Aluminum alloy power cable aging prediction method and system for offshore wind power

Abstract

The application discloses an aluminum alloy power cable aging prediction method and system for offshore wind power, relates to the technical field of cable aging prediction, and comprises the following steps: acquiring a mechanical vibration sequence, an electric-thermal temperature and humidity sequence and a shielding layer dielectric loss current; a double-branch physical constraint time sequence network including a mechanical damage branch and an electrochemical corrosion branch is constructed, the mechanical vibration sequence and the electric-thermal temperature and humidity sequence are input respectively, the conductor surface oxide film rupture degree and the conductor corrosion degree are output, the reference dielectric loss current is calculated, and the residual error between the calculated dielectric loss current and the measured shielding layer dielectric loss current is further calculated; when the residual error exceeds a preset threshold value, the electrochemical corrosion branch is corrected; and based on the corrected double-branch physical constraint time sequence network, an aging prediction result is output. The application solves the technical problems of inconsistency of multi-source data, lack of physical coupling of the mechanical and corrosion branches and unreliable residual error correction in the aging prediction of the aluminum alloy power cable for offshore wind power, and thus the prediction accuracy is low.

Description

Technical Field

[0001] This invention relates to the field of cable aging prediction technology, specifically to a method and system for predicting the aging of aluminum alloy power cables for offshore wind power. Background Technology

[0002] In the offshore wind power environment, aluminum alloy power cables are subjected to complex conditions such as mechanical vibration, salt spray corrosion, alternating damp heat, and temperature cycling over long periods. Their aging and deterioration are caused by the combined effects of mechanical damage and electrochemical corrosion. Accurately predicting the aging state of cables is crucial for ensuring the safe operation of offshore wind farms, reducing operation and maintenance costs, and avoiding unplanned downtime. However, existing aging prediction results often deviate significantly from actual operating conditions, making it difficult to meet the accuracy requirements for long-term reliable operation of offshore wind power.

[0003] In existing aging prediction methods, data on mechanical vibration, electrothermal temperature and humidity, and shielding layer dielectric loss current are inconsistent in terms of time reference, frequency domain range, and physical dimensions. There is a lack of physical correlation between mechanical damage and electrochemical corrosion, making it difficult to reflect competing effects such as high-temperature evaporation inhibition and corrosion product densification passivation. Measurement noise in the measured signals is easily confused with actual aging deviations, and the correction process lacks physical boundary constraints and convergence verification, resulting in low accuracy, poor interpretability, and insufficient reliability of the prediction results. Therefore, improving the accuracy and reliability of aging prediction for offshore wind power aluminum alloy power cables has become an urgent technical problem to be solved. Summary of the Invention

[0004] This invention provides a method and system for predicting the aging of aluminum alloy power cables for offshore wind power. This invention solves the technical problems of low prediction accuracy caused by inconsistencies in multi-source data, lack of physical coupling between mechanical and corrosion branches, and unreliable residual correction in the aging prediction of aluminum alloy power cables for offshore wind power.

[0005] In view of the above problems, the present invention provides a method for predicting the aging of aluminum alloy power cables for offshore wind power, comprising: Obtain the mechanical vibration sequence, electrothermal temperature and humidity sequence, and shielding dielectric loss current of the target aluminum alloy power cable; A dual-branch physical constraint time series network is constructed, which includes a mechanical damage branch and an electrochemical corrosion branch. The mechanical vibration sequence is input into the mechanical damage branch, and the electrothermal temperature and humidity sequence is input into the electrochemical corrosion branch. The degree of oxide film rupture on the conductor surface and the degree of conductor corrosion are output respectively. The reference dielectric loss current of the aluminum alloy power cable is calculated based on the degree of oxide film rupture and conductor corrosion. The residual between the reference dielectric loss current and the measured dielectric loss current of the shielding layer is calculated. When the residual exceeds the preset threshold, the electrochemical corrosion branch is corrected. Based on the corrected dual-branch physical constraint time series network, the aging prediction results of aluminum alloy power cables are output.

[0006] This invention also provides an aging prediction system for aluminum alloy power cables for offshore wind power, comprising: The multi-source data acquisition module is used to acquire the mechanical vibration sequence, electrothermal temperature and humidity sequence, and shielding layer dielectric loss current of the target aluminum alloy power cable. The dual-branch network construction module is used to construct a dual-branch physical constraint temporal network. The dual-branch physical constraint temporal network includes a mechanical damage branch and an electrochemical corrosion branch. The mechanical vibration sequence is input into the mechanical damage branch and the electrothermal temperature and humidity sequence is input into the electrochemical corrosion branch, which outputs the degree of oxide film rupture on the conductor surface and the degree of conductor corrosion, respectively. The reference current correction module is used to calculate the reference dielectric loss current of aluminum alloy power cables based on the degree of oxide film rupture and conductor corrosion on the conductor surface, and to calculate the residual between the reference dielectric loss current and the measured dielectric loss current of the shielding layer. When the residual exceeds a preset threshold, the electrochemical corrosion branch is corrected. The aging prediction output module is used to output the aging prediction results of aluminum alloy power cables based on the corrected dual-branch physical constraint time series network.

[0007] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention constructs a dual-branch physically constrained time-series network by synchronously acquiring time-aligned, frequency-domain consistent, and physically meaningful mechanical vibration sequences, electrothermal humidity sequences, and shielding layer dielectric loss current. This network couples the mechanical damage branch and the electrochemical corrosion branch across channels. It utilizes the dynamic fracture area ratio as a nonlinear amplification multiplier for chloride ion permeation channels and integrates positive temperature acceleration, moisture evaporation suppression, and the dense passivation effect of corrosion products to make the conductor corrosion prediction more consistent with actual operating conditions. Furthermore, it maps the conductor surface oxide film rupture degree and conductor corrosion degree to the interface polarization parameter and porous dielectric leakage conductance parameter in the equivalent insulation impedance physical topology, respectively, achieving the separation and quantification of the contributions of mechanical damage and electrochemical damage. A closed-loop prediction mechanism is formed through fluctuation standard deviation normalization, weight updates under physical upper limit constraints, synchronous correction of the mechanical damage branch, and residual convergence verification. This significantly improves the accuracy, robustness, and interpretability of aging prediction for offshore wind power aluminum alloy power cables and can output the oxide film rupture degree and conductor corrosion degree, providing a reliable basis for cable operation and maintenance and replacement decisions.

[0008] In summary, this invention solves the technical problems of low prediction accuracy caused by inconsistent multi-source data, lack of physical coupling between mechanical and corrosion branches, and unreliable residual correction in the aging prediction of aluminum alloy power cables for offshore wind power. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating the aging prediction method for aluminum alloy power cables for offshore wind power provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the residual calculation and correction process in the aging prediction method for aluminum alloy power cables for offshore wind power provided in this embodiment of the invention. Figure 3 This is a schematic diagram of the aging prediction system for aluminum alloy power cables for offshore wind power provided in an embodiment of the present invention.

[0011] The components represented by each number in the attached diagram are explained below: Multi-source data acquisition module 11, dual-branch network construction module 12, reference current correction module 13, aging prediction output module 14. Detailed Implementation

[0012] This invention provides a method and system for predicting the aging of aluminum alloy power cables for offshore wind power, specifically addressing the technical problems of inconsistent multi-source data, lack of physical coupling between mechanical and corrosion branches, and unreliable residual correction in predicting the aging of aluminum alloy power cables for offshore wind power, which lead to low prediction accuracy.

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0014] Example 1, as Figure 1 As shown, this invention provides a method for predicting the aging of aluminum alloy power cables for offshore wind power, including: S100: Obtain the mechanical vibration sequence, electrothermal temperature and humidity sequence, and shielding dielectric loss current of the target aluminum alloy power cable.

[0015] Currently, in the aging prediction of aluminum alloy cables for offshore wind power, mechanical vibration, electrothermal humidity, and shielding dielectric loss current are often collected independently. The sampling frequency, time reference, and physical dimensions are inconsistent. Vibration signals are mostly limited to the tower or local nodes, and temperature, humidity, and salt spray concentration lack synchronization along the line. Furthermore, the grounding wire leakage current contains a mixture of capacitive reactive power, fundamental harmonics, and electromagnetic interference. This results in subsequent mechanical damage and electrochemical corrosion branches failing to obtain time-aligned, frequency-matched, and physically meaningful inputs, easily distorting aging predictions. Therefore, how to synchronously acquire multi-source time-series data covering mechanical vibration, electrothermal humidity, and shielding dielectric loss current, with time alignment, frequency consistency, and clear physical meaning, for the complex operating conditions of offshore wind power, has become an urgent technical problem to be solved.

[0016] Step S100 in the method provided in this embodiment of the invention includes: The mechanical vibration acceleration signals of the submarine cable monitoring nodes deployed at the seabed contact points of the offshore wind turbine base are acquired to form a high-frequency mechanical vibration sequence. By using a distributed fiber optic temperature measurement system and environmental sensors laid inside the cable, the conductor temperature, relative humidity and salt spray concentration along the line are collected simultaneously to form a low-frequency electrothermal temperature and humidity sequence. The leakage current on the cable grounding wire is collected by a high-frequency current transformer, and the fundamental and higher harmonic components are filtered out. The resistive leakage current in the characteristic frequency band is extracted as the dielectric loss current of the shielding layer.

[0017] In this embodiment of the invention, the mechanical vibration sequence, electrothermal temperature and humidity sequence, and shielding dielectric loss current of the target aluminum alloy power cable are acquired, providing time-aligned, frequency-domain consistent, and physically meaningful multi-source input data for aging prediction of the target aluminum alloy power cable. The target aluminum alloy power cable is laid in an offshore wind power environment, where parameters such as mechanical vibration, temperature, humidity, and salt spray concentration are unevenly distributed in space and time. Therefore, monitoring nodes, sensors, and sampling parameters need to be set up for key sections and key parameters respectively.

[0018] Specifically, monitoring nodes are set at the seabed contact section between the offshore wind turbine foundation and the seabed, including the tower foundation transition section, the J-tube outlet, both ends of the bend limiter, the seabed contact section, and the anchoring section. The node spacing is set between 20m and 100m, preferably 50m, because these locations are areas of significant vibration, bending, and salt spray deposition. A 50m spacing can cover typical bending sections and high-risk corrosion sections. Too large a spacing may lead to missed detection of local damage, while too small a spacing will increase the number of sensors, wiring complexity, and maintenance costs.

[0019] Furthermore, each mechanical vibration monitoring node is equipped with a triaxial accelerometer, which can be a MEMS accelerometer or a fiber optic accelerometer. The measurement range is set to ±2g to ±20g, preferably ±5g, with a resolution better than 0.001g, and the frequency response range is set to 0.5Hz to 2kHz. The principle behind setting this measurement range and frequency response range is that the mechanical vibrations of concern for offshore wind power cables are mainly concentrated in the 0.5Hz to 2kHz range. The ±5g measurement range can cover common wave vibrations and wind turbine vibrations, while retaining sufficient resolution to prevent small vibration signals from being drowned out by noise.

[0020] Meanwhile, the sampling rate of the mechanical vibration acceleration signal is set to 1kHz to 20kHz, preferably 10kHz. Since the upper limit of the vibration frequency band of interest is 2kHz, according to the Nyquist sampling theorem, the sampling rate should be at least 4kHz. Considering anti-aliasing filtering and spectral analysis margin, 10kHz is preferred to avoid frequency aliasing and preserve high-frequency damage characteristics. The acquired signal undergoes anti-aliasing filtering, detrending, and bandpass filtering, with the bandpass filtering range set to 0.5Hz to 2kHz.

[0021] Furthermore, a sliding time window is used to extract the mechanical vibration acceleration signal. The window length is set to 10s to 60s, preferably 30s, and the sliding step size is set to 1s to 5s, preferably 5s. The principle behind setting this window length and step size is that the wave period of offshore wind power is typically 5s to 15s, and the wind turbine rotation period is typically 3s to 10s. A 30s window can cover multiple periods, stabilizing the root mean square value, peak value, kurtosis, and spectral statistics; a 5s step size can balance real-time performance and computational complexity. Within each window, the root mean square value, peak value, kurtosis, and spectral features are extracted to construct a high-frequency mechanical vibration sequence.

[0022] It should be noted that if the effective value of vibration exceeds 80% of the accelerometer's measurement range, it will be marked as abnormal and resampling will be triggered. Accelerometers are prone to clipping distortion when approaching full scale; the clipped waveform introduces spurious high-frequency components, affecting subsequent stress fatigue cycle feature extraction. During resampling, data quality can be ensured by reducing gain, changing the measurement range, or checking the installation status.

[0023] Furthermore, a distributed fiber optic temperature measurement system is laid along the cable. The spatial resolution of the distributed fiber optic temperature measurement system is set to 0.5m to 1m, preferably 1m, and the sampling interval is set to 10s to 60s, preferably 30s. The principle behind setting this spatial resolution and sampling interval is that the temperature change of the cable conductor along the axial direction is typically on the meter scale. A spatial resolution of 1m is sufficient to locate local hot spots, and a 30s time interval can track slow temperature changes while avoiding excessive data volume. The fiber optic temperature measurement results are inverted into conductor temperature using a thermal circuit model, and calibration is performed using built-in thermocouples or the thermal balance relationship of the operating current.

[0024] The environmental sensors are installed at the tower foundation and the J-tube outlet to collect ambient relative humidity and salt spray concentration. The sampling rate of the environmental sensors is set to 0.1Hz to 1Hz, preferably 1Hz, because the changes in ambient relative humidity and salt spray concentration are slower than those caused by mechanical vibration, and 1Hz sampling is sufficient to describe their changing trends. A higher sampling rate would increase power consumption and storage burden. The salt spray concentration range is set to 0 to 100 mg / m³, with an accuracy of ±5%, which can be obtained through conductivity or salt spray deposition rate conversion. The conductor temperature, ambient relative humidity, and salt spray concentration together constitute a low-frequency electrothermal temperature and humidity sequence.

[0025] Furthermore, all monitoring nodes are synchronized using GPS, BeiDou, or the IEEE 1588 precision time protocol, with a time error of less than 1ms. Phase and event alignment are required between mechanical vibration and grounding wire leakage current; millisecond-level synchronization meets the requirements for subsequent cross-branch physical gating and residual calculation. Low-frequency electrothermal temperature and humidity sequences are interpolated to a unified time grid, which can be set to 1s or 30s. High-frequency mechanical vibration sequences are aligned with low-frequency timestamps after calculating statistics within a window.

[0026] Meanwhile, during data quality checks, conditions where the conductor temperature exceeds the long-term allowable operating temperature threshold for cable insulation are marked as high-temperature conditions; salt spray concentrations exceeding 50 mg / m³ are marked as high-salt spray conditions; and ambient relative humidity exceeds 95% are marked as condensation conditions. This is because conductor temperatures exceeding the long-term allowable operating temperature accelerate the deterioration of insulation and the oxide film on the conductor surface; salt spray concentrations exceeding 50 mg / m³ significantly accelerate aluminum alloy corrosion due to marine atmospheric salt spray deposition; and ambient relative humidity exceeding 95% approaches saturation, easily forming a liquid film on the cable surface and promoting electrochemical corrosion.

[0027] Furthermore, the characteristic frequency band is determined based on the dielectric relaxation characteristics of the cable insulation material. The peak frequency of the characteristic frequency band is taken as the frequency within the 10Hz to 1kHz range where the dielectric loss factor is highest. To determine this center frequency, samples are taken from the same batch of insulation material or from the insulation layer of retired cables. The sample thickness is set to 0.1mm to 1mm, preferably 0.5mm, and the sample diameter is set to 25mm to 50mm, preferably 40mm. Excessive thickness will result in uneven electric field distribution and reduced test sensitivity, while insufficient thickness can easily introduce surface leakage and edge effects. A 40mm diameter ensures that the test electrodes cover a uniform electric field area.

[0028] Broadband dielectric spectroscopy tests were performed on the samples, covering a frequency range of 10 Hz to 1 kHz, extended to 0.1 Hz to 1 MHz, and a temperature range of 20 °C to 120 °C, to obtain the curve of dielectric loss factor as a function of frequency. The peak frequency of the dielectric loss factor in the 10 Hz to 1 kHz band was taken as the center frequency of the characteristic frequency band. If multiple peaks exist, the main relaxation peak with the largest dielectric loss factor amplitude and the most sensitive to changes in temperature and aging was selected, because the dielectric relaxation peak corresponds to the dipole polarization and interface polarization processes inside the insulating material. Aging changes the number of polarization units and the interface state, causing the peak frequency and amplitude to shift. Using the peak frequency as the center frequency can improve the sensitivity of the dielectric loss current of the shielding layer to the aging state. If there is no obvious peak in the 10 Hz to 1 kHz band, the frequency corresponding to the maximum dielectric loss factor in this band is taken as the center frequency.

[0029] The bandwidth of the characteristic frequency band is set to 5% to 20% of the center frequency, preferably 10%. For example, when the center frequency is 120Hz, the characteristic frequency band is set to 108Hz to 132Hz. This is because the peak value of the dielectric loss factor usually has a certain half-width at half-maximum (HWHM). If the bandwidth is too narrow, the peak energy will be missed; if the bandwidth is too wide, interference from adjacent frequencies will be introduced. If the dielectric spectrum test can provide the peak HWHM, then the HWHM is directly taken as the characteristic frequency band bandwidth, ensuring that the bandwidth does not exceed 100Hz. The principle behind setting this upper limit is that a bandwidth exceeding 100Hz is prone to covering multiple relaxation processes or power frequency interference, reducing the characteristic frequency band's ability to distinguish a single aging mechanism.

[0030] Furthermore, leakage current on the cable grounding wire is acquired using a broadband current transformer. The bandwidth of the broadband current transformer is set to 0.1Hz to 10MHz, or a combination of a high-frequency current transformer and a power-frequency current transformer can be used. The sampling rate is set to 1MHz to 10MHz, preferably 5MHz, and the analog-to-digital conversion bit depth is not less than 16 bits. The principle behind setting this bandwidth and sampling rate is that a wide bandwidth can cover the characteristic frequency band and high-frequency interference, a higher sampling rate can preserve leakage current details, and a 16-bit analog-to-digital conversion can improve the resolution of weak resistive currents.

[0031] Furthermore, the original leakage current signal is filtered by anti-aliasing filtering and digital filtering to remove high-order harmonics, switching noise and external electromagnetic interference; when the characteristic frequency band does not cover the power frequency, the power frequency fundamental wave is also filtered out; when the characteristic frequency band covers the power frequency, only the power frequency reactive component and high-order harmonics are filtered out, while the active component in phase with the operating voltage or test voltage within the characteristic frequency band is retained.

[0032] When collecting data while the cable is in operation, a voltage reference signal within the characteristic frequency band is simultaneously acquired. The real part of the cross-power spectrum is extracted as the resistive leakage current through cross-power spectrum calculation, and this resistive leakage current is used as the dielectric loss current of the shielding layer. If the characteristic frequency band component is absent in the operating voltage, an active injection test method is used. During the active injection test, an AC test voltage with a center frequency equal to the center frequency of the characteristic frequency band is applied to the cable conductor. The amplitude is set to 10% to 20% of the rated phase voltage of the cable, preferably 10%, and the test time is set to 10 to 60 seconds, preferably 30 seconds. The low voltage avoids additional damage to the cable insulation, and 30 seconds coincides with the mechanical vibration window, which can suppress random noise and improve the signal-to-noise ratio of resistive current extraction. If the test time is too short, noise suppression will be insufficient; if the test time is too long, the dynamic response will be slowed down. The grounding wire leakage current is collected through a broadband current transformer, and the active component in phase with the test voltage is extracted as the dielectric loss current of the shielding layer. The cross-power spectrum integration time is set to 10s to 60s, preferably 30s. 30s is consistent with the mechanical vibration window, which can suppress random noise and improve the signal-to-noise ratio of resistive current extraction.

[0033] When extracting within the characteristic frequency band, a wideband current transformer is used to collect the grounding wire leakage current. After digital filtering, the active component in phase with the operating voltage or test voltage within the characteristic frequency band is extracted as the shielding layer dielectric loss current. The extraction results are timestamped together with the mechanical vibration sequence and the electrothermal temperature and humidity sequence to form a data set that can be input into a dual-branch physical constraint time series network.

[0034] When handling missing data, adjacent node interpolation is used when the missing proportion of the mechanical vibration sequence exceeds 5%, forward padding is used when the missing proportion of the electrothermal temperature and humidity sequence exceeds 10%, and the test time or cross-power spectrum integration time is extended to 60s when the signal-to-noise ratio of the shielding layer dielectric loss current is below 20dB. The principle behind setting these processing methods and thresholds is to ensure the integrity of the input sequence while avoiding excessive interpolation that introduces false physical features. Extending the phase-locked integration time can improve the accuracy of resistive current extraction under weak signal conditions.

[0035] This invention employs vibration monitoring nodes deployed in key sections and distributed optical fibers and environmental sensors to synchronously collect conductor temperature, humidity, and salt spray concentration along the route. It also uses broadband current transformers and resistive component extraction to obtain the dielectric loss current of the shielding layer. This forms a synchronous data foundation for high-frequency mechanical vibration sequences, low-frequency electrothermal temperature and humidity sequences, and dielectric loss current. This enables the mechanical damage branch to extract stress fatigue cycle characteristics, and the electrochemical corrosion branch to perform corrosion accumulation analysis using salt spray, humidity, and temperature. Furthermore, it provides reliable input for subsequent cross-branch physical gating, baseline dielectric loss current calculation, and residual correction, thereby improving the accuracy and robustness of aging prediction for offshore wind power aluminum alloy power cables.

[0036] S200: Construct a dual-branch physical constraint time-series network, which includes a mechanical damage branch and an electrochemical corrosion branch. Input the mechanical vibration sequence into the mechanical damage branch and the electrothermal temperature and humidity sequence into the electrochemical corrosion branch, and output the degree of oxide film rupture on the conductor surface and the degree of conductor corrosion, respectively.

[0037] Currently, in the aging prediction of aluminum alloy power cables for offshore wind power, the rupture of the conductor surface oxide film caused by mechanical vibration and electrochemical corrosion are often modeled separately. There is a lack of physical coupling between the mechanical damage branch and the electrochemical corrosion branch. The accelerating effect of chloride ion permeation channel expansion when the amplitude of mechanical vibration increases cannot be reflected. Simultaneously, the competitive effect of moisture evaporation on electrochemical corrosion inhibition under high-temperature conditions is not considered, and the dense passivation layer formed after corrosion products accumulate to a certain extent is not described. This leads to either overestimation or underestimation of conductor corrosion degree, and a significant deviation between the aging prediction results and the actual operating conditions. Therefore, how to construct a cross-branch physical gating coupling between the mechanical damage branch and the electrochemical corrosion branch, and simultaneously reflect the competitive effects of positive temperature acceleration, moisture evaporation inhibition, and dense passivation of corrosion products in the electrochemical corrosion branch, has become an urgent technical problem to be solved.

[0038] In step S200 of the method provided in this embodiment of the invention, the dual-branch physical constraint temporal network specifically includes: The mechanical damage branch includes a temporal convolution module, which is used to extract stress fatigue cycle features from the mechanical vibration sequence and outputs a dynamic fracture area ratio that characterizes the proportion of exposed oxide film area on the conductor surface, as the degree of oxide film fracture on the conductor surface. During the feature extraction process, the electrochemical corrosion branch receives the electrothermal temperature and humidity sequence as the main input, and at the same time introduces the dynamic fracture area ratio output by the mechanical damage branch as the cross-branch physical gate signal. The cross-branch physical gate signal is a nonlinear amplification multiplier of the chloride ion permeation channel. In the electrochemical corrosion branch, the dynamic rupture area ratio is used as a nonlinear amplification multiplier of the chloride ion permeation channel to dynamically scale the input salt spray concentration characteristics. This allows for a higher effective salt spray concentration to participate in the reaction under conditions with greater mechanical vibration amplitude, thereby accelerating the cross-channel coupling of mechanical vibration to electrochemical corrosion.

[0039] In this embodiment of the invention, when constructing the dual-branch physical constraint temporal network, the mechanical damage branch and the electrochemical corrosion branch are set up side by side. The mechanical vibration sequence is input into the mechanical damage branch, and the electrothermal temperature and humidity sequence is input into the electrochemical corrosion branch.

[0040] Specifically, the mechanical damage branch includes a temporal convolution module, which employs a one-dimensional dilated causal convolution structure. The number of layers in the one-dimensional dilated causal convolution is set to 3 to 8, preferably 5; the kernel size of each layer is set to 3 to 5, preferably 3; the dilation coefficients are set sequentially to 1, 2, 4, 8, and 16; the number of channels per layer is set to 16 to 128, preferably 64; and the activation function is ReLU or GELU. The one-dimensional dilated causal convolution can expand the receptive field while maintaining temporal causality, enabling the mechanical damage branch to cover multiple wave cycles and wind turbine rotation cycles, thereby extracting stress fatigue cycle characteristics.

[0041] Furthermore, the input to the mechanical damage branch is a high-frequency mechanical vibration sequence. This sequence includes the root mean square value, peak value, kurtosis, and spectral features within each sliding window. The temporal convolution module performs layer-by-layer convolution and pooling on these features to extract stress fatigue cycle features. These features include cycle amplitude, number of cycles, peak factor, and kurtosis rate of change. The output of the temporal convolution module is connected to a fully connected layer. The output of this layer is mapped to the 0-1 range using a sigmoid function to obtain the dynamic fracture area ratio. The dynamic fracture area ratio represents the proportion of exposed oxide film area on the conductor surface. A dynamic fracture area ratio of 0 indicates an intact oxide film, while a ratio of 1 indicates complete oxide film fracture. The dynamic fracture area ratio serves as the measure of oxide film fracture on the conductor surface.

[0042] Furthermore, the training label for the dynamic fracture area ratio can be obtained through accelerated aging tests. Specifically, in accelerated aging tests, mechanical vibrations of different amplitudes and cycles are applied to aluminum alloy conductor samples, and then the proportion of exposed oxide film area is measured using scanning electron microscopy or electrochemical impedance spectroscopy as a monitoring label. If direct measurement is not possible, the fatigue damage degree can be calculated using the rainflow counting method and Miner's linear cumulative damage theory, and then mapped to the dynamic fracture area ratio. The dynamic fracture area ratio is set to a range of 0 to 1. A dynamic fracture area ratio less than 0.05 is considered a minor fracture, a dynamic fracture area ratio between 0.05 and 0.3 is considered a slight fracture, a dynamic fracture area ratio between 0.3 and 0.7 is considered a moderate fracture, and a dynamic fracture area ratio greater than 0.7 is considered a severe fracture. This is because an oxide film fracture area ratio below 5% has little impact on chloride ion permeation channels, and a ratio above 70% essentially results in the loss of oxide film protection.

[0043] Furthermore, the electrochemical corrosion branch receives the electrothermal temperature and humidity sequence as its main input. This sequence includes conductor temperature, ambient relative humidity, and salt spray concentration. The electrochemical corrosion branch first extracts the temporal features of the electrothermal temperature and humidity sequence using a one-dimensional convolutional layer or a gated recurrent unit. The salt spray concentration feature is dynamically scaled by a cross-branch physical gating signal before being used in subsequent cumulative calculations. This cross-branch physical gating signal originates from the dynamic fracture area ratio output by the mechanical damage branch, which serves as a nonlinear amplification multiplier for the chloride ion permeation channel.

[0044] Specifically, the nonlinear amplification multiplier = 1 + coupling strength coefficient × nonlinear exponent of dynamic fracture area ratio. The coupling strength coefficient is set to 0.5 to 5, preferably 2, and the nonlinear exponent is set to 0.5 to 2, preferably 1. The principle behind setting this range is that if the coupling strength coefficient is too small, the accelerating effect of mechanical vibration on corrosion is not significant; if the coupling strength coefficient is too large, the effective salt spray concentration is likely to exceed the physically reasonable range. A nonlinear exponent less than 1 indicates that even slight fractures produce large penetration channels, while a nonlinear exponent greater than 1 indicates that penetration is significantly amplified only after the fracture area reaches a certain extent. Effective salt spray concentration = measured salt spray concentration × nonlinear amplification multiplier. The larger the dynamic fracture area ratio, the larger the nonlinear amplification multiplier, and the higher the effective salt spray concentration allowed to participate in the reaction, thereby achieving cross-channel coupling acceleration of electrochemical corrosion by mechanical vibration.

[0045] In the electrochemical corrosion branch, the dynamic fracture area ratio is used as a nonlinear amplification multiplier for chloride ion permeation channels to dynamically scale the input salt spray concentration characteristics. This also includes: The positive acceleration factor of temperature on the chemical reaction rate is determined based on the conductor temperature, and the positive acceleration factor is positively correlated with the conductor temperature. The amount of local liquid water loss due to water evaporation is determined based on the conductor temperature, and the electrolyte participation factor is determined based on the amount of local liquid water loss. The electrolyte participation factor is negatively correlated with the amount of local liquid water loss. Thermodynamic competition constraint results are obtained by fusion calculation based on the positive acceleration factor and the electrolyte participation factor. These results are used to characterize the inhibitory competition effect of high-temperature evaporation on corrosion acceleration. Based on the salt spray concentration and thermodynamic competition constraint results, the cumulative amount of historical corrosion products was obtained through time-cumulative analysis. When the cumulative amount of historical corrosion products has not reached the densification threshold, the degree of conductor corrosion is determined based on the salt spray concentration and thermodynamic competition constraint results. When the accumulated amount of historical corrosion products reaches the density threshold, the passivation inhibition factor is activated in the electrochemical corrosion branch, and the degree of conductor corrosion is determined based on the passivation inhibition factor, salt spray concentration, and thermodynamic competitive constraint results.

[0046] In this embodiment of the invention, a positive acceleration factor of temperature on the chemical reaction rate is determined based on the conductor temperature, and the positive acceleration factor is positively correlated with the conductor temperature. The positive acceleration factor is calculated in exponential form, i.e., positive acceleration factor = e^(-π / 2) / π / 2. a Where a = temperature acceleration factor × (conductor temperature - reference temperature). The reference temperature is set to 25°C, and the temperature acceleration factor is set to 0.02 to 0.1 per degree Celsius, preferably 0.05 per degree Celsius, because the electrochemical corrosion rate of aluminum alloy increases with temperature, but the increase is constrained by the Arrhenius relation. If the temperature acceleration factor is too small, it will not reflect the high-temperature acceleration, and if it is too large, it will cause the prediction results to deviate from reality.

[0047] Furthermore, the amount of localized liquid water loss due to moisture evaporation is determined based on the conductor temperature. The amount of localized liquid water loss increases with increasing conductor temperature. The amount of localized liquid water loss can be calculated using a normalized form: localized liquid water loss = temperature coefficient × (conductor temperature - reference temperature), and is limited to the range of 0 to 1. The temperature coefficient is set to 0.005 to 0.02 degrees Celsius, preferably 0.01 degrees Celsius, because increasing conductor temperature accelerates the evaporation of the surface liquid film, but the amount of liquid film loss will not increase indefinitely and will not increase further after complete drying.

[0048] Furthermore, the electrolyte participation factor is determined based on the amount of local liquid water loss. The electrolyte participation factor is negatively correlated with the amount of local liquid water loss. The electrolyte participation factor is calculated using a reciprocal form, i.e., electrolyte participation factor = 1 / (1 + evaporation inhibition coefficient × local liquid water loss). The evaporation inhibition coefficient is set to 1 to 5, preferably 2. The principle behind setting this range is that the greater the amount of local liquid water loss, the less liquid water is available on the cable surface to dissolve chloride ions and form an electrolyte solution, and the more the electrochemical corrosion reaction is inhibited.

[0049] Furthermore, the thermodynamic competition constraint result is obtained by fusing the positive acceleration factor and the electrolyte participation factor. The fusing method adopts multiplicative fusion, that is, thermodynamic competition constraint result = positive acceleration factor × electrolyte participation factor. The thermodynamic competition constraint result is used to characterize the inhibitory competitive effect of high-temperature evaporation on corrosion acceleration. The principle of setting multiplicative fusion is that temperature both promotes corrosion by accelerating the chemical reaction rate and reduces the amount of electrolyte participation by accelerating liquid film evaporation. These two effects exist simultaneously and compete with each other. Multiplicative fusion can simultaneously reflect positive acceleration and negative inhibition.

[0050] Furthermore, based on the salt spray concentration and thermodynamic competition constraint results, a time-cumulative analysis was performed to obtain the cumulative amount of historical corrosion products. The time-cumulative analysis adopted a recursive accumulation method, with the cumulative amount of historical corrosion products updated step-by-step according to the time step. The cumulative amount of historical corrosion products at the current moment = the cumulative amount of historical corrosion products at the previous moment × aging recovery factor + effective salt spray concentration × thermodynamic competition constraint result × time step × cumulative calibration coefficient. Aging recovery factor = e b Where b = negative time step / aging recovery time constant, the aging recovery time constant is set to 1 to 30 days, preferably 7 days. The time step is set to 1 to 24 hours, preferably 24 hours. The cumulative calibration coefficient is set to 0.0001 to 0.01, preferably 0.001. Corrosion products accumulate over time, but some loose corrosion products may detach or transform. The aging recovery factor is used to describe this decay process, and the cumulative calibration coefficient is used to normalize the physical dimensions to the range of 0 to 3.

[0051] Furthermore, the densification threshold is set to 0.6 to 0.9, preferably 0.8, because when the accumulated amount of historical corrosion products reaches 0.8, the corrosion product coverage area is close to 80%, the corrosion product layer gradually becomes dense, the channels for chloride ions and moisture to penetrate into the conductor surface are partially blocked, and the passivation effect begins to be significant. When the accumulated amount of historical corrosion products does not reach the densification threshold, the degree of conductor corrosion is determined based on the salt spray concentration and thermodynamic competition constraints. The degree of conductor corrosion can be calculated using a saturated growth form, i.e., conductor corrosion degree = 1 - e c c = cumulative negative historical corrosion products / corrosion calibration constant. The corrosion calibration constant is set to 0.5 to 2, preferably 1. The principle behind this setting is that the degree of corrosion increases with the increase of the cumulative amount of corrosion products, but the growth rate gradually decreases, which conforms to the saturation characteristics of electrochemical corrosion.

[0052] Finally, when the accumulated amount of historical corrosion products reaches the density threshold, a passivation inhibition factor is activated in the electrochemical corrosion branch. The passivation inhibition factor is calculated using a linear decay form, i.e., passivation inhibition factor = 1 - passivation decay coefficient × (accumulated amount of historical corrosion products - density threshold) / (maximum accumulation amount - density threshold). The passivation decay coefficient is set to 0.1 to 1, preferably 0.5. The maximum accumulation amount is set to 2 to 3, preferably 2.5, because the denser the corrosion product layer, the stronger the inhibition of chloride ion penetration and moisture transport, and the slower the increase in conductor corrosion degree. The degree of conductor corrosion is determined based on the passivation inhibition factor, salt spray concentration, and thermodynamic competition constraints. Conductor corrosion degree = density threshold + (1 - passivation inhibition factor) × (accumulated amount of historical corrosion products - density threshold). This calculation method can reduce the rate of corrosion growth after reaching the density threshold, preventing the degree of corrosion from increasing indefinitely.

[0053] The dual-branch physical constraint time-series network is trained using a loss function that includes data fitting terms and physical constraint terms. The total loss = data fitting loss + first weight coefficient × monotonicity constraint loss + second weight coefficient × boundary constraint loss + third weight coefficient × residual constraint loss. The first weight coefficient is between 0.1 and 0.5, preferably 0.2. The second weight coefficient is between 0.1 and 0.5, preferably 0.2. The third weight coefficient is between 0.5 and 2.0, preferably 1.0. The principle behind setting this weight range is that the data fitting loss dominates the network's approximation of the measured aging label, the physical constraint term acts as a regularization term to suppress outputs that violate physical laws, and the residual constraint loss is directly related to the measured dielectric loss current of the shielding layer; therefore, the third weight coefficient is slightly higher than the other physical constraint terms.

[0054] The data fitting loss is expressed as mean squared error. The squares of the differences between the predicted dynamic fracture area ratio and the dynamic fracture area ratio label, and the squares of the differences between the predicted conductor corrosion degree and the conductor corrosion degree label are calculated separately. These are then averaged across all training samples and summed. The dynamic fracture area ratio label is obtained through accelerated aging tests. Different amplitudes and cycles of mechanical vibration are applied to aluminum alloy conductor samples, and the proportion of exposed oxide film area is measured using scanning electron microscopy or electrochemical impedance spectroscopy as the label. The conductor corrosion degree label is obtained through salt spray corrosion tests. A 5% sodium chloride solution is set in a salt spray chamber at 35°C for 24 to 3000 hours. The porous corrosion depth is measured using metallographic microscopy or electrochemical impedance spectroscopy and normalized to form the conductor corrosion degree label.

[0055] The monotonicity constraint loss is implemented using a hinge loss mechanism. Two time windows with different mechanical vibration amplitudes are selected from the training set. If the mechanical vibration amplitude in the first window is smaller than that in the second window, the predicted dynamic fracture area ratio in the first window must not exceed that in the second window. Similarly, two time windows with different salt spray concentrations are selected from the training set. If the salt spray concentration in the first window is smaller than that in the second window, the predicted conductor corrosion degree in the first window must not exceed that in the second window. When monotonicity is violated, the larger of the difference between the predicted value in the second window and the predicted value in the first window, and zero, is taken as the penalty term. The penalty is zero when monotonicity is satisfied. The average of the penalty terms for all sample pairs is used as the monotonicity constraint loss.

[0056] The boundary constraint loss adopts the form of hinge loss. When the predicted dynamic fracture area ratio is less than zero, the larger value of zero minus the predicted value and zero is taken, and then the square is used as a penalty term. When the predicted dynamic fracture area ratio is greater than one, the larger value of the predicted value minus one and zero is taken, and then the square is used as a penalty term. The predicted conductor corrosion degree is processed in the same way. The average of the four penalty terms is taken as the boundary constraint loss. The principle of setting this form is that when the predicted value is in the range of zero to one, the penalty is zero, and when it exceeds the range, a positive penalty is generated, thereby ensuring that the output has a clear physical meaning.

[0057] The residual constraint loss is expressed as the mean square error of the normalized residual. The dynamic fracture area ratio and conductor corrosion degree of the network output are mapped to incremental parameters in the physical topology of the equivalent insulation impedance, and the reference dielectric loss current is calculated. The measured shield dielectric loss current sequence and the reference dielectric loss current sequence are extracted by a sliding time window, and the normalized residual is obtained according to the aforementioned normalization method. The square of the normalized residual is averaged over all training samples and used as the residual constraint loss.

[0058] The following are sample parameters for the training process. The training, validation, and test sets are divided chronologically, with proportions of 70%, 15%, and 15%, respectively. This chronological division prevents future data leakage into the training set. The validation set is used for model selection and early stopping, while the test set is used for final evaluation. The batch size is set to 32 to 256, preferably 64. The number of iterations is set to 100 to 1000, preferably 500. The optimizer uses Adam or AdamW. The initial learning rate is set to 0.0001 to 0.01, preferably 0.001. The learning rate decay uses cosine annealing or step decay, decreasing by 0.1 times every 100 iterations. The gradient clipping threshold is set to 0.5 to 2.0, preferably 1.0. This threshold prevents excessively large gradients in the residual constraint loss on individual outliers, which could cause training oscillations. The early stopping condition is set to stop training if the total loss on the validation set does not decrease for 20 consecutive iterations. After training is complete, save the model parameters that minimize the total loss of the validation set.

[0059] This invention, through the construction of a dual-branch physical constraint time-series network and the refinement of the electrochemical corrosion branch, uses the dynamic fracture area ratio output by the mechanical damage branch as a nonlinear amplification multiplier for the chloride ion permeation channel. This results in a higher effective salt spray concentration when the mechanical vibration amplitude is greater, achieving cross-channel coupling acceleration of electrochemical corrosion by mechanical vibration. Simultaneously, the thermodynamic competition constraint result is obtained by multiplying the temperature positive acceleration factor with the electrolyte participation factor. Furthermore, the accumulation of historical corrosion products and the densification threshold activate the passivation inhibition factor, which can accurately describe the nonlinear influence of high-temperature evaporation inhibition and corrosion product densification passivation on the degree of conductor corrosion. This improves the physical consistency and accuracy of aging prediction for offshore wind power aluminum alloy power cables.

[0060] S300: Calculate the reference dielectric loss current of aluminum alloy power cable based on the degree of oxide film rupture and conductor corrosion, and calculate the residual between the reference dielectric loss current and the measured shielding dielectric loss current. When the residual exceeds the preset threshold, the electrochemical corrosion branch is corrected.

[0061] Currently, in the aging prediction of aluminum alloy power cables for offshore wind power, the baseline dielectric loss current is often calculated using a fixed empirical model. The contributions of oxide film rupture caused by mechanical vibration and porous corrosion layer caused by electrochemical corrosion to dielectric loss cannot be separated. Measurement noise in the measured dielectric loss current of the shielding layer is easily confused with actual aging deviations. When residuals exceed limits, there is a lack of clear physical correction directions. The branch weight parameters of electrochemical corrosion may be adjusted indefinitely, violating the laws of material physics, leading to large and unexplainable deviations in aging prediction results. Therefore, how to establish a baseline dielectric loss current calculation model that can separate mechanical and electrochemical damage, and how to perform noise normalization, trend judgment, and branch correction under physical upper limit constraints on the measured residuals, have become urgent technical problems to be solved.

[0062] Step S300 of the method provided in this embodiment of the invention, which calculates the reference dielectric loss current of the aluminum alloy power cable based on the degree of oxide film rupture and conductor corrosion, includes: Obtain the nominal operating voltage of the aluminum alloy power cable; The physical topology of equivalent insulation impedance for aluminum alloy power cables is constructed. The physical topology of equivalent insulation impedance includes parallel interface polarization parasitic capacitance branches and porous dielectric leakage current conduction branches. The degree of oxide film rupture on the conductor surface output by the mechanical damage branch is mapped to the capacitance increment parameter of the interface polarization parasitic capacitance branch. The degree of conductor corrosion output from the electrochemical corrosion branch is mapped to the conductivity increment parameter of the porous medium leakage conductivity branch; The nominal operating voltage is applied across the equivalent insulation impedance physical topology. The total AC leakage current flowing through the interface polarization parasitic capacitance branch and the porous dielectric leakage conduction branch is calculated. The active component that is in phase with the nominal operating voltage is extracted and used as the reference dielectric loss current of the aluminum alloy power cable.

[0063] In this embodiment of the invention, when calculating the reference dielectric loss current based on the degree of oxide film rupture and conductor corrosion on the conductor surface, the nominal operating voltage of the target aluminum alloy power cable is first obtained. Specifically, the nominal operating voltage can be read from the cable nameplate parameters, the operation monitoring system, or the dispatching system. When the center frequency of the characteristic frequency band is equal to the power frequency, the nominal operating voltage is taken as the power frequency phase voltage. When the center frequency of the characteristic frequency band deviates from the power frequency, the nominal operating voltage is taken as the equivalent test voltage corresponding to the center frequency of the characteristic frequency band, or the equivalent voltage is obtained according to the voltage-frequency conversion relationship. The principle of this processing method is that the reference dielectric loss current and the measured shielding layer dielectric loss current must be in the same frequency band; otherwise, the residual between the two will contain spurious deviations caused by frequency domain inconsistencies.

[0064] Furthermore, an equivalent insulation impedance physical topology for aluminum alloy power cables is constructed. This equivalent insulation impedance physical topology includes parallel interfacial polarization parasitic capacitance branches and porous dielectric leakage conductance branches. The interfacial polarization parasitic capacitance branch is equivalent to polarization capacitance and polarization loss conductance in parallel, and the porous dielectric leakage conductance branch is equivalent to porous dielectric leakage conductance. The principle behind this topology is that the rupture of the oxide film on the conductor surface enhances the interfacial polarization effect, which manifests as both an increase in polarization capacitance and an increase in polarization loss conductance. Conductor corrosion forms a porous corrosion layer, increasing the electrolyte connectivity paths within the porous corrosion layer, which manifests as an increase in porous dielectric leakage conductance. After both are connected in parallel, the active component in phase with the voltage in the total AC leakage current can simultaneously reflect both mechanical and electrochemical damage.

[0065] Furthermore, the degree of oxide film rupture on the conductor surface output by the mechanical damage branch is mapped to the capacitance increment parameter of the interfacial polarization parasitic capacitance branch, and simultaneously to the polarization loss conductance increment parameter of the interfacial polarization parasitic capacitance branch. The higher the degree of oxide film rupture on the conductor surface, the larger the proportion of exposed oxide film area, the stronger the interfacial polarization effect, and the larger both the capacitance increment parameter and the polarization loss conductance increment parameter. The capacitance increment parameter is positively correlated with the proportion of exposed oxide film area. The polarization loss conductance increment parameter is also positively correlated with the proportion of exposed oxide film area. Pure capacitance increment only changes the reactive current, not the active current. Only after introducing the polarization loss conductance increment can mechanical damage contribute to the active component of the reference dielectric loss current.

[0066] Simultaneously, the conductor corrosion degree output from the electrochemical corrosion branch is mapped to the conductivity increment parameter of the porous medium leakage conductivity branch. The higher the conductor corrosion degree, the greater the porous corrosion depth, the greater the increase in electrolyte connectivity paths within the porous medium, and the larger the conductivity increment parameter. The conductivity increment parameter is positively correlated with the porous corrosion depth. The porous corrosion layer provides more conductive channels for chloride ions and moisture; the more severe the corrosion, the greater the leakage conductivity and the greater the active leakage current in phase with the voltage.

[0067] Furthermore, the nominal operating voltage is applied across the equivalent insulation impedance physical topology, and the total AC leakage current flowing through the interface polarization parasitic capacitance branch and the porous dielectric leakage conductance branch is calculated. The total AC leakage current = nominal operating voltage × total admittance. Total admittance = admittance of the interface polarization parasitic capacitance branch + admittance of the porous dielectric leakage conductance branch. Admittance of the interface polarization parasitic capacitance branch = polarization loss conductance + polarization capacitance admittance. Admittance of the porous dielectric leakage conductance branch equals porous dielectric leakage conductance. The active component in phase with the nominal operating voltage is extracted from the total AC leakage current and used as the reference dielectric loss current for the aluminum alloy power cable. The reference dielectric loss current = nominal operating voltage × total conductance. The dielectric loss current is essentially an active leakage current in phase with the voltage; capacitive reactive current does not generate dielectric loss.

[0068] Among them, the degree of oxide film rupture on the conductor surface output by the mechanical damage branch is mapped to the capacitance increment parameter of the interface polarization parasitic capacitance branch, including: The proportion of exposed oxide film area is determined based on the degree of oxide film rupture on the conductor surface. The degree of enhancement of the interfacial polarization effect is determined based on the proportion of exposed oxide film area, and the capacitance increment parameter is positively correlated with the proportion of exposed oxide film area. Among them, mapping the conductor corrosion degree output by the electrochemical corrosion branch to the conductivity increment parameter of the porous medium leakage conductivity branch includes: The depth of porous corrosion is determined based on the degree of conductor corrosion. The degree of increase in electrolyte connectivity pathways in porous media is determined by the depth of porous corrosion, and the conductivity increment parameter is positively correlated with the depth of porous corrosion.

[0069] In this embodiment of the invention, the proportion of exposed oxide film area is determined based on the degree of oxide film rupture on the conductor surface. Specifically, the proportion of exposed oxide film area can be obtained directly from the dynamic rupture area ratio, or it can be obtained by multiplying the dynamic rupture area ratio by a calibration coefficient and limiting it to the range of 0 to 1.

[0070] Simultaneously, the enhancement degree of interfacial polarization effect is determined based on the proportion of exposed oxide film area. The capacitance increment parameter is calculated as: capacitance calibration coefficient × exposed oxide film area proportion × basic polarization capacitance. The capacitance calibration coefficient is set to 0.1 to 1, preferably 0.3. This coefficient represents the ratio of the interfacial polarization capacitance increment to the basic polarization capacitance when the oxide film is completely ruptured. If it is less than 0.1, the capacitance increment is easily drowned out by dielectric spectrum measurement noise; if it is greater than 1, local oxide film rupture is easily misjudged as insulation dampness or overall degradation. The preferred value of 0.3 aligns with the rule that the interfacial polarization capacitance increment when the aluminum alloy oxide film is completely ruptured is typically 20% to 40% of the basic value.

[0071] The polarization loss conductivity increment parameter is calculated as follows: polarization loss calibration coefficient × oxide film exposed area ratio × basic polarization loss conductivity. The polarization loss calibration coefficient is set to 0.5 to 5, preferably 2. This coefficient represents the ratio of the polarization loss conductivity increment to the basic polarization loss conductivity when the oxide film is completely ruptured. If it is less than 0.5, the contribution of mechanical damage to the active component in phase with the voltage is insufficient; if it is greater than 5, it is easily confused with penetrating conductive channels or severe moisture absorption. The preferred value of 2 conforms to the rule that the polarization loss conductivity increment when the oxide film is completely ruptured is usually one to three times the basic value.

[0072] Furthermore, the porosity corrosion depth is determined based on the degree of conductor corrosion. Porosity corrosion depth = corrosion depth calibration coefficient × conductor corrosion degree. The corrosion depth calibration coefficient is set to 0.01 mm to 1 mm, preferably 0.1 mm, because the conductor corrosion degree is 0 to 1, corresponding to a porosity corrosion depth from zero to hundreds of micrometers, which conforms to the physical scale of long-term corrosion of aluminum alloys in marine environments.

[0073] Simultaneously, the increase in electrolyte connectivity paths within the porous medium is determined based on the porous corrosion depth. The conductivity increment parameter = conductivity calibration coefficient × porous corrosion depth / reference corrosion depth × basic porous medium leakage conductance. The reference corrosion depth is set to 0.1 mm. The conductivity calibration coefficient is set to 0.5 to 5, preferably 2. This coefficient represents the ratio of the porous medium conductivity increment to the basic porous medium leakage conductance when the porous corrosion depth reaches the reference corrosion depth. If less than 0.5, the contribution of electrochemical damage to the loss current of the reference medium is too small; if greater than 5, it is easily confused with macroscopic cracks or insulation water ingress. The preferred value of 2 aligns with the general rule that the conductivity increment of the porous corrosion layer is typically one to three times the base value.

[0074] The aforementioned calibration coefficients can also be calibrated through accelerated aging tests. Specifically, using aluminum alloy conductor samples from the same batch, mechanical vibrations of different amplitudes and cycles are applied, and the proportion of exposed oxide film area, polarization capacitance, and polarization loss conductivity are measured. Then, a salt spray corrosion test is conducted to measure the porous corrosion depth and porous medium leakage conductivity. Each calibration coefficient is obtained through linear fitting. The sum of the products of the slope and the corresponding width of each interval in the piecewise linear mapping should equal the calibration coefficient. If the cable structure, insulation thickness, sheath material, or operating temperature range differs, those skilled in the art can adjust within the above range, but should not exceed it to avoid deviating from the physical laws of aluminum alloy conductor interface polarization and porous corrosion.

[0075] Furthermore, step S300 in the method provided in this embodiment of the invention further includes: Under the control condition of keeping the incremental conductivity parameter of the porous dielectric leakage current branch zero, the contribution of the incremental capacitance parameter to the loss current of the reference dielectric is determined as the contribution of mechanical damage. Under the control condition of keeping the capacitance increment parameter of the interfacial polarization parasitic capacitance branch zero, the contribution of the conductivity increment parameter to the reference dielectric loss current is determined as the electrochemical loss contribution. The mechanical and electrochemical damage contributions are output to quantify the absolute proportion of the impact of mechanical vibration and wet heat corrosion on cable aging and deterioration under current complex sea conditions.

[0076] In this embodiment of the invention, under the control condition of maintaining the incremental conductance parameter of the porous dielectric leakage conductance branch at zero, the contribution of the incremental capacitance parameter and the incremental polarization loss conductance parameter to the reference dielectric loss current is determined as the mechanical damage contribution. Specifically, the incremental conductance parameter is set to zero, while the incremental capacitance parameter and the incremental polarization loss conductance parameter are retained, and the first reference dielectric loss current is calculated at this time. The reference dielectric loss current under the base state is used as a reference value. Mechanical damage contribution = |difference between first reference dielectric loss current and reference value| / |current reference dielectric loss current - reference value|. The principle of this calculation method is that by setting the electrochemical damage channel to zero, the independent contribution of mechanical damage to the reference dielectric loss current is separated.

[0077] Furthermore, under the control condition of maintaining the incremental parameters of capacitance and polarization loss conductance of the interface polarization parasitic capacitance branch at zero, the contribution of the incremental conductance parameter to the reference dielectric loss current is determined as the electrochemical damage contribution. Specifically, the incremental parameters of capacitance and polarization loss conductance are set to zero, while the incremental conductance parameter is retained, and the second reference dielectric loss current is calculated at this time. Electrochemical damage contribution = |second reference dielectric loss current - reference value| / |current reference dielectric loss current - reference value|. The mechanical damage contribution and electrochemical damage contribution are output to quantify the absolute proportion of mechanical vibration and wet heat corrosion on cable aging and deterioration under the current complex sea conditions. The principle of setting this output method is that the sum of the two contributions is close to one, which can intuitively show whether the current aging state is mainly dominated by mechanical damage or electrochemical corrosion.

[0078] Further, the residual between the reference dielectric loss current and the measured shielding dielectric loss current is calculated, including: The measured dielectric loss current sequence of the shielding layer and the predicted baseline dielectric loss current sequence are extracted using a sliding time window. The error base is determined based on the degree of deviation between the shielding layer dielectric loss current sequence and the reference dielectric loss current sequence within the same time window; The standard deviation of the dielectric loss current sequence of the shielding layer within the time window is extracted as the uncertainty coefficient. The error base is normalized according to the uncertainty coefficient to obtain the true physical deviation after eliminating measurement noise interference, which is used as the residual.

[0079] In this embodiment of the invention, when calculating the residual between the reference dielectric loss current and the measured shielding dielectric loss current, a sliding time window is used to extract the measured shielding dielectric loss current sequence and the predicted reference dielectric loss current sequence. The sliding time window length is set to 10s to 60s, preferably 30s; the sliding step size is set to 1s to 5s, preferably 5s. This is because a 30s window can cover multiple wave cycles and wind turbine rotation cycles, stabilizing the current statistics, while a 5s step size balances real-time performance and computational complexity. The error base is determined based on the degree of deviation between the shielding dielectric loss current sequence and the reference dielectric loss current sequence within the same time window. The error base can be the root mean square error or the mean absolute error. In this embodiment, the mean absolute error is used: mean absolute error = sum of |measured values ​​within the same time window - reference values| / number of sampling points.

[0080] Furthermore, the standard deviation of the fluctuation of the shielding layer dielectric loss current sequence within the time window is extracted as the uncertainty coefficient. The fluctuation standard deviation is equal to the standard deviation of the measured shielding layer dielectric loss current sequence within the same time window. The error base is normalized based on the uncertainty coefficient to obtain the true physical deviation after eliminating measurement noise interference, which is then used as the residual. When the measurement noise is large, the fluctuation standard deviation is large, and the residual decreases after normalization, thus avoiding misjudging noise as true aging deviation; when the measurement noise is small, the fluctuation standard deviation is small, and the residual increases after normalization, thus improving the sensitivity to true physical deviation.

[0081] The normalized residual is calculated as the larger of the error base and the uncertainty coefficient / lower fluctuation limit. The lower fluctuation limit is set to 0.01 μA to 0.1 μA, preferably 0.05 μA. This limit is set because the uncertainty coefficient is the standard deviation of the current. When the measured dielectric loss current of the shielding layer within the window has almost no fluctuation, this standard deviation is close to zero. Direct division would drastically amplify the normalized residual, potentially causing sensor jamming or sampling anomalies to be misjudged as true aging deviations. The lower fluctuation limit is determined based on 2 to 5 times the minimum resolvable current of the current transformer, or 5% to 20% of the standard deviation of the measured dielectric loss current of the shielding layer during the normal aging stage, preferably 10%. When the uncertainty coefficient is less than or equal to the lower fluctuation limit, the measured current within that time window is considered to have no effective fluctuation, marking the data quality as abnormal. Correction of the electrochemical corrosion branch for that window is suspended, or the normalized residual is calculated based on the lower fluctuation limit while simultaneously triggering a data quality check. The error base, uncertainty coefficient, and lower fluctuation limit use the same current unit, and the normalized residual is dimensionless.

[0082] Furthermore, when the residual exceeds a preset threshold, the electrochemical corrosion branch is corrected, such as... Figure 2 As shown, it includes: When the residual shows a continuously expanding positive divergence trend and exceeds the preset threshold, it is determined that the actual chloride ion penetration rate and water molecule diffusion rate inside the cable have exceeded the theoretical benchmark setting of the electrochemical corrosion branch. Determine the first weighted parameter characterizing the chloride ion diffusion rate and the second weighted parameter characterizing the water molecule condensation rate in the electrochemical corrosion branch; With the goal of minimizing the residual, the measured dielectric loss current of the shielding layer is used as a monitoring signal to calculate the adjustment direction of the residual on the first weighting parameter and the second weighting parameter, respectively. The first and second weight parameters are incrementally updated according to the adjustment direction. The updated first weight parameter does not exceed the physical upper limit of the chloride ion diffusion coefficient of aluminum alloy in seawater environment, and the updated second weight parameter does not exceed the theoretical maximum condensation rate corresponding to the current ambient humidity. When the updated first or second weight parameter reaches the corresponding physical upper limit and the residual has not yet converged, the synchronous correction signal of the mechanical damage branch is triggered.

[0083] In this embodiment of the invention, the preset threshold is initially set to 2 to 5, preferably 3, because based on the historical residual statistics of the normal aging stage of the cable, the mean of the normal residual plus three times the standard deviation usually does not exceed 3. Therefore, a residual exceeding 3 can be judged as abnormal. If the residual exceeds the preset threshold, the correction of the electrochemical corrosion branch is triggered.

[0084] Furthermore, when the residual exhibits a continuously expanding positive divergence trend and exceeds a preset threshold, it is determined that the actual chloride ion penetration rate and water molecule diffusion rate inside the cable have exceeded the theoretical benchmark settings of the electrochemical corrosion branch. The theoretical benchmark settings of the electrochemical corrosion branch refer to the set of initial parameters set for chloride ion penetration rate, water molecule diffusion rate, liquid film coverage, corrosion product accumulation, and passivation inhibition in this branch under standard environmental conditions and initial material conditions. Standard environmental conditions are: conductor temperature 25℃, relative humidity 75%, salt spray concentration 5mg / m³, no mechanical vibration, and an intact oxide film on the conductor surface. The initial values ​​of the first weighting parameter and the second weighting parameter are both set to 1, and the benchmark chloride ion diffusion coefficient is set to 5×10⁻⁶. -12 The baseline values ​​are m² / s, water condensation rate (0.3), effective liquid film coverage (0.6), thermodynamic competition constraint (1), corrosion product accumulation rate (0.0005 h / h), and passivation inhibition factor (1). Under standard conditions, the corrosion rate is low but measurable; a weighting parameter of 1 facilitates subsequent incremental updates. Too high a baseline value would cause the initial model to over-predict corrosion, while too low a value would make the initial model insensitive to corrosion. The first weighting parameter characterizes the chloride ion diffusion rate, and the second weighting parameter characterizes the water condensation rate.

[0085] When the residual shows a continuously expanding positive divergence trend and exceeds the preset threshold, it indicates that the measured dielectric loss current of the shielding layer is continuously greater than the reference dielectric loss current, and the deviation increases window by window. In the absence of a synchronous increase in the dynamic fracture area ratio, the positive divergence indicates that the conductor corrosion degree output by the electrochemical corrosion branch is low, and the actual chloride ion penetration rate or water molecule diffusion rate is higher than the theoretical benchmark. Specific manifestations include an actual chloride ion diffusion rate higher than the reference chloride ion diffusion coefficient, an actual water molecule condensation rate higher than the reference water molecule condensation rate, an actual effective liquid film coverage rate higher than the reference effective liquid film coverage rate, an actual corrosion product accumulation rate higher than the reference corrosion product accumulation rate, or an actual passivation inhibition factor lower than the reference passivation inhibition factor.

[0086] The continuously expanding positive divergence trend is determined by the residuals of multiple consecutive time windows. The number of consecutive windows is 3 to 5, preferably 3. The residual of each window exceeds a preset threshold, which is 2 to 5, preferably 3, and the residual of the subsequent window is greater than that of the previous window, with a residual increment greater than 0.1 to 0.2, preferably 0.15. The reason for setting this number and increment is that exceeding the limit in a single window may be caused by instantaneous interference from the sensor or load fluctuations, and increasing the residual increment over three consecutive windows can eliminate random noise. The residual increment threshold is used to confirm that the actual corrosion rate is consistently higher than the model baseline.

[0087] Furthermore, with the goal of minimizing the residual, the measured dielectric loss current of the shielding layer is used as a monitoring signal to calculate the adjustment direction of the residual with respect to the first and second weighting parameters. The adjustment direction is determined by the partial derivative of the residual with respect to the weighting parameters. The partial derivative of the residual with respect to the first weighting parameter = residual × partial derivative of the reference dielectric loss current with respect to the first weighting parameter. The partial derivative of the residual with respect to the second weighting parameter = residual × partial derivative of the reference dielectric loss current with respect to the second weighting parameter. The first and second weighting parameters are incrementally updated according to the adjustment direction. The update step size is set to 0.01 to 0.1, preferably 0.05, because a step size that is too small results in slow correction speed, while a step size that is too large can easily cause oscillations.

[0088] Furthermore, the updated first weighting parameter does not exceed the physical upper limit of the chloride ion diffusion coefficient of aluminum alloys in seawater environments. This physical upper limit is obtained through a normalized mapping of the baseline diffusion coefficient. The upper limit of the chloride ion diffusion coefficient of aluminum alloys in seawater environments is set to 1 × 10⁻⁶. -10 m² / s to 1×10 -9 m² / s, preferably 5×10 -10The m² / s range covers the common upper limit of chloride ion diffusion coefficient for aluminum alloys in marine environments; exceeding this range violates the laws of material physics. The updated second weighting parameter does not exceed the theoretical maximum condensation rate corresponding to the current ambient humidity. The theoretical maximum condensation rate is calculated based on the ambient relative humidity and conductor temperature. When the ambient relative humidity is 95%, the theoretical maximum condensation rate is set to 1. The principle behind setting this upper limit is that the condensation rate of water molecules cannot exceed the saturated condensation capacity under the current humidity conditions.

[0089] When the updated first or second weight parameter reaches the corresponding physical upper limit and the residual still has not converged, the synchronous correction signal for the mechanical damage branch is triggered. Synchronous correction of the mechanical damage branch is achieved by adjusting the dynamic fracture area ratio output scaling factor. The initial value of the output scaling factor is set to 1, and the adjustment step size is set to 0.05 to 0.2, preferably 0.1. The adjustment direction is set to increase the dynamic fracture area ratio, thereby increasing the nonlinear amplification multiplier, increasing the effective salt spray concentration, and bringing the reference dielectric loss current closer to the measured shielding layer dielectric loss current. The principle behind this synchronous correction is that when the weight parameter of the electrochemical corrosion branch reaches the physical upper limit and still cannot explain the residual, it indicates that the mechanical damage branch underestimates the degree of oxide film rupture, and the input needs to be corrected from the mechanical damage side.

[0090] This invention employs a baseline dielectric loss current calculation, contribution separation, residual normalization, and branch correction scheme. By mapping oxide film rupture to interface polarization branch parameters and conductor corrosion to porous dielectric leakage conductance parameters through equivalent insulation impedance physical topology, it can separate mechanical and electrochemical damage contributions from in-phase active components. Measurement noise interference is eliminated through sliding window and fluctuation standard deviation normalization. Electrochemical corrosion branches are corrected through residual trend judgment and weight updates under physical upper limit constraints. Synchronous correction of mechanical damage branches is triggered when the weight reaches the physical upper limit, thereby improving the physical consistency, noise resistance, and long-term reliability of aging prediction for offshore wind power aluminum alloy power cables.

[0091] S400: Based on the corrected dual-branch physical constraint time-series network, outputs the aging prediction results of aluminum alloy power cables.

[0092] Currently, in the aging prediction of aluminum alloy power cables for offshore wind power, after correcting for the electrochemical corrosion and mechanical damage branches, the prediction results are often directly output without recalculating the baseline dielectric loss current or verifying whether the residuals converge. Insufficient correction results in predictions that deviate from reality, while excessive correction may introduce false aging signals, leading to unreliable output results. Therefore, how to recalculate the baseline dielectric loss current and update the residuals after correction, and output aging prediction results when the residuals converge to within a preset threshold, has become a pressing technical problem that needs to be solved.

[0093] In this embodiment of the invention, the reference dielectric loss current is recalculated based on the corrected dual-branch physical constraint time-series network. The corrected first weight parameter, second weight parameter, and adjusted dynamic fracture area ratio output scaling factor are substituted into each branch to re-obtain the conductor surface oxide film fracture degree and conductor corrosion degree. These are then mapped to incremental parameters in the equivalent insulation impedance physical topology to calculate the updated reference dielectric loss current. The residual between the measured shielding layer dielectric loss current and the reference dielectric loss current is recalculated within the same time window and normalized as described above to obtain the updated residual. When the updated residual converges to within a preset threshold, the aging prediction result of the aluminum alloy power cable is output. The aging prediction result includes the conductor surface oxide film fracture degree, conductor corrosion degree, mechanical damage contribution, and electrochemical damage contribution. If convergence is not achieved, iterative correction continues, or a prediction result with a low-confidence label is output after reaching the maximum number of iterations.

[0094] This invention recalculates the reference dielectric loss current and updates the residual after correction. It uses a preset threshold and continuous window convergence condition to determine whether the correction is sufficient. After convergence, it outputs aging prediction results including the degree of oxide film rupture, conductor corrosion degree, and contribution. If convergence fails, it continues to iterate or outputs a low confidence label, forming a closed-loop correction and verification mechanism. This avoids prediction bias caused by insufficient or excessive correction and improves the accuracy, robustness, and interpretability of aging prediction for offshore wind power aluminum alloy power cables.

[0095] In summary, the embodiments of the present invention have at least the following technical effects: This invention constructs a dual-branch physically constrained time-series network by synchronously acquiring time-aligned, frequency-domain consistent, and physically meaningful mechanical vibration sequences, electrothermal humidity sequences, and shielding layer dielectric loss current. This network couples the mechanical damage branch and the electrochemical corrosion branch across channels. It utilizes the dynamic fracture area ratio as a nonlinear amplification multiplier for chloride ion permeation channels and integrates positive temperature acceleration, moisture evaporation suppression, and the dense passivation effect of corrosion products to make the conductor corrosion prediction more consistent with actual operating conditions. Furthermore, it maps the conductor surface oxide film rupture degree and conductor corrosion degree to the interface polarization parameter and porous dielectric leakage conductance parameter in the equivalent insulation impedance physical topology, respectively, achieving the separation and quantification of the contributions of mechanical damage and electrochemical damage. A closed-loop prediction mechanism is formed through fluctuation standard deviation normalization, weight updates under physical upper limit constraints, synchronous correction of the mechanical damage branch, and residual convergence verification. This significantly improves the accuracy, robustness, and interpretability of aging prediction for offshore wind power aluminum alloy power cables and can output the oxide film rupture degree and conductor corrosion degree, providing a reliable basis for cable operation and maintenance and replacement decisions.

[0096] In summary, this invention solves the technical problems of low prediction accuracy caused by inconsistent multi-source data, lack of physical coupling between mechanical and corrosion branches, and unreliable residual correction in the aging prediction of aluminum alloy power cables for offshore wind power.

[0097] Example 2, as Figure 3 As shown, based on the same inventive concept as the aluminum alloy power cable aging prediction method for offshore wind power provided in Embodiment 1, this embodiment of the invention also provides an aluminum alloy power cable aging prediction system for offshore wind power, including: The multi-source data acquisition module 11 is used to acquire the mechanical vibration sequence, electrothermal temperature and humidity sequence, and shielding layer dielectric loss current of the target aluminum alloy power cable. The dual-branch network construction module 12 is used to construct a dual-branch physical constraint temporal network. The dual-branch physical constraint temporal network includes a mechanical damage branch and an electrochemical corrosion branch. The mechanical vibration sequence is input into the mechanical damage branch and the electrothermal temperature and humidity sequence is input into the electrochemical corrosion branch, which outputs the degree of oxide film rupture on the conductor surface and the degree of conductor corrosion, respectively. The reference current correction module 13 is used to calculate the reference dielectric loss current of the aluminum alloy power cable based on the degree of oxide film rupture and conductor corrosion, and to calculate the residual between the reference dielectric loss current and the measured shielding layer dielectric loss current. When the residual exceeds the preset threshold, the electrochemical corrosion branch is corrected. The aging prediction output module 14 is used to output the aging prediction results of aluminum alloy power cables based on the corrected dual-branch physical constraint time series network.

[0098] In one embodiment, the multi-source data acquisition module 11 is specifically used for: The mechanical vibration acceleration signals of the submarine cable monitoring nodes deployed at the seabed contact points of the offshore wind turbine base are acquired to form a high-frequency mechanical vibration sequence. By using a distributed fiber optic temperature measurement system and environmental sensors laid inside the cable, the conductor temperature, relative humidity and salt spray concentration along the line are collected simultaneously to form a low-frequency electrothermal temperature and humidity sequence. The leakage current on the cable grounding wire is collected by a high-frequency current transformer, and the fundamental and higher harmonic components are filtered out. The resistive leakage current in the characteristic frequency band is extracted as the dielectric loss current of the shielding layer.

[0099] In one embodiment, the dual-branch network building module 12 is specifically used for: Constructing a two-branch physically constrained temporal network includes: The mechanical damage branch includes a temporal convolution module, which is used to extract stress fatigue cycle features from the mechanical vibration sequence and outputs a dynamic fracture area ratio that characterizes the proportion of exposed oxide film area on the conductor surface, as the degree of oxide film fracture on the conductor surface. During the feature extraction process, the electrochemical corrosion branch receives the electrothermal temperature and humidity sequence as the main input, and at the same time introduces the dynamic fracture area ratio output by the mechanical damage branch as the cross-branch physical gate signal. The cross-branch physical gate signal is a nonlinear amplification multiplier of the chloride ion permeation channel. In the electrochemical corrosion branch, the dynamic rupture area ratio is used as a nonlinear amplification multiplier of the chloride ion permeation channel to dynamically scale the input salt spray concentration characteristics. This allows for a higher effective salt spray concentration to participate in the reaction under conditions with greater mechanical vibration amplitude, thereby accelerating the cross-channel coupling of mechanical vibration to electrochemical corrosion.

[0100] In the electrochemical corrosion branch, the dynamic fracture area ratio is used as a nonlinear amplification multiplier for chloride ion permeation channels to dynamically scale the input salt spray concentration characteristics. This also includes: The positive acceleration factor of temperature on the chemical reaction rate is determined based on the conductor temperature, and the positive acceleration factor is positively correlated with the conductor temperature. The amount of local liquid water loss due to water evaporation is determined based on the conductor temperature, and the electrolyte participation factor is determined based on the amount of local liquid water loss. The electrolyte participation factor is negatively correlated with the amount of local liquid water loss. Thermodynamic competition constraint results are obtained by fusion calculation based on the positive acceleration factor and the electrolyte participation factor. These results are used to characterize the inhibitory competition effect of high-temperature evaporation on corrosion acceleration. Based on the salt spray concentration and thermodynamic competition constraint results, the cumulative amount of historical corrosion products was obtained through time-cumulative analysis. When the cumulative amount of historical corrosion products has not reached the densification threshold, the degree of conductor corrosion is determined based on the salt spray concentration and thermodynamic competition constraint results. When the accumulated amount of historical corrosion products reaches the density threshold, the passivation inhibition factor is activated in the electrochemical corrosion branch, and the degree of conductor corrosion is determined based on the passivation inhibition factor, salt spray concentration, and thermodynamic competitive constraint results.

[0101] In one embodiment, the reference current correction module 13 is specifically used for: The reference dielectric loss current of aluminum alloy power cables is calculated based on the degree of oxide film rupture and conductor corrosion on the conductor surface, including: Obtain the nominal operating voltage of the aluminum alloy power cable; The physical topology of equivalent insulation impedance for aluminum alloy power cables is constructed. The physical topology of equivalent insulation impedance includes parallel interface polarization parasitic capacitance branches and porous dielectric leakage current conduction branches. The degree of oxide film rupture on the conductor surface output by the mechanical damage branch is mapped to the capacitance increment parameter of the interface polarization parasitic capacitance branch. The degree of conductor corrosion output from the electrochemical corrosion branch is mapped to the conductivity increment parameter of the porous medium leakage conductivity branch; The nominal operating voltage is applied across the equivalent insulation impedance physical topology. The total AC leakage current flowing through the interface polarization parasitic capacitance branch and the porous dielectric leakage conduction branch is calculated. The active component that is in phase with the nominal operating voltage is extracted and used as the reference dielectric loss current of the aluminum alloy power cable.

[0102] The degree of oxide film rupture on the conductor surface output by the mechanical damage branch is mapped to the capacitance increment parameter of the interface polarization parasitic capacitance branch, including: The proportion of exposed oxide film area is determined based on the degree of oxide film rupture on the conductor surface. The degree of enhancement of the interfacial polarization effect is determined based on the proportion of exposed oxide film area, and the capacitance increment parameter is positively correlated with the proportion of exposed oxide film area. Among them, mapping the conductor corrosion degree output by the electrochemical corrosion branch to the conductivity increment parameter of the porous medium leakage conductivity branch includes: The depth of porous corrosion is determined based on the degree of conductor corrosion. The degree of increase in electrolyte connectivity pathways in porous media is determined by the depth of porous corrosion, and the conductivity increment parameter is positively correlated with the depth of porous corrosion.

[0103] Under the control condition of keeping the incremental conductivity parameter of the porous dielectric leakage current branch zero, the contribution of the incremental capacitance parameter to the loss current of the reference dielectric is determined as the contribution of mechanical damage. Under the control condition of keeping the capacitance increment parameter of the interfacial polarization parasitic capacitance branch zero, the contribution of the conductivity increment parameter to the reference dielectric loss current is determined as the electrochemical loss contribution. The mechanical and electrochemical damage contributions are output to quantify the absolute proportion of the impact of mechanical vibration and wet heat corrosion on cable aging and deterioration under current complex sea conditions.

[0104] Calculate the residual between the reference dielectric loss current and the measured shield dielectric loss current, including: The measured dielectric loss current sequence of the shielding layer and the predicted baseline dielectric loss current sequence are extracted using a sliding time window. The error base is determined based on the degree of deviation between the shielding layer dielectric loss current sequence and the reference dielectric loss current sequence within the same time window; The standard deviation of the dielectric loss current sequence of the shielding layer within the time window is extracted as the uncertainty coefficient. The error base is normalized according to the uncertainty coefficient to obtain the true physical deviation after eliminating measurement noise interference, which is used as the residual.

[0105] When the residual exceeds a preset threshold, the electrochemical corrosion branch is corrected, including: When the residual shows a continuously expanding positive divergence trend and exceeds the preset threshold, it is determined that the actual chloride ion penetration rate and water molecule diffusion rate inside the cable have exceeded the theoretical benchmark setting of the electrochemical corrosion branch. Determine the first weighted parameter characterizing the chloride ion diffusion rate and the second weighted parameter characterizing the water molecule condensation rate in the electrochemical corrosion branch; With the goal of minimizing the residual, the measured dielectric loss current of the shielding layer is used as a monitoring signal to calculate the adjustment direction of the residual on the first weighting parameter and the second weighting parameter, respectively. The first and second weight parameters are incrementally updated according to the adjustment direction. The updated first weight parameter does not exceed the physical upper limit of the chloride ion diffusion coefficient of aluminum alloy in seawater environment, and the updated second weight parameter does not exceed the theoretical maximum condensation rate corresponding to the current ambient humidity. When the updated first or second weight parameter reaches the corresponding physical upper limit and the residual has not yet converged, the synchronous correction signal of the mechanical damage branch is triggered.

[0106] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0107] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0108] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. A method for predicting the aging of aluminum alloy power cables for offshore wind power, characterized in that, include: Obtain the mechanical vibration sequence, electrothermal temperature and humidity sequence, and shielding dielectric loss current of the target aluminum alloy power cable; A dual-branch physical constraint time-series network is constructed, which includes a mechanical damage branch and an electrochemical corrosion branch. The mechanical vibration sequence is input into the mechanical damage branch, and the electrothermal temperature and humidity sequence is input into the electrochemical corrosion branch. The degree of oxide film rupture on the conductor surface and the degree of conductor corrosion are output respectively. The reference dielectric loss current of the aluminum alloy power cable is calculated based on the degree of oxide film rupture and the degree of conductor corrosion. The residual between the reference dielectric loss current and the measured dielectric loss current of the shielding layer is calculated. When the residual exceeds a preset threshold, the electrochemical corrosion branch is corrected. Based on the corrected dual-branch physical constraint time-series network, the aging prediction results of the aluminum alloy power cable are output.

2. The method for predicting the aging of aluminum alloy power cables for offshore wind power according to claim 1, characterized in that, Obtain the mechanical vibration sequence, electrothermal temperature and humidity sequence, and shielding dielectric loss current of the target aluminum alloy power cable, including: The mechanical vibration acceleration signals of the submarine cable monitoring nodes deployed at the seabed contact points of the offshore wind turbine base are acquired to form a high-frequency mechanical vibration sequence. By using a distributed fiber optic temperature measurement system and environmental sensors laid inside the cable, the conductor temperature, relative humidity and salt spray concentration along the line are collected simultaneously to form a low-frequency electrothermal temperature and humidity sequence. The leakage current on the cable grounding wire is collected by a high-frequency current transformer, and the fundamental and higher harmonic components are filtered out. The resistive leakage current in the characteristic frequency band is extracted as the dielectric loss current of the shielding layer.

3. The method for predicting the aging of aluminum alloy power cables for offshore wind power according to claim 1, characterized in that, The mechanical damage branch includes a temporal convolution module, which is used to extract stress fatigue cycle features from the mechanical vibration sequence and output a dynamic fracture area ratio that characterizes the proportion of exposed oxide film area on the conductor surface, as the degree of oxide film fracture on the conductor surface. During the feature extraction process, the electrochemical corrosion branch receives the electrothermal temperature and humidity sequence as the main input, and simultaneously introduces the dynamic fracture area ratio output by the mechanical damage branch as a cross-branch physical gate signal. The cross-branch physical gate signal is a nonlinear amplification multiplier of the chloride ion permeation channel. In the electrochemical corrosion branch, the dynamic rupture area ratio is used as a nonlinear amplification multiplier of the chloride ion permeation channel to dynamically scale the input salt spray concentration characteristics, so that the effective salt spray concentration allowed to participate in the reaction is higher under the condition of greater mechanical vibration amplitude, thereby realizing the cross-channel coupling acceleration of mechanical vibration on electrochemical corrosion.

4. The method for predicting the aging of aluminum alloy power cables for offshore wind power according to claim 3, characterized in that, In the electrochemical corrosion branch, the dynamic rupture area ratio is used as a nonlinear amplification multiplier for the chloride ion permeation channel to dynamically scale the input salt spray concentration characteristics, and the method further includes: The positive acceleration factor of temperature on the chemical reaction rate is determined based on the conductor temperature, and the positive acceleration factor is positively correlated with the conductor temperature. The amount of local liquid water loss due to water evaporation is determined based on the conductor temperature, and the electrolyte participation factor is determined based on the amount of local liquid water loss. The electrolyte participation factor is negatively correlated with the amount of local liquid water loss. Thermodynamic competition constraint results are obtained by fusing the positive acceleration factor and the electrolyte participation factor. These results are used to characterize the inhibitory competition effect of high-temperature evaporation on corrosion acceleration. Based on the salt spray concentration and thermodynamic competition constraint results, the cumulative amount of historical corrosion products was obtained through time-cumulative analysis. When the cumulative amount of historical corrosion products has not reached the densification threshold, the degree of conductor corrosion is determined based on the salt spray concentration and thermodynamic competition constraint results. When the accumulated amount of historical corrosion products reaches the density threshold, the passivation inhibition factor is activated in the electrochemical corrosion branch, and the degree of conductor corrosion is determined based on the passivation inhibition factor, salt spray concentration, and thermodynamic competition constraint results.

5. The method for predicting the aging of aluminum alloy power cables for offshore wind power according to claim 1, characterized in that, The reference dielectric loss current of the aluminum alloy power cable is calculated based on the degree of oxide film rupture and the degree of conductor corrosion, including: Obtain the nominal operating voltage of the aluminum alloy power cable; Construct an equivalent insulation impedance physical topology for aluminum alloy power cables, wherein the equivalent insulation impedance physical topology includes parallel interface polarization parasitic capacitance branches and porous dielectric leakage current conduction branches. The degree of oxide film rupture on the conductor surface output by the mechanical damage branch is mapped to the capacitance increment parameter of the interface polarization parasitic capacitance branch. The degree of conductor corrosion output from the electrochemical corrosion branch is mapped to the conductivity increment parameter of the porous medium leakage conductivity branch; The nominal operating voltage is applied across the equivalent insulation impedance physical topology. The total AC leakage current flowing through the interface polarization parasitic capacitance branch and the porous dielectric leakage current branch is calculated. The active component that is in phase with the nominal operating voltage is extracted and used as the reference dielectric loss current of the aluminum alloy power cable.

6. The method for predicting the aging of aluminum alloy power cables for offshore wind power according to claim 5, characterized in that, Mapping the degree of oxide film rupture on the conductor surface output by the mechanical damage branch to the capacitance increment parameter of the interface polarization parasitic capacitance branch includes: The proportion of exposed oxide film area is determined based on the degree of oxide film rupture on the conductor surface. The degree of enhancement of the interface polarization effect is determined based on the proportion of the exposed area of ​​the oxide film, and the capacitance increment parameter is positively correlated with the proportion of the exposed area of ​​the oxide film. The mapping of the conductor corrosion degree output by the electrochemical corrosion branch to the conductivity increment parameter of the porous medium leakage conductivity branch includes: The depth of porous corrosion is determined based on the degree of corrosion of the conductor. The degree of increase in electrolyte connectivity paths in the porous medium is determined based on the porous corrosion depth, and the conductivity increment parameter is positively correlated with the porous corrosion depth.

7. The method for predicting the aging of aluminum alloy power cables for offshore wind power according to claim 5, characterized in that, Also includes: Under the control condition of keeping the conductivity increment parameter of the porous dielectric leakage conductance branch at zero, the contribution of the capacitance increment parameter to the reference dielectric loss current is determined as the mechanical damage contribution. Under the control condition of keeping the capacitance increment parameter of the interface polarization parasitic capacitance branch at zero, the contribution of the conductivity increment parameter to the reference dielectric loss current is determined as the electrochemical loss contribution. The mechanical damage contribution and the electrochemical damage contribution are output to quantify the absolute proportion of the impact of mechanical vibration and wet heat corrosion on cable aging and deterioration under the current complex sea conditions.

8. The method for predicting the aging of aluminum alloy power cables for offshore wind power according to claim 1, characterized in that, Calculating the residual between the reference dielectric loss current and the measured dielectric loss current of the shielding layer includes: The measured dielectric loss current sequence of the shielding layer and the predicted reference dielectric loss current sequence are extracted using a sliding time window. The error base is determined based on the degree of deviation between the shielding layer dielectric loss current sequence and the reference dielectric loss current sequence within the same time window; The standard deviation of the dielectric loss current sequence of the shielding layer within the time window is extracted as the uncertainty coefficient. The error base is normalized according to the uncertainty coefficient to obtain the true physical deviation after eliminating measurement noise interference, which is used as the residual.

9. The method for predicting the aging of aluminum alloy power cables for offshore wind power according to claim 8, characterized in that, When the residual exceeds a preset threshold, the electrochemical corrosion branch is corrected, including: When the residual shows a continuously expanding positive divergence trend and exceeds the preset threshold, it is determined that the actual chloride ion penetration rate and water molecule diffusion rate inside the cable have exceeded the theoretical benchmark setting of the electrochemical corrosion branch. Determine the first weighted parameter characterizing the chloride ion diffusion rate and the second weighted parameter characterizing the water molecule condensation rate in the electrochemical corrosion branch; With the goal of minimizing the residual, the measured dielectric loss current of the shielding layer is used as a monitoring signal to calculate the adjustment direction of the residual on the first weighting parameter and the second weighting parameter, respectively. The first weight parameter and the second weight parameter are incrementally updated according to the adjustment direction. The updated first weight parameter does not exceed the physical upper limit of the chloride ion diffusion coefficient of aluminum alloy in seawater environment, and the updated second weight parameter does not exceed the theoretical maximum condensation rate corresponding to the current ambient humidity. When the updated first or second weight parameter reaches the corresponding physical upper limit and the residual has not yet converged, the synchronous correction signal of the mechanical damage branch is triggered.

10. An aging prediction system for aluminum alloy power cables for offshore wind power, characterized in that, The method for predicting the aging of aluminum alloy power cables for offshore wind power as described in any one of claims 1 to 9 includes: The multi-source data acquisition module is used to acquire the mechanical vibration sequence, electrothermal temperature and humidity sequence, and shielding layer dielectric loss current of the target aluminum alloy power cable. A dual-branch network construction module is used to construct a dual-branch physical constraint temporal network. The dual-branch physical constraint temporal network includes a mechanical damage branch and an electrochemical corrosion branch. The mechanical vibration sequence is input into the mechanical damage branch, and the electrothermal temperature and humidity sequence is input into the electrochemical corrosion branch. The degree of oxide film rupture on the conductor surface and the degree of conductor corrosion are output respectively. The reference current correction module is used to calculate the reference dielectric loss current of the aluminum alloy power cable based on the degree of oxide film rupture and the degree of conductor corrosion, and to calculate the residual between the reference dielectric loss current and the measured dielectric loss current of the shielding layer. When the residual exceeds a preset threshold, the electrochemical corrosion branch is corrected. The aging prediction output module is used to output the aging prediction results of the aluminum alloy power cable based on the corrected dual-branch physical constraint time series network.

Images