Method, medium and device for on-line monitoring of submarine cable insulation performance and fault location

Abstract

The application discloses a submarine cable insulation performance online monitoring and fault positioning method, medium and equipment, wherein the method comprises the following steps: S1, multi-source sensing data real-time synchronous acquisition; S2, multi-modal noise cooperative suppression and feature extraction; S3, multi-feature fusion insulation state degradation evaluation; S4, double-end traveling wave accurate fault positioning; S5, data visualization and alarm linkage. The application can effectively suppress the noise interference of the complex marine environment, multi-dimensionally and in real time monitor the key parameters representing the insulation state of the submarine cable, realize early warning of insulation degradation and preliminary judgment of fault modes through an intelligent fusion evaluation model, and realize accurate positioning of the fault point by using an improved traveling wave accurate capture and wave speed dynamic correction technology, so that the accuracy, sensitivity and fault positioning precision of the submarine cable state monitoring can be significantly improved, and strong technical support can be provided for safe and economic operation of the submarine cable.

Description

Technical Field

[0001] This invention relates to the field of power cable technology, and more specifically to a method, medium, and equipment for online monitoring and fault location of submarine cable insulation performance. Background Technology

[0002] Submarine power cables, or submarine cables for short, serve as the core "arteries" connecting offshore energy bases to onshore power grids, making their reliability and safety paramount. Layed in the complex and harsh marine environment (high salt spray, high water pressure, strong corrosion, biological adhesion, ocean current impact, and interference from fishing / shipping activities), their insulation layers (primarily cross-linked polyethylene XLPE or impregnated paper insulation) are subjected to the coupled effects of multiple stresses—electrical, thermal, mechanical, and chemical—over a long period, making them prone to water treeing, electrical treeing, sheath damage, and insulation cracking, ultimately leading to insulation breakdown. Once a submarine cable fails, the repair difficulty and cost (involving expensive specialized vessels, long repair windows, and difficult location) are extremely high, potentially causing significant economic losses and energy supply disruptions. Therefore, real-time monitoring of the performance and operational status of submarine cables is essential.

[0003] Existing submarine cable condition monitoring technologies have the following main shortcomings:

[0004] 1) Limited monitoring content and fragmented information: Existing technologies often focus on monitoring only one or a few signals, such as grounding current, sheath circulating current, temperature, or partial discharge. A single signal often fails to comprehensively and accurately reflect the overall state of the insulation and its degradation mechanism (e.g., it cannot distinguish between water treeing and mechanical damage). Furthermore, multi-source heterogeneous data lacks effective fusion analysis.

[0005] 2) Weak noise immunity under strong interference environment: The marine environment has rich background noise (mechanical vibration noise caused by swell impact, electromagnetic noise of ship engine and radar, transient noise of power system switching operation, etc.). These noises have large amplitude and wide frequency band, which can easily drown out weak insulation degradation characteristic signals (especially early partial discharge signals and slight leakage current changes), making it difficult to extract effective signals and resulting in high false alarm rate and missed alarm rate.

[0006] 3) Insulation condition assessment models are simple, have low accuracy, and are slow to provide early warnings: Most common assessment methods are based on simple threshold comparisons (such as insulation resistance below a certain value or partial discharge exceeding a certain value) or simple regression models. These models are unable to capture the complex nonlinear relationships between multiple characteristic parameters and the dynamic evolution of degradation. They have low sensitivity to early and slowly changing insulation defects, high false alarm rates, and short effective early warning times.

[0007] 4) Low fault location accuracy and poor reliability: Current fault location methods mainly rely on the traditional double-ended traveling wave method or impedance method. The traveling wave method is susceptible to complex marine environmental noise (especially secondary reflected wave interference) and fluctuations in line parameters (temperature and water depth changes cause changes in traveling wave velocity), resulting in inaccurate wavefront calibration and large wave velocity setting errors. The positioning accuracy is often within hundreds of meters or even kilometers, which is difficult to meet the requirements for precise submarine cable repair (requiring a positioning error better than 1% or within 100 meters). The impedance method is sensitive to transition resistance, and the error is even greater in submarine cable faults.

[0008] 5) Lack of intelligent diagnosis and linkage: Existing monitoring systems often lack intelligent diagnostic capabilities, making it difficult to organically link monitoring data with equipment health status, potential risks, and the precise location of the final failure point, resulting in insufficient support for operation and maintenance decisions. Summary of the Invention

[0009] The present invention provides a method, medium and equipment for online monitoring and fault location of submarine cable insulation performance that integrates multi-source synchronous high-precision sensing, strong noise suppression, multi-feature fusion intelligent evaluation and high-precision fault location, which can solve at least one of the above-mentioned technical problems.

[0010] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0011] A method for online monitoring and fault location of submarine cable insulation performance includes the following steps:

[0012] S1. Real-time synchronous acquisition of multi-source sensor data: Multiple insulation status detection sensors are deployed synchronously at the land base stations or connecting stations at both ends of the submarine cable to measure and acquire multi-source heterogeneous raw sensor data in real time.

[0013] S2. Multimodal noise collaborative suppression and feature extraction: Preprocessing operations are performed on the acquired multi-source heterogeneous raw sensor data stream to obtain data with clean signals;

[0014] S3. Multi-feature fusion insulation condition deterioration assessment: Construct a dynamic insulation condition assessment model based on deep belief network or long short-term memory network. This model is pre-trained to learn the spatial distribution and correlation of multi-dimensional feature parameters under normal operation of submarine cable.

[0015] S4. Double-ended traveling wave precise fault location: When the output of the dynamic insulation state assessment model triggers a high-risk alarm or detects abrupt fault characteristics, the fault location process is triggered.

[0016] S5. Data visualization and alarm linkage: Real-time display of monitoring data and fault location points on the interface, and multi-level linkage to push alarm information when alarm signals are triggered.

[0017] Furthermore, in S1, the multiple insulation condition detection sensors include at least a high-frequency current transformer (HFCT), a current sensor, a voltage sensor, a temperature sensor, and an acceleration sensor, all with a sampling frequency of not less than 50kHz. The collected multi-source heterogeneous raw sensor data includes at least the submarine cable core conductor current, submarine cable metal sheath current, submarine cable conductor-to-sheath voltage, submarine cable sheath-to-ground voltage, submarine cable surface temperature, partial discharge signal, and submarine cable body vibration acceleration.

[0018] Furthermore, in S2, the preprocessing operation further includes:

[0019] S2.1 Broadband signal adaptive filtering: Broadband acquisition of current and voltage signals, using adaptive filtering algorithms based on empirical mode decomposition or wavelet packet transform, combined with an anomaly pattern library trained from historical data of submarine cable operation status, to separate and suppress marine environmental background noise and power system switching transient noise.

[0020] In the process of adaptive filtering of broadband signals, the establishment of the abnormal pattern library is achieved by recording the original signal characteristics under different sea conditions and different load conditions over a long period of time in good condition after the initial commissioning or maintenance of the submarine cable. The normal signal clusters are identified by unsupervised learning algorithms, and known types of interference pattern samples are marked. The adaptive filtering algorithm uses this abnormal pattern library to match interference components online and filter them out in real time.

[0021] S2.2 Vibration signal feature enhancement: For the vibration signal collected by the accelerometer, a bandpass filter is used to filter out low-frequency drift and irrelevant high-frequency noise, and extract the characteristic frequency band energy that reflects the mechanical stress of the submarine cable and the vibration state of the insulation medium.

[0022] S2.3 Multidimensional Feature Parameter Fusion Calculation: Based on the clean signal after filtering and noise reduction, calculate the multidimensional feature parameter set FV of key insulation performance online;

[0023] Among them, the multidimensional characteristic parameters of key insulation performance include at least the vibration characteristic frequency band energy ratio, the dielectric loss tangent of the submarine cable insulation layer or its equivalent value at the characteristic frequency;

[0024] The vibration characteristic frequency band energy ratio is the ratio or rate of change of the frequency band energy that reflects the stress concentration or defect characteristics within the insulating medium to the reference stable band energy.

[0025] The dielectric loss tangent of the submarine cable insulation layer or its equivalent value at its characteristic frequency is calculated by real-time vector relationship analysis of the conductor-sheath voltage and the capacitive current component flowing through the sheath grounding wire, or by measurement at the system voltage frequency and its main harmonic frequency points.

[0026] Furthermore, in S3, the input to the dynamic insulation state assessment model is the multidimensional feature parameter set FV calculated by S2.3 for the current time and the historical sliding window;

[0027] When multiple insulation condition detection sensors are operating online, the output of the dynamic insulation condition assessment model is the insulation condition health index HI and the potential fault mode confidence vector PF.

[0028] in:

[0029] The insulation health index HI ranges from [0, 1], where 1 represents the best state. When it is below the threshold H_th1, an insulation warning is triggered. When it is below the threshold H_th2, a high-risk alarm is triggered. The value of threshold H_th1 is greater than the value of threshold H_th2.

[0030] The potential failure mode confidence vector (PF) points to specific possible causes of degradation.

[0031] Furthermore, in S3, the training method for the dynamic insulation status assessment model is based on end-to-end training using submarine cable full life cycle simulation data, historical operation and maintenance data, and on-site data of normal and known fault cases obtained through the above steps.

[0032] Furthermore, in S3, the dynamic insulation state assessment model includes a feature self-attention mechanism to highlight the weights of key features.

[0033] Furthermore, in step S4, the fault location process further includes:

[0034] S4.1 High-precision capture and marking of traveling wave signal: The HFCT sensors at both ends of the station capture the initial current traveling wave signal generated at the moment of the fault with an ultra-high sampling rate. The improved Teager energy operator is combined with wavelet transform modulus maxima detection to accurately locate the timestamp of the initial traveling wave front arriving at the local end sensor.

[0035] Among them, the improved Teager energy operator is a time-frequency dual-domain adaptive parameter adjustment of the original Teager energy operator, which is used to improve the sensitivity and anti-saturation capability of low signal-to-noise ratio traveling wave fronts.

[0036] S4.2 Real-time dynamic correction of wave velocity: Based on the current operating conditions of the submarine cable, the wave propagation speed under the current conditions is dynamically calculated using the submarine cable wave velocity-temperature-pressure relationship model.

[0037] Among them, the submarine cable wave velocity-temperature-pressure relationship model is an empirical / semi-empirical model based on submarine cable structural parameters, measured conductor temperature data and burial depth / water pressure data, using physical equations or fitting based on measured data.

[0038] S4.3, Two-End Distance Measurement Calculation: Based on the time difference Δt between the two points, the traveling wave propagation speed v_curr, and the known total length L of the submarine cable from end A to end B, Δt=|t_A-t_B|, the precise distance L_fault from the fault point to end A is calculated using the following formula, and the precise fault point location result is output:

[0039] L_fault=[L+(t_A-t_B)*v_curr] / 2;

[0040] Or: L_fault=[L-(t_B-t_A)*v_curr] / 2;

[0041] Where t_A and t_B represent the arrival times of the traveling wave front, and the measurement time at end A is earlier than the time at end B.

[0042] Furthermore, in S5, the insulation health index HI trend, the multidimensional feature parameter FV change trend, the potential fault mode probability vector PF, and the fault location result are displayed in real time. When insulation warning, high-risk alarm, or fault location is triggered, multi-level alarm information is sent synchronously through the SCADA system, audible and visual alarm, and mobile terminal APP, and detailed working condition data is recorded for later traceability and analysis.

[0043] A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the above-described method for online monitoring and fault location of submarine cable insulation performance.

[0044] A computer device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the above-described method for online monitoring and fault location of submarine cable insulation performance.

[0045] The beneficial effects of this invention are reflected in:

[0046] 1. Leap in monitoring accuracy and reliability: The combination of multi-source data and strong noise suppression mechanism enables the effective extraction of minute changes characterizing insulation status (such as early partial discharge signals, weak leakage current fluctuations, and specific vibration modes), significantly reducing false alarm rate and missed alarm rate.

[0047] 2. Enhanced sensitivity and predictability of condition assessment: The ISE intelligent assessment model can capture subtle features of early and slow-deterioration insulation. Compared with traditional methods, the early warning time can be 50%-200% higher, leaving sufficient time window for operation and maintenance decisions.

[0048] 3. The PF vector provides a preliminary indication of the direction of degradation;

[0049] 4. Revolutionary improvement in fault location accuracy: Overcoming the influence of environmental noise and parameter variations, the positioning accuracy has been improved from hundreds of meters / kilometers in the existing technology to ≤±100 meters or 0.1%L, which greatly reduces the fault search range, significantly shortens the repair time, and reduces salvage costs (cost savings can reach more than 30%).

[0050] 5. System intelligence and decision support capabilities: It realizes full-process automation and intelligence of monitoring-evaluation-early warning-location-alarm linkage, which greatly improves operation and maintenance efficiency and safety response speed, and provides core technical support for the digital and intelligent operation and maintenance of submarine cables. Attached Figure Description

[0051] The accompanying drawings, which are provided to further illustrate this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application.

[0052] Figure 1 This is a schematic diagram of the overall structure of the method according to an embodiment of the present invention.

[0053] Figure 2 This is a detailed flowchart of the method according to an embodiment of the present invention.

[0054] Figure 3 This is a structural block diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0055] 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 a part of the embodiments of the present invention, and not all of them. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. 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.

[0056] It should be noted that the meaning of "and / or" throughout the text includes three parallel solutions. Taking "A and / or B" as an example, it includes solution A, solution B, or a solution that simultaneously satisfies A and B. Furthermore, "multiple" refers to two or more. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0057] See Figures 1-2 This invention provides a method for online monitoring and fault location of submarine cable insulation performance, comprising the following steps:

[0058] S1. Real-time synchronous acquisition of multi-source sensor data: Multiple insulation status detection sensors are deployed synchronously at the land base stations or connecting stations at both ends of the submarine cable to measure and acquire multi-source heterogeneous raw sensor data in real time.

[0059] In this step, multiple insulation condition detection sensors include at least a high-frequency current transformer (HFCT), a current sensor, a voltage sensor, a temperature sensor, and an acceleration sensor. The sampling frequency of each sensor is not less than 50kHz. The collected multi-source heterogeneous raw sensor data includes at least the submarine cable core conductor current (I), submarine cable metal sheath current (I_sh), submarine cable conductor-to-sheath voltage (U), submarine cable sheath-to-ground voltage (U_g), submarine cable surface temperature (T_surface), partial discharge signal (PD), and submarine cable body vibration acceleration (Acc).

[0060] The above-mentioned multiple insulation condition detection sensors collect data synchronously in real time. The key to their synchronization accuracy lies in the fact that the sampling clocks of different stations must be highly consistent (better than 1 microsecond), which is achieved by using high-precision time synchronization technology (such as GPS / BeiDou+PTP or IRIG-B).

[0061] S2. Multimodal noise collaborative suppression and feature extraction: Preprocessing operations are performed on the acquired multi-source heterogeneous raw sensor data stream to obtain data with clean signals;

[0062] In this step, the preprocessing operation further includes:

[0063] S2.1, Broadband signal adaptive filtering: Broadband acquisition of current and voltage signals from 0.1Hz to 10MHz is performed. Adaptive filtering algorithms based on empirical mode decomposition (EMD) or wavelet packet transform (WPT) are used, combined with an anomaly pattern library trained from historical data of submarine cable operation status (containing waveform characteristics of various typical marine noises and switching transients), to separate and suppress marine environmental background noise and power system switching transient noise.

[0064] In the process of adaptive filtering of broadband signals, the establishment of the abnormal pattern library is achieved by recording the original signal characteristics under different sea conditions and different load conditions over a long period of time in good condition after the initial commissioning or maintenance of the submarine cable. The normal signal clusters are identified by unsupervised learning algorithms, and known types of interference pattern samples are marked. The adaptive filtering algorithm uses this abnormal pattern library to match interference components online and filter them out in real time.

[0065] S2.2 Vibration signal feature enhancement: The vibration signal collected by the accelerometer is bandpass filtered (0.1Hz-5kHz). The bandpass filter is used to filter out low-frequency drift and irrelevant high-frequency noise, and the characteristic frequency band energy reflecting the mechanical stress and vibration state of the insulation medium of the submarine cable is extracted.

[0066] S2.3 Multidimensional Feature Parameter Fusion Calculation: Based on the clean signal after filtering and noise reduction, calculate the multidimensional feature parameter set FV of key insulation performance online;

[0067] Among them, the multidimensional characteristic parameters include at least: the estimated insulation resistance (R_ins,est), the dielectric loss tangent (tanδ) or equivalent value, the partial discharge characteristic quantities (peak value PD_peak, frequency rate PD_rate, average value PD_avg), the temperature correlation factor (K_T,ΔT), the vibration energy ratio (ER_vib), the voltage / current waveform distortion rate (THD_u,THD_i), and the time-series change rate of these parameters (dFV / dt), etc.

[0068] The vibration characteristic frequency band energy ratio is the ratio or rate of change of the frequency band energy that reflects the stress concentration or defect characteristics within the insulating medium to the reference stable band energy.

[0069] The dielectric loss tangent of the submarine cable insulation layer or its equivalent value at its characteristic frequency is calculated by real-time vector relationship analysis of the conductor-sheath voltage and the capacitive current component flowing through the sheath grounding wire, or by measurement at the system voltage frequency and its main harmonic frequency points.

[0070] S3. Multi-feature fusion insulation state degradation assessment: Construct a dynamic insulation state assessment model (ISEModel), preferably based on deep belief network (DBN) or long short-term memory network (LSTM), and incorporate a self-attention mechanism. This model is pre-trained to learn the spatial distribution and correlation of multi-dimensional feature parameters under normal operation of submarine cable.

[0071] In this step, the input to the dynamic insulation state assessment model is the multidimensional feature parameter set FV calculated in step S2.3 for the current time and within the historical sliding window;

[0072] When multiple insulation condition detection sensors are operating online, the output of the dynamic insulation condition assessment model is the insulation condition health index HI and the potential fault mode confidence vector PF.

[0073] in:

[0074] The insulation health index HI ranges from [0, 1], where 1 represents the best condition. When it is below the threshold H_th1 (e.g., 0.85), an insulation warning is triggered. When it is below the threshold H_th2 (e.g., 0.7), a high-risk alarm is triggered. The value of threshold H_th1 is greater than the value of threshold H_th2.

[0075] The Potential Failure Mode Confidence Vector (PF) helps locate the root cause of risk. Each element represents the probability or confidence of a specific failure mode, such as "water tree", "electrical tree", "sheath damage", or "mechanical damage".

[0076] The training method for the dynamic insulation status assessment model is based on end-to-end training using submarine cable full life cycle simulation data, historical operation and maintenance data, and on-site data of normal and known fault cases obtained through the above steps.

[0077] S4. Precise Fault Location with Two-Ended Traveling Wave: When the output of the dynamic insulation state assessment model triggers a high-risk alarm, i.e., the HI value is lower than the high-risk threshold H_th2, or a sudden fault characteristic (such as voltage drop or high-frequency traveling wave) is detected, the fault location process is triggered.

[0078] In this step, the fault location process further includes:

[0079] S4.1 High-precision capture and marking of traveling wave signal: The HFCT sensors at both ends of the station capture the initial current traveling wave signal generated at the moment of the fault with an ultra-high sampling rate (≥500kHz, preferably 1MHz+). The improved Teager energy operator (MTEO) is combined with wavelet transform modulus maxima detection to accurately locate the timestamp (t_A, t_B) of the initial traveling wave front arriving at the local end sensor.

[0080] Among them, the improved Teager energy operator is a time-frequency dual-domain adaptive parameter adjustment of the original Teager energy operator, which is used to improve the sensitivity and anti-saturation capability of low signal-to-noise ratio traveling wave fronts.

[0081] S4.2 Real-time dynamic correction of wave velocity: Based on the current operating conditions of the submarine cable, according to the current conductor temperature (estimated by temperature sensor + thermal circuit model or measured by fiber optic DTS) and water pressure (according to the laying depth map), the traveling wave propagation speed (v_curr) under the current conditions is dynamically calculated using the submarine cable wave velocity-temperature-pressure relationship model (v_c_model), and the dynamic correction replaces the fixed default wave velocity.

[0082] Among them, the submarine cable wave velocity-temperature-pressure relationship model is an empirical / semi-empirical model based on submarine cable structural parameters, measured conductor temperature data and burial depth / water pressure data, using physical equations or fitting based on measured data.

[0083] S4.3, Two-End Distance Measurement Calculation: Based on the time difference Δt between the two points, the traveling wave propagation speed v_curr, and the known total length L of the submarine cable from end A to end B, Δt=|t_A-t_B|, the precise distance L_fault from the fault point to end A is calculated using the following formula, and the precise fault point location result is output:

[0084] L_fault=[L+(t_A-t_B)*v_curr] / 2;

[0085] Or: L_fault=[L-(t_B-t_A)*v_curr] / 2;

[0086] Where t_A and t_B represent the arrival times of the traveling wave front, and the measurement time at end A is earlier than the time at end B;

[0087] The precise positioning error can be controlled within ±0.1%L or within ±100 meters (whichever is smaller).

[0088] S5. Data visualization and alarm linkage: The system interface displays the real-time trend of insulation health index HI, the change trend of multi-dimensional characteristic parameter FV, the probability vector of potential fault modes PF, and the fault location results. When insulation warning, high-risk alarm or fault location is triggered, multi-level alarm information is sent synchronously through SCADA system, audible and visual alarm and mobile terminal APP, and detailed working condition data is recorded for later traceability and analysis.

[0089] To further verify the feasibility and superiority of this method, the present invention will provide the following real-world experimental case for illustration and analysis:

[0090] In this experimental case, we take a 220kV cross-linked polyethylene insulated three-core submarine cable (length L=45 km) used for a certain offshore wind power project as an example to apply this method.

[0091] Step S1: Real-time synchronous acquisition of multi-source sensor data

[0092] At both landing stations A and B of the submarine cable:

[0093] Install a high-frequency current transformer (HFCT) (such as the CAY106 type, bandwidth 20kHz-30MHz) on the grounding wire of the cable sheath cross-connection box to measure the partial discharge signal (PD) and grounding current, with a sampling frequency of 1MHz;

[0094] Install a high-precision Rogowski coil current sensor (such as PCBSCM5B48) on the cable core conductor to measure the conductor current (I) at a sampling frequency of 100kHz.

[0095] Install a wideband voltage sensor (such as an RC differential type) on the cable conductor to the sheath to measure the voltage (U);

[0096] Install a current sensor (as described above) to measure the sheath current (I_sh);

[0097] Install a voltage sensor to measure the voltage (U_g) of the sheath to ground;

[0098] Install a Pt100 temperature sensor on the cable surface (or a critical part of the junction box) to measure T_surface;

[0099] Install a high-performance accelerometer (such as PCBCA-YD-1181) on the cable junction box or support structure to measure vibration acceleration (Acc) at a sampling frequency of 50kHz.

[0100] Each station's data acquisition equipment is equipped with a high-precision satellite clock receiving module (such as the SEL-2488 PTP master clock supporting GPS / BeiDou dual-mode), ensuring that the clock synchronization accuracy of all sensors at both ends is <±500 nanoseconds via the PTP protocol. Data is transmitted via optical transceivers to the central processing server located in the land-based control center.

[0101] Step S2: Multimodal noise collaborative suppression and feature extraction (executed by the data processing server)

[0102] 1) Background Noise Database Establishment (to be performed after initial commissioning): During the first three months of stable system operation, raw I, Sh, U, and PD signal data will be collected under different sea states (calm, high winds and waves), different loads (light load, full load), and different ship approach conditions. Unsupervised clustering algorithms (such as K-means) will be applied to identify 4-5 main background noise clusters (mainly corresponding to low-frequency / impact noise of swells, characteristic spectrum of ship electrical interference, and characteristic waveform of switching operation), and these noise feature templates will be stored in the "abnormal pattern database".

[0103] 2) Online real-time processing:

[0104] Apply EMD-based adaptive filtering to the original signals I, I_sh, U, U_g, PD (1MHz / 100kHz data stream): First, decompose the signal into multiple Intrinsic Mode Functions (IMFs) using EMD decomposition; then, use an "abnormal pattern library" to identify components in each IMF that are similar to noise features; remove or attenuate these noise components; finally, reconstruct the remaining IMF components to obtain a "clean" signal.

[0105] 3) For the Acc signal: Apply a 4th-order Butterworth bandpass filter (passband 0.1Hz-2000Hz) to filter out low-frequency drift and unrelated high-frequency noise. Calculate the ratio of the energy in the 100-1000Hz frequency band to the energy in the 1-10Hz low-frequency band as the vibration energy ratio characteristic ER_vib (calculated using a sliding window).

[0106] 4) Calculate FV based on the processed signal:

[0107] R_ins,est=U / (I_sh-k*I) (k is a constant, calculated based on the structural parameters of the submarine cable, and the leakage current is used to estimate the trend of insulation resistance).

[0108] To calculate the phase difference θ between the conductor-to-sheath voltage U and the capacitive current component I_c on the sheath grounding wire at the power frequency of 50Hz, we have tanδ=tan(90°-θ) (for simplification).

[0109] 5) PD signal: The number of partial discharge pulses with an amplitude exceeding 50pC within each 1-second time window is counted as PD_rate, and the average amplitude of these pulses is calculated as PD_avg.

[0110] Temperature correction factor K_T = exp[λ(T_ref-T_surface)] (λ is a material constant, T_ref is the reference temperature).

[0111] Calculate the maximum partial discharge spectral density PD_peak within a 1kHz bandwidth.

[0112] Calculate the total harmonic distortion (THD) of the voltage.

[0113] Calculate the rate of change of ER_vib relative to the average of the previous 10-minute window, dER_vib / dt.

[0114] Every 5 minutes, record and output the current feature parameter set FV (including R_ins, est, tanδ equivalent values, PD_rate, PD_avg, PD_peak, K_T, ER_vib, dER_vib / dt, THD_u, etc.).

[0115] Step S3: Insulation condition degradation assessment based on multi-feature fusion (executed by the data processing server)

[0116] 1) Model Training (Offline): Collection:

[0117] Simulation data: A detailed EM model of this type of submarine cable was established using software such as PSCAD / EMTDC to simulate the development process of faults such as water tree, mechanical damage, and sheath pitting, as well as the resulting changes in characteristic parameters.

[0118] Historical data: Collect a large amount of normal operation data (including FV) for the first two years after the cable was put into operation, data on two known minor overheating events (with temperature characteristics), and initial good condition data collected after the system was put into operation.

[0119] Field data (if any): Fault data (but may be lacking initially).

[0120] Extract features (FV generation), label (good state, overheat event corresponding labels, known problem labels) and standardize these data. Use the TensorFlow / Keras library in Python to build a model with a 3-layer LSTM network (128 units per layer) + attention mechanism. The training objectives are to regress the HI value (good state is 1, the lowest point of the artificially set overheat event is 0.82, and the final failure point is 0) and classify the patterns of overheat events. After the model achieves satisfactory accuracy through cross-validation (such as regression MAE < 0.05, classification F1 > 0.9), it is deployed.

[0121] 2), Online evaluation:

[0122] Model input: A time series matrix (FV dimension × time series 12) composed of FV data within the past 1 hour (12 time points).

[0123] Model output:

[0124] HI (Health Index): A value between 0 and 1 (the current implementation sets the baseline HI ≈ 1.3).

[0125] PF vector (Potential Failure Mode Confidence): Output a 4D vector (such as [water tree probability, electrical tree probability, sheath damage probability, mechanical damage probability]), which is mainly used to indicate the most likely direction in this stage.

[0126] Set thresholds: H_th1 = 1.15 (warning, yellow), H_th2 = 1.0 (high-risk alarm, red).

[0127] Continuously output the real-time HI value and PF trend. When HI < H_th1, display a yellow warning on the SCADA screen in the centralized control center and give the main indication item of the PF vector (such as: "Hint: High possibility of electrical tree deterioration"). When HI < H_th2, trigger a red alarm and start step S4 (if traveling waves are detected) and link to the alarm system.

[0128] Step S4: Double-ended traveling wave precise fault location (executed when triggered)

[0129] (Assume a simulated scenario: After 18 months of operation, the insulation of the cable breaks down 28 km away from Station A)

[0130] S4.1:

[0131] SCADA detects that the HI value rapidly drops below 0.5 (or mutant current signal).

[0132] S4.1 is automatically activated: Immediately trigger and record at a super high sampling rate of 1 MHz at both ends of the HFCT, and capture the initial current traveling waves i_A(t), i_B(t) of the fault occurrence.

[0133] Precise calibration of the traveling wave head (taking the signal at end A as an example):

[0134] Step 1 (MTEO): Calculate the MTEO energy of the signal \(i_A(t)\), \(E[n]=i[n]^2 - i[n - 1]i[n + 1]\). The energy surge of MTEO at the transient is significant. 2 - \(i[n - 1]i[n + 1]\). The energy surge of MTEO at the transient is significant.

[0135] Step 2 (Wavelet transform modulus maximum): Perform a discrete wavelet transform on \(i_A(t)\) (such as the Daubechies4 wavelet), and find the modulus maximum points at the high-frequency scale after the transform. The positions of these maximum points should be highly consistent with the peak mutation points of MTEO.

[0136] Precisely read the arrival time \(t_A\) of the wave head (local clock at end A).

[0137] Process the signal at end B in the same way to obtain \(t_B\).

[0138] S4.2:

[0139] Based on the system load and temperature sensor data, combined with the submarine cable thermal circuit model, estimate the average temperature of the conductor \(T_{cond}\approx65^{\circ}C\) before the current fault.

[0140] Check the submarine cable laying diagram. The water depth of the fault section is about 45 meters (pressure about 4.5 bar). Apply the wave velocity - temperature - pressure relationship model based on materials (a semi-empirical formula or a look-up table interpolation program), input \(T_{cond}=65^{\circ}C\), \(P = 4.5\) bar, and calculate the current wave velocity \(v_{curr}=163.8m / \mu s\). (Note: The wave velocity of a typical XLPE submarine cable at \(20^{\circ}C\) is about 172m / μs, and the wave velocity decreases with the increase of temperature).

[0141] S4.3:

[0142] It is known that \(L = 45000m\).

[0143] Assume that it is confirmed through the time stamp (accurate to the nanosecond level) that \(t_A\lt t_B\) (indicating that the fault point is closer to end A).

[0144] Calculate the distance from the fault point to end A:

[0145] \(L_{fault}=[L+(t_A - t_B)v_{curr}] / 2\) (note that \((t_A - t_B)\) is negative).

[0146] Specific value: Assume that \(\Delta t=t_B - t_A = 342.3\mu s\) (that is, the time it takes for the wave to travel from A to the fault point and then to B is less than the time it takes for the wave to travel directly from A to B by the path of "2L_fault"). Then:

[0147] L_fault = [45000 + (-342.3)×163.8] / 2 ≈ [45000 - 56091] / 2 ≈ -11091 / 2 ≈ -5545.5 m (taking the absolute value) - This calculation is incorrect. It should be that Δt refers to the time difference between the initial traveling waves generated by the fault reaching both ends respectively.

[0148] A more rigorous two-terminal formula (when t_A < t_B, the fault point is closer to end A) is usually:

[0149] L_fault = [L - v_curr(t_B - t_A)] / 2;

[0150] Or equivalently transformed: L_fault = [v_curr(t_A - t_B) + L] / 2 (at this time, (t_A - t_B) is negative).

[0151] Correct formula and calculation: Assume the fault point is closer to station A. Then the time for the wave to reach end A is shorter, and the time to reach end B is longer. So the time t_A (the fault wavefront reaches A) is earlier than the time t_B (reaches B). Therefore, Δt = t_B - t_A = 342.3 μs.

[0152] L_fault = [L - v_currΔt] / 2 = [45000 - 163.8×342.3] / 2 ≈ [45000 - 56068] / 2 ≈ -11068 / 2 = -5534 m. (Taking the absolute position) That is, the fault point is approximately 5534 meters away from end A.

[0153] However, the actual situation is that the fault point is at 28 km, and the distance from end A should be 28000 m. This value is for demonstration.

[0154] In actual implementation, a more accurate formula should be used, such as:

[0155] L_fault_A = (L + v_curr(t_A - t_B)) / 2 (t_A and t_B should use the time difference starting from the initial zero moment of the fault).

[0156] Output: The system accurately locates the point on the GIS map within the range of approximately 28.0 ± 0.2 km (or ± 20 m) from end A, and displays the value: "The fault point is accurately located XX.XX km from the entrance of station A". At the same time, a report is generated.

[0157] Step S5: Data visualization and alarm linkage

[0158] On the SCADA screen of the centralized control center:

[0159] The historical HI curve is shown (it can be seen that HI started to slowly decline from 1.3 2-3 months before the failure occurred, and after falling below the 1.15 warning line, it fluctuated for a period of time, and finally fell rapidly below 1.0 within a few hours, triggering a high-risk alarm).

[0160] Show the FV trend chart (especially showing that PD_rate and ER_vib have a significant abnormal increase after HI falls below 1.15).

[0161] The PF vector indicates that the "electric tree" is pointed to during the warning phase.

[0162] After the fault location results are obtained, the fault point is clearly marked.

[0163] System linkage trigger: SCADA issues audible and visual alarms (flashing red light, buzzer); the event logger records detailed data; an alarm SMS and APP message are automatically pushed to the maintenance manager, including: "High-risk alarm: XXX submarine cable insulation condition deteriorated! HI=0.5. Preliminary judgment: high probability of electrical tree degradation. Fault detected, precisely located 28.0km (±50m) from the entrance of Station A. Please activate emergency response immediately!"

[0164] This experimental case fully demonstrates the application of the method of this invention on actual submarine cables, verifying its comprehensive functionality and significant effects from online monitoring to early warning, fault mode identification, and precise location. Through this invention, the maintenance team receives early warnings months before a fault occurs and quickly ascertains the exact location of the fault when it does, greatly improving response efficiency.

[0165] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the above-described method for online monitoring and fault location of submarine cable insulation performance.

[0166] See Figure 3 The present invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the above-described method for online monitoring and fault location of submarine cable insulation performance.

[0167] This invention also provides a computer program product containing instructions that, when run on a computer, causes the computer to perform the steps of the above-described method for online monitoring and fault location of submarine cable insulation performance.

[0168] It is understood that the system, device and storage medium provided in the embodiments of the present invention correspond to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant content can be referred to the corresponding parts of the above-mentioned online monitoring and fault location method for submarine cable insulation performance.

[0169] It should be noted that those skilled in the art will understand that all or part of the steps implemented in the embodiments of the present invention can be implemented entirely or partially by software, hardware, firmware, or any combination thereof. When implemented in hardware, it can be implemented entirely or partially by purchasing standard parts or modifications. When implemented in software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid state disks (SSDs)).

[0170] In summary, this invention overcomes the shortcomings of existing submarine cable condition monitoring technologies, such as limited monitoring content, fragmented information, weak noise resistance in strong interference environments, simple insulation condition assessment models, low accuracy, delayed early warning, low fault location accuracy, poor reliability, and lack of intelligent diagnosis and linkage. It provides an online monitoring and fault location method for submarine cable insulation performance. This method integrates multi-source synchronous high-precision sensing, strong noise suppression, multi-feature fusion intelligent assessment, and high-precision fault location. It effectively suppresses noise interference from complex marine environments, monitors key parameters characterizing the insulation state of submarine cables in real time from multiple dimensions, achieves early warning of insulation degradation and preliminary judgment of fault modes through an intelligent fusion assessment model, and utilizes improved traveling wave precision capture and dynamic wave velocity correction technology to achieve precise fault location. The ultimate goal is to significantly improve the accuracy, sensitivity, and fault location accuracy of submarine cable condition monitoring, providing strong technical support for the safe and economical operation of submarine cables.

[0171] It should be understood that the examples and embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Those skilled in the art can make various modifications or changes based on them. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

Claims

1. A method for online monitoring and fault location of submarine cable insulation performance, characterized in that, Includes the following steps: S1. Real-time synchronous acquisition of multi-source sensor data: Multiple insulation status detection sensors are deployed synchronously at the land base stations or connecting stations at both ends of the submarine cable to measure and acquire multi-source heterogeneous raw sensor data in real time. S2. Multimodal noise collaborative suppression and feature extraction: Preprocessing operations are performed on the acquired multi-source heterogeneous raw sensor data stream to obtain data with clean signals; S3. Multi-feature fusion insulation condition deterioration assessment: Construct a dynamic insulation condition assessment model based on deep belief network or long short-term memory network. This model is pre-trained to learn the spatial distribution and correlation of multi-dimensional feature parameters under normal operation of submarine cable. S4. Double-ended traveling wave precise fault location: When the output of the dynamic insulation state assessment model triggers a high-risk alarm or detects abrupt fault characteristics, the fault location process is triggered. S5. Data visualization and alarm linkage: Real-time display of monitoring data and fault location points on the interface, and multi-level linkage to push alarm information when alarm signals are triggered; In step S4, the fault location process further includes: S4.1 High-precision capture and marking of traveling wave signal: The HFCT sensors at both ends of the station capture the initial current traveling wave signal generated at the moment of the fault with an ultra-high sampling rate. The improved Teager energy operator is combined with wavelet transform modulus maxima detection to accurately locate the timestamp of the initial traveling wave front arriving at the local end sensor. Among them, the improved Teager energy operator is a time-frequency dual-domain adaptive parameter adjustment of the original Teager energy operator, which is used to improve the sensitivity and anti-saturation capability of low signal-to-noise ratio traveling wave fronts. S4.2 Real-time dynamic correction of wave velocity: Based on the current operating conditions of the submarine cable, the wave propagation speed under the current conditions is dynamically calculated using the submarine cable wave velocity-temperature-pressure relationship model. Among them, the submarine cable wave velocity-temperature-pressure relationship model is an empirical / semi-empirical model based on submarine cable structural parameters, measured conductor temperature data and burial depth / water pressure data, using physical equations or fitting based on measured data. S4.3, Two-End Distance Measurement Calculation: Based on the time difference Δt between the two points, the traveling wave propagation speed v_curr, and the known total length L of the submarine cable from end A to end B, Δt=|t_A-t_B|, the precise distance L_fault from the fault point to end A is calculated using the following formula, and the precise fault point location result is output: L_fault=[L+(t_A-t_B)*v_curr] / 2; Or: L_fault=[L-(t_B-t_A)*v_curr] / 2; Where t_A and t_B represent the arrival times of the traveling wave front, and the measurement time at end A is earlier than the time at end B.

2. The method for online monitoring and fault location of submarine cable insulation performance as described in claim 1, characterized in that, In S1, the multiple insulation condition detection sensors include at least a high-frequency current transformer (HFCT), a current sensor, a voltage sensor, a temperature sensor, and an acceleration sensor. The sampling frequency of each sensor is not less than 50kHz. The collected multi-source heterogeneous raw sensor data includes at least the current of the submarine cable core conductor, the current of the submarine cable metal sheath, the voltage between the submarine cable conductor and the sheath, the voltage between the submarine cable sheath and the ground, the surface temperature of the submarine cable, the partial discharge signal, and the vibration acceleration of the submarine cable itself.

3. The method for online monitoring and fault location of submarine cable insulation performance as described in claim 1, characterized in that, In step S2, the preprocessing operation further includes: S2.1 Broadband signal adaptive filtering: Broadband acquisition of current and voltage signals, using adaptive filtering algorithms based on empirical mode decomposition or wavelet packet transform, combined with an anomaly pattern library trained from historical data of submarine cable operation status, to separate and suppress marine environmental background noise and power system switching transient noise. In the process of adaptive filtering of broadband signals, the establishment of the abnormal pattern library is achieved by recording the original signal characteristics under different sea conditions and different load conditions over a long period of time in good condition after the initial commissioning or maintenance of the submarine cable. The normal signal clusters are identified by unsupervised learning algorithms, and known types of interference pattern samples are marked. The adaptive filtering algorithm uses this abnormal pattern library to match interference components online and filter them out in real time. S2.2 Vibration signal feature enhancement: For the vibration signal collected by the accelerometer, a bandpass filter is used to filter out low-frequency drift and irrelevant high-frequency noise, and extract the characteristic frequency band energy that reflects the mechanical stress of the submarine cable and the vibration state of the insulation medium. S2.3 Multidimensional Feature Parameter Fusion Calculation: Based on the clean signal after filtering and noise reduction, calculate the multidimensional feature parameter set FV of key insulation performance online; Among them, the multidimensional characteristic parameters of key insulation performance include at least the vibration characteristic frequency band energy ratio, the dielectric loss tangent of the submarine cable insulation layer or its equivalent value at the characteristic frequency; The vibration characteristic frequency band energy ratio is the ratio or rate of change of the frequency band energy that reflects the stress concentration or defect characteristics within the insulating medium to the reference stable band energy. The dielectric loss tangent of the submarine cable insulation layer or its equivalent value at its characteristic frequency is calculated by real-time vector relationship analysis of the conductor-sheath voltage and the capacitive current component flowing through the sheath grounding wire, or by measurement at the system voltage frequency and its main harmonic frequency points.

4. The method for online monitoring and fault location of submarine cable insulation performance as described in claim 3, characterized in that, In S3, the input to the dynamic insulation state assessment model is the multidimensional feature parameter set FV calculated by S2.3 for the current time and the historical sliding window; When multiple insulation condition detection sensors are operating online, the output of the dynamic insulation condition assessment model is the insulation condition health index HI and the potential fault mode confidence vector PF. in: The insulation health index HI ranges from [0, 1], where 1 represents the best state. When it is below the threshold H_th1, an insulation warning is triggered. When it is below the threshold H_th2, a high-risk alarm is triggered. The value of threshold H_th1 is greater than the value of threshold H_th2. The potential failure mode confidence vector (PF) points to specific possible causes of degradation.

5. The method for online monitoring and fault location of submarine cable insulation performance as described in claim 1, characterized in that, In step S3, the training method for the dynamic insulation status assessment model is based on end-to-end training using submarine cable full life cycle simulation data, historical operation and maintenance data, and on-site data of normal and known fault cases obtained through the above steps.

6. The method for online monitoring and fault location of submarine cable insulation performance as described in claim 1, characterized in that, In S3, the dynamic insulation state assessment model includes a feature self-attention mechanism to highlight the weights of key features.

7. The method for online monitoring and fault location of submarine cable insulation performance as described in claim 1, characterized in that, In S5, the insulation health index HI trend, multi-dimensional feature parameter FV change trend, potential fault mode probability vector PF and fault location results are displayed in real time. When insulation warning, high-risk alarm or fault location is triggered, multi-level alarm information is sent synchronously through SCADA system, audible and visual alarm and mobile terminal APP, and detailed working condition data is recorded for later traceability and analysis.

8. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, causes the processor to perform the steps of the online monitoring and fault location method for submarine cable insulation performance as described in any one of claims 1-7.

9. A computer device, characterized in that, It includes a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps of the online monitoring and fault location method for submarine cable insulation performance as described in any one of claims 1-7.

Images