Offshore wind power cable line fault early warning method and system

Abstract

The invention discloses an offshore wind power cable line fault early warning method and system, and relates to the technical field of cable line fault early warning, and the method comprises the following steps: laying a three-dimensional distributed optical fiber temperature array along an offshore wind power cable, and monitoring the cable thermal field space distribution in combination with an active acoustic dispersion imaging technology; acquiring initial evolution trajectory data of the edge high-temperature region; and on the basis of the initial evolution trajectory data, constructing a coupling relationship among the ocean current parameter, the sediment disturbance characteristic and the cable heat conduction parameter, forming an ocean current-sediment-heat conduction coupled digital twin model, and dynamically correcting a virtual scene of the cable burial depth in the model through a recursive filtering algorithm. Through combination of thermal field monitoring, digital twin modeling and intelligent intervention, accurate early warning and dynamic response of offshore wind power cable faults are realized, the operation and maintenance efficiency is improved, system self-optimization and cross-scene migration are supported, and the safety and intelligent level of cable operation are significantly enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cable line fault early warning, in particular to a kind of offshore wind power cable line fault early warning method and system. BACKGROUND

[0002] The offshore wind power cable line fault early warning system refers to an intelligent system specially used for monitoring, analyzing and predicting the running state of transmission cable in offshore wind farm. The system collects real-time multi-dimensional parameters such as current, voltage, temperature, partial discharge, vibration, tidal scour of the cable through optical fiber sensing units, electrical monitoring modules and environmental monitoring equipment deployed along the cable, and combines historical operation data to identify and trend predict abnormal features by using big data analysis, machine learning or physical model algorithm. When the system judges that there is potential overheating, insulation aging, mechanical damage or electrical breakdown risk of the cable, it will give an early warning signal in advance to guide the maintenance personnel to take maintenance or isolation measures in time, thereby significantly reducing the risk of sudden power failure and improving the safety, stability and operation efficiency of offshore wind power system.

[0003] The prior art has the following disadvantages: Under strong tidal flow and complex seabed environment conditions, periodic disturbance of seabed sediment structure is easy to occur, which leads to dynamic change of the burial depth state of the laid cable, and further causes nonlinear mutation of the heat dissipation path of the cable. When the local burial depth of the cable is reduced or the exposed area is directly exposed to the fluid disturbance interface, the heat conduction mechanism of the cable will change from the original uniform diffusion mode to a highly asymmetric edge migration process, which causes the heat field distribution center to shift to the edge of the cable. Since the traditional cable temperature monitoring system usually adopts a fixed-point linear layout method, it is difficult to effectively sense the dynamic edge hot spot distribution caused by burial disturbance, which leads to long-term abnormal high temperature state of the local area of the cable, which easily induces the premature aging of the insulation material, and even causes local dielectric breakdown, thereby seriously threatening the long-term operation safety of the cable and the stability of the entire system.

[0004] The above information disclosed in the BACKGROUND section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide a method and system for offshore wind power cable line fault early warning, which utilizes a three-dimensional distributed optical fiber array and acoustic scatter imaging technology to accurately capture the evolution trajectory of high temperature areas, and through the construction of a digital twin model coupled with sea current-sediment-thermal conduction and a thermal potential vector solving mechanism, realizes dynamic modeling and prediction of thermal field risk trends. At the same time, the introduction of a multi-dimensional risk tensor driven robot plug-in backfilling and reinforcement learning based backfilling strategy optimization not only improves the timeliness and accuracy of thermal anomaly intervention, but also significantly reduces the power generation loss caused by load reduction of the wind turbine, and finally through the structured data sedimentation and knowledge graph construction, realizes the continuous self-optimization and cross-wind farm intelligent migration ability of the fault early warning system, to solve the problems in the above background technology.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme: a method for offshore wind power cable line fault early warning, comprising the following steps: S100, a three-dimensional distributed optical fiber temperature array is arranged along the offshore wind power cable, and combined with an active acoustic scatter imaging technology, the cable thermal field spatial distribution is monitored to obtain initial evolution trajectory data of the edge high temperature area; S200, based on the initial evolution trajectory data, a coupling relationship between sea current parameters, sediment disturbance characteristics and cable thermal conduction parameters is constructed to form a digital twin model coupled with sea current-sediment-thermal conduction, and a recursive filtering algorithm is used to dynamically correct the virtual scene of the cable burial depth in the model; S300, based on the thermal field information output by the digital twin model, an adaptive thermal potential vector solver is used to calculate the dynamic drift rate of the thermal field distribution center of gravity of the cable, and an advance load reduction decision curve of the wind turbine power output is generated according to the drift trend; S400, the dynamic change data of the thermal potential vector and the spatial distribution information of the digital twin model are fused to construct a multi-dimensional risk tensor, and according to the change result of the risk tensor, a local robot patrol node deployed on the cable path is triggered to perform plug-in sediment backfilling operation; S500, the local thermal field change result after the backfilling operation is collected, and based on the heat island duration, the cable local temperature change rate and the wind turbine power loss information, a hybrid reinforcement learning algorithm is used to optimize the trigger frequency and backfilling depth of the sediment backfilling; S600, the sensing data, simulation results, load reduction strategy, backfilling response and optimization parameters involved in the fault early warning process are structured and collected, and are deposited into a hierarchical knowledge graph to realize continuous learning optimization of the fault early warning process and rapid deployment migration to new offshore wind farm sites.

[0007] Preferably, step S100 comprises: A three-dimensional distributed optical fiber temperature array composed of main optical fibers, branch optical fibers and plug-in probe optical fibers is laid at key nodes and potential risk areas of the offshore wind power cable path, and is installed by a shipborne remote control laying system; After the temperature array is laid, an active acoustic scattering imaging unit is deployed synchronously to generate a spatial map of sediment disturbance and burial depth change through water acoustic pulse excitation and echo reception; The temperature data and acoustic scattering images are compared to identify high-risk hotspot initiation areas where temperature rise and structural disturbance coexist; A hotspot evolution trajectory database is constructed, and hotspot migration direction, speed and stability indicators are extracted for subsequent risk identification.

[0008] Preferably, step S200 comprises: The temperature peak value change, heat flow diffusion speed and thermal gradient offset direction of the edge high temperature area at different time nodes are extracted, and a three-dimensional joint feature data set is formed in combination with the sediment echo density change in the acoustic scattering imaging data; Based on the current parameters, sediment disturbance characteristics and cable heat conduction parameters, a coupled physical model is established, and the finite element method is used to realize dynamic simulation of the cable heat conduction path in a virtual three-dimensional seabed scene; The extended Kalman filter algorithm is used to recursively correct the cable burial depth state, realizing a two-step cycle of prediction and correction; The corrected burial depth state and thermal field results are fed back to the system for hotspot prediction and strategy formulation.

[0009] Preferably, step S300 comprises: Based on the three-dimensional temperature field data output by the digital twin model, a thermal potential distribution function is constructed and a thermal potential vector field is generated; The thermal potential vector field is spatially integrated and time-differenced using an adaptive thermal potential vector solver to calculate the drift path and rate of the thermal potential distribution center of gravity; The thermal potential drift rate is compared with historical high temperature fault samples to identify risk trends, and a coupling model of wind turbine power output and thermal potential diffusion is constructed; An advance load reduction decision curve is generated and sent to the dispatching system to dynamically adjust the wind turbine power output, realizing hierarchical control based on thermal potential response.

[0010] Preferably, step S400 comprises: The dynamic change data of the thermal potential vector and the spatial distribution information of the digital twin model are fused to form a multi-dimensional fusion data set; A four-dimensional risk tensor is constructed based on the fusion data set, and the risk level is calculated according to the thermal potential drift rate, temperature rise speed, burial depth reduction rate and sea current disturbance frequency in the tensor unit; When the risk tensor meets the early warning threshold, trigger the local robot patrol node to perform navigation positioning and plug-and-play sediment backfilling operation in the target area; After backfilling is completed, the thermal potential change and burial depth feedback data are returned, and it is determined whether to perform secondary backfilling or coordinate with other nodes based on the evaluation results.

[0011] Preferably, step S500 comprises: Collecting temperature time series data and fan power output change data after backfilling operation, and extracting thermal island duration, local cable temperature change rate and power loss information; Constructing a reinforcement learning state space with three types of feedback data as state variables, and taking backfilling trigger frequency and backfilling depth as action space, and using a hybrid reinforcement learning algorithm to optimize the control strategy; Through experience replay and action discretization mechanism to improve learning efficiency, define a reward function that balances thermal island dissipation efficiency and power loss; Synchronize the optimal backfilling strategy parameters learned to the patrol node, and deposit the learning process data to the system knowledge graph.

[0012] Preferably, step S600 comprises: Collecting and structuring the sensor data, simulation results, load reduction strategies, backfilling responses and optimization parameters in the fault early warning process to form an original data asset library; Construct a hierarchical knowledge graph containing a sensing layer, a model layer, a strategy layer and a feedback layer based on domain semantics, and establish semantic associations between attribute edges and instance nodes; Based on the graph neural network algorithm, the strategy nodes and response result nodes are clustered and analyzed to generate an initial strategy set suitable for new wind farm sites; Continuously update the graph content during operation to realize the evolution iteration and cross-site migration application of knowledge.

[0013] A marine wind power cable line fault early warning system, comprising a thermal field monitoring module, a digital twin modeling module, a thermal potential analysis and decision module, a risk identification and intervention module, a backfilling strategy optimization module, and a knowledge graph deposition module; The thermal field monitoring module, a three-dimensional distributed optical fiber temperature array is arranged along the marine wind power cable, and combined with active acoustic diffraction imaging technology, the spatial distribution of the cable thermal field is monitored, and the initial evolution trajectory data of the edge high temperature area is obtained; The digital twin modeling module, based on the initial evolution trajectory data, constructs the coupling relationship between the sea current parameters, the sediment disturbance characteristics and the cable heat conduction parameters, forms a digital twin model of sea current-sediment-heat conduction coupling, and dynamically corrects the virtual scene of cable burial depth in the model through recursive filtering algorithm; The thermal potential analysis decision module calculates the dynamic drift rate of the thermal field distribution center of gravity of the cable based on the thermal field information output by the digital twin model using an adaptive thermal potential vector solver, and generates an advance load reduction decision curve for the fan power output according to the drift trend. The risk identification intervention module fuses the dynamic change data of the thermal potential vector with the spatial distribution information of the digital twin model, constructs a multi-dimensional risk tensor, and triggers the local robot patrol node deployed on the cable path to perform the plug-in sand backfill operation according to the change result of the risk tensor. The backfill strategy optimization module collects the local thermal field change result after the backfill operation, and optimizes the trigger frequency and backfill depth of the sand backfill based on the heat island duration, cable local temperature change rate and fan power loss information through a hybrid reinforcement learning algorithm. The knowledge graph sedimentation module structures and collects the sensing data, simulation results, load reduction strategies, backfill responses and optimization parameters involved in the fault early warning process, and deposits them into a hierarchical knowledge graph, realizing continuous learning optimization and rapid deployment migration of the fault early warning process for new offshore wind farms.

[0014] In the above technical solutions, the present application provides technical effects and advantages: The present application precisely captures the evolution trajectory of the high-temperature area by using a three-dimensional distributed optical fiber array and acoustic scatter imaging technology, and realizes dynamic modeling and prediction of the thermal field risk trend by constructing a digital twin model coupled with sea current-sand-heat conduction and a thermal potential vector solving mechanism. At the same time, the robot plug-in backfill driven by the multi-dimensional risk tensor and the backfill strategy optimization based on reinforcement learning not only improve the timeliness and accuracy of the thermal anomaly intervention, but also significantly reduce the power generation loss caused by the load reduction of the fan. Finally, through structured data sedimentation and knowledge graph construction, the continuous self-optimization and cross-wind farm intelligent migration ability of the fault early warning system are realized, significantly improving the safety, stability and intelligent level of the offshore wind power cable operation. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments or prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0016] Figure 1 The method flowchart of the offshore wind power cable line fault early warning method of the present application.

[0017] Figure 2 The module schematic diagram of the offshore wind power cable line fault early warning system of the present application. DETAILED DESCRIPTION

[0018] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations can be implemented in any

[0019] The present application provides a method for offshore wind power cable line fault early warning as shown in Figure 1 The method for offshore wind power cable line fault early warning comprises the following steps: S100, a three-dimensional distributed optical fiber temperature array is arranged along the offshore wind power cable, and a spatial distribution of the cable thermal field area is monitored in combination with an active acoustic scattering imaging technology to obtain initial evolution track data of an edge high temperature area; For accurate identification of the initial abnormal area of the thermal field, the three-dimensional distributed optical fiber temperature array and the active acoustic scattering imaging technology are cooperatively arranged and operated. The specific implementation steps are as follows: A three-dimensional distributed optical fiber temperature array is arranged at key nodes and potential risk areas of the offshore wind farm cable path. The three-dimensional distributed optical fiber temperature array is composed of a main fiber longitudinally arranged on the surface of the cable outer sheath, a plurality of branch fibers transversely buried above and below and on both sides of the cable, and a deep insertion probe fiber, which can collect temperature change data of different positions around the cable in real time with a centimeter-level resolution, and construct a three-dimensional temperature distribution map of the thermal field. The temperature array arrangement process is performed by a shipborne remote control laying system to ensure that the optical fiber temperature sensor is closely attached to the cable and is least affected by external seawater disturbance. This arrangement not only guarantees the integrity of the data, but also significantly improves the response sensitivity to edge temperature fluctuations.

[0020] After the distributed optical fiber temperature array is arranged, an active acoustic scattering imaging unit is synchronously deployed around the cable. The active acoustic scattering imaging unit works in combination with high-frequency underwater acoustic pulse excitation and array receiver, actively emits acoustic waves to the sand interface and seabed structure area near the cable, and receives the echo signals. The acoustic waves are affected by scattering effects caused by local sand density changes, cable external structure disturbances, and temperature gradients during propagation. Through high-dimensional modeling and time-frequency demodulation of the echo signals, a real-time spatial map of the cable surrounding sand disturbance and burial depth change can be generated. The map and the optical fiber temperature distribution map jointly constitute an "electric-thermal-acoustic" joint spatial perception data set, which enhances the recognition accuracy of early edge abnormal changes.

[0021] The deployed optical fiber temperature data and acoustic scattering imaging results are used to form a dynamic comparison and abnormal trend extraction in the initial stage of the edge high temperature area. By setting a reference thermal field distribution template, combining a sliding window algorithm and a dual-domain discrimination mechanism, areas with temperature deviation greater than the preset threshold and with lateral diffusion trend are marked, and whether the acoustic echo change shows sediment structure disturbance or burial depth reduction phenomenon is analyzed. When the optical fiber temperature monitoring data and acoustic scattering signals simultaneously show "temperature rise + structure change" resonance characteristics in time and space dimensions, the area is defined as a "high-risk hotspot emerging area" and is numbered and prioritized for monitoring.

[0022] Based on the identification results of the above high-risk hotspot emerging area, the system constructs its evolution trajectory database to track the time-space evolution trend of the hotspot in the thermal field distribution. Combined with environmental current parameters, cable load history, and acoustic scattering image evolution, the direction, speed, and stability indicators of the hotspot migration to the edge are further extracted. These indicators are used to train the subsequent thermal field prediction model, and are used as the input parameters of the risk warning system trigger mechanism to realize early identification, trend judgment, and warning activation of the edge hotspot area, laying a high-precision sensing foundation for the entire fault warning process.

[0023] This embodiment combines three-dimensional temperature collection and acoustic space modeling, not only significantly improving the capture efficiency of submarine cable thermal anomaly emerging points, but also realizing structural analysis of the initial evolution behavior of hotspots, providing key data support for subsequent thermal field modeling and intelligent decision-making, and embodying significant technical innovation and engineering practical value.

[0024] S200, based on the initial evolution trajectory data of the edge high temperature area, a coupling relationship between the sea current parameters, sediment disturbance characteristics, and cable heat conduction parameters is constructed to form a sea current-sediment-heat conduction coupled digital twin model, and a recursive filtering algorithm is used to dynamically correct the virtual scene of the cable burial depth in the model; To realize high-precision prediction and dynamic correction of the influence of cable burial depth change on heat conduction path, a multi-physical field coupling relationship between sea current parameters, sediment disturbance characteristics, and cable heat conduction parameters is constructed to form a sea current-sediment-heat conduction coupled digital twin model, and a recursive filtering algorithm is used to dynamically correct the virtual scene of the cable burial depth. The specific implementation steps are as follows: According to the initial evolution trajectory data of the edge high-temperature area, the temperature peak change, thermal gradient offset direction and thermal flow diffusion speed of the edge high-temperature area at different time nodes are extracted, and combined with the corresponding silt echo density change and reflection mode anomaly in the acoustic scattering imaging data, a set of joint feature data sets that can be mapped in three-dimensional space is formed. Subsequently, the local sea current parameters of the edge high-temperature area in the preset time window are obtained, including the flow velocity vector, fluid shear force, sea current disturbance frequency, etc., and together with the structural information of the cable (insulation layer thickness, conductor material, cross-sectional area) as input feature variables.

[0025] Based on the above multi-source input data, a coupled physical relationship model is constructed. In this model, the sea current parameters and the silt disturbance characteristics are connected through the bottom bed shear critical model, the silt disturbance and the burial depth change are coupled through the deposition-erosion evolution equation, and the cable thermal conduction characteristics are modeled through the non-steady-state heat conduction partial differential equation. The triple coupling model uses the finite element method to establish a calculation grid in a virtual three-dimensional seabed scene, and realizes the simulation of the dynamic influence of the burial depth change caused by silt disturbance on the cable thermal conduction path at different time steps.

[0026] In order to improve the adaptability and robustness of the model in real complex environment, a recursive filtering algorithm is introduced to dynamically correct the modeling results. The filter used is an extended Kalman filter, and the state transition equation is based on the difference between the historical simulation results and the measured thermal field data to update the estimated value of the cable burial depth scene in each time period. Specifically, after receiving the new optical fiber temperature distribution and acoustic scattering spectrum, the system automatically performs a two-step prediction-correction process to predict the current burial depth state and then correct the prior estimate with a new round of sensor observations, thereby realizing high-frequency dynamic fitting of the cable burial depth virtual scene.

[0027] After each recursive correction, the coupled digital twin model feeds back the updated cable burial depth state and thermal field simulation results to the system to guide subsequent hot spot drift trend prediction and intervention response strategy formulation. At the same time, the evolution law between the sea current disturbance pattern and the silt response behavior in the model is continuously recorded to enrich the virtual scene library and improve the system's migration ability in other wind power scenarios. Through the above multi-layer coupled modeling and dynamic correction process, high precision, high dynamics and high generalization ability are realized in the cable thermal environment modeling, providing a solid data foundation and evolution support for the trend identification and precise intervention of cable faults, and significantly improving the intelligence and engineering applicability of the entire fault warning system.

[0028] S300, based on the thermal field information output by the digital twin model, the dynamic drift rate of the thermal field distribution center of gravity of the cable is calculated using an adaptive thermal potential vector solver, and an advance load reduction decision curve for wind turbine power output is generated according to the drift trend; To realize the trend perception and risk response control of the cable thermal abnormal area, on the basis of the dynamic thermal field information output by the digital twin model, an adaptive thermal potential vector solver is introduced to accurately calculate the dynamic drift rate of the thermal field distribution center of gravity of the cable, and a pre-load reduction decision curve of the fan power output is constructed accordingly, so as to realize intelligent and pre-operational operation control. The specific implementation steps include the following four stages: Based on the three-dimensional temperature field data output by the digital twin model, the temperature spatial distribution of the entire cable at each time node is extracted, and a thermal potential distribution function is constructed through the heat density weighting method. The thermal potential distribution function is based on the temperature value in each calculation unit and the heat capacity parameter of the cable material, forming a continuous thermal potential field expression. Through mathematical modeling, the thermal potential field is mapped into a three-dimensional vector tensor, where the direction of the thermal potential vector at each point represents the trend of heat diffusion, and the vector size reflects the heat energy concentration degree.

[0029] The adaptive thermal potential vector solver is used to perform spatial integration and time series difference analysis on the above thermal potential field. The solver uses a local guidance algorithm based on convolution kernel fitting, which can dynamically adjust the sampling window and weight function of the thermal potential vector according to the disturbance intensity of the cable surrounding environment and the heat field change speed, avoiding the problem of insufficient recognition ability of traditional fixed grid method for mutation points and edge features. In each sampling period, the system calculates the drift path of the thermal potential field distribution center of gravity through the solver and generates a time series form of the thermal potential center of gravity trajectory, thereby obtaining the drift rate and direction change trend of the thermal potential center of gravity.

[0030] After obtaining the dynamic drift rate of the thermal potential center of gravity, the system compares and analyzes it with the high-temperature fault samples in the cable operation history to identify whether the current drift rate has reached the risk critical interval. If the thermal potential center of gravity continuously moves towards the edge of the cable and shows an accelerating trend, the system automatically enters the risk response calculation stage. Based on the empirical heat capacity model and the cable load thermal balance formula, a function relationship describing the influence of thermal potential diffusion trend on temperature rise rate is constructed, and a coupling model between fan power output and temperature rise is further derived. Based on this, the system automatically generates a pre-load reduction decision curve, which specifically includes power reduction gradient, time window length and load control threshold.

[0031] The load reduction decision curve is sent to the wind farm dispatching system, and is connected with the SCADA system to realize signal docking. On the premise of meeting the safe operation condition, the power output of the corresponding wind turbine group is dynamically adjusted, and the load of the wind turbine on the hot spot drift path is preferentially reduced. At the same time, the load reduction strategy has the ability of backtracking and self-adaptive adjustment, and the strategy parameters can be adjusted according to the real-time feedback results of the thermal potential field, so as to realize hierarchical response and gradual recovery. Through the embodiment, not only the cable insulation failure caused by local high temperature diffusion can be effectively prevented, but also the economic loss caused by emergency power-off or unplanned shutdown can be maximally reduced, the dynamic balance between the operation safety and the economy of the whole wind power system is ensured, and the embodiment has outstanding engineering innovation and intelligent control value.

[0032] The SCADA system (Supervisory Control and Data Acquisition) is a comprehensive automation system for remotely monitoring equipment in an industrial process, collecting real-time data, automatically controlling and visually managing the running state, and is widely used in the fields of power, water, oil and other infrastructure.

[0033] S400, the dynamic change data of the thermal potential vector is fused with the spatial distribution information of the digital twin model, a multi-dimensional risk tensor is constructed, and a local robot patrol node deployed in the cable path is triggered according to the change result of the risk tensor to perform the plug-in type silt backfilling operation; In order to realize the structured identification and active response control of the hot spot diffusion risk, the dynamic change data of the thermal potential vector is fused and analyzed with the spatial distribution information of the digital twin model, a multi-dimensional risk tensor is constructed, and a local robot patrol node deployed in the cable path is intelligently triggered based on the tensor evolution trend to perform the plug-in type silt backfilling operation. The specific implementation steps of the process are as follows: Based on the cable thermal field distribution gravity center drift path and speed data obtained by the thermal potential vector solver, the evolution trajectory data set of the thermal potential vector field in the time dimension is constructed, and is taken as a dynamic input variable. At the same time, the spatial distribution information at the corresponding time node is extracted from the sea current-silt-thermal conduction coupled digital twin model, including the local burial depth state, the silt density gradient, the sea current disturbance intensity, the seabed shear force distribution and the like. The above-mentioned thermal potential change and spatial scene information are rasterized through a unified spatial index system to form a time-space consistent multi-dimensional fusion data set.

[0034] Based on the multi-source fusion data set, a multi-dimensional risk tensor is constructed. The risk tensor is a four-dimensional tensor structure, with dimensions corresponding to spatial position, time node, heat potential change and burial depth disturbance intensity. The risk score of the region is calculated in each tensor cell. The scoring model uses a weighted aggregation method, with heat potential drift rate, temperature rise speed, burial depth reduction rate, and sea current disturbance frequency as input factors. The discriminant function outputs the risk level (e.g. low, medium, high, critical) through training. By continuously monitoring the risk evolution trend of each cell in the tensor, the system can determine whether the hotspot region is in a superimposed risk state of continuous temperature rise and continuous burial depth reduction.

[0035] When the risk tensor evolution trend meets the early warning threshold condition, the system automatically generates a position locking instruction and triggers the local robot patrol node deployed at the key node of the cable path through the communication protocol. The patrol node is a modular autonomous mobile device with the ability to quickly navigate to the specified coordinates, deploy the sand conveying arm, and accurately backfill. After receiving the task, the robot first performs path optimization and navigation positioning, hovers above the target area and starts the fine positioning module, then deploys the plug-in sand backfill device, and accurately implements backfilling in the cable hotspot exposed area according to the set backfill depth, density and slope angle parameters, forming a closed protective layer and effectively blocking the asymmetric dissipation of heat to the outside.

[0036] After the backfilling operation is completed, the robot returns the heat potential change and burial depth state of the operation area to the system center as feedback data, and the system updates and evaluates the risk tensor to determine whether the risk level has decreased and whether there is a residual heat island phenomenon. If there is still a temperature rise trend, the robot can automatically adjust the backfilling parameters for secondary correction backfilling or notify other patrol nodes for collaborative work. The entire process has high autonomy, high response speed and micro-scale intervention precision, significantly improving the intelligent level and real-time safety guarantee capability of cable operation and maintenance, and can operate in a closed loop with the front-end monitoring and subsequent learning modules, demonstrating the system integration and innovation value of the scheme in the active intervention layer.

[0037] S500, collect the local thermal field change results after backfilling operation, based on the heat island duration, cable local temperature change rate and fan power loss information, optimize the trigger frequency and backfill depth of sand backfilling through a hybrid reinforcement learning algorithm; To realize the continuous optimization and systematic self-evolution after sand backfilling operation, a hybrid reinforcement learning algorithm is introduced to dynamically optimize the two key control parameters of backfilling trigger frequency and backfilling depth. This optimization process is based on the local thermal field change results after backfilling operation, combined with the heat island duration, cable local temperature change rate and fan power loss information to establish a multi-dimensional feedback mechanism, thereby achieving the optimal balance between thermal environment recovery effect and system operation efficiency. The specific implementation steps include the following stages: After the completion of the plug-and-play sediment backfilling operation, the three-dimensional distributed optical fiber temperature array deployed in the cable path immediately resumes high-frequency monitoring of the local thermal field. The system collects the temperature time series data of the target area after the backfilling operation in real time, and extracts characteristic indicators representing the trend of thermal anomaly changes, including but not limited to temperature peak decay rate, thermal gradient recovery rate, edge temperature zone diffusion degree, etc. At the same time, the system records the total time experienced by the heat island from being identified as a high-risk state to recovering to a safe temperature threshold, defined as "heat island duration". In addition, the power output change curve of the fan during this stage is also collected synchronously, and the total power loss caused by the unloading operation triggered by backfilling is calculated as an energy cost evaluation index of thermal intervention.

[0038] A reinforcement learning state space is constructed with "heat island duration", "cable local temperature change rate" and "fan power loss" as state variables, and "backfill trigger frequency" and "backfill depth" as action space. On this basis, a hybrid reinforcement learning algorithm is introduced, combining deep Q learning (DQN) and policy gradient method (Policy Gradient), to construct an intelligent agent model with dual channels of policy optimization and value evaluation. The system automatically inputs the collected data as a learning sample after each backfilling operation to train the model's ability to judge the causal relationship between different behavior combinations and environmental responses.

[0039] To accelerate learning efficiency and improve generalization ability in complex marine environments, the experience replay mechanism and priority sampling mechanism are used to make the model preferentially learn backfill behavior trajectories with significant state changes during training. In addition, to overcome the dimensionality explosion problem of continuous state space and action space, the action discretization strategy is adopted, dividing the backfill depth into several standard levels (such as 5 cm, 10 cm, 15 cm, etc.), and the trigger frequency is quantified in hours. The reward function is defined as "heat island dissipation efficiency improvement rate minus unit power loss value", so that the reinforcement learning system has a clear goal and enhanced convergence.

[0040] After several rounds of backfill-feedback-learning cycles, the system obtains a set of optimal backfill strategy parameters under the current seabed environment, cable load characteristics and thermal response rules, including backfill response thresholds under different risk levels, recommended values of sediment thickness for each operation, and response strategies when continuous heat islands appear. The optimization results will be used as the basis for updating the control strategy library and automatically synchronized to the execution module of subsequent patrol nodes to realize the inheritance and sharing of strategy knowledge. At the same time, all state transitions, action decisions and reward records in the reinforcement learning process are structured as structured data and incorporated into the system knowledge graph to support strategy migration for subsequent new sites. Through this implementation, not only is the response efficiency and energy economy of the sand backfill intervention significantly improved, but also an intelligent thermal intervention system with self-optimization ability is constructed, truly realizing the transition from "passive operation and maintenance" to "self-adaptive prevention".

[0041] S600, the sensing data, simulation results, load reduction strategies, backfill responses and optimization parameters involved in the fault warning process are structured and collected, and are deposited in a hierarchical knowledge graph, realizing continuous learning optimization of the fault warning process and rapid deployment and migration of new offshore wind farms; To realize the long-term evolution, self-learning and cross-scene adaptation of the fault warning process, it is proposed to structure and collect multiple types of core data involved in the whole process, and to build a multi-level and extensible knowledge graph, realizing continuous optimization of the fault warning process and rapid migration and deployment for new offshore wind farms. The specific implementation steps of this part include the following stages: During the operation of the fault warning system, data output from each sub-module is continuously collected and collected, including temperature time series data collected by the three-dimensional distributed optical fiber temperature array, sand disturbance maps generated by the active acoustic imaging unit, dynamic simulation results output by the digital twin model, thermal potential drift path and rate calculated by the adaptive thermal potential vector solver, power control curve generated by the load reduction strategy engine, and optimal backfill parameter set (trigger frequency, backfill depth, backfill response time) formed in the hybrid reinforcement learning module. All data are labeled, cleaned and structured according to unified data standards and formats to form a raw data asset library for semantic modeling.

[0042] The structured data is abstracted and ontology designed in the field semantics, and a fault early warning hierarchical knowledge graph is constructed. The knowledge graph adopts a multi-layer structure, which is divided into a "sensing layer", a "model layer", a "strategy layer" and a "feedback layer": the sensing layer describes various environmental and cable state parameters and their relationships; the model layer encapsulates the thermal-force-flow multi-physical field coupling mechanism and the digital twin framework logic; the strategy layer covers the load reduction control logic, intervention trigger conditions and response instructions; the feedback layer is used to record the system response performance and optimization results after various intervention behaviors. The association between layers is realized through the definition of clear attribute edges and instance nodes, which can support upstream and downstream semantic reasoning and knowledge recall.

[0043] After the knowledge graph is built, the feature embedding and similarity clustering analysis of various strategy nodes and response result nodes in the graph are performed based on the graph neural network algorithm (such as GraphSAGE, GAT), so as to mine the potential laws and strategy evolution paths under different fault scenarios. The system builds an initial strategy set that can adapt to new environments, and when a new offshore wind farm site is deployed, by inputting its basic geographic environment parameters, tidal flow distribution characteristics and preliminary cable layout, the most similar historical scenario can be quickly matched in the knowledge graph, and the corresponding model parameters and strategy templates can be migrated, greatly reducing the initial configuration cost and response delay.

[0044] During the operation of the system, the whole process of perception, judgment, intervention and feedback experienced by each fault early warning event will be automatically archived and converted into new knowledge fragments in the graph, realizing the continuous iteration and intelligent expansion of the graph content. Through the "perception-modeling-decision-feedback" closed-loop learning ability of the graph, the system not only has self-adaptation ability in a single site, but also has knowledge transfer ability and strategy promotion ability between cross-regional sites, forming an intelligent knowledge hub for the wide-area distributed scenario of offshore wind farms. This implementation significantly improves the intelligence level and deployment efficiency of the fault early warning system, and has high creativity and industry scalability.

[0045] The offshore wind power cable line fault early warning method provided by the application can realize intelligent perception, prediction and intervention control of the dynamic change of cable burial depth caused by seabed silt disturbance and the asymmetric evolution process of the heat field caused thereby. The method breaks through the limitation of traditional cable temperature monitoring means that can only perceive temperature changes at fixed positions and cannot identify the drift of hot spot edges, accurately captures the evolution trajectory of high temperature areas by using a three-dimensional distributed fiber array and acoustic tomography technology, and realizes dynamic modeling and prediction of the risk trend of the heat field by constructing a digital twin model coupled with sea current-silt-heat conduction and a thermal potential vector solving mechanism. At the same time, the introduction of a multi-dimensional risk tensor driven robot plug-in backfilling and a backfilling strategy optimization based on reinforcement learning not only improves the timeliness and accuracy of the thermal anomaly intervention, but also significantly reduces the power generation loss caused by the load reduction of the wind turbine. Finally, through the sedimentation of structured data and the construction of a knowledge graph, the system realizes continuous self-optimization and cross-wind farm intelligent migration capability, significantly improving the safety, stability and intelligent level of offshore wind power cable operation.

[0046] The application provides an offshore wind power cable line fault early warning system as shown in Figure 2 The application provides an offshore wind power cable line fault early warning system as shown in The heat field monitoring module is used for laying a three-dimensional distributed fiber temperature array along the offshore wind power cable, and combining with an active acoustic tomography technology, monitoring the spatial distribution of the cable heat field, and acquiring initial evolution trajectory data of the edge high temperature area. The digital twin modeling module is used for constructing a coupling relationship between sea current parameters, silt disturbance characteristics and cable heat conduction parameters based on the initial evolution trajectory data, forming a digital twin model coupled with sea current-silt-heat conduction, and dynamically correcting the virtual scene of the cable burial depth in the model through a recursive filtering algorithm. The thermal potential analysis and decision module is used for calculating the dynamic drift rate of the heat field distribution center of gravity of the cable by using an adaptive thermal potential vector solver based on the heat field information output by the digital twin model, and generating an advance load reduction decision curve of the wind turbine power output according to the drift trend. The risk identification and intervention module is used for fusing the dynamic change data of the thermal potential vector and the spatial distribution information of the digital twin model, constructing a multi-dimensional risk tensor, and triggering a local robot patrol node deployed in the cable path to perform plug-in silt backfilling operation according to the change result of the risk tensor. The backfilling strategy optimization module is used for collecting the local heat field change result after the backfilling operation, and optimizing the trigger frequency and backfilling depth of the silt backfilling by a hybrid reinforcement learning algorithm based on the heat island duration, the local temperature change rate of the cable and the wind turbine power loss information. The knowledge graph sinking module structures and collects the sensing data, simulation results, load reduction strategies, backfill responses and optimization parameters involved in the fault early warning process, and sinks them into a hierarchical knowledge graph, so as to realize continuous learning optimization of the fault early warning process and rapid deployment and migration to new offshore wind farm sites.

[0047] The offshore wind power cable line fault early warning method provided by the embodiment of the application is realized by the offshore wind power cable line fault early warning system, and the specific method and process of the offshore wind power cable line fault early warning system are described in the embodiment of the offshore wind power cable line fault early warning method, which will not be described here again.

[0048] The above only describes some exemplary embodiments of the application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the application. Therefore, the above drawings and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the application.

[0049] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0050] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here again.

[0051] In several embodiments provided in the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0052] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0053] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0054] If the functions are realized in the form of software functional units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the part of the present application that essentially contributes to the prior art or the part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0055] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for early warning of faults in offshore wind power cable lines, characterized in that, Includes the following steps: S100: Deploy a three-dimensional distributed fiber optic temperature array along the offshore wind power cable and combine it with active acoustic dispersion imaging technology to monitor the spatial distribution of the cable's thermal field and obtain the initial evolution trajectory data of the high-temperature edge region. S200. Based on the initial evolution trajectory data, the coupling relationship between ocean current parameters, sediment disturbance characteristics and cable thermal conduction parameters is constructed to form a digital twin model of ocean current-sediment-thermal conduction coupling. The virtual scene of cable burial depth in the model is dynamically corrected through a recursive filtering algorithm. S300: Based on the thermal field information output by the digital twin model, the dynamic drift rate of the centroid of the cable thermal field distribution is calculated using an adaptive thermal potential vector solver, and an early load reduction decision curve for the power output of the fan is generated according to the drift trend. S400: The dynamic change data of the thermal potential vector is fused with the spatial distribution information of the digital twin model to construct a multidimensional risk tensor. Based on the change results of the risk tensor, the local robot inspection nodes deployed along the cable path are triggered to perform plug-in silt backfilling operations. S500 collects local thermal field change results after backfilling operations. Based on the heat island duration, local temperature change rate of cables, and wind turbine power loss information, it optimizes the trigger frequency and backfilling depth of silt backfilling through a hybrid reinforcement learning algorithm. S600 collects and stores the sensor data, simulation results, load reduction strategies, backfill responses and optimization parameters involved in the fault early warning process in a structured manner, and deposits them into a hierarchical knowledge graph to realize continuous learning and optimization of the fault early warning process and rapid deployment and migration of new offshore wind farm sites.

2. The method for early warning of faults in offshore wind power cable lines according to claim 1, characterized in that, Step S100 includes: A three-dimensional distributed fiber temperature array consisting of trunk fiber, branch fiber and insert probe fiber is deployed at key nodes and potential risk areas along the offshore wind power cable route, and is installed by a shipborne remote-controlled deployment system. After the temperature array is deployed, an active acoustic dispersion imaging unit is deployed simultaneously to generate a spatial map of sediment disturbance and burial depth changes through underwater acoustic pulse excitation and echo reception. By comparing temperature data with acoustic dispersion images, high-risk hotspot incubation areas where temperature rise and structural disturbance coexist can be identified. A database of hotspot evolution trajectories was constructed, and indicators of hotspot migration direction, speed, and stability were extracted for subsequent risk identification.

3. The method for early warning of faults in offshore wind power cable lines according to claim 1, characterized in that, Step S200 includes: The temperature peak changes, heat flow diffusion rate and thermal gradient offset direction of the high-temperature edge region at different time points are extracted, and combined with the changes in sediment echo density in acoustic imaging data to form a three-dimensional joint feature dataset. Based on ocean current parameters, sediment disturbance characteristics, and cable heat conduction parameters, a coupled physical model was established, and the finite element method was used to realize the dynamic simulation of the cable heat conduction path in a virtual three-dimensional seabed scene. An extended Kalman filter algorithm is used to recursively correct the cable burial depth status, realizing a two-step cycle of prediction and correction. The corrected burial depth and thermal field results are fed back to the system for hotspot prediction and strategy formulation.

4. The method for early warning of faults in offshore wind power cable lines according to claim 1, characterized in that, Step S300 includes: Based on the three-dimensional temperature field data output by the digital twin model, a thermal potential distribution function is constructed and a thermal potential vector field is generated; An adaptive thermal potential vector solver is used to perform spatial integration and temporal difference on the thermal potential vector field to calculate the drift path and rate of the centroid of the thermal potential distribution. By comparing the thermal potential drift rate with historical high-temperature fault samples, risk trends are identified, and a coupled model of wind turbine power output and thermal potential diffusion is constructed. The system generates an early load reduction decision curve and sends it to the scheduling system to dynamically adjust the fan power output, thereby achieving hierarchical control based on thermal potential response.

5. The method for early warning of faults in offshore wind power cable lines according to claim 1, characterized in that, Step S400 includes: The dynamic change data of the thermal potential vector is fused with the spatial distribution information of the digital twin model to form a multidimensional fused dataset; A four-dimensional risk tensor is constructed based on the fused dataset, and the risk level is calculated based on the thermal potential drift rate, temperature rise rate, burial depth reduction rate and ocean current disturbance frequency in the tensor unit. When the risk tensor meets the warning threshold, the local robot inspection node is triggered to perform navigation and positioning of the target area and plug-in silt backfilling operation. After backfilling is completed, the thermal changes and burial depth feedback data will be transmitted back, and a decision will be made on whether to backfill a second time or coordinate with other nodes based on the evaluation results.

6. The method for early warning of faults in offshore wind power cable lines according to claim 1, characterized in that, Step S500 includes: Collect temperature time-series data and fan power output change data after backfilling operations, and extract information on heat island duration, local temperature change rate of cables, and power loss. A reinforcement learning state space is constructed with three types of feedback data as state variables, and the backfilling trigger frequency and backfilling depth are used as the action space. A hybrid reinforcement learning algorithm is used to optimize the control strategy. Learning efficiency is improved by using experience playback and action discretization mechanisms, and a reward function is defined that balances heat island dissipation efficiency and power loss. The optimal backfill strategy parameters learned are synchronized to the inspection nodes, and the learning process data is stored in the system knowledge graph.

7. The method for early warning of faults in offshore wind power cable lines according to claim 1, characterized in that, Step S600 includes: Collect and structure sensor data, simulation results, load reduction strategies, backfill responses, and optimization parameters from the fault early warning process to form an original data asset library; A hierarchical knowledge graph containing a sensing layer, a model layer, a strategy layer, and a feedback layer is constructed based on domain semantics, and semantic associations between attribute edges and instance nodes are established. Cluster analysis of strategy nodes and response result nodes is performed based on graph neural network algorithm to generate an initial strategy set suitable for new wind farm sites; The map content is continuously updated during operation to enable knowledge evolution and cross-site migration applications.

8. A fault early warning system for offshore wind power cable lines, used to implement the fault early warning method for offshore wind power cable lines according to any one of claims 1-7, characterized in that, It includes a thermal field monitoring module, a digital twin modeling module, a thermal potential analysis and decision-making module, a risk identification and intervention module, a backfilling strategy optimization module, and a knowledge graph accumulation module; The thermal field monitoring module deploys a three-dimensional distributed fiber optic temperature array along the offshore wind power cable and combines it with active acoustic dispersion imaging technology to monitor the spatial distribution of the cable's thermal field and obtain the initial evolution trajectory data of the high-temperature edge region. The digital twin modeling module, based on the initial evolution trajectory data, constructs the coupling relationship between ocean current parameters, sediment disturbance characteristics and cable thermal conduction parameters, forming a digital twin model of ocean current-sediment-thermal conduction coupling, and dynamically corrects the virtual scene of cable burial depth in the model through a recursive filtering algorithm; The thermal potential analysis and decision module, based on the thermal field information output by the digital twin model, uses an adaptive thermal potential vector solver to calculate the dynamic drift rate of the centroid of the cable thermal field distribution, and generates an early load reduction decision curve for the fan power output based on the drift trend. The risk identification and intervention module integrates the dynamic change data of the thermal potential vector with the spatial distribution information of the digital twin model to construct a multidimensional risk tensor. Based on the change results of the risk tensor, it triggers the local robot inspection nodes deployed along the cable path to perform plug-in silt backfilling operations. The backfill strategy optimization module collects the results of local thermal field changes after backfilling operations. Based on the heat island duration, the rate of local temperature change of the cable, and the power loss information of the wind turbine, it optimizes the trigger frequency and backfill depth of silt backfilling through a hybrid reinforcement learning algorithm. The knowledge graph accumulation module collects and accumulates the sensor data, simulation results, load reduction strategies, backfill responses and optimization parameters involved in the fault early warning process into a hierarchical knowledge graph, enabling continuous learning and optimization of the fault early warning process and rapid deployment and migration of new offshore wind farm sites.

Images