A fan noise reduction muffler parameter optimization method and system

Abstract

The application relates to a fan noise reduction muffler parameter optimization method and system, the method first selects a benchmark overall configuration meeting heat dissipation requirements from a plurality of candidate configurations through CFD flow resistance simulation and threshold comparison; then, based on the benchmark configuration, a high-quality parameter-performance dataset is constructed through Latin hypercube sampling and acoustic simulation; further, a high-precision acoustic performance proxy model is constructed by adopting a stacked ensemble model combined with Bayesian hyperparameter optimization; finally, an optimization problem is constructed by taking the maximum transfer loss predicted by the proxy model as the target and taking flow resistance and the like as the constraint, and the marine predator algorithm is used for solving, and an optimal parameter combination is output. The application realizes source collaborative design of heat dissipation and noise reduction performance, greatly improves optimization precision, efficiency and automation level, and forms a reusable intelligent design closed loop.

Description

Technical Field

[0001] This application relates to the field of noise reduction technology, and in particular to a method and system for optimizing the parameters of a noise reduction hood for wind turbines. Background Technology

[0002] Fans are key pieces of equipment in industrial and civil sectors, and their operating noise is one of the major sources of environmental pollution. Silencers are an effective means of reducing aerodynamic and mechanical noise from fans. However, the design of silencers faces a core contradiction: to improve noise reduction (usually measured by transmission loss), parameters such as opening ratio and internal structure need to be adjusted, but this often increases airflow resistance (flow resistance), leading to poor ventilation, reduced heat dissipation efficiency, and affecting the safe and stable operation of the equipment.

[0003] Currently, the design and optimization of soundproof enclosures in the industry generally suffer from the following problems: First, the design process lacks systematic collaboration. Most solutions focus on optimizing noise reduction performance through experience or acoustic simulation, while placing flow resistance verification in the testing phase after physical prototype fabrication. This easily leads to the disruptive design flaw of "noise reduction meeting standards but heat dissipation failing," resulting in a huge waste of time and economic costs. Second, optimization methods are inefficient and yield poor results. Parameter adjustments rely on engineers' personal experience for repeated trial and error, or the use of single empirical formulas or simulation models with limited predictive accuracy. Even when traditional optimization algorithms (such as genetic algorithms and particle swarm optimization) are introduced, they often fail to efficiently obtain the global optimum due to insufficient accuracy of the objective function (performance prediction model) or the algorithm's tendency to get trapped in local optima, resulting in lengthy design cycles. Finally, the technology chain is fragmented. Configuration concept design, fluid simulation analysis, acoustic performance simulation, and parameter optimization decisions are usually completed independently by different tools or personnel. Data and processes are disconnected, failing to form a rapid iterative feedback automated design closed loop, resulting in poor reusability of technical solutions.

[0004] Therefore, there is an urgent need for a soundproof enclosure design method that can systematically balance heat dissipation and noise reduction requirements and achieve efficient, accurate, and automated parameter optimization. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and system for optimizing the parameters of a noise reduction hood for wind turbines, in order to overcome the defects mentioned in the background art.

[0006] A method for optimizing the parameters of a noise-reducing enclosure for wind turbines includes the following steps: S1. Overall configuration feasibility verification: Set the flow resistance threshold according to the fan heat dissipation requirements, and based on the flow resistance threshold, select the configuration with flow resistance lower than the threshold from a variety of candidate overall configurations as the benchmark overall configuration. S2. Construction of acoustic performance dataset: Based on the baseline overall configuration, determine multiple core structural parameters and their value ranges, and use the Latin hypercube sampling method to generate multiple sets of parameter combination samples within the value range. For each set of parameter combination samples, construct a three-dimensional model of the soundproof enclosure and perform acoustic simulation to obtain the corresponding transmission loss data, forming a parameter-transmission loss dataset. S3. Proxy Model Construction and Optimization: Based on the parameter-transfer loss dataset, a stacked ensemble model is trained as an acoustic performance proxy model. The stacked ensemble model includes multiple base learners and a meta learner. The hyperparameters of the stacked ensemble model are tuned using a Bayesian optimization algorithm to improve the prediction accuracy of the model. S4. Multi-constraint optimization solution: Taking the maximization of the transmission loss value predicted by the acoustic performance surrogate model as the objective function, and taking the flow resistance not exceeding the flow resistance threshold and the core structural parameters being within their value range as constraints, an optimization problem is established, and the ocean predator algorithm is used to solve the optimization problem, outputting the optimal combination of core structural parameters that satisfies all constraints.

[0007] As a preferred embodiment of the method for optimizing the parameters of the noise reduction cover for wind turbines in this invention, step S1, in which the configuration with flow resistance lower than the threshold is selected as the baseline overall configuration, specifically includes: calculating the flow resistance value of each candidate overall configuration through computational fluid dynamics simulation, comparing it with the flow resistance threshold, and determining the configuration with the flow resistance value lower than the flow resistance threshold as the baseline overall configuration.

[0008] As a preferred embodiment of the method for optimizing the parameters of the noise reduction hood for wind turbines in this invention, after determining the baseline overall configuration, the method further includes: verifying or optimizing the compression ratio of the baseline overall configuration, wherein the compression ratio is the ratio of the maximum cross-sectional area of ​​the throat of the airflow channel of the noise reduction hood to the minimum cross-sectional area of ​​the inlet.

[0009] As a preferred embodiment of the method for optimizing the parameters of the noise reduction hood for wind turbines in this invention, in step S2, the core structural parameters include the opening ratio, opening diameter, compression ratio, and hood length.

[0010] As a preferred embodiment of the method for optimizing the parameters of the noise reduction cover for wind turbines in this invention, step S3 involves constructing the stacked ensemble model by dividing the parameter-transfer loss dataset into a training set and a test set, training the multiple base learners using the training set, using the prediction results of each base learner on the training set as input features, and training the meta-learner to fuse the prediction results of each base learner.

[0011] As a preferred embodiment of the method for optimizing the parameters of the noise reduction shield for wind turbines in this invention, the base learner includes at least two different models among random forest, gradient boosting tree, support vector machine and BP neural network.

[0012] As a preferred embodiment of the method for optimizing the parameters of the noise reduction cover for wind turbines in this invention, step S3, which involves using a Bayesian optimization algorithm to perform hyperparameter tuning, includes: modeling the relationship between hyperparameters and model prediction error using a Gaussian process model, and iteratively selecting the combination of hyperparameters to be evaluated based on the acquisition function to minimize the prediction error of the model on the test set.

[0013] As a preferred embodiment of the noise reduction hood parameter optimization method for wind turbines in this invention, in step S4, the ocean predator algorithm adaptively switches between a global exploration strategy and a local exploitation strategy based on the ratio of the current iteration number to the maximum iteration number during the solution process. The global exploration strategy is based on Brownian motion, and the local exploitation strategy is based on Levy flight.

[0014] As a preferred method for optimizing the parameters of the noise reduction cover for wind turbines in this invention, the marine predator algorithm introduces the fish aggregation device effect during the iteration process and enables the algorithm to escape local optima through a random perturbation strategy.

[0015] A smart design system for a fan noise reduction enclosure, used to execute the above-mentioned parameter optimization method for fan noise reduction enclosures, the system comprising: The configuration screening module is used to perform the overall configuration feasibility verification; The data construction module is used to perform the construction of the acoustic performance dataset; The model training module is used to perform the construction and optimization of the proxy model; The optimization solution module is used to perform the multi-constraint optimization solution; The modules are connected in sequence to form an automated parameter optimization closed loop.

[0016] The beneficial effects of this invention are: This invention, by making CFD flow resistance analysis a mandatory entry point for configuration screening and using flow resistance value as a hard constraint throughout the process, fundamentally ensures that the final optimized solution simultaneously meets the dual physical requirements of noise reduction and heat dissipation, thus completely solving the disruptive risks and cost waste caused by the lag in performance verification in the traditional design process.

[0017] This invention integrates the advantages of multiple machine learning algorithms through a stacked ensemble model and utilizes Bayesian optimization for fine-tuning of hyperparameters. The constructed surrogate model exhibits prediction accuracy and stability far exceeding that of a single model, providing a highly reliable objective function estimate for the optimization process. Furthermore, the marine predator algorithm, with its unique adaptive search mechanism and powerful global exploration capabilities, can efficiently and robustly find the global optimum, significantly shortening the design cycle and substantially improving the performance ceiling of the final solution.

[0018] This invention seamlessly integrates multiple heterogeneous processes, including conceptual design, simulation analysis, machine learning, and intelligent optimization, forming a fully automated, closed-loop intelligent design workflow from configuration input to parameter output. This system significantly reduces reliance on manual experience and the cost of trial and error. Its core framework is highly versatile; for different wind turbine models or operating conditions, only basic input conditions (such as flow resistance thresholds and parameter ranges) need to be adjusted to quickly initiate and complete the optimized design of new products. It possesses strong engineering reuse value and potential for large-scale application, driving the transformation of wind turbine noise reduction design towards a data-driven, intelligent decision-making modern model. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the method for optimizing the parameters of a fan silencer cover according to an embodiment of this application; Figure 2 This is a schematic diagram of the opening structure of the silencer cover according to an embodiment of this application; Figure 3 This is a schematic diagram of the compression ratio structure of the silencer cover according to an embodiment of this application; Figure 4 A schematic diagram illustrating the process of constructing an integrated model using the stacking method in this application embodiment; Figure 5 This is a schematic diagram of the Bayesian optimization algorithm for hyperparameter tuning of a surrogate model according to an embodiment of this application. Figure 6 This is a schematic diagram of the ocean predator algorithm for final parameter optimization in an embodiment of this application; Figure 7 This is a cross-sectional structural diagram of the fully enclosed shell configuration according to an embodiment of this application; Figure 8 This is a cross-sectional structural diagram of an open shell configuration with a flow guide channel according to an embodiment of this application; Figure 9 This is a schematic diagram showing the comparison of transmission loss performance between the optimized anechoic chamber and the initial baseline configuration according to an embodiment of this application. Detailed Implementation

[0021] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0022] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicating the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0023] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0024] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "joining," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0025] In this application, unless otherwise expressly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "on top of," and "over" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0026] It should be noted that when an element is referred to as being "fixed to" or "set on" another element, it can be directly on the other element or there may be an intervening element. When an element is considered to be "connected to" another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "upper," "lower," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0027] According to one aspect of this application, embodiments of this application provide a method and system for optimizing the parameters of a noise-reducing hood for wind turbines, which can be found in conjunction with this description. Figures 1 to 9 The method for optimizing the parameters of the noise reduction enclosure for this fan includes the following steps: Step 1: Overall Configuration Feasibility Verification. Based on the fan's spatial dimensions, air intake and exhaust directions, and other specific installation environment and heat dissipation system design requirements, determine a clear flow resistance threshold R0. This threshold is used to judge whether heat dissipation meets the standards. Subsequently, conceive at least two theoretically feasible candidate overall configurations for the silencer enclosure, such as... Figure 7 The fully enclosed shell configuration shown is similar to... Figure 8 The open-shell configuration with flow guide channels shown is used to establish fluid domain models for each candidate overall configuration using computational fluid dynamics simulation software. Boundary conditions such as inlet velocity and outlet pressure, which match the rated operating conditions of the fan, are applied to perform steady-state or transient flow field calculations. Finally, the static pressure difference between the outlet and the inlet is extracted as the flow resistance value R. The calculated flow resistance R of each configuration is compared with the threshold R0, and all configurations that satisfy R≤R0 are selected. The one with the best overall performance is selected as the benchmark overall configuration for subsequent steps. The above steps can ensure the heat dissipation feasibility of the design from the source.

[0028] Step Two: Constructing the Acoustic Performance Dataset. After determining the baseline overall configuration, parameterize it to explore its noise reduction potential. Extract and define the core structural parameters that have the most significant impact on its transmission loss. These parameters typically include, but are not limited to: Aperture ratio α: The percentage of the area of ​​the perforated portion of a perforated plate or louver to the total area.

[0029] Opening diameter d: The diameter of a circular opening, or the equivalent hydraulic diameter.

[0030] Compression ratio β: such as Figure 2 and 3 As shown, β is defined as the ratio of the maximum cross-sectional area A at the narrowest point of the airflow channel inside the silencer to the minimum cross-sectional area B at the airflow channel inlet, i.e., β=A / B. This parameter has a significant impact on flow resistance and acoustic performance.

[0031] Silencer length L: The overall structural length along the airflow direction.

[0032] A reasonable physical value range is set for each parameter. Then, Latin hypercube sampling is used to generate N sets of uniformly distributed, non-repeating parameter combination samples in the multidimensional parameter space. For each set of samples, parametric modeling software is used to automatically generate the corresponding three-dimensional geometric model of the ambient cover. All models are imported into acoustic simulation software in batches for finite element or boundary element acoustic simulation to calculate the transmission loss TL spectrum in the target frequency band. The average transmission loss can be further calculated as a scalar performance index. Finally, a dataset containing N rows of data is formed, where each row represents the mapping relationship from (α, d, β, L) to TL, providing training materials for machine learning.

[0033] Step 3: Proxy Model Construction and Optimization. This involves building a mathematical model that can quickly and accurately predict TL (Time To Live) to replace time-consuming direct simulation. Specific implementation includes: Data preparation: The dataset obtained above is randomly divided into training set and test set according to a preset ratio.

[0034] Stacked integration model construction: such as Figure 4 As shown, a two-layer model is constructed.

[0035] Base Learner Layer: Multiple different types of basic machine learning models are selected as base learners to learn data features from different hypothesis spaces, as shown in Table 1. Multiple base learners are used, and preferred combinations may include: Random Forest, which is good at handling high-dimensional features and nonlinear relationships; Gradient Boosting Tree, which improves prediction accuracy through iteration; Support Vector Machine, which is suitable for small sample and high-dimensional pattern recognition; Backpropagation Neural Network, which has a strong nonlinear fitting ability. These base learners are trained independently using training set data.

[0036] Table 1 Stacked Model Architecture

[0037] Meta-learner layer: A relatively simple model, such as linear regression or logistic regression, is selected as the meta-learner. The prediction results of all base learners from the previous step on the training set samples are collected to form a new M-dimensional feature matrix. This new feature matrix and the original true TL values ​​are used as training data to train the meta-learner, enabling it to learn how to optimally weight and combine or fuse the prediction results of each base learner, thereby forming an initial surrogate model.

[0038] Hyperparameter optimization: To improve model accuracy, the following methods are employed: Figure 5 The Bayesian optimization algorithm shown is automatically tuned as follows: The search space for hyperparameters is defined as shown in Table 2. The prediction error of the initial surrogate model on the test set is used as the objective function. A Gaussian process model is used to probabilistically model the complex relationship between hyperparameters and the objective function value. In each iteration, the next most promising hyperparameter combination is determined and evaluated based on the acquisition function constructed by the Gaussian process model. Through multiple iterations, the hyperparameter combination that minimizes the error on the test set is efficiently found, thus obtaining the final high-precision acoustic surrogate model. Table 2 Range of Hyperparameter Values

[0039] Step 4: Multi-constraint optimization solution: Transform the design problem into a mathematical optimization problem and solve it.

[0040] Optimization model establishment: Taking the maximization of the transmission loss TL of the silencer as the core objective, the transmission loss is a key evaluation index of the acoustic performance of the silencer. Its value is strongly correlated with the core structural parameters of the silencer. The mathematical expression of the optimization objective is obtained as shown in Equation 1, and the flow resistance and structural parameters are limited to a certain range as needed.

[0041] (1) Where TL is the propagation loss. The high-precision surrogate model obtained in step three is α, where α is the opening ratio, d is the opening diameter, β is the compression ratio, and L is the length of the silencer.

[0042] Constraints: R(α,d,β,L)≤R0 flow resistance constraint, R can be quickly estimated by a simplified computational fluid dynamics response surface model, empirical formula or interpolation rule based on the results of step one.

[0043] Intelligent optimization solution: The marine predator algorithm is used to solve the above-mentioned constrained optimization problem. The algorithm flow is as follows: Figure 6 As shown.

[0044] Initialization phase: A set of candidate solutions is randomly generated within the parameter domain to form the initial predator population. The mathematical description of the optimization process of the marine predator algorithm is shown in Equation 2: (2) In the formula: , The search space range; It is a random number within the range [0,1].

[0045] Iterative optimization phase: The marine predator algorithm simulates predator behavior and iteratively updates the population position.

[0046] In the initial stage of the first iteration, when the predator's speed is faster than the prey's speed, the mathematical description of the optimization process of the marine predator algorithm based on the exploration strategy is shown in Equation 3: (3) in: The movement step size, Let be a normally distributed Brownian motion random vector. An elite matrix constructed by apex predators, To be a prey matrix with the same dimension as the elite matrix, The term-by-term multiplication operator is P=0.5, R is a uniform random vector within [0,1], n is the population size, and Iter and Max_Iter are the current and maximum number of iterations, respectively.

[0047] In the middle of the second iteration, when the predator and prey have the same speed, the prey is responsible for development based on local refinement of Levy flight, while the predator is responsible for exploration based on Brownian motion strategy. Gradually, the predator shifts from exploration strategy to development strategy. The mathematical descriptions of development and exploration are shown in Equations 4, 5, and 6: (4) (5) Where: R L It is a random vector that follows a Lévy distribution; (6) CF is an adaptive parameter used to control the predator's step size.

[0048] At the end of the third iteration, when the predator's speed is slower than the prey's speed, the predator adopts an exploitation strategy based on the Levy line, the mathematical description of which is shown in Equation 7: (7) Escaping local optima: During the iteration process, a fish aggregation device effect is introduced with a certain probability, which applies random mutations to the positions of some individuals to help the algorithm escape the local optimum trap. Its mathematical description is shown in Equation 8: (8) Where: FADs=0.2 is the influence probability, U is a binary vector, and r is a random number within [0,1]. , These are two different random indices in the prey matrix.

[0049] Evaluation and Convergence: In each generation, the surrogate model trained in step three is used to quickly evaluate the propagation loss of each individual in the population, guiding the population to evolve towards high-performance regions. Iteration continues until the stopping condition is met, and the optimal combination of core structural parameters that maximizes the propagation loss and satisfies all constraints is output.

[0050] The above method was verified using a specific wind turbine silencer design project.

[0051] Project Background: A certain model of centrifugal fan has a rated speed of 3280 r / min and a rated air volume of 20000 m³ / h. The sound pressure level of its inlet and outlet noise sources is about 80 dB. The design requirements are: the flow resistance of the added silencer must be less than 100 Pa, and the average transmission loss must be greater than 15 dB.

[0052] Step 1 Set the flow resistance threshold R0 = 100 Pa.

[0053] Two candidate configurations were designed: Configuration 1, a closed anechoic enclosure, such as... Figure 7 As shown, it is an all-metal sealed housing; the configuration is a two-open silencer, as... Figure 8 As shown, an air inlet is provided on the side of the casing.

[0054] Computational fluid dynamics simulation was performed using fluid dynamics simulation software. A polyhedral mesh was used, with the inlet set as a velocity inlet and the outlet set as a pressure outlet. The simulation results showed that for configuration 1, the flow resistance R1 = 135.7 Pa, which exceeded the allowable flow resistance threshold for fan heat dissipation and was therefore unacceptable. For configuration 2, the flow resistance R2 = 67.6 Pa, which met the flow resistance requirements and was therefore acceptable.

[0055] Conclusion: Configuration 2 was selected as the baseline overall configuration. Further analysis of the compression ratio β of Configuration 2 revealed that when β=4, the flow resistance was 94.6 Pa; when β=2.5, the flow resistance stabilized at 67.6 Pa. To allow for design margin, β=2.5 was determined as the fixed parameter for subsequent configurations.

[0056] Step Two For configuration 2, the following optimization parameters are determined: the opening ratio α is 5%~45%, the opening diameter d is 3~8mm, and the silencer length L is 0.5~1m.

[0057] Using Latin hypercube sampling, 200 samples containing α, d, and L were generated.

[0058] Generate a 3D model for each set of samples and export it in a format compatible with acoustic simulation software.

[0059] The acoustic simulation software was used to set up the simulation: the sound source was a diffused sound field applied at the inlet of the fan, and the receiving point was the outlet. The sound-absorbing material used for filling was glass wool, and its parameters are shown in Table 3. Batch calculations were performed to obtain 200 corresponding average transmission loss values ​​from 20Hz to 6000Hz, forming a dataset.

[0060] Table 3 Parameters of Sound Absorbing Materials

[0061] Execute step three The dataset is divided into training and testing sets in a 140:60 ratio.

[0062] Constructing a stacked model: A stacked architecture of three base learners and one meta-learner is adopted. The aperture ratio, aperture size, and silencer length of the training set are used as input features, and the corresponding average transmission loss from 20Hz to 6000Hz is used as the output label. The four base learners are trained separately. The prediction results of the four base learners on the training set are extracted and used as input features of the meta-learner. The average transmission loss of the original training set is used as the output label to train a logistic regression model. L2 regularization is used to avoid overfitting. Finally, the prediction error of the meta-learner on the training set is controlled within 0.8dB, forming a complete stacked proxy model.

[0063] Bayesian hyperparameter optimization: The search space is shown in Table 2. After 50 iterations, the optimal combination of hyperparameters is found.

[0064] Model validation: The TL predicted by the optimized stacked model on the test set was compared with the simulation value of the acoustic simulation software, and the mean absolute error (MAE) was calculated to be 0.9 dB, indicating that the model has extremely high accuracy.

[0065] Step 4: Component optimization problem: (9) (10) Given a population size of 50 and a maximum of 200 iterations, write an MPA algorithm program that calls the pre-trained stacked model in each generation to evaluate the propagation loss of all individuals.

[0066] Optimization results: The optimization objective shown in Equation 9 was solved using the marine predator algorithm under the constraints of Equation 10. During the iteration process, the combination of structural parameters was dynamically adjusted based on the transmission loss value. The algorithm converged in about 150 iterations and output the optimal parameter combination: opening ratio α=45%, opening diameter d=3mm, and silencing cover length L=0.6m, with a transmission loss of 24.59dB.

[0067] To verify the authenticity and validity of the optimization results, a final review will be conducted: Flow resistance verification: Substituting the optimal parameters (45%, 3mm, 0.6m, β=2.5) into the computational fluid dynamics model of configuration 2 for simulation, the flow resistance R_final=67.6Pa was obtained, which still meets the requirement of less than 100Pa.

[0068] Acoustic performance verification: High-fidelity acoustic simulation was performed on the model corresponding to the optimal parameters, and the average transmission loss TL_true = 24.5 dB was obtained, which is basically consistent with the prediction value of the surrogate model.

[0069] Performance Comparison: Compared to the unoptimized initial configuration 2, with intermediate parameters α=25%, d=5.5mm, and L=0.75m, the initial configuration 2 achieves a TL of approximately 15dB, representing a noise reduction performance improvement of approximately 63.9%, far exceeding the design requirement of 15dB. A comparison of the transmission loss spectrum before and after optimization is provided. Figure 9 As shown, the optimized curve shifts significantly upward across the entire frequency band.

[0070] According to one aspect of this application, an embodiment of this application provides an intelligent design system for a wind turbine noise reduction enclosure, used to automatically execute the above-described method for optimizing the parameters of a wind turbine noise reduction enclosure. This system modularizes and integrates the various steps of the method to form a collaborative hardware and software platform. Specifically, the system includes the following modules connected in sequence: Configuration Selection Module: This module is used to perform feasibility verification of the overall configuration. It integrates or encapsulates a computational fluid dynamics solver and is configured with a human-computer interface or configuration file input interface. Users or upstream programs can input flow resistance thresholds and basic geometric parameters or sketches of candidate overall configurations through this interface. Internally, the module automatically completes fluid domain modeling, mesh generation, boundary condition setting, flow field solving, and flow resistance value extraction. It then compares the calculation results with the thresholds and finally outputs the benchmark overall configuration identifier and related initial parameters that meet the flow resistance requirements.

[0071] Data Construction Module: This module is used to construct the acoustic performance dataset. It receives baseline configuration information from the configuration selection module. Internally, it integrates or links the batch processing control interface of the parametric modeling engine and the acoustic simulation software. After the user or system presets the value range of the core structural parameters, the module automatically calls the Latin hypercube sampling algorithm to generate samples, drives the modeling engine to generate batch 3D geometric models, and schedules the acoustic simulation software to perform automated simulation calculations. Finally, it collects, parses, and stores the transmission loss data corresponding to each parameter combination, forming a structured parameter-transmission loss dataset.

[0072] Model Training Module: This module is used to construct and optimize the surrogate model. It integrates various machine learning algorithm libraries and provides a configuration interface for stacking ensemble models. It reads the dataset generated by the data construction module and automatically performs data partitioning, training of base and meta-learners, and preliminary evaluation of model performance. More importantly, this module implements a Bayesian optimization process, automatically defining the hyperparameter space, constructing a Gaussian process surrogate model, and running optimization iterations to find the hyperparameter combination that minimizes the error on the test set. Finally, it produces and saves the optimized, high-precision acoustic surrogate model file.

[0073] The optimization solution module is used to perform the multi-constraint optimization solution. This module is the core of the system's decision-making process and integrates the optimization engine of the Marine Predator Algorithm (MPA). Users or the system set the final optimization objective and constraints in this module. The module loads the surrogate model produced by the model training module and uses it as a fast objective function calculator. During the MPA iteration process, this module generates candidate parameter combinations, calls the surrogate model for instantaneous performance prediction, and updates the population based on the prediction results and algorithm rules until the convergence condition is met. Finally, this module outputs the optimal core structure parameter combination and can provide historical data and performance curves of the optimization process for analysis.

[0074] System Workflow and Closed Loop: The four modules described above are sequentially connected through standardized data interfaces (such as databases, shared file systems, application programming interfaces, or message queues), forming a complete automated workflow. After a design task is initiated, data and instructions flow along the path of configuration selection → data construction → model training → optimization solution, requiring no manual intervention in intermediate data transformations or step connections. The results obtained from the optimization solution module can be further fed back to the parametric modeling engine to generate final production drawings, or directly output for review, thus achieving a fully automated parameter optimization closed loop from design requirement input to optimal solution output. This system can be deployed on a single high-performance workstation or in a distributed manner based on a client-server architecture to handle large-scale computing demands.

[0075] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0076] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for optimizing the parameters of a noise-reducing enclosure for wind turbines, characterized in that, Includes the following steps: S1. Overall configuration feasibility verification: Set the flow resistance threshold according to the fan heat dissipation requirements, and based on the flow resistance threshold, select the configuration with flow resistance lower than the threshold from a variety of candidate overall configurations as the benchmark overall configuration. S2. Construction of acoustic performance dataset: Based on the baseline overall configuration, determine multiple core structural parameters and their value ranges, and use the Latin hypercube sampling method to generate multiple sets of parameter combination samples within the value range. For each set of parameter combination samples, construct a three-dimensional model of the soundproof enclosure and perform acoustic simulation to obtain the corresponding transmission loss data, forming a parameter-transmission loss dataset. S3. Proxy Model Construction and Optimization: Based on the parameter-transfer loss dataset, a stacked ensemble model is trained as an acoustic performance proxy model. The stacked ensemble model includes multiple base learners and a meta learner. The hyperparameters of the stacked ensemble model are tuned using a Bayesian optimization algorithm to improve the prediction accuracy of the model. S4. Multi-constraint optimization solution: Taking the maximization of the transmission loss value predicted by the acoustic performance surrogate model as the objective function, and taking the flow resistance not exceeding the flow resistance threshold and the core structural parameters being within their value range as constraints, an optimization problem is established, and the ocean predator algorithm is used to solve the optimization problem, outputting the optimal combination of core structural parameters that satisfies all constraints.

2. The method for optimizing the parameters of a noise-reducing enclosure for wind turbines according to claim 1, characterized in that, In step S1, the step of selecting configurations with flow resistance lower than the threshold as the baseline overall configuration specifically includes: calculating the flow resistance value of each candidate overall configuration through computational fluid dynamics simulation, comparing it with the flow resistance threshold, and determining the configuration with the flow resistance value lower than the flow resistance threshold as the baseline overall configuration.

3. The method for optimizing the parameters of a noise-reducing enclosure for wind turbines according to claim 2, characterized in that, After determining the baseline overall configuration, the process further includes: verifying or optimizing the compression ratio of the baseline overall configuration, wherein the compression ratio is the ratio of the maximum cross-sectional area of ​​the throat of the silencer airflow channel to the minimum cross-sectional area of ​​the inlet.

4. The method for optimizing the parameters of a noise-reducing enclosure for wind turbines according to claim 1, characterized in that, In step S2, the core structural parameters include the opening ratio, opening diameter, compression ratio, and silencer length.

5. The method for optimizing the parameters of a noise-reducing enclosure for wind turbines according to claim 1, characterized in that, In step S3, the construction of the stacked ensemble model includes: dividing the parameter-transfer loss dataset into a training set and a test set; training the multiple base learners using the training set; using the prediction results of each base learner on the training set as input features; and training the meta-learner to fuse the prediction results of each base learner.

6. The method for optimizing the parameters of a noise-reducing enclosure for wind turbines according to claim 5, characterized in that, The base learner includes at least two different models among random forest, gradient boosting tree, support vector machine and backpropagation neural network.

7. The method for optimizing the parameters of a noise-reducing enclosure for wind turbines according to claim 5, characterized in that, In step S3, the hyperparameter tuning using the Bayesian optimization algorithm includes: modeling the relationship between hyperparameters and model prediction error using a Gaussian process model, and iteratively selecting the hyperparameter combination to be evaluated based on the acquisition function to minimize the prediction error of the model on the test set.

8. The method for optimizing the parameters of a noise-reducing enclosure for wind turbines according to claim 1, characterized in that, In step S4, during the solution process, the ocean predator algorithm adaptively switches between a global exploration strategy and a local exploitation strategy based on the ratio of the current iteration number to the maximum iteration number. The global exploration strategy is based on Brownian motion, and the local exploitation strategy is based on Levy flight.

9. The method for optimizing the parameters of a noise-reducing enclosure for wind turbines according to claim 8, characterized in that, The marine predator algorithm introduces a fish aggregation device effect during the iteration process, and uses a random perturbation strategy to enable the algorithm to escape local optima.

10. A smart design system for a fan silencer cover, characterized in that, The system is used to perform the method for optimizing parameters of a noise-reducing hood for wind turbines as described in any one of claims 1 to 9, the system comprising: The configuration screening module is used to perform the overall configuration feasibility verification; The data construction module is used to perform the construction of the acoustic performance dataset; The model training module is used to perform the construction and optimization of the proxy model; The optimization solution module is used to perform the multi-constraint optimization solution; The modules are connected in sequence to form an automated parameter optimization closed loop.

Images