1603results about How to "Improve robustness" patented technology

The invention belongs to the field of industrial control and computer data security, and particularly relates to a millisecond-level power failure data protection system for an industrial control panel. The invention discloses a millisecond power failure data protection system for an industrial control panel. The millisecond power failure data protection system comprises a core processing component, a real-time synchronous shadow memory component, a bus voltage monitoring component, a pre-entropy reduction logic control component, a nonvolatile storage component and a flight control bus arbitration component. The bus voltage monitoring assembly monitors a voltage drop trend and performs early warning, the pre-entropy reduction logic control assembly calculates a minimum data survival set according to energy entropy, the real-time synchronous shadow memory assembly constructs a synchronous shadow control domain and atomizes data, and the flight control bus arbitration assembly compulsively takes over a bus and freezes processor access at the moment of power failure. Atomized writing of data into the nonvolatile storage component is guaranteed. According to the invention, zero-delay data protection and millisecond-level hot start recovery at the moment of power failure are realized, and the data security and reliability of the industrial control equipment in a severe environment are enhanced.

This application discloses a method and apparatus for constructing a multi-scale adaptive gating MambaPlus network, belonging to the field of artificial intelligence and machine learning technology. The method includes: initializing the network configuration and constructing the basic structure; preprocessing the input data to generate a standard dataset; mapping the input data to the hidden space via an input mapping layer, and extracting backbone features from the Mamba backbone; constructing at least two parallel scale branches in the hidden space to obtain a multi-scale feature set; inputting the backbone features and multi-scale features into an adaptive gating module, dynamically allocating weights and adaptively fusing them through a hierarchical gating mechanism to generate fused features; further enhancing the features through cross-scale attention and feedforward enhancement, and then superimposing the residuals to generate the final discriminative features; finally, completing category prediction and model training evaluation. This application, while retaining the advantages of Mamba's long-range dependency modeling, addresses the problems of insufficient utilization of multi-scale information, poor adaptive feature fusion, and low robustness in complex scenarios.

This invention proposes a method for predicting the health status of lithium batteries based on multidimensional features and neural network constant differential equations. The method includes the following steps: S1, preprocessing the capacity data and charging stage operation data collected during lithium battery operation, and constructing features from historical health status data; S2, constructing multidimensional feature inputs for health status prediction based on the charging stage operation data; S3, inputting the multidimensional features into a gated recurrent unit network to fuse and encode the historical health status sequence and constant current charging stage features to obtain a potential feature representation characterizing the battery degradation state; S4, comparing the predicted health status value output by the neural network constant differential equation model with the corresponding actual health status value, calculating the prediction error, and evaluating the prediction accuracy. This application achieves high-precision prediction of lithium battery health status by integrating a multidimensional feature screening mechanism and a continuous-time state evolution modeling method.

The application provides a logistics transportation management method and system based on artificial intelligence, relates to the field of logistics transportation, and solves the technical problems that the prior art is difficult to balance real-time performance, distribution feasibility and resource utilization efficiency in the sudden interruption scene of a logistics vehicle, resulting in slow scheduling response, frequent secondary conflicts or failure of high-value order fulfillment. The method comprises the following steps: obtaining a set of orders to be delivered by a target vehicle at present and real-time position information; based on a preset hard constraint condition, screening candidate vehicles and constructing a task migration feasibility graph; inputting the task migration feasibility graph into a pre-trained lightweight graph neural network model to output the compatibility scores of each candidate vehicle for each order to be delivered in the order set; based on the compatibility scores, evaluating the urgency of the order to be delivered; and based on the urgency of the order to be delivered, updating the delivery task scheme of each candidate vehicle. The application is used in the process of logistics transportation.

A storage array, a memory, a reading method thereof, and an electronic device are disclosed. The storage array includes multiple storage cell subarrays, a reference cell column, and a set of reference voltage generation circuits, reference cell switching circuits, and read word line control circuits. The storage cell subarrays and reference cell column in two storage arrays are electrically connected via read bit lines and symmetrically distributed around a read decision unit. The transistors of the reference cells and storage cells are isomorphic and have the same circuit structure. The memory includes at least one pair of open storage arrays, each pair comprising a first storage array, a second storage array, and a set of read decision units. Each storage array is electrically connected to one of the two input terminals of the read decision unit. During reading, the two storage arrays provide a read voltage and a reference voltage, respectively. The two input terminals of the read decision unit have similar RC parameters, and the drift generated by the voltages at the two input terminals is also matched. This application improves storage density and data reading reliability.

A robustness measurement method for LeNet-5 networks based on adversarial boundary constraints is presented, relating to the field of deep learning model testing. The main steps include: for each training sample, dynamically generating adversarial examples based on the model's current state during training iterations; constructing a composite loss function based on standard cross-entropy loss and dynamic boundary constraint loss; performing end-to-end training on all parameters of the LeNet-5 network; and using the overall approximate robustness boundary as a measure of model robustness after training. This method effectively improves the resistance of the LeNet-5 model to fast gradient sign-based adversarial attacks without altering the basic structure of the LeNet-5 network by designing a new loss function.

The present application relates to the technical field of short-circuit current calculation, and particularly relates to a flexible DC system short-circuit current calculation method based on a physical information neural network, comprising: setting a model input feature vector, the model input feature vector being used to represent a system operating state before a fault and fault information, and performing data preprocessing on the model input feature vector; constructing a hybrid driving calculation model, the hybrid driving calculation model comprising a physical calculation module and a neural network module, and being coupled based on a preset fusion architecture; performing end-to-end training and optimization on the hybrid driving calculation model by using a preset composite loss function; and performing flexible DC system short-circuit current calculation based on the optimized hybrid driving calculation model, so that the problems of poor precision, speed and convergence, poor interpretability and weak generalization ability in the prior art are solved.

The application belongs to the technical field of positioning and measurement, and particularly relates to a target positioning method and device combining CSI coarse positioning and Y-type CCD fine correction, which comprises the following steps: S1, CSI acquisition equipment is arranged in a target area, and CSI data between a sending end and a receiving end is acquired; S2, a Y-type CCD sensor is geometrically calibrated and a coordinate system is corrected, and coordinate reconstruction is completed; S3, the CSI data is preprocessed, noise is removed, and input is normalized; S4, time domain features of the CSI are extracted by using a time domain branch; S5, frequency domain features of the CSI are extracted by using a frequency domain branch; S6, a gating fusion module is used to fuse the time domain and frequency domain features, and the approximate position coordinates of a target are output; S7, the Y-type CCD sensor is controlled to optically acquire a target area, and point light source coordinate information is acquired; and S8, error correction is performed on the data acquired by the Y-type CCD, and the accurate position coordinates of the target are output.

This invention relates to the field of atmospheric particulate matter analysis and discloses a deep learning-based method for calibrating organic aerosol concentrations. This invention addresses the technical problems in existing organic aerosol concentration calibration techniques, such as insufficient utilization of multi-source information, inadequate use of temporal evolution information, and fixed feature fusion methods lacking dynamic weight adjustment. By leveraging multi-source observation vectors, it fully utilizes observation information from different physical properties to establish a stable nonlinear mapping relationship under complex atmospheric conditions and mixed aerosol backgrounds. Through multi-source observation features and historical state features, the model can characterize the dynamic evolution of organic aerosol concentrations. Employing a bidirectional long short-term memory network and SE attention mechanism to process multi-source observation features and historical state features, it can automatically learn and dynamically adjust the importance weights of feature channels, effectively suppressing noise and redundant feature interference. Furthermore, it utilizes forward and backward time-series modeling to fully capture the contextual relationships between consecutive time points.

The application discloses a data latch circuit and an integrated circuit comprising the same, which comprises a sampling module and a level conversion module, the sampling module comprising a data end, an enable end and a capacitor C2, the data end being used for receiving backup data, the enable end being used for receiving an enable signal, the sampling module being configured to selectively charge or discharge the capacitor C2 according to the backup data when the enable signal is at a high level, and cut off the charging and discharging path of the capacitor C2 when the enable signal is at a low level, and the level conversion module being used for converting the level state stored by the capacitor C2 into a standard digital level for reading by a digital processing module in a power supply module power-on stage. The data latch circuit utilizes the charge storage characteristics of the capacitor C2 itself, locks and keeps the data before power failure, and can directly read the backup data after the power supply is re-powered, so that the whole system can quickly recover to the stable working state before power failure.

The present application relates to the technical field of feed processing, and particularly relates to a preparation method of Maotai-flavor distiller's grains fermented feed, which comprises the following steps: drying treatment based on the moisture content of Maotai-flavor distiller's grains to obtain basic distiller's grains; determining the strain ratio of brewer's yeast, bacillus subtilis and aspergillus niger based on the tannin content of the basic distiller's grains to obtain a compound fermentation inoculant; inoculating and then fermenting; determining state evaluation parameters based on the total polyphenol accumulation rate, the dissolved oxygen consumption rate and the microbial activity index during the fermentation process, and performing qualified evaluation, and dynamically adjusting when unqualified; determining the fermentation end point based on the tannin degradation rate and the total polyphenol accumulation rate; and mixing the fermented distiller's grains with the basic feed at a mass ratio of 1% after low-temperature drying of the fermented distiller's grains. The present application realizes three-layer control of raw material evaluation, process control and end point determination, significantly improves the stability and content of functional active ingredients in the product, reduces the tannin and safety risk, and has a good intestinal health improvement effect on animals.

The application discloses a vehicle aerodynamic simulation model correction method and device, electronic equipment and medium, and relates to the technical field of vehicle simulation. Wherein, by establishing the mapping relationship between the actual flow field and the simulation flow field and calculating the difference data set, then using the physical field inversion algorithm to obtain the target correction factor distribution of the to-be-corrected parameter in space, and combining the flow state characteristic quantity to train the machine learning model to obtain the target correction function, finally embedding the function into the solver to realize the bottom correction of the vehicle aerodynamic simulation model. In this way, the problem of inaccurate vehicle simulation of the vehicle aerodynamic simulation model is fundamentally improved, the corrected simulation model has spatial adaptive ability, can automatically apply differential correction according to local flow characteristics, significantly reduces the error between simulation and test, avoids the problems of product performance not meeting the standard, design change and cost waste caused by simulation deviation, and greatly improves the accuracy and robustness of simulation.

The application discloses a kind of unmanned aerial vehicle target detection methods based on feature fusion DINO, belong to target detection technical field. Including the following steps: obtaining original unmanned aerial vehicle image and pre-processing;The data after pre-processing is sent into feature extraction network Backbone and image feature extraction is carried out, output multi-scale feature map, and it is input into integrated collection feature Neck module to strengthen spatial feature and semantic feature fusion, output enhanced feature map;Fusion enhanced feature map is sent into encoder, and global feature information is highlighted through the multi-scale receptive field of hollow convolution path, finally, the class and boundary box prediction of unmanned aerial vehicle target are generated after decoder generation.The present application is based on DINO architecture, and a high-precision unmanned aerial vehicle target detection model is constructed by a multi-scale hollow convolution fusion method.The detection accuracy and robustness of small-size unmanned aerial vehicle targets are improved by effectively extracting multi-scale features and recursively fusing.

This invention provides a polarization-based saliency defect detection method, comprising the following steps: (1) constructing a polarization image dataset based on polarization images; (2) building a polarization-based saliency defect detection network structure; and (3) inputting data for network training. This invention designs a novel network structure based on an optical model to extract the physical features of polarization images, utilizing the differences in polarization characteristics of different defects for saliency detection, thus providing a new solution for defect detection in industrial scenarios.

The application discloses a robot map updating method, which comprises the following steps: step 1, a robot establishes a current map in a current environment; step 2, the current map is updated according to the angle relationship between the current map and a map stored in a historical map library last time, the updated current map is stored in the historical map library, and the updated current map is updated as the map stored in the historical map library last time, so that the current map established in step 1 is updated; wherein each frame of map stored in the same historical map library represents the current environment.

This invention discloses a lightweight few-shot incremental learning method for power line defects. The method includes: first, constructing an inspection image dataset containing basic and incremental categories, and designing a lightweight feature extraction network; training the network using the basic category data to generate prototype vectors for each category, forming an initial prototype classification model; in the incremental stage, constructing incremental prototypes for a small number of samples of the new category, and introducing a prototype transfer mechanism based on inter-class tension, adaptively adjusting the new prototype based on the geometric similarity and discrimination conflict strength between the new and existing prototypes to alleviate category conflicts; simultaneously, introducing an adaptive feature fusion mechanism that perceives discrimination uncertainty, dynamically fusing the discrimination results of the main and auxiliary branches to improve recognition stability. This invention, employing the above-mentioned lightweight few-shot incremental learning method for power line defects, can achieve efficient and stable power line defect recognition in scenarios with few samples and continuously expanding categories.

This application provides a method for adding watermarks to an image. Belonging to the field of information security, this method involves dividing an image into blocks to obtain a target number of non-overlapping blocks. The texture complexity is determined based on the number of pixels at the block boundaries and the size of the blocks. The watermark embedding amount is determined based on the texture complexity of each block, the average watermark embedding capacity, and the initial texture complexity. The average watermark embedding capacity is determined based on the total watermark information capacity and the target number. Singular value decomposition is performed on each block to obtain the maximum singular value. The singular value after watermark embedding is determined based on the maximum singular value of each block and the watermark information corresponding to the watermark embedding amount. Inverse singular value decomposition is performed on the singular values ​​after watermark embedding in each block, and each block after watermark embedding is reconstructed to obtain a watermarked image. This method can improve the amount of watermark information and the quality of the watermarked image.

This invention relates to the interdisciplinary field of mechanical engineering and optical inspection, and discloses a method and system for quality sorting of automotive stamping parts combined with optical inspection. The method includes: synchronous imaging by a high-speed linear array camera and a structured light projector via an encoder and photoelectric triggering device; fusing 3D point clouds and 2D images, and using a deep convolutional neural network for defect identification and quality grading; calculating the six-degree-of-freedom pose of the workpiece using an iterative nearest-point algorithm and principal component analysis; calculating the spatiotemporal parameters of sorting based on a predictive instruction generation module, and issuing dynamic control commands to multi-degree-of-freedom pneumatic push rods via a hard real-time communication bus to execute sorting. The system includes a synchronous triggering unit, an imaging unit, a defect identification unit, a pose calculation unit, a predictive instruction generation module, and a dynamic sorting execution unit. This invention achieves millisecond-level accurate identification and adaptive sorting, significantly improving detection accuracy, sorting efficiency, and system robustness.

The application discloses an abnormal driving behavior recognition method and system based on multi-modal data fusion, and particularly relates to the technical field of multi-vehicle cooperative driving safety control, and is used for solving the problem that the existing cooperative driving system cannot convert the abnormal behavior risk of a driver into a cooperative control instruction at the vehicle fleet level; the abnormal state of the driver is recognized by collecting and fusing multi-modal sensor data of the vehicle; when the vehicle is a lead vehicle of a vehicle fleet, the comprehensive risk of the abnormal state to the safety of the vehicle fleet is evaluated based on the abnormal state through probabilistic evolution simulation; the expected effect of different vehicle fleet control strategies is simulated according to the risk evaluation result, and the best strategy is selected; vehicle fleet risk information containing the strategy is generated and sent to all following vehicles through vehicle-to-vehicle communication; the whole process from single-vehicle driver state monitoring to vehicle fleet level cooperative risk prevention and control is realized, and the overall safety of the vehicle fleet system is improved.

The present application relates to the technical field of precipitation forecast, solves the problem that the prior art is prone to underestimate the peak value of precipitation or produce false alarm under strong convective weather, and particularly relates to a short-impending precipitation forecast method fusing GNSS and water conservancy rain measurement radar, GNSS observation data and water conservancy rain measurement radar base data in a monitoring area are acquired, pretreated, corresponding atmospheric precipitable water distribution field and radar reflectivity factor distribution field are generated, based on a preset time sliding window, time sequence characteristics of the atmospheric precipitable water distribution field and the radar reflectivity factor distribution field are extracted, and the two are superimposed in channel dimension to construct a multi-channel spatio-temporal fusion input tensor. The present application overcomes the defect that the traditional extrapolation method rapidly decays with the prolongation of the forecast time, significantly improves the prediction accuracy of the rainstorm center intensity and the falling area, reduces the missed report and underestimation of the precipitation event, and solves the problem of low prediction accuracy of the traditional linear extrapolation algorithm under strong convective weather.

The application discloses a sensor fault robust multi-output soft measurement method and system based on adversarial learning, relates to a sensor fault robust multi-output soft measurement method and system, and belongs to the technical field of soft measurement. The application discloses a sensor fault robust multi-output soft measurement method and system based on adversarial learning, relates to a sensor fault robust multi-output soft measurement method and system, and belongs to the technical field of soft measurement. The application discloses a sensor fault robust multi-output soft measurement method and system based on adversarial learning, relates to a sensor fault robust multi-output soft measurement method and system, and belongs to the technical field of soft measurement. The application discloses a sensor fault robust multi-output soft measurement method and system based on adversarial learning, relates to a sensor fault robust multi-output soft measurement method and system, and belongs to the technical field of soft measurement. The application discloses a sensor fault robust multi-output soft measurement method and system based on adversarial learning, relates to a sensor fault robust multi-output soft measurement method and system, and belongs to the technical field of soft measurement. The application discloses a sensor fault robust multi-output soft measurement method and system based on adversarial learning, relates to a sensor fault robust multi-output soft measurement method and system,

The application relates to the field of radiation source identification, and specifically provides a high-explainability radiation source identification method and system, a training set radar pulse signal is acquired, a multi-domain joint time-varying fingerprint feature is extracted based on physical layer modeling, the fingerprint feature is input into a time-frequency double-branch feature fusion network to obtain time-frequency fusion features; the time-frequency fusion features are subjected to signal-level physical consistency enhancement and feature-level mixed enhancement; based on the enhanced features, an end-to-end training is conducted on the time-frequency double-branch feature fusion network through a hierarchical contrast learning framework containing intra-class contrast and inter-class prototype contrast, a radiation source identification model is obtained, and the trained radiation source identification model is used to identify the individual class of a radar pulse signal to be identified. Through the introduction of the physical layer explainable time-varying fingerprint feature, the adaptive double enhancement mechanism and the hierarchical contrast learning framework, the application realizes high-credibility, strong-adaptation and stable-identification radiation source individual identification.

A kind of distributed attack-resistant domain management and control method and system for giant constellation, the present application relates to satellite constellation management and control technical field, the present application aims at solving the problem of cooperative management and control and attack-resistant resilience maintenance of multi-layer heterogeneous giant constellation in dynamic topology environment.Technical points: the present application takes the management and control node satellite as core, combines with the cluster head satellite and ordinary satellite in domain, forms distributed management and control domain structure.The structure undertakes cross-orbit communication switching and distributed cooperative management and control function, effectively deals with the cooperative management and control challenge brought by the difference of multi-orbit satellite in operation cycle, communication topology and communication time delay;Introduce multi-element maximum entropy trust model: based on fuzzy C clustering algorithm and maximum entropy principle, construct dynamic trust evaluation system, and comprehensively quantitatively analyze the trust degree of distributed nodes.Through real-time identification of high-risk nodes and implementation of isolation measures, significantly reduce the interference of malicious satellite to inter-satellite data transmission, improve system security.The present application realizes the efficient cooperative management and control and attack-resistant ability enhancement of multi-layer heterogeneous giant constellation, provides key technical support for stable operation of giant constellation under complex threat environment.

This invention relates to a robust image semantic perception adaptive masking transmission method for vehicle-to-everything (V2X) networks, comprising: Step 1: The transmitting end performs semantic analysis on the original input image, extracts high-precision semantic features, and identifies the regions of interest (ROI) and non-ROI regions of interest (NROI); Step 2: A Boolean mask is dynamically generated based on the semantic segmentation result and SNR; Step 3: The ROI features are directly masked based on the Boolean mask; Subsequently, the processed feature map is fed into a semantic encoder to extract multi-scale semantic features; Step 4: Power normalization is performed, dynamic feature encoding is performed, and then the encoded signal is transmitted through a typical analog wireless channel, and noise is superimposed before reception and recovery; Step 5: The receiving end receives the noisy feature signal, performs channel fading compensation and preliminary denoising, and sends the decoded features into a semantic decoder; Through progressive upsampling and multi-level semantic reconstruction, multi-scale features are fused, and finally, a high-quality reconstructed image is recovered and output.

This invention discloses a pollen identification method based on dual-model decision arbitration and multimodal database association, belonging to the field of intelligent identification technology. The method includes: acquiring pollen images; inputting them into a first model based on Visual Transformer and a second model based on YOLO for parallel inference, outputting candidate categories and predicted probabilities; inputting the results into an arbitration engine; if the highest confidence categories match, a direct determination is made; otherwise, dynamic weighted arbitration is performed to obtain the final determination category; subsequently, the corresponding morphological feature text descriptions are retrieved in real time from the multimodal database, and the determination category, text description, feature heatmap, and alternative references are simultaneously pushed to the interactive interface. This invention overcomes the "black box" defect of models, achieves macro- and micro-feature complementarity and intelligent arbitration, constructs an "instant recognition and interpretation" auxiliary identification system, and improves the accuracy and interpretability of pollen identification.

The application discloses a power maintenance action prediction method based on target consistency screening and a bidirectional state space, first constructs an observation side video feature sequence, then carries out action semantic token coding, probability weighted mapping and time sequence modeling, obtains action semantic context representation, constructs a joint input vector based on the action semantic context representation, carries out time sequence modeling and distribution construction through a gate recurrent unit, obtains a feature side target distribution, and extracts an observation action representation; subsequently, a future action prediction model based on a bidirectional selective state space is used for prediction, and multiple candidate future action sequences are output; finally, an action side posterior distribution is constructed, a target consistency score is calculated, and the candidate future action sequence with the minimum score is selected as an optimal action prediction result. The application explicitly models statistical correlation and potential target distribution of maintenance actions, introduces a bidirectional selective state space diffusion generation architecture, and realizes accurate and logical prediction of power maintenance actions.

The application discloses a projection knowledge distillation method based on adaptive mask weighting, first, a relationship matrix of the features extracted by the student module based on the student network is constructed to make the information contained between adjacent pixels more diverse, second, adaptive mask matrices are respectively constructed based on the relationship matrix and the feature map of the student network for adaptive mask weighting, then a projection layer is constructed and the features weighted by the mask are projected under the guidance of the teacher network to obtain complete features approximating the teacher features, finally, the corresponding feature layer of the teacher network is used to supervise the corresponding feature layer of the student network and update the student model, the application improves the expression ability of the student network model for the learned rich information, solves the problem of limited representation ability of the student network and insufficient information utilization caused by the limited receptive field of the adjacent pixels of the student features and the student features with random mask, and improves the robustness and generalization ability of the knowledge distillation model.

This invention relates to the field of power distribution system resilience enhancement technology, specifically a method for enhancing the resilience of a mobile hydrogen energy storage power distribution system. This method includes data acquisition and parameter initialization; constructing physical constraints for the power-transportation coupled system based on a multi-microgrid reconfiguration strategy; expanding the original typhoon scenario using an enhancement strategy to generate an expanded sample set including wind and solar power output prediction errors; constructing a fuzzy uncertainty set covering multi-source uncertainties; establishing a sub-Bruker optimal scheduling model with the objective of minimizing the overall system operating cost; and performing dual transformation and linearization on the sub-Bruker optimal scheduling model to obtain the trajectory of the mobile hydrogen energy storage vehicle, charging and discharging commands, and the action sequences of distribution network segment switches and tie switches. This invention effectively solves the problems of high difficulty in multi-grid resource coordination and difficulty in accurately obtaining data distribution characteristics under extreme disasters, significantly improving the economic efficiency of post-disaster system recovery and achieving an effective balance between robustness and economy.

The application discloses a LIBS element quantitative analysis method based on double-branch feature fusion and electronic equipment, the method comprises the following steps: preprocessing the collected LIBS spectrum signal, including baseline correction and spectrum resampling; inputting the spectrum signal data after preprocessing into a double-branch feature extraction network, extracting local features through a CNN branch, and extracting global features through an MLP branch in parallel; performing adaptive weighted fusion on the local features and the global features through a gating fusion module to obtain fusion features; inputting the fusion features into a WMA-MLP model, the WMA-MLP model is based on MLP, integrates a multi-head self-attention mechanism and a residual module, is used for modeling the global dependency relationship between features, and outputs a final element quantitative analysis result. Through the parallelly arranged CNN and MLP branch feature extraction structures, more comprehensive spectrum feature representation can be obtained, and the accuracy and robustness of the spectrum analysis model are improved.

This invention relates to the field of computer-aided design technology, and more particularly to a method and system for constructing wind dams for concentrated wind energy capture. The method includes the following steps: acquiring multi-source wind field data; processing the multi-source wind field data to obtain wind field structure data; generating a guiding structure from the wind field structure data to obtain wind dam geometric structure data; performing fluid-structure interaction (FSI) simulation based on the wind field structure data and the wind dam geometric structure data to obtain FSI data; and performing multi-constraint dynamic optimization on the FSI data to obtain construction control strategy data. This invention achieves proactive guidance and efficient matching of the wind dam structure to airflow convergence behavior, improving the efficiency of concentrated wind energy capture and the reliability of structural design. By using a surrogate model and multi-constraint dynamic optimization, the highly complex simulation process is transformed into a rapidly reasoning-based decision-making process, effectively reducing computational costs.