Noise damping control system of offshore wind turbine generator

Abstract

The invention relates to a noise damping control system of an offshore wind turbine generator, and belongs to the technical field of offshore wind power generation. The coupling state evaluation module is used for determining a coupling state evaluation index through a preset deep neural network model based on the platform motion information, the vibration spectrum data and the sound pressure signal; the cooperative control strategy generation module is used for determining target rigidity and target damping according to the coupling state evaluation index and the vibration spectrum data; according to the invention, the coupling state evaluation module processes platform motion information, vibration spectrum data and sound pressure signals acquired by the data acquisition module through a preset deep neural network model; the preset model is endowed with strong nonlinear fitting capability through offline supervised learning of massive historical working condition data; and an accurate and quantitative risk basis which cannot be reached by the prior art is provided for subsequent control decisions.

Description

Technical Field

[0001] This invention relates to the field of offshore wind power generation, specifically to a noise reduction and vibration control system for offshore wind turbines. Background Technology

[0002] In deep-sea areas, floating wind turbines are subjected to complex dynamic loads such as wind, waves, and currents, triggering frequency-varying coupling effects between the turbine structure and the floating platform. Existing vibration reduction technologies mostly employ fixed control strategies, which are ill-suited to these frequency-varying characteristics, resulting in insufficient vibration reduction performance. When the platform's tilt exceeds a certain threshold, such as 15°, the performance of conventional vibration reduction systems degrades drastically. This nonlinear coupling effect can also generate new vibration modes, creating unexpected underwater noise sources. Existing technologies face a technical contradiction between pursuing broadband and efficient vibration reduction and ensuring stability under extreme operating conditions, failing to fundamentally solve the problems of nonlinear vibration and newly generated noise caused by frequency-varying coupling.

[0003] To address the aforementioned technical problems, this invention provides an intelligent variable stiffness-variable damping coordinated control system. This system aims to address the vibration and noise generated by floating wind turbines under complex sea conditions due to frequency-varying coupling effects, and to resolve the technical contradiction between wide-band high-efficiency vibration reduction and stability under extreme conditions.

[0004] This technical solution is implemented through an integrated control process. First, the real-time operating status of the unit is acquired through multi-dimensional sensor fusion. This status information is used as input to a deep neural network model, which generates an evaluation index that quantifies the current coupling severity. Next, a nonlinear adaptive control strategy dynamically and collaboratively adjusts the equivalent stiffness and equivalent damping of the system based on this evaluation index. The system has a preset platform tilt angle threshold. Once this threshold is exceeded, the system switches to an emergency stabilization mode to prioritize platform attitude stability. Finally, the system integrates a continuous optimization module. This module uses reinforcement learning algorithms to continuously fine-tune the core parameters of the control strategy based on long-term operating data feedback to achieve adaptive optimization of vibration reduction and noise reduction performance.

[0005] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a noise reduction and vibration control system for offshore wind turbines to solve the problems mentioned in the background art.

[0007] The technical solution of the present invention includes:

[0008] The data acquisition module is used to acquire platform motion information, vibration spectrum data, and sound pressure signals that characterize the operating status of the wind turbine.

[0009] The coupling state assessment module is used to determine the coupling state assessment index based on platform motion information, vibration spectrum data and sound pressure signal through a preset deep neural network model.

[0010] The collaborative control strategy generation module is used to determine the target stiffness and target damping based on the coupling state evaluation index and vibration spectrum data.

[0011] The control mode switching module is used to obtain the maximum tilt angle of the platform in the platform motion information, and when the maximum tilt angle of the platform exceeds the preset tilt angle threshold, the preset maximum safety value is used as the system stiffness and damping.

[0012] The performance continuous optimization module is used to adjust the preset control parameters adopted by the cooperative control strategy generation module based on the coupling state evaluation index, the underwater noise level determined by the sound pressure signal, and the system control energy consumption.

[0013] Preferably, the coupling state evaluation module is used to determine the coupling state evaluation index, including:

[0014] The platform motion information, vibration spectrum data, and sound pressure signal are integrated into a multi-dimensional feature vector. This multi-dimensional feature vector is then input into a deep neural network model. Through the nonlinear fitting operation of the deep neural network model, a coupling state evaluation index is output.

[0015] Preferably, the cooperative control strategy generation module is used to determine the target stiffness, including:

[0016] The dominant vibration frequency is extracted from the vibration spectrum data. Based on the dominant vibration frequency and the preset system reference natural frequency, the frequency detuning factor is determined. The coupling state evaluation index and the frequency detuning factor are combined, and the target stiffness is determined according to the preset reference stiffness.

[0017] Preferably, the cooperative control strategy generation module is used to determine the target damping, including:

[0018] The dominant vibration amplitude is extracted from the vibration spectrum data. Based on the dominant vibration amplitude and the preset maximum safe vibration amplitude, the normalized vibration amplitude is calculated. The target damping is determined by combining the coupling state evaluation index and the normalized vibration amplitude, and according to the preset benchmark damping.

[0019] Preferably, the frequency detuning factor is used to quantify the proximity between the dominant vibration frequency and the system reference natural frequency, and is calculated by substituting the difference between the dominant vibration frequency and the system reference natural frequency into a preset exponential function model.

[0020] Preferably, the control mode switching module is specifically used for:

[0021] Monitor the platform's maximum tilt angle in real time and compare it with a preset tilt angle threshold.

[0022] When the maximum tilt angle of the platform exceeds the preset tilt angle threshold, the emergency stabilization mode is triggered, and the preset maximum safety value is used as the system stiffness and damping.

[0023] When the maximum tilt angle of the platform does not exceed the preset tilt angle threshold, the normal collaborative control mode is maintained, and the target stiffness and target damping determined by the collaborative control strategy generation module are adopted.

[0024] Preferably, the performance continuous optimization module is specifically used for:

[0025] The sound pressure signal is processed to determine the normalized underwater noise level, and the monitoring system is used to determine the normalized control energy consumption. Based on the coupled state evaluation index, the normalized underwater noise level, and the normalized control energy consumption, and combined with the preset weight coefficients, the real-time reward is calculated. Based on the real-time reward, the preset control parameters are adjusted through a preset reinforcement learning algorithm.

[0026] Preferably, the instant reward is obtained by weighted summation of the coupling state evaluation index, the normalized underwater noise level, and the normalized control energy consumption, in order to unify the optimization objectives of high stability, low noise, and low energy consumption.

[0027] Preferably, the preset control parameter is the dimensionless gain in the collaborative control strategy, and the adjustment output of the performance continuous optimization module is used as the input parameter of the collaborative control strategy generation module to close the intelligent control loop of the system.

[0028] This invention provides an improved noise reduction and vibration control system for offshore wind turbines, which has the following improvements and advantages compared with the prior art:

[0029] 1. The coupling state assessment module in this solution processes the platform motion information, vibration spectrum data, and sound pressure signal acquired by the data acquisition module through a pre-set deep neural network model. This pre-set model is endowed with powerful nonlinear fitting capabilities through offline supervised learning of massive historical working condition data. It provides accurate and quantitative risk basis that was previously unattainable for subsequent control decisions, enabling the system to transform from blind response to intelligent control based on precise assessment.

[0030] 2. The collaborative control strategy generation module of this solution shows a fundamental difference, accurately configuring energy-dissipating damping according to the actual vibration energy and system risk level; this synergy and dynamic adjustment of stiffness and damping enables the system to effectively suppress vibration under broadband and time-varying excitation, solving the problem of insufficient vibration reduction performance of existing technologies under frequency-varying loads.

[0031] 3. The control mode switching module in this solution is designed to address this deficiency. This module monitors the platform's maximum tilt angle in real time. Once it exceeds the preset tilt angle threshold determined by hydrodynamic simulation and instability risk assessment, the system will unconditionally switch to emergency stability mode and forcibly adopt the preset maximum safety value as the system stiffness damping. This design ensures that in extreme sea conditions that may lead to disaster, the control objective shifts from performance optimization to ensuring structural survival, providing a solid safety redundancy for the unit.

[0032] 4. The performance continuous optimization module of this solution uses reinforcement learning algorithms to change this situation. This enables the control system of this solution to have the ability to adapt and evolve throughout its life cycle, and to continuously maintain the operating performance at the optimal state. Attached Figure Description

[0033] The present invention will be further explained below with reference to the accompanying drawings and embodiments;

[0034] Figure 1 This is a flowchart of a noise reduction and vibration control system for offshore wind turbines according to the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0036] Example 1:

[0037] Please see Figure 1 The present invention provides a noise reduction and vibration control system for offshore wind turbines, comprising: a data acquisition module for acquiring platform motion information, vibration spectrum data and sound pressure signal characterizing the operating status of the wind turbine;

[0038] The coupling state assessment module is used to determine the coupling state assessment index based on platform motion information, vibration spectrum data and sound pressure signal through a preset deep neural network model.

[0039] The collaborative control strategy generation module is used to determine the target stiffness and target damping based on the coupling state evaluation index and vibration spectrum data.

[0040] The control mode switching module is used to obtain the maximum tilt angle of the platform in the platform motion information, and when the maximum tilt angle of the platform exceeds the preset tilt angle threshold, the preset maximum safety value is used as the system stiffness and damping.

[0041] The performance continuous optimization module is used to adjust the preset control parameters adopted by the cooperative control strategy generation module based on the coupling state evaluation index, the underwater noise level determined by the sound pressure signal, and the system control energy consumption.

[0042] In this embodiment, the noise reduction and vibration control system of the offshore wind turbine is configured to cope with the complex dynamic loads in deep sea areas. Through the precise coordination of five major modules, it achieves intelligent response to the frequency-varying coupling effect of the floating wind turbine. The data acquisition module, as the perception foundation of the system, captures comprehensive information in real time, including the platform's six-degree-of-freedom motion, vibration of key components, underwater sound pressure, and even wind, waves, and currents, providing high-dimensional real-time data input for subsequent decision-making. The coupling state assessment module, collaborative control strategy generation module, control mode switching module, and performance continuous optimization module, based on this data flow, construct a complete closed loop from state quantification, strategy generation, safe switching, and self-optimization. This integrated design aims to address the limitations of traditional fixed control strategies and solve the problems of nonlinear vibration and new underwater noise sources caused by frequency-varying coupling between the wind turbine and the platform. Thus, while ensuring wide-band and efficient vibration reduction performance, it also ensures the structural stability and survivability of the unit under extreme sea conditions.

[0043] Example 2

[0044] The coupling state assessment module is used to determine coupling state assessment metrics, including:

[0045] The platform motion information, vibration spectrum data and sound pressure signal are integrated into a multi-dimensional feature vector. The multi-dimensional feature vector is then input into a deep neural network model. Through the nonlinear fitting operation of the deep neural network model, the coupling state evaluation index is output.

[0046] The collaborative control strategy generation module is used to determine the target stiffness, including:

[0047] The dominant vibration frequency is extracted from the vibration spectrum data. Based on the dominant vibration frequency and the preset system reference natural frequency, the frequency detuning factor is determined. The coupling state evaluation index and the frequency detuning factor are combined, and the target stiffness is determined according to the preset reference stiffness.

[0048] To enable those skilled in the art to better implement the present invention, a preferred embodiment of the preset deep neural network model is described in detail below. Considering that the input data contains signals with temporal characteristics such as platform motion and vibration, the present invention preferably employs a long short-term memory network capable of processing time-series data. For example, the model can be configured as an LSTM network containing two hidden layers, each containing 64 neurons. After the LSTM layer, a fully connected layer is connected, and the output is passed through a sigmoid activation function to ensure the coupling state evaluation index S. c The values ​​are normalized to the [0,1] interval;

[0049] In this embodiment, the function of the coupled state evaluation module is to accurately quantify the state of the highly nonlinear, time-varying coupled system of wind turbine-platform in real time; the module integrates the multi-source heterogeneous data provided by the data acquisition module into a multi-dimensional feature vector X after time synchronization and normalization processing.

[0050] As a concrete example, a multidimensional feature vector X can be composed of the following synchronized and normalized features:

[0051] Platform motion information; statistical characteristics of the platform's six degrees of freedom motion over the past one-minute time window, including sway, pitch, roll, heave, sway, and roll, such as mean, standard deviation, and maximum value;

[0052] Vibration spectrum data; current dominant vibration frequency ω v The corresponding vibration amplitude A v And the total vibrational energy within the key frequency band;

[0053] Sound pressure signal; the total sound pressure level acquired and processed by the hydrophone array;

[0054] By concatenating the above features together, an instantaneous multidimensional feature vector X is formed and input into the deep neural network model.

[0055] This vector is then fed into a deep neural network model trained offline. This pre-built model uses its nonlinear fitting capabilities to learn autonomously from complex data and extract core features that can profoundly characterize the intensity of system coupling, thus overcoming the limitations of traditional analytical models in accurately describing the process.

[0056] The technical background is that the coupling effect between wind turbines and floating platforms is a highly nonlinear and time-varying process. Traditional analytical models are unable to accurately describe its inherent laws, while deep neural networks can learn from massive historical operating data and corresponding risk labels to construct an accurate mapping from multidimensional inputs to risk assessment.

[0057] The module's structure and parameter sources are as follows: The output is a dimensionless coupling state evaluation index S. c The calculation process is represented by the following formula:

[0058] S c =F θ (X)

[0059] Among them, S c This represents the coupling state assessment index, with values ​​normalized to the [0,1] interval. Larger values ​​indicate stronger system coupling effects and a higher risk of vibration and instability. F θ This represents a deep neural network model with a specific network layer and activation function; θ is the model weight parameter, obtained by offline supervised learning on a training dataset containing historical working condition data and corresponding coupled risk labels; X is the input multidimensional feature vector generated by the data acquisition module.

[0060] The underlying logic lies in the calculated coupling state evaluation index S c It is directly transmitted to the collaborative control strategy generation module as one of the key decision bases for dynamically adjusting the system stiffness; the collaborative control strategy generation module receives S c Subsequently, in order to effectively suppress the frequency-varying coupling effect, the target stiffness K was determined. t This module considers both vibration severity and frequency matching; it is based on the reference stiffness K. b Superimposed with a coupling index S c The adjustment amount, driven by the frequency detuning factor ζ, enables intelligent adjustment of the system stiffness.

[0061] The theoretical basis of this target stiffness calculation formula is active structural control theory. The technical solution is to actively change the natural frequency of the system by significantly increasing the stiffness when the system is close to resonance, thereby avoiding the risk of resonance. At the same time, the severity of the coupling state is also taken into consideration to ensure that the increase in stiffness matches the actual risk level and achieve precise control.

[0062] Target stiffness K t The composition and parameter sources are as follows:

[0063] K t =K b +(α s S c +α ω ζ)K r

[0064] Among them, K t K represents the target stiffness. b The preset reference stiffness is determined based on the unit design specifications and offline simulation; α s ,αω It is a dimensionless gain dynamically optimized by the performance continuous optimization module; S c ζ is the coupling index output by the coupling state assessment module; K is the frequency detuning factor used to quantify the resonance risk. r It is a reference stiffness determined according to the unit design specifications;

[0065] Based on the above results, the application of this formula makes the adjustment of system stiffness sensitive to both the severity of coupling and the proximity of resonance; on the one hand, when the coupling index S c When the frequency increases, the system will correspondingly increase its stiffness to enhance stability. On the other hand, when the dominant vibration frequency approaches the system's natural frequency, i.e., when ζ approaches 1, the stiffness will also be greatly increased to change the natural frequency and effectively avoid resonance. This coordinated adjustment mechanism enables the system to not only suppress existing vibrations when facing complex sea conditions, but also to proactively avoid potential resonance risks. Its vibration reduction effect far exceeds that of control strategies that rely on a single index.

[0066] This invention achieves precise quantification and forward-looking assessment of system risks; existing technologies struggle to effectively characterize the complex nonlinear coupling between the wind turbine and the platform; the coupling state assessment module in this solution processes platform motion information, vibration spectrum data, and sound pressure signals acquired by the data acquisition module using a pre-defined deep neural network model; this pre-defined model, through offline supervised learning of massive historical operating data, is endowed with powerful nonlinear fitting capabilities; its core operation can be abstracted into a function S. c =F θ (X), the practical significance of this formula lies in mapping a high-dimensional, multi-source real-time feature vector X to a dimensionless coupling state evaluation index S in the interval [0,1]. c The indicator S c It provides a precise and quantifiable risk basis for subsequent control decisions that was previously unattainable with technology, enabling the system to shift from blind response to intelligent control based on accurate assessment.

[0067] Example 3

[0068] The collaborative control strategy generation module is used to determine the target damping, including:

[0069] The dominant vibration amplitude is extracted from the vibration spectrum data. Based on the dominant vibration amplitude and the preset maximum safe vibration amplitude, the normalized vibration amplitude is calculated. The coupling state evaluation index and the normalized vibration amplitude are combined, and the target damping is determined according to the preset benchmark damping.

[0070] The frequency detuning factor is used to quantify the proximity between the dominant vibration frequency and the system's reference natural frequency. It is calculated by substituting the difference between the dominant vibration frequency and the system's reference natural frequency into a preset exponential function model.

[0071] To further clarify, the collaborative control strategy generation module is not only adept at stiffness control but also at damping management, with the core objective of efficiently dissipating vibration energy. This module extracts the dominant vibration amplitude A from the vibration spectrum data. v And based on the maximum safe vibration amplitude A allowed in the unit design specifications. m After normalization, the normalized vibration amplitude A is obtained. ′ This normalization process transforms the absolute vibration amplitude into a relative index directly related to the safety margin, making damping adjustment more relevant to engineering practice. At the same time, the frequency detuning factor ζ, which is used to assist in stiffness calculation, is also clearly defined mathematically here. It accurately quantifies the resonance risk through a pre-defined exponential function model.

[0072] The frequency detuning factor ζ is defined based on resonance theory. Its purpose is to create a continuous, smooth index to describe the system's proximity to the resonance point, allowing for weighting in the control law and thus enhancing control as the system approaches resonance. The logic of damping calculation follows the principle of energy dissipation, with the increase in damping reflected by amplitude and the system coupling risk by S. c This embodies direct linkage, enabling precise energy dissipation;

[0073] The composition and parameter sources of the frequency detuning factor ζ are as follows:

[0074]

[0075] Where ζ represents the frequency detuning factor, with a value between 0 and 1; exp is an exponential function; ω v It is the dominant vibration frequency extracted from the real-time vibration spectrum; ω b It is based on the reference stiffness K b The preset reference natural frequency is determined by the equivalent mass of the system; σ is a preset frequency sensitivity parameter used to define the degree of frequency proximity, and its dimension is consistent with the frequency, such as rad / s;

[0076] To clarify, the parameter σ physically characterizes the natural frequency ω around the reference that the system needs to focus on. b The frequency bandwidth; its value determines the sensitivity range of the stiffness control system to resonance risk; a small σ value means that only when the dominant vibration frequency ω v Very close to ω bThe system will only respond violently when the frequency is low; a larger σ value will allow the system to start adjusting its stiffness over a wider frequency range. In practice, the value of σ can be determined based on statistical analysis of the wind and wave spectrum of the target sea area or offline simulation results. Its goal is to cover the main range in which the dominant excitation frequency of the unit may drift during operation. For example, σ can be set to 0.1 rad / s based on historical data analysis.

[0077] Target damping D t The composition and parameter sources are as follows:

[0078] D t =D b +βS c (A ′ ) 2 D r

[0079] Where D t Indicates target damping; D b The preset reference damping is determined according to the unit design specifications and energy consumption analysis; β is the dimensionless damping gain set by the continuous performance optimization module; S c It is an indicator for assessing coupling risk or coupling status; A ′ It is the normalized vibration amplitude, calculated as follows:

[0080] A ′ =A v / A m

[0081] Among them, A v To control the vibration amplitude in real time, A m To preset the maximum safe vibration amplitude; D r It is a preset reference damping determined according to the unit design specifications;

[0082] Based on the above calculations, the system achieves intelligent coordinated control of stiffness and damping; when the coupling state intensifies, S c When the stiffness is increased, the system not only avoids the risk of resonance by adjusting the stiffness, but also simultaneously increases the damping to dissipate vibrational energy; the two complement each other. The precise quantification of the frequency detuning factor ζ makes the stiffness adjustment highly forward-looking, while the target damping D... t The calculations ensure that the efficiency of energy dissipation is directly related to the actual vibration amplitude. This variable stiffness-variable damping synergistic strategy, compared with any single control method, can more effectively suppress the complex vibration of floating wind turbines under broadband and time-varying excitation, thereby significantly reducing structural fatigue and underwater noise.

[0083] This invention constructs an active broadband vibration reduction mechanism that coordinates stiffness and damping adjustment; existing vibration reduction technologies mostly employ a single fixed control strategy, which is difficult to adapt to the time-varying characteristics of the excitation frequency; the coordinated control strategy generation module of this scheme exhibits a fundamental difference; when determining the target stiffness K... t When, its calculation formula K t =K b +(α s S c +α ω ζ)K r It possesses profound physical implications; this formula does not simply increase rigidity, but rather reduces the coupling risk S. c Combined with a frequency detuning factor ζ characterizing the resonant proximity; frequency detuning factor It quantifies the real-time dominant vibration frequency ω using an exponential function. v With the system's reference natural frequency ω b The degree of deviation; when ω v Approaching ω b When ζ approaches 1, the stiffness adjustment increases sharply, its function being to actively change the system's natural frequency to avoid resonance; simultaneously, the target damping D... t Calculation of D t =D b +βS c (A ′ ) 2 D r Then the coupling risk S c With the normalized dominant vibration amplitude A ′ The correlation aims to precisely configure energy-dissipating damping based on actual vibration energy and system risk level; this synergy and dynamic adjustment of stiffness and damping enables the system to effectively suppress vibrations under broadband and time-varying excitation, solving the problem of insufficient vibration reduction performance of existing technologies under frequency-varying loads.

[0084] Example 4

[0085] The control mode switching module is specifically used for:

[0086] Monitor the platform's maximum tilt angle in real time and compare it with a preset tilt angle threshold.

[0087] When the maximum tilt angle of the platform exceeds the preset tilt angle threshold, the emergency stabilization mode is triggered, and the preset maximum safety value is used as the system stiffness and damping.

[0088] When it is determined that the maximum tilt angle of the platform does not exceed the preset tilt angle threshold, the normal cooperative control mode is maintained, and the target stiffness and target damping determined by the cooperative control strategy generation module are adopted.

[0089] In this embodiment, the control mode switching module, as part of the system's safety redundancy design, has a simple and reliable operating logic that does not rely on complex model calculations, ensuring response speed and reliability under extreme operating conditions. The module's function is to monitor in real time the platform's maximum tilt angle φ provided by the data acquisition module. p and compare it with a preset physical threshold φ th Perform continuous comparisons; this threshold φ th The method for determining this is, for example, setting it to 15°, which, based on a comprehensive consideration of unit design specifications, hydrodynamic simulation analysis, and instability risk assessment, represents the critical point for maintaining the platform's attitude safely; when the system determines the platform's maximum tilt angle φ... p When the critical safety angle is not exceeded, the system will remain in normal cooperative control mode, with the cooperative control strategy generation module dynamically outputting the target stiffness and damping; conversely, once φ is detected... p >φ th The module will immediately trigger the emergency stabilization mode; in this mode, the original cooperative control strategy is forcibly suspended, and the system stiffness and damping are directly set to the preset maximum safety value K. m and D m These two emergency parameters are derived from the structural ultimate load analysis, with the goal of providing the platform with the maximum restoring torque and the strongest motion suppression capability. This clear switching mechanism ensures that when the platform attitude enters an extremely dangerous state, the control objective can instantly switch from performance optimization to ensuring survival safety, thereby protecting the structural integrity of the floating wind turbine platform at the most critical moment.

[0090] This invention establishes a rigid switching logic to ensure survival under extreme working conditions; conventional shock absorption systems experience a sharp performance degradation or even failure when the platform's tilt exceeds a certain threshold; the control mode switching module in this solution is designed to address this deficiency; this module monitors the platform's maximum tilt angle φ in real time. p Once it exceeds the preset tilt angle threshold φ determined based on hydrodynamic simulation and instability risk assessment, th The system will unconditionally switch to emergency stability mode, forcibly adopting the preset maximum safety value K. m and D m As system stiffness and damping, this design ensures that in extreme sea conditions that could lead to disaster, the control objective shifts from performance optimization to ensuring structural survival, providing robust safety redundancy for the unit.

[0091] Example 5

[0092] The continuous performance optimization module is specifically used for:

[0093] The sound pressure signal is processed to determine the normalized underwater noise level, and the monitoring system is used to determine the normalized control energy consumption. Based on the coupling state evaluation index, the normalized underwater noise level and the normalized control energy consumption, and combined with the preset weight coefficients, the real-time reward is calculated. Based on the real-time reward, the preset control parameters are adjusted through the preset reinforcement learning algorithm.

[0094] The immediate reward is obtained by weighted summation of the coupling state evaluation index, normalized underwater noise level, and normalized control energy consumption, and is used to unify the optimization objectives of high stability, low noise, and low energy consumption.

[0095] The preset control parameter is the dimensionless gain in the collaborative control strategy. The adjustment output of the performance continuous optimization module is used as the input parameter of the collaborative control strategy generation module to close the intelligent control loop of the system.

[0096] To ensure the feasibility of this invention, a preferred preset reinforcement learning algorithm—the Deep Deterministic Policy Gradient Algorithm (DDPG)—is provided here. DDPG is an algorithm based on the Actor-Critic framework and applicable to continuous action spaces, making it very suitable for the continuous dimensionless gain α required in this invention. s ,α ω ,β is adjusted in specific scenarios;

[0097] Actor network; responsible for determining the current state from the S... c N ′ ,P ′ This constitutes a direct output of a specific action, that is, determining the gain parameter α. s ,α ω The specific value of β;

[0098] The Critic network is responsible for evaluating the quality of actions selected by the Actor network. Its inputs are the state and the action output by the Actor, and its output is the Q-value of the state-action pair, which is the expected long-term cumulative reward.

[0099] By training a Critic network to more accurately evaluate the value of actions and using the evaluation results of the Critic network to guide the updates of the Actor network, the Actor learns to select actions that can obtain higher long-term rewards. This process often combines techniques such as experience replay and target networks to improve the stability and efficiency of learning.

[0100] In the top-level intelligent design of this embodiment, the performance continuous optimization module uses a reinforcement learning algorithm to optimize the core dimensionless gain parameter α in the cooperative control strategy. s ,α ωThe module performs long-term, adaptive online optimization; its theoretical basis is the Markov decision process, and its design goal is to find a control strategy that maximizes the long-term cumulative reward for the control system. To achieve this goal, the module first processes the sound pressure signal collected by the hydrophone array to determine the normalized underwater noise level N. ′ And monitor the system itself to obtain normalized control energy consumption P ′ An instant reward function is defined to unify the three mutually constraining optimization objectives of high stability, low noise, and low energy consumption into a single scalar reward.

[0101] The design of this reward function draws on multi-objective optimization theory. The solution is to provide a clear and quantitative feedback signal for the reinforcement learning agent to guide its learning process. By adjusting the weight coefficients, the operator can flexibly guide the system's self-optimization direction according to the operational priorities of different periods, such as environmental protection priority or energy efficiency priority.

[0102] Instant Rewards R t The composition and parameter sources are as follows:

[0103] R t =w s (1-S c )-w n N ′ -w p P ′

[0104] Among them, R t This represents an immediate reward, which is a dimensionless scalar; w s ,w n ,w p These are the top-level strategy weighting coefficients set by the operator, corresponding to the importance of stability, noise, and energy consumption, respectively; S c N ′ ,P ′ These are normalized state variables that characterize coupling risk, underwater noise, and control energy consumption, respectively. The data are derived from the coupling state assessment module, sound pressure signal processing, and system energy consumption monitoring.

[0105] The actions of the reinforcement learning agent are the dimensionless gain α in the cooperative control strategy. s ,α ω β is fine-tuned; at each decision point, the agent adjusts the current state S. c N ′ ,P ′ Based on historical experience, perform an action to adjust the gain and obtain a reward R from the environment. tThrough continuous state-action-reward cycles, the agent's internal policy network gradually learns a parameter adjustment strategy that maximizes long-term cumulative rewards; the output of this module is the continuously optimized gain α. s ,α ω The β value is directly fed back to the collaborative control strategy generation module as an input parameter when calculating the target stiffness and target damping. This feedback mechanism closes the intelligent control loop of perception-evaluation-decision-execution-optimization of the entire system. This enables the entire noise reduction and vibration control system to have the ability to adapt and learn, and to continuously fine-tune its control logic based on the feedback of its long-term operation data in a specific sea area, so as to achieve adaptive optimization of vibration reduction and noise reduction performance and achieve the best global operating effect.

[0106] This invention introduces a long-term performance optimization closed loop with adaptive optimization capabilities; existing technologies, once deployed, have fixed control strategies that cannot adapt to environmental changes or aging; this solution's continuous performance optimization module uses reinforcement learning algorithms to change this situation; and it utilizes an instantaneous reward function R aimed at unifying high stability, low noise, and low energy consumption optimization objectives. t =w s (1-S c )-w n N ′ -w p P ′ This module addresses the dimensionless gain α in the collaborative control strategy. s α ω β is continuously fine-tuned online; the practical significance of this process is that the system can learn from its own long-term operating data and autonomously find the combination of control parameters that maximizes the long-term cumulative reward; the adjustment output of this module serves as the input parameter of the collaborative control strategy generation module, closing the intelligent control loop of the entire system; this enables the control system of this scheme to have the ability to adapt and evolve throughout its life cycle, and to continuously maintain its operating performance at the optimal state.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A noise reduction and vibration control system for offshore wind turbines, characterized in that, include: The data acquisition module is used to acquire platform motion information, vibration spectrum data, and sound pressure signals that characterize the operating status of the wind turbine. The coupling state assessment module is used to determine the coupling state assessment index based on platform motion information, vibration spectrum data and sound pressure signal through a preset deep neural network model. The collaborative control strategy generation module is used to determine the target stiffness and target damping based on the coupling state evaluation index and vibration spectrum data. The control mode switching module is used to obtain the maximum tilt angle of the platform in the platform motion information, and when the maximum tilt angle of the platform exceeds the preset tilt angle threshold, the preset maximum safety value is used as the system stiffness and damping. The performance continuous optimization module is used to adjust the preset control parameters adopted by the cooperative control strategy generation module based on the coupling state evaluation index, the underwater noise level determined by the sound pressure signal, and the system control energy consumption.

2. The noise reduction and vibration control system for offshore wind turbines according to claim 1, characterized in that, The coupling state evaluation module is used to determine coupling state evaluation indicators, including: The platform motion information, vibration spectrum data, and sound pressure signal are integrated into a multi-dimensional feature vector. This multi-dimensional feature vector is then input into a deep neural network model. Through the nonlinear fitting operation of the deep neural network model, a coupling state evaluation index is output.

3. The noise reduction and vibration control system for offshore wind turbines according to claim 1, characterized in that, The collaborative control strategy generation module is used to determine the target stiffness, including: The dominant vibration frequency is extracted from the vibration spectrum data. Based on the dominant vibration frequency and the preset system reference natural frequency, the frequency detuning factor is determined. The coupling state evaluation index and the frequency detuning factor are combined, and the target stiffness is determined according to the preset reference stiffness.

4. The noise reduction and vibration control system for offshore wind turbines according to claim 1, characterized in that, The collaborative control strategy generation module is used to determine the target damping, including: The dominant vibration amplitude is extracted from the vibration spectrum data. Based on the dominant vibration amplitude and the preset maximum safe vibration amplitude, the normalized vibration amplitude is calculated. The target damping is determined by combining the coupling state evaluation index and the normalized vibration amplitude, and according to the preset benchmark damping.

5. A noise reduction and vibration control system for offshore wind turbines according to claim 3, characterized in that, The frequency detuning factor is used to quantify the proximity between the dominant vibration frequency and the system's reference natural frequency. It is calculated by substituting the difference between the dominant vibration frequency and the system's reference natural frequency into a preset exponential function model.

6. The noise reduction and vibration control system for offshore wind turbines according to claim 1, characterized in that, The control mode switching module is specifically used for: Monitor the platform's maximum tilt angle in real time and compare it with a preset tilt angle threshold. When the maximum tilt angle of the platform exceeds the preset tilt angle threshold, the emergency stabilization mode is triggered, and the preset maximum safety value is used as the system stiffness and damping. When the maximum tilt angle of the platform does not exceed the preset tilt angle threshold, the normal collaborative control mode is maintained, and the target stiffness and target damping determined by the collaborative control strategy generation module are adopted.

7. The noise reduction and vibration control system for offshore wind turbines according to claim 1, characterized in that, The continuous performance optimization module is specifically used for: The sound pressure signal is processed to determine the normalized underwater noise level, and the monitoring system is used to determine the normalized control energy consumption. Based on the coupled state evaluation index, the normalized underwater noise level, and the normalized control energy consumption, and combined with the preset weight coefficients, the real-time reward is calculated. Based on the real-time reward, the preset control parameters are adjusted through a preset reinforcement learning algorithm.

8. A noise reduction and vibration control system for offshore wind turbines according to claim 7, characterized in that, The instant reward is obtained by weighted summation of the coupling state evaluation index, normalized underwater noise level, and normalized control energy consumption, and is used to unify the optimization objectives of high stability, low noise, and low energy consumption.

9. A noise reduction and vibration control system for offshore wind turbines according to claim 7, characterized in that, The preset control parameter is the dimensionless gain in the collaborative control strategy. The adjustment output of the performance continuous optimization module is used as the input parameter of the collaborative control strategy generation module to close the intelligent control loop of the system.

Images