set to generate effective environmental parameter data; The current waveform data after power frequency removal, temperature field analysis result data, partial discharge pulse characteristic data set and effective environmental parameter data are collected and processed to generate a multi-source state data set; the multi-source state data set is normalized to obtain a standardized state data packet.

3. The comprehensive online monitoring system for power cables according to claim 2, characterized in that: The standardized status data packets are transmitted to the cloud platform via a dual-channel redundant network of TCP / IP and power carrier. A dynamic time warping algorithm is used to align the temporal and spatial dimensions of current, temperature, and partial discharge data. Environmental parameters are integrated to generate a multi-dimensional feature fusion data set, including: The standardized status data packet is transmitted to the cloud platform in parallel through the TCP / IP network channel and the power carrier communication channel; the data packets received by the dual channels are verified to be correct through redundancy verification and then a cloud-side reception verification data packet is generated; The verification data packets received on the cloud are processed to obtain the current time series, temperature time series, and partial discharge pulse time series. Dynamic time warping algorithm and nonlinear time axis transformation are then used to generate the current series, temperature series, and partial discharge series aligned in time and space. According to the current sequence, temperature alignment sequence and partial discharge alignment sequence aligned in the time and space dimensions, the normalized environmental parameter data is extracted from the verification data packet received through the cloud, and the current alignment sequence, temperature alignment sequence, partial discharge alignment sequence and environmental parameter data are integrated according to the time point to generate a multi-dimensional feature fusion dataset.

4. The comprehensive online monitoring system for power cables according to claim 3, characterized in that: Extract the PD distribution point set from the multi-dimensional feature fusion dataset and calculate the weight factor of each distribution point. Based on the weight factor, use the weighted point set concave hull generation algorithm to construct the PD source probability distribution boundary and obtain the boundary feature point set. The boundary feature point set and the multi-dimensional feature fusion dataset are input into the convolution-long short-term memory hybrid neural network to generate the diagnosis results, including: The spatiotemporal distribution point set of partial discharge pulses is extracted from the multi-dimensional feature fusion data set. Based on the spatiotemporal distribution point set of partial discharge pulses, the time aggregation weight factor of each point's timestamp and the relative intensity weight factor of each point's amplitude are calculated, and the two are fused to generate a joint weight factor to obtain a weighted partial discharge site set. The PD distribution point set and the corresponding weight factors of each point are generated by using the weighted point set concave hull generation algorithm to calculate a set of convex boundary points that can surround all PD distribution points, construct the probability distribution boundary of the high-weight area, and obtain the boundary feature point set. The boundary feature point set is mapped to generate a spatial probability matrix and fused with the time series data. The matrix is ​​input into the convolution-LSTM hybrid neural network to extract spatiotemporal features and generate diagnostic results.

5. The comprehensive online monitoring system for power cables according to claim 4, characterized in that: Based on the diagnostic results, a three-level warning signal is generated and pushed to the remote operation and maintenance terminal and mobile app via the MQTT protocol. This automatically triggers the digital work order system to generate maintenance tasks and perform location verification, including: Analyze the cable insulation status risk level value in the diagnosis result; match the risk level value with the preset three-level warning threshold range to obtain the warning signal level; extract the spatial location coordinates and timestamp data of the partial discharge source in the diagnosis result; encapsulate the warning signal level, location coordinates, and timestamp into a structured MQTT protocol message; Push warning information in parallel via the cloud platform to remote operation and maintenance terminals and mobile apps using structured MQTT protocol messages. Generate maintenance priority identifiers based on warning signal levels. Bind location coordinates, warning signal levels, and priority identifiers to generate structured maintenance instructions. Automatically trigger the digital work order service to generate electronic maintenance work orders. The spatial positioning service of the on-site verification terminal is driven by the positioning coordinates in the maintenance work order to execute local discharge source positioning and receive feedback on the verification results.

6. The comprehensive online monitoring system for power cables according to claim 5, characterized in that: Transmit the standardized status data packet to the cloud platform in parallel through the TCP / IP network channel and the power carrier communication channel; After the data packets received by the dual channels are verified as correct through redundancy check, a cloud reception verification data packet is generated, including: The standardized status data packet is copied into two copies at the same time and transmitted to the cloud platform in parallel via the TCP / IP network channel and the power carrier communication channel respectively. The cloud platform receives the data packet copies of the two channels respectively, which are recorded as copy A and copy B, and performs preliminary data integrity checks on each. Based on the status of replicas A and B after preliminary verification, a preset time window is started for waiting; if both replicas are valid within the window period, valid replicas A and B are obtained; otherwise, an exception is triggered and the process is terminated; Apply the same preset verification algorithm to valid replicas A and B, respectively, to generate verification identifiers A and B, and compare the two. If they match, generate replicas A and B with the same content. Otherwise, trigger data inconsistency exception processing and terminate this process. Based on the confirmed content of replicas A and B, one of the replicas is selected according to the preset selection strategy, and a cloud-received verification data packet is generated and output.

7. The comprehensive online monitoring system for power cables according to claim 6, characterized in that: Process the verification data packets received on the cloud to obtain the current time series, temperature time series and partial discharge pulse time series; The dynamic time warping algorithm and nonlinear time axis transformation are used to generate current sequences, temperature sequences, and partial discharge sequences aligned in time and space, including: Based on the cloud, it receives verification data packets, performs predefined protocol parsing operations, and generates original current time series, temperature time series, and partial discharge pulse time series; The temperature time series is designated as the reference series, and a dynamic time warping algorithm is applied between the current time series and the temperature time series. The nonlinear time warping path between the two series is calculated to generate a current-temperature warping mapping relationship. Apply a dynamic time warping algorithm between the PD pulse time series and the temperature time series; calculate the tortuous path and generate a PD-temperature warping mapping relationship; Based on the extracted original time series and mapping relationship, nonlinear time axis transformation is performed to generate current alignment sequence and partial discharge alignment sequence that are completely aligned with the time axis of the temperature series, and the original temperature series is used as the temperature alignment sequence.

8. The comprehensive online monitoring system for power cables according to claim 7, characterized in that: The PD distribution point set and the weight factors corresponding to each point are used to generate the weighted point set concave hull algorithm. A set of convex boundary points that can surround all PD distribution points is calculated to construct the probability distribution boundary of the high-weight area and obtain the boundary feature point set, including: Receive a set of partial discharge distribution points and the joint weight factors corresponding to each point; perform a descending sort operation on the entire point set based on the weight factor value of each point; after sorting, the point with the largest weight factor is at the front of the sequence, select the first three non-collinear points with the largest weight factors, and calculate and construct an initial convex hull; Traverse the remaining point set in descending order of weight, determine the position relationship between each point and the convex hull, and if it is outside, dynamically reconstruct the convex hull and update the boundary vertex set to complete the ordered iteration; After the ordered iteration is completed, the convex hull vertex set formed by high-weight points is processed first based on the weight descending mechanism, and the boundary feature point set representing the boundary of the local discharge source probability distribution is output.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the system according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which, when executed by a processor, implements the system according to any one of claims 1 to 8.

Images