Offshore wind power single pile foundation energy storage-structure-intelligent self-repairing integrated system

Abstract

The invention relates to an offshore wind power single pile foundation energy storage-structure-intelligent self-repairing integrated system which comprises a single pile foundation and a concrete disc-shaped expansion structure. A converter cabin, a honeycomb battery cabin and a counterweight cabin are arranged in the single pile foundation; a plurality of areas are evenly divided in the circumferential direction, and a seawater battery, a fiber grating strain sensor, a piezoelectric actuator, a shape memory alloy driver and a broadband acoustic emission sensor are installed in each area. The truss of the honeycomb battery cabin in each area is connected with the inner wall of the single pile foundation through a prestressed steel strand and an anchorage device, and a resistance wire is arranged on the surface of the shape memory alloy driver; the bacillus microcapsules are doped into the concrete; an impedance instrument probe, a self-priming pump and an automatic control valve are mounted on the seawater battery; an AI framework for outputting a three-dimensional fault probability is integrated in the converter cabin; and triggering fault self-repairing of the corresponding dimension of the corresponding region according to the probability threshold. And full-life-cycle perception-prediction-restoration closed-loop management of the offshore wind turbine is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of offshore wind power, and particularly relates to an integrated energy storage - structure - intelligent self - repair system for offshore wind power monopile foundations. This system integrates energy storage, structural strengthening, AI fault prediction, and intelligent self - repair, and is particularly suitable for the high - reliability and intelligent operation and maintenance requirements of deep - sea and far - sea wind farms. Background Art

[0002] The volatility and randomness of wind power generation result in significant fluctuations in power supply at different times. Especially in offshore wind farms far from onshore power grids, the problems of power transmission and storage are particularly prominent. Existing energy storage technologies (such as pumped - storage energy storage, compressed - air energy storage, etc.) have problems such as large space occupation, complex installation, and high maintenance costs in the marine environment, and are difficult to meet the actual needs of offshore wind farms. As a low - cost, high - efficiency, and sustainable energy storage technology, seawater batteries have a wide range of inexpensive raw material sources, can achieve low battery prices, and are suitable for large - scale applications.

[0003] Monopile foundations have significant problems of insufficient structural reliability under extreme sea conditions. Especially when facing strong winds, waves, and tidal changes, their stability is easily threatened. The health monitoring of offshore wind turbines faces many challenges. Traditional monitoring methods are not only costly but also difficult to cope with the complex and changing marine environment. In deep - sea and far - sea areas, if AI - driven fault prediction technology is used for health monitoring, the monitoring efficiency can be effectively improved and the cost can be reduced.

[0004] In view of the technical pain points of the electrical energy storage system in offshore wind farms, the present invention proposes an innovative solution for the coordination of basic structure and battery function, establishes a coupling structure between the honeycomb battery compartment and the monopile foundation, effectively combines the honeycomb battery compartment with the monopile foundation structure, enables these two parts to cooperate with each other physically and functionally, support each other, and work together, reduces the separated layout between the energy storage system and the basic structure in traditional wind farms, reduces the vibration transmission between the wind turbine unit and the energy storage system, improves the operation efficiency and safety of the overall system, and realizes integrated design and dynamic response optimization; realizes the health monitoring and self - repair of the internal and external structures of offshore wind turbines through AI - driven fault monitoring. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the technical problem to be solved by the present invention is: to provide an integrated energy storage - structure - intelligent self - repair system for offshore wind power monopile foundations, which improves the foundation stability through the coupling of the honeycomb battery compartment and the monopile foundation, breaks through the structural - functional separation design limitation of the energy storage system and the wind turbine foundation, conducts zoning control on the structure, realizes fault prediction using the AI architecture, and drives the actuator for self - repair, so as to realize the closed - loop management of "perception - prediction - repair" throughout the life cycle of offshore wind turbines.

[0006] The technical solution adopted by the present invention to solve the above technical problems is as follows: An integrated energy storage-structure-intelligent self-repair system for an offshore wind power monopile foundation, comprising a monopile foundation and a concrete disk-shaped expansion structure cast at the junction of the monopile foundation and the seabed mud surface; an inverter cabin and a honeycomb battery cabin are arranged inside the monopile foundation; the honeycomb battery cabin and the concrete disk-shaped expansion structure are at the same height; The monopile foundation is evenly divided into multiple regions in the circumferential direction. Seawater batteries are installed in the honeycomb battery cabins in each region. At the same time, fiber Bragg grating strain sensors and piezoelectric actuators are installed in each region. The piezoelectric actuators are located in the gap between the honeycomb battery cabin and the inner wall of the monopile foundation; the fiber Bragg grating strain sensors are located at the connection interface between the concrete disk-shaped expansion structure and the monopile foundation for monitoring fiber strain; The truss of the honeycomb battery cabin in each region is connected to the inner wall of the monopile foundation through prestressed steel strands and anchors. Shape memory alloy actuators are installed on the steel strands. Resistance wires are arranged on the surface of the shape memory alloy actuators. Wideband acoustic emission sensors are installed at the anchors for acquiring acoustic emission signals; Bacillus microcapsules are incorporated into the concrete of the concrete disk-shaped expansion structure; An impedance probe for monitoring the change rate of the battery internal resistance, a self-priming pump for controlling the circulation of the electrolyte in and out, and an automatic control valve are installed on the seawater battery; An AI architecture for multi-modal intelligent diagnosis is integrated in the inverter cabin; the inputs of the AI architecture are the fiber strain, acoustic emission signal, and battery internal resistance change rate of each region, and the output is a three-dimensional failure probability P composed of the crack level probability, prestress loss level probability, and electrolyte deterioration level probability; A probability threshold is set. When a certain dimension of the three-dimensional failure probability P of a certain region output by the AI architecture is greater than the probability threshold, a fault self-repair instruction corresponding to the corresponding dimension of the corresponding region is triggered; The fault self-repair instructions include a crack self-repair instruction for causing the Bacillus microcapsules in the region to be directionally damaged through the piezoelectric actuator, a prestress self-repair instruction for causing the shape memory alloy actuator to perform tensioning at high temperature through the resistance wire, and an electrolyte deterioration self-repair instruction for replacing the electrolyte of the seawater battery through the self-priming pump and the automatic control valve.

[0007] Furthermore, the honeycomb battery cabin includes a truss and a number of seawater batteries evenly arranged in the circumferential direction inside the truss. A set of four-wire AC impedance probes are installed at the anode and cathode ends of each seawater battery; the trusses are connected by flanges; Each area uses a group of steel strands, and a shape memory alloy driver is installed in the middle of each group of steel strands through a connector. The connector adopts a bidirectional threaded sleeve, one end of the bidirectional threaded sleeve is anchored to the shape memory alloy driver through a thread, and the other end is anchored to the steel strand through a thread.

[0008] Furthermore, anchors are provided at both ends of the steel strand, and a broadband acoustic emission sensor is installed at each anchor. The broadband acoustic emission sensors at all anchors in each area are used together to capture the acoustic emission signals of the steel strand slippage in real time.

[0009] Furthermore, the concrete disc-shaped expansion structure comprises a disc surface, a number of ribs arranged on the disc surface, and concrete poured outside the disc surface, and the concrete poured in the concrete disc-shaped expansion structure is mixed with Bacillus microcapsules and pre-buried with slow-release nutrients; The action range of the sound waves generated by each piezoelectric actuator can cause directional damage to the Bacillus microcapsules in the corresponding area of ​​the concrete disk-shaped expansion structure.

[0010] Furthermore, the converter cabin also integrates a converter, an MPC controller, a SOC equalizer and a programmable logic controller; The MPC controller collects grid frequency, grid voltage, grid load, wind turbine output power and battery state of charge SOC in real time, and predicts the state of the grid in the future. According to the grid demand in the future period, the MPC controller generates power allocation instructions and charge and discharge scheduling instructions, and transmits the power allocation instructions and charge and discharge scheduling instructions to the converter and seawater battery through the programmable logic controller to stabilize the grid output power.

[0011] Furthermore, all seawater batteries are electrically connected to the SOC equalizer. During the stage of charging the seawater batteries by the offshore wind turbine, the SOC equalizer actively discharges the seawater batteries with SOC>55% to the power grid or other seawater batteries except itself. During the discharge stage of the seawater batteries, the offshore wind turbine boosts the voltage of the seawater batteries with SOC<45%.

[0012] Furthermore, the threshold judger in the programmable logic controller is used to monitor whether the three-dimensional fault probability output by the AI ​​architecture has a value exceeding the probability threshold, and the logic arbitration module in the programmable logic controller is used to independently judge the three types of faults; the programmable logic controller outputs a fault self-repair instruction.

[0013] Furthermore, the AI ​​architecture adopts a CNN-BiLSTM model based on an attention mechanism, a Transformer-based multimodal fusion architecture, a spatial topological architecture driven by a graph neural network, a multi-task learning and dynamic gating architecture, a self-supervised contrastive learning pre-training architecture, or a lightweight edge computing architecture.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: In the present invention, the honeycomb battery compartment is coupled with the monopile foundation. The honeycomb battery compartment adopts a truss structure and is embedded with seawater batteries. The truss is connected to the inner wall of the monopile foundation through prestressed steel strands to form a collaborative bearing system, optimizing the dynamic response of the monopile foundation. The AI architecture integrates acoustic emission signals, the change rate of battery internal resistance, and fiber optic strain data to achieve the joint prediction of the prestress loss level, electrolyte deterioration level, and crack level. When the prediction result exceeds the set threshold, the self-repair mechanism is triggered. Through the integrated structural design, the electric energy storage, the foundation of the offshore wind turbine, and the intelligent diagnosis are closely combined to form an efficient, stable, and intelligent comprehensive solution, which not only improves the reliability and flexibility of the overall system, but also reduces the maintenance cost and operation risk, providing a solid technical guarantee for the sustainable development of the offshore wind farm.

[0015] The present invention uses the AI technology framework for multimodal intelligent diagnosis, integrating three-dimensional data of fiber optic strain, acoustic emission signals, and the change rate of battery internal resistance, and outputs the diagnosis results of the crack level, prestress loss level, and electrolyte deterioration level, that is, outputs the three-dimensional fault probability, and automatically determines whether to trigger the corresponding actuator for self-repair according to the probability of the fault prediction. When the probability of the fault prediction exceeds the set probability threshold, the self-repair of the corresponding fault is triggered. This method can minimize the system downtime, ensure the high availability of the system, and reduce the fault risk.

[0016] In the present invention, through the prestress coupling of the honeycomb battery compartment and the monopile foundation, and in combination with the concrete disk-shaped expansion structure, compared with the traditional monopile foundation, the anti-overturning ability of the foundation is significantly enhanced, and the dynamic response characteristics of the system are optimized.

[0017] The present invention uses seawater batteries for electric energy storage, which can play a role in suppressing the power fluctuation of the power grid. Compared with the traditional battery technology, seawater batteries have rich resource sources, lower costs, and less environmental pollution. By combining seawater batteries with offshore wind turbines, the power fluctuation caused by environmental factors can be effectively smoothed, ensuring the stability and reliability of the power supply.

[0018] In the present invention, by adopting an MPC (Model Predictive Control) controller to monitor the grid frequency, grid voltage, grid load, wind turbine output power, and battery charge state (SOC) in real time, predict the future power demand, and dynamically adjust the output of the seawater battery to suppress the power fluctuation. At the same time, through the battery charge state (SOC) equalizer, the capacity decay of the seawater battery is delayed, ensuring the long-term stability and efficient operation of the system.

[0019] In summary, the system of the present invention is particularly suitable for the specific requirements of offshore wind farms, demonstrating significant technical advantages and innovative achievements. This system not only effectively improves the stability of the offshore wind farm's infrastructure but also solves the problem of health monitoring for offshore wind farms, reflecting excellent technological foresight and broad application potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram showing the internal structure of a monopile foundation and the relative position arrangement with a concrete disk-shaped expansion structure in an embodiment of the present invention.

[0021] Figure 2 It is a schematic diagram showing the structure of an AI architecture - a CNN-BiLSTM model based on an attention mechanism in an embodiment of the present invention.

[0022] Figure 3 It is a schematic diagram showing the arrangement of a shape memory alloy actuator and a steel strand in an embodiment of the present invention.

[0023] In the figure, 1 is a monopile foundation, 2 is a converter cabin, 3 is a concrete disk-shaped expansion structure, 4 is a honeycomb battery cabin, 5 is a counterweight cabin; 6 is a resistance wire, 7 is a shape memory alloy actuator, 8 is a steel strand, 9 is a bidirectional threaded sleeve. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The following further explains the present invention in conjunction with embodiments and the accompanying drawings, but does not limit the protection scope of this application thereby.

[0025] The integrated energy storage-structure-intelligent self-repair system for the offshore wind monopile foundation of the present invention realizes multi-element coupling by hierarchically arranging batteries, control components, and counterweights inside the monopile foundation. The interior of the monopile foundation is divided into a counterweight cabin, a honeycomb battery cabin, and a converter cabin. The lower counterweight cabin is filled with concrete to enhance the anti-overturning ability. The middle honeycomb battery cabin is embedded with seawater batteries. The upper converter cabin integrates a converter, an MPC controller, a programmable logic controller, etc. to complete grid power compensation. The SOC equalizer is used to regulate the charge and discharge balance of the seawater batteries. The AI architecture processes acoustic emission signals, the battery internal resistance change rate, and fiber optic strain data in real time, predicts concrete cracks, steel strand relaxation, and electrolyte deterioration faults, triggers the rupture of Bacillus microcapsules to release Bacillus repair agents, the tensioning of steel strands by shape memory alloy actuators, and the electrolyte circulation replacement mechanism, realizes the self-repair of structural damage and the performance of seawater batteries, and ensures the long-term stable operation of the system in a complex marine environment.

[0026] The seawater battery of the present invention can be, but is not limited to, a rechargeable chloride ion battery, a membraneless rechargeable seawater battery, a sodium metal seawater battery, a water-based zinc battery, etc. The AI ​​architecture of the present invention is aimed at multimodal intelligent diagnosis tasks (joint prediction of crack level, prestress loss level, and electrolyte degradation level), and can adopt a CNN-BiLSTM model based on an attention mechanism, a multimodal fusion architecture based on a Transformer, a spatial topology architecture driven by a graph neural network (GNN), a multi-task learning (MTL) and dynamic gating architecture, a self-supervised comparative learning pre-training architecture, a lightweight edge computing architecture, etc.

[0027] Embodiment 1: The offshore wind power monopile foundation energy storage-structure-intelligent self-repairing integrated system of the present invention comprises a monopile foundation 1 and a concrete disc-shaped expansion structure 3 cast at the junction of the monopile foundation and the seabed mud surface; a converter cabin, a honeycomb battery cabin and a counterweight cabin are arranged from top to bottom in the monopile foundation; The honeycomb battery compartment includes a truss and a number of seawater batteries evenly arranged in the truss along the circumferential direction. The circumference of the monopile foundation is divided into multiple areas according to the number of seawater batteries. A set of four-wire AC impedance meter probes are installed at the positive and negative ends of each seawater battery to monitor the battery internal resistance change rate; a piezoelectric actuator is installed in the gap between the honeycomb battery compartment and the inner wall of the monopile foundation in each area; The concrete disc-shaped extension structure includes a disc surface, a number of ribs arranged on the disc surface, and concrete poured outside the disc surface, and the area surrounded by adjacent ribs corresponds to the number and position of the areas divided according to the number of seawater batteries; in each area, a fiber grating strain sensor is arranged at the connection interface between the concrete disc-shaped extension structure and the single pile foundation, and bacillus microcapsules are mixed into the concrete poured by the concrete disc-shaped extension structure 3, and a slow-release nutrient is pre-buried; The trusses are connected by flanges, and the trusses are connected to the inner wall of the single pile foundation by prestressed steel strands and anchors. A group of steel strands is used in each area, and a shape memory alloy driver is installed in the middle of each group of steel strands through a connector. A resistance wire 6 is arranged on the surface of the shape memory alloy driver. In this embodiment, a nickel-chromium alloy resistance wire is used. Anchors are arranged at both ends of the steel strands, and a broadband acoustic emission sensor is installed at each anchor. The broadband acoustic emission sensors at all anchors in each area are used together to capture the acoustic emission signals of steel strand slippage in real time.

[0028] The range of sound waves generated by each piezoelectric actuator can cause directional damage to the bacillus microcapsules in the corresponding area of ​​the concrete disc-shaped expansion structure. Each seawater battery is controlled by an automatic control valve to control whether to replenish or replace the electrolyte; the seawater battery provides power for the system.

[0029] The AI ​​architecture for multi-modal intelligent diagnosis is integrated in the converter compartment 2; The input of the AI architecture is the optical fiber strain of each area, all acoustic emission signals, and the change rate of the battery internal resistance, and the output is the three-dimensional fault probability P = [P c , P l , P d for each area. where P c is the crack grade probability, P l is the prestress loss grade probability, and P d is the electrolyte deterioration grade probability; Set a probability threshold. When a certain dimension of the three-dimensional fault probability P of a certain area output by the AI architecture is greater than the probability threshold, a fault self-repair instruction corresponding to the corresponding dimension of the corresponding area is triggered.

[0030] Example 2: The integrated energy storage-structure-intelligent self-repair system for the monopile foundation of offshore wind power in this embodiment includes: a monopile foundation 1, a converter cabin 2, a concrete disk-shaped expansion structure 3 cast at the junction of the monopile foundation and the seabed mud surface, a honeycomb battery cabin 4, and a counterweight cabin 5 below the honeycomb battery cabin.

[0031] The monopile foundation 1 uses a Q355 steel pipe pile with a diameter of 8 - 12 meters and a wall thickness of 60 - 80 mm, and the pile top elevation is 3 - 5 meters higher than the extreme wave height. The inside of the pile body is hollow and divided into three layers. The lower layer is the counterweight cabin 5, which is filled with concrete to provide anti-overturning counterweight for the foundation. The height of the counterweight cabin is 3 - 5 m, and the distance from the pile bottom is 0.4 - 0.6 times the pile length; the middle layer is the honeycomb battery cabin 4, which houses seawater batteries and transfers structural loads. The height of the honeycomb battery cabin is 2 - 4 meters, adjacent to the cushion layer of the counterweight cabin, and is connected to the counterweight cabin by bolts; the upper layer is the converter cabin 2, with a height of 2 - 3 meters, and the distance from the honeycomb battery cabin 4 is 0.3 times the pile length.

[0032] The converter cabin 2 integrates a converter, an MPC controller, an SOC equalizer, an AI architecture for multi-modal intelligent diagnosis, a programmable logic controller, and a communication interface, etc. Among them, MPC is model predictive control, and SOC is state of charge.

[0033] After the installation of the single pile foundation 1 is completed, a concrete disc-shaped extension structure 3 is cast at the junction of the single pile foundation and the seabed mud surface. The concrete disc-shaped extension structure includes a disc surface, a number of ribs arranged on the disc surface, and concrete cast outside the disc surface. The disc surface diameter is 20-25m, the inclination angle between the ribs and the disc surface is 15°, and a rib is arranged every 60° with a thickness of 15cm. The entire disc surface is divided into six areas by the ribs. At the same time, fiber grating strain sensors are arranged at the connection interface between the concrete disc-shaped extension structure 3 and the single pile foundation 1, with one group arranged every 60°, and a total of 6 groups arranged to monitor the concrete optical fiber strain in real time. Bacillus microcapsules are added to the cast concrete of the concrete disc-shaped extension structure 3, and pH-sensitive slow-release nutrients are pre-embedded for concrete self-repair.

[0034] The honeycomb battery compartment 4 includes a truss and a number of seawater batteries installed in an array in the truss. In this embodiment, 6 seawater batteries are arranged. The truss is made of S420 high-strength steel with a wall thickness of 12mm. The trusses are connected by flanges and coupled to the inner wall of the single pile foundation 1 through prestressed steel strands and anchors. The steel strand 8 has a diameter of 32mm, a preload of ≥800kN, and a group of 60° per group, with a total of 6 groups of steel strands. At the same time, a shape memory alloy driver 7 is installed in the middle of each group of steel strands through a connector. The connector adopts a bidirectional threaded sleeve 9, one end of the bidirectional threaded sleeve is anchored to the shape memory alloy driver 7 through a thread, and the other end is anchored to the steel strand 8 through a thread; the installation positions of the two ends of the shape memory alloy driver 7 are set at two 1 / 3 positions of the total length of the steel strand 8. The nickel-chromium alloy resistance wire is evenly wound on the surface of the shape memory alloy driver 7, and the nickel-chromium alloy resistance wire (resistance is 30Ω) is electrically connected to the seawater battery. After the steel strand is tensioned, the circumferential gap between the honeycomb battery compartment 4 and the monopile foundation is ≤100mm (this distance can be adjusted). One end of the steel strand is connected to the inner wall of the monopile foundation through an anchor, and the other end is connected to the truss through an anchor. Broadband acoustic emission sensors are arranged at all anchors. The broadband acoustic emission sensors at all anchors in each area are used together to capture the acoustic emission signals of steel strand slip in real time, with a sampling rate of 1MHz.

[0035] The power transmission between the converter cabin 2 and the honeycomb battery cabin 4 is achieved through cables. Six piezoelectric actuators are evenly arranged and installed in the gap between the honeycomb battery cabin and the inner wall of the single pile foundation. The installation height of the piezoelectric actuator is consistent with the height of the concrete disc-shaped expansion structure outside the pile. Each piezoelectric actuator generates sound waves with a range of 60°. The piezoelectric actuator is electrically connected to the seawater battery to supply power to it. According to the range of action of the sound waves generated by the piezoelectric actuator, they are evenly arranged in the circumferential direction, which can cause directional damage to the Bacillus microcapsules in the corresponding area inside the concrete disc-shaped expansion structure.

[0036] The seawater battery is embedded in the honeycomb battery compartment. The electrolyte of the seawater battery is natural seawater electrolyte after filtering impurities, and the specific internal structure of the seawater battery can be realized according to the existing technology. A set of impedance probe is installed at the anode and cathode ends of each seawater battery to monitor the change rate of battery internal resistance. In this embodiment, the impedance probe is a four-wire AC impedance probe. A total of two self-priming pumps are placed on the top of all seawater batteries for the replenishment or replacement of the electrolyte. Each seawater battery is equipped with two automatic control valves to control the entry and outflow of natural seawater respectively. The self-priming pumps are electrically connected to the seawater batteries to provide electrical energy for the self-priming pumps. This embodiment configures 6 seawater batteries. The middle area of the honeycomb battery compartment does not place seawater batteries to ensure the heat dissipation space.

[0037] The MPC controller predicts the state of the power grid in the future for a period of time based on the real-time collected grid frequency, grid voltage, grid load, fan output power, and battery SOC status (10 μs sampling period), and then combines the grid demand in the future period to generate the optimal control strategy for the converter through rolling horizon optimization. The control strategy includes power distribution instructions and charge and discharge scheduling instructions.

[0038] Among them, the power distribution instruction refers to controlling the voltage / current waveform output by the converter by adjusting the duty cycle and switching frequency. The charge and discharge scheduling instruction means that when the grid frequency mutation or battery SOC anomaly is detected, the MPC controller adjusts the charging and discharging power of the battery to restore the stability of the power grid and ensure the safe operation of the battery. The above instructions are transmitted from the programmable logic controller to the converter and the seawater battery.

[0039] All seawater batteries are electrically connected to the SOC equalizer. The working process of the SOC equalizer in the charging stage and the discharging stage is as follows: Charging stage When the state of charge (SOC) of a certain seawater battery exceeds 55%, the SOC equalizer will actively discharge it (the maximum discharge current is 10 A) to avoid overcharging.

[0040] Discharging stage When the state of charge (SOC) of a certain seawater battery is lower than 45%, the SOC equalizer boosts the voltage to compensate for it to prevent insufficient output power caused by over-discharge.

[0041] The input of the AI architecture is the fiber strain of each area, all acoustic emission signals, and the change rate of battery internal resistance, and the output is the three-dimensional fault probability P = [P c , P l , P d of each area. where P c is the crack level probability, P l is the prestress loss level probability, P dis the probability of electrolyte degradation level.

[0042] Set a probability threshold. In this embodiment, the probability threshold is set to 0.8. When there is a value in a certain dimension of the three-dimensional fault probability P output by the AI architecture that is greater than the probability threshold, the corresponding fault self-repair instruction for that dimension in the corresponding area is triggered. Transmit the three-dimensional fault probability to the programmable logic controller, and the programmable logic controller makes a judgment. If the judgment is successful, it outputs a fault self-repair instruction and transmits the instruction to the execution end (piezoelectric brake, nickel-chromium alloy resistance wire, or self-priming pump and automatic control valve).

[0043] Embodiment 3: The AI architecture for multi-modal intelligent diagnosis in this embodiment is implemented by a CNN-BiLSTM deep learning model based on the attention mechanism. Considering the three aspects of "efficiency-accuracy-interpretability" for the single-pile diagnosis scenario of offshore wind power, the CNN-BiLSTM serial structure adapts to edge computing resources to meet real-time requirements; the attention mechanism dynamically focuses on key modalities and temporal segments to improve the fine-grained diagnosis ability.

[0044] The CNN-BiLSTM deep learning model based on the attention mechanism includes a CNN module, a BiLSTM layer, an attention mechanism, and a multi-task classifier connected in sequence. According to the layout angles of various sensors (60° in a group), it can be divided into 6 areas, and the three types of signals in each area are used as a group of inputs. Among them, the CNN module uses 5 layers of depthwise separable convolution to extract 128-dimensional time-frequency features. The BiLSTM layer realizes the cross-modal temporal correlation of a 500ms time window through a gated recurrent unit, and adopts dynamic weight allocation. According to the contribution degree of different types of sensor data to the current fault diagnosis, the weight coefficients of each modal feature are automatically adjusted. Through the attention mechanism, feature fusion is carried out using the dynamically generated weights, and the joint diagnosis results of the crack level (level I-IV), prestress loss level (level I-III), and electrolyte degradation level (level I-V) are output by the multi-task classifier, that is, the three-dimensional fault probability P = [P c , P l , P d for each area is output. The implementation method is as follows: 1. Feature extraction: After preprocessing (normalization), three types of signals, namely fiber optic strain, acoustic emission signal, and battery internal resistance change rate, enter the CNN module and output 128-dimensional feature vectors, which are the acoustic emission signal feature vector F AE , the battery internal resistance change rate feature vector F IR , and the fiber optic strain feature vector F FS . Then, the 128-dimensional feature vector output by the CNN module is input into the BiLSTM layer to output cross-modal temporal features H AE , H IR , H FS。

[0045] 2. Dynamic weight assignment: First, calculate the attention score :

[0046] is a learnable parameter matrix, b a is a learnable parameter, H i is the cross-modal temporal feature of the i-th modality, and tanh is the hyperbolic tangent function.

[0047] Normalize the attention score to obtain the dynamic weight w i , where the subscript i takes the dynamic weights corresponding to the acoustic emission signal, the battery internal resistance change rate, and the fiber optic strain for AE, IR, and FS respectively;

[0048] Use the dynamic weights to perform feature fusion on the cross-modal temporal features to obtain the fused feature F fusion :

[0049] 3. Three-dimensional fault probability output Map the fused feature F fusion to the dimensions required for each classification task through three fully connected layers,

[0050]

[0051]

[0052] where, V c , V l , V d are mapped to 4-dimensional vectors for the crack level task, 3-dimensional vectors for the prestress loss level task, and 5-dimensional vectors for the electrolyte deterioration level task respectively; W c , W l , W d are the weight matrices of the fully connected layers under different tasks; b c , b l , b d are the bias terms of the fully connected layers under different tasks.

[0053] After that, use three Softmax classifiers to classify the mapped feature vectors V c , V l , V dPerform probability normalization and output the combined diagnosis results of crack levels (Level I - IV), prestress loss levels (Level I - III), and electrolyte degradation levels (Level I - V), that is, output the three - dimensional fault probability P = [P c ,P l ,P d for each area.

[0054]

[0055]

[0056]

[0057] Among them, j represents the component corresponding to a certain level of different disease types; that is Figure 2 Softmax classifier 1 in is used for outputting the crack level probability, Softmax classifier 2 is used for outputting the prestress loss level probability, and Softmax classifier 3 is used for outputting the electrolyte degradation level probability.

[0058] Example: The three - dimensional fault probability P = [P c ,P l ,P d for a certain area P c = [0.1, 0.6, 0.2, 0.1] indicates that the probability of crack level I is 10%, the probability of level II is 60%, the probability of level III is 20%, and the probability of level IV is 10%. The maximum value of the four - level probabilities is used as the crack fault level of this area.

[0059] = [0.3, 0.5, 0.2] indicates that the probability of prestress loss level I is 30%, the probability of level II is 50%, and the probability of level III is 20%. The maximum value of the three - level probabilities is used as the prestress fault level of this area.

[0060] P d = [0.1, 0.2, 0.4, 0.2, 0.1] indicates that the probability of electrolyte degradation level I is 10%, the probability of level II is 20%, the probability of level III is 40%, the probability of level IV is 20%, and the probability of level V is 10%. The maximum value of the five - level probabilities is used as the electrolyte degradation fault level of this area.

[0061] A probability threshold is set. In this embodiment, the probability threshold is set to 0.8. When the three-dimensional fault probability P of a certain area output by the AI ​​architecture contains a value greater than the probability threshold, the corresponding fault self-repair instruction of the corresponding area is triggered. In this example, there are no faults in the three dimensions. The three-dimensional fault probability is transmitted to the programmable logic controller, which makes a judgment. If the judgment is successful, the fault self-repair instruction is output and transmitted to the execution end (piezoelectric brake, nickel-chromium alloy resistance wire or self-priming pump and automatic control valve).

[0062] If P c If there is a value greater than the probability threshold in , it is a crack fault. The piezoelectric actuator is controlled to apply vibration waves to break the microcapsule wall in this area to release Bacillus for concrete self-repair. l If there is a value greater than the probability threshold, it is a prestress loss fault, driving the nickel-chromium alloy resistance wire to convert current into heat energy to provide a working environment for the shape memory alloy driver. The shape memory alloy driver 7 performs tensioning at high temperature to restore the previous prestress; if P d If there is a value greater than the probability threshold, it is an electrolyte degradation fault, and the self-priming pump and the automatic control valve of the seawater battery in the corresponding area are driven to open to change the electrolyte of the seawater battery. In this embodiment, the piezoelectric actuator installed on the inner wall of the pile applies a 10kHz vibration wave to break the capsule wall in the area to release Bacillus for concrete self-repair; when the prestress loss rate is large, the shape memory alloy driver 7 is used to perform tensioning at high temperature to restore the previous prestress, and the nickel-chromium alloy resistance wire provides a working environment for the shape memory alloy driver by converting current into heat energy. Each time it is powered on with a 1A DC current, each time for no more than 15 minutes; when the electrolyte is replaced, a self-priming pump starts to pump seawater from the bottom, and a self-priming pump starts to pump seawater from the top; when the electrolyte is supplemented, the self-priming pump is turned on to pump seawater from the top. The directional transportation of natural seawater electrolyte can be achieved through existing automatic control valve technology.

[0063] The system of the present invention monitors the status and structural health of seawater batteries in real time through a variety of sensors, and makes predictions and decisions based on the data sensed by the sensors to ensure that the system operates in the optimal state. The seawater battery performs charging and discharging operations according to the instructions of the SOC equalizer, and the converter adjusts the operating mode according to the instructions of the MPC controller to achieve efficient energy management and stable system operation.

[0064] Embodiment 4: In this embodiment, the converter cabin 2 integrates a converter, an MPC controller, an SOC equalizer, an AI architecture for multimodal intelligent diagnosis, a programmable logic controller, and a communication interface. The alternating current generated by the offshore wind turbine is converted into direct current for seawater battery storage through the converter. The direct current of the seawater battery can be converted into alternating current for the power grid by the converter. The rated power of the converter is 2 MW. The MPC controller is connected to the power grid, the seawater battery, and the SOC equalizer through the communication interface, and at the same time realizes data interaction with the converter. It samples the grid frequency, grid voltage, grid load, wind turbine output power, and battery charge state (SOC) in real time with a sampling period of 10 μs, predicts the state of the system in the next 200 μs, generates power distribution instructions and charge-discharge scheduling instructions through rolling horizon optimization, and transmits them to the converter and the seawater battery by the programmable logic controller to ensure stable power output. All seawater batteries are electrically connected to the SOC equalizer. The SOC equalizer actively discharges the seawater battery with SOC > 55% with a maximum of 10 A during the charging stage, and boosts the voltage of the seawater battery with SOC < 45% during the discharging stage.

[0065] The power grid transmits the grid frequency, grid voltage, and grid load signals to the MPC controller; The offshore wind turbine transmits the wind turbine output power to the MPC controller; The SOC equalizer transmits the SOC signal to the MPC controller; The MPC controller transmits the power distribution instructions and charge-discharge scheduling instructions to the programmable logic controller; The programmable logic controller transmits the power distribution instructions to the converter; The programmable logic controller transmits the charge-discharge scheduling instructions to the seawater battery; The converter and the seawater battery transmit feedback signals to the MPC controller.

[0066] Embodiment 5: The construction process of the system in this embodiment is as follows: 1. Construction of the monopile foundation: According to the design requirements, calculate the bearing capacity and pile body size of the monopile foundation.

[0067] Q355 steel pipe piles are used, with a diameter of 8-12 meters and a wall thickness of 60-80mm. The interior is divided into a lower counterweight cabin filled with concrete, with a cabin height of 3-5m, providing anti-overturning counterweight; the middle honeycomb battery cabin is arranged with S420 steel trusses, with a wall thickness of 12mm, arranged in a ring, coupled with the pile body of the single pile foundation through steel strands, with a gap of ≤100mm (adjustable), and the steel strands are equipped with shape memory alloy drivers; the upper converter cabin is equipped with converters, MPC controllers, AI architectures for multi-modal intelligent diagnosis, programmable logic controllers, and SOC equalizers, and a concrete disc-shaped extension structure with a diameter of 20-25m is cast outside the pile. The inner rib plate of the concrete disc-shaped extension structure has an inclination angle of 15° and a thickness of 15cm. Fiber Bragg grating strain sensors are arranged at the connection interface between the concrete disc-shaped extension structure and the single pile foundation. At the same time, Bacillus microcapsules are pre-embedded in the concrete, and pH-sensitive slow-release nutrients are added.

[0068] 2. Battery compartment and structure coupling: The honeycomb battery compartment truss is embedded with seawater batteries. The upper part of the seawater battery is equipped with two self-priming pumps and automatic control valves. The inlet and outlet of the natural seawater of each seawater battery are each provided with an automatic control valve. All seawater batteries share two self-priming pumps to realize the circulation of natural seawater between the seawater batteries. The electrolyte of the seawater battery uses filtered natural seawater. In this embodiment, 6 seawater batteries are arranged. At the same time, a four-wire AC impedance meter probe is installed at the positive and negative ends of each seawater battery to monitor the internal resistance change rate of the seawater battery in real time.

[0069] 3. Implementation of grid power compensation: The AC power generated by the wind turbine is converted into DC power by the converter and stored in the seawater battery. After the DC power of the seawater battery is converted into AC power by the converter, it can be used for grid power compensation. The grid power compensation control is realized by the MPC controller, which has a sampling period of 10μs, can predict the grid state within 200μs, and generate the optimal control strategy. The SOC equalizer discharges the seawater battery with SOC>55% (≤10A) during the charging stage, and boosts the seawater battery with SOC<45% during the discharging stage.

[0070] 4. AI-driven fault prediction: Data acquisition is carried out through multiple sensors. A broadband acoustic emission sensor is used to capture the steel strand slip signal; the state of the seawater battery is monitored by the probe of a four-wire AC impedance meter to detect the change rate of the battery internal resistance; the concrete strain is detected in real time by a fiber Bragg grating strain sensor for the concrete disk-shaped structure. A CNN-BiLSTM deep learning model based on the attention mechanism is adopted to dynamically allocate the weights of acoustic emission, internal resistance, and strain. During the feature extraction process, the CNN module extracts 128-dimensional time-frequency features through 5 layers of depthwise separable convolution, and the BiLSTM layer performs cross-modal temporal correlation through a 500ms time window. The multi-task classifier is used for fault classification, and outputs the crack level (Level I - IV), the prestress loss level (Level I - III), and the electrolyte deterioration level (Level I - V), that is, outputs the three-dimensional fault probability.

[0071] 5. Self-repair implementation: When there is a value in a certain dimension of the three-dimensional fault probability P of a certain area in the output that is greater than the probability threshold, the actuator at the corresponding area is triggered to perform corresponding fault self-repair. The self-repair methods are as follows: The piezoelectric actuator emits a 10kHz vibration wave to directionally break the microcapsules and release the Bacillus repair material; the stress recovery of the steel strand is achieved by heating the shape memory alloy actuator to perform tension compensation; the electrolyte maintenance relies on the self-priming pump to supplement or replace the filtered seawater as needed.

[0072] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0073] The parts not described in the present invention are applicable to the prior art.

Claims

1. An offshore wind power monopile foundation energy storage-structure-intelligent self-repairing integrated system, comprising a monopile foundation and a concrete disc-shaped expansion structure cast at the junction of the monopile foundation and the seabed mud surface; characterized in that: A converter cabin and a honeycomb battery cabin are arranged in the monopile foundation; the honeycomb battery cabin and the concrete disc-shaped expansion structure are located at the same height; The monopile foundation is evenly divided into multiple areas in the circumferential direction, and seawater batteries are installed in the honeycomb battery compartment in each area. At the same time, a fiber grating strain sensor and a piezoelectric actuator are installed in each area, and the piezoelectric actuator is located in the gap between the honeycomb battery compartment and the inner wall of the monopile foundation; the fiber grating strain sensor is located at the connection interface between the concrete disc-shaped expansion structure and the monopile foundation, and is used to monitor the optical fiber strain; The trusses of the honeycomb battery compartment in each area are connected to the inner wall of the monopile foundation through prestressed steel strands and anchors, a shape memory alloy driver is installed on the steel strands, a resistance wire is set on the surface of the shape memory alloy driver, and a broadband acoustic emission sensor is installed at the anchor to obtain acoustic emission signals; Bacillus microcapsules were incorporated into the concrete of the concrete disk-shaped expansion structure; An impedance meter probe for monitoring the rate of change of the battery's internal resistance, a self-priming pump for controlling the circulation of electrolyte in and out, and an automatic control valve are installed on the seawater battery; The converter cabin integrates an AI architecture for multi-modal intelligent diagnosis; the input of the AI ​​architecture is the optical fiber strain, acoustic emission signal and battery internal resistance change rate of each area, and the output is a three-dimensional fault probability P composed of crack level probability, prestress loss level probability and electrolyte degradation level probability; Set a probability threshold. When a dimension of the three-dimensional fault probability P of a certain area output by the AI ​​architecture has a value greater than the probability threshold, the fault self-repair instruction of the corresponding dimension of the corresponding area is triggered. The fault self-repair instructions include crack self-repair instructions for causing directional damage to Bacillus microcapsules in an area through a piezoelectric actuator, prestressed self-repair instructions for causing shape memory alloy drivers to perform tensioning at high temperatures through resistance wires, and electrolyte deterioration self-repair instructions for replacing electrolyte in seawater batteries through self-priming pumps and automatic control valves.

2. The offshore wind power monopile foundation energy storage-structure-intelligent self-repairing integrated system according to claim 1 is characterized in that: The honeycomb battery cabin includes a truss and a number of seawater batteries evenly arranged in the circumferential direction inside the truss, and a set of four-wire AC impedance meter probes are installed at the positive and negative ends of each seawater battery; the trusses are connected by flanges; Each area uses a group of steel strands, and a shape memory alloy driver is installed in the middle of each group of steel strands through a connector. The connector adopts a bidirectional threaded sleeve, one end of the bidirectional threaded sleeve is anchored to the shape memory alloy driver through a thread, and the other end is anchored to the steel strand through a thread.

3. The offshore wind power monopile foundation energy storage-structure-intelligent self-repairing integrated system according to claim 2 is characterized in that: Anchors are set at both ends of the steel strand, and a broadband acoustic emission sensor is installed at each anchor. The broadband acoustic emission sensors at all anchors in each area are used together to capture the acoustic emission signals of steel strand slippage in real time.

4. The offshore wind power monopile foundation energy storage-structure-intelligent self-repairing integrated system according to claim 1 is characterized in that: The concrete disc-shaped expansion structure comprises a disc surface, a number of ribs arranged on the disc surface, and concrete poured outside the disc surface. Bacillus microcapsules are mixed into the concrete poured for the concrete disc-shaped expansion structure, and a slow-release nutrient is pre-embedded. The action range of the sound waves generated by each piezoelectric actuator can cause directional damage to the Bacillus microcapsules in the corresponding area of ​​the concrete disk-shaped expansion structure.

5. The offshore wind power monopile foundation energy storage-structure-intelligent self-repairing integrated system according to claim 1 is characterized in that: The converter cabin also integrates a converter, an MPC controller, a SOC equalizer and a programmable logic controller; The MPC controller collects grid frequency, grid voltage, grid load, wind turbine output power and battery state of charge SOC in real time, and predicts the state of the grid in the future. According to the grid demand in the future period, the MPC controller generates power allocation instructions and charge and discharge scheduling instructions, and transmits the power allocation instructions and charge and discharge scheduling instructions to the converter and seawater battery through the programmable logic controller to stabilize the grid output power.

6. The offshore wind power monopile foundation energy storage-structure-intelligent self-repairing integrated system according to claim 5 is characterized in that: All seawater batteries are electrically connected to the SOC equalizer. During the stage of charging the seawater batteries by offshore wind turbines, the SOC equalizer actively discharges the seawater batteries with SOC>55% to the grid or other seawater batteries except itself. During the discharge stage of the seawater batteries, the offshore wind turbines boost the voltage to compensate for the seawater batteries with SOC<45%.

7. The offshore wind power monopile foundation energy storage-structure-intelligent self-repairing integrated system according to claim 5 is characterized in that: The threshold judger in the programmable logic controller is used to monitor whether the three-dimensional fault probability output by the AI ​​architecture has a value exceeding the probability threshold. The logic arbitration module in the programmable logic controller is used to independently judge the three types of faults; the programmable logic controller outputs a fault self-repair instruction.

8. The offshore wind power monopile foundation energy storage-structure-intelligent self-repairing integrated system according to claim 1 is characterized in that: The AI ​​architecture adopts a CNN-BiLSTM model based on an attention mechanism, a Transformer-based multimodal fusion architecture, a spatial topological architecture driven by a graph neural network, a multi-task learning and dynamic gating architecture, a self-supervised contrastive learning pre-training architecture, or a lightweight edge computing architecture.

Images