Hydrodynamic effect intelligent prediction method and system based on multi-factor full-level condition embedding network

Abstract

This invention relates to the field of hydrodynamic prediction technology, specifically to an intelligent prediction method and system for hydrodynamic effects based on a multi-factor, full-level conditional embedding network. First, key influencing factor parameters of hydrodynamic effects are obtained, and standardized conditional input vectors are constructed and preprocessed. Then, an MHCE-Net is built, comprising conditional input, encoding, and mapping modules, as well as encoders, bottleneck layers, decoders, and output modules. The standardized vectors are mapped to conditional feature layers through feature enhancement and dimension matching, and the corresponding layers of the full-level embedding network are fused with convolutional feature maps. Finally, the model is trained by extracting and reconstructing the fused features, and the actual spatial distribution field of physical quantities is obtained through inverse normalization. This invention enhances feature representation capabilities through pre-encoding enhancement of multi-factor conditional vectors, enabling rapid and accurate prediction of spatial fields related to hydrodynamic effects without complex numerical simulations, significantly reducing computational costs and improving prediction efficiency and accuracy. It is suitable for hydrodynamic effect prediction needs in groundwater environments, deformation fields of water-related engineering structures, and eco-hydrological scenarios.

Description

Technical Field

[0001] This invention relates to the field of hydrodynamic prediction technology, specifically to an intelligent prediction method and system for hydrodynamic effects based on a multi-factor, full-level conditional embedding network. Background Technology

[0002] Hydrodynamic effect prediction is one of the core technologies in water-related engineering design, eco-hydrological regulation, and groundwater environment management. Its prediction accuracy and efficiency directly determine the rationality of engineering schemes, the effectiveness of ecological environment protection, and the timeliness of on-site decision-making. Hydrodynamic effects are influenced by multiple factors such as flow velocity, water level, and media blockage, exhibiting complex nonlinear spatial distribution characteristics. Traditional hydrodynamic effect prediction methods mainly rely on numerical simulation techniques such as the finite difference method and the finite element method.

[0003] Traditional methods of this kind require iterative calculations based on complex hydrodynamic control equations and numerous boundary and initial conditions. This results in high computational complexity, low efficiency, and stringent hardware requirements, making it difficult to meet the practical needs of rapid decision-making and real-time prediction in engineering projects. With the development of deep learning technology, convolutional neural networks (CNNs) and U-Net structures have gradually been applied to the field of hydrodynamic prediction. U-Net, a special type of CNN architecture, achieves multi-scale feature extraction and spatial reconstruction through an encoder-decoder framework, becoming the mainstream model for hydrodynamic spatial field prediction.

[0004] However, the traditional U-Net simply concatenates the multi-factor condition vectors at the model input layer, failing to achieve deep integration of multi-factor information with the features of each layer of the network, resulting in the inability of multi-factor coupling information to play its full role; at the same time, there is a problem of dimensionality adaptation between low-dimensional condition information and high-dimensional spatial features, and the model's ability to model complex multi-factor coupled hydrodynamic problems is insufficient, ultimately making it difficult to meet the actual engineering requirements in terms of prediction accuracy and generalization ability. Summary of the Invention

[0005] The purpose of this invention is to propose an intelligent prediction method and system for hydrodynamic effects based on a multi-factor, full-level conditional embedding network. By constructing the MHCE-Net network architecture, it achieves full-level deep fusion of multi-factor conditional information and hydrodynamic spatial characteristics, which significantly improves the prediction accuracy while reducing computational complexity, thus meeting the needs of rapid and accurate prediction in engineering sites.

[0006] According to a first aspect of the present disclosure, a smart prediction method for hydrodynamic effects based on a multi-factor, full-level conditional embedding network is provided, comprising the following steps: We obtain parameters of multiple key factors affecting hydrodynamic effects, construct a multi-factor conditional input vector, and preprocess the conditional input vector to obtain a standardized conditional input vector. A multi-factor full-level conditional embedding network MHCE-Net is constructed, which includes a conditional input module, a conditional encoding module, a conditional mapping module, an encoder, a bottleneck layer, a decoder, and an output module. The standardized conditional input vector is input into the conditional input module of MHCE-Net, and the conditional encoding module performs feature enhancement on the conditional input vector to obtain the enhanced conditional feature vector. The enhanced conditional feature vectors are mapped to conditional feature layers that match the spatial and channel dimensions of the feature maps at each level of the encoder, bottleneck layer, and decoder through the conditional mapping module. The mapped conditional feature layer is embedded into the corresponding layers of the encoder, bottleneck layer, and decoder, and fused with the convolutional feature maps of each layer. The fused features are extracted and reconstructed layer by layer, and the MHCE-Net model is trained to obtain the prediction results of the hydrodynamic effect spatial field. The prediction results are inversely normalized to obtain the spatial distribution field of the actual physical quantities, thereby enabling rapid and intelligent prediction of hydrodynamic effects.

[0007] In one embodiment, the preprocessing includes dataset partitioning and standardization; the standardization maps key factor parameters to the same numerical range; the key factors include flow rate, water level, and media blockage status.

[0008] In one embodiment, the conditional coding module includes multiple fully connected layers. These fully connected layers employ non-linear activation functions to extract and enhance features from the conditional input vector, mapping the low-dimensional conditional input vector to a high-dimensional conditional feature vector; and enhancing the conditional features. e The method of obtaining it is: In the formula, W 1. W 2. b 1. b 2 represents the parameters of the fully connected layer; c Input for multi-factor conditions; , This is the ReLU activation function.

[0009] In one embodiment, the mapping process of the conditional mapping module includes: first, mapping the enhanced conditional feature vector to a dimension consistent with the number of channels in the target feature map through a fully connected layer; then, expanding it to a dimension consistent with the spatial dimension of the target feature map through repeated vector operations; and finally, obtaining a conditional feature map that perfectly matches the dimension of the target feature map through a reshaping operation. Specifically: In the formula, E (k) This is the conditional feature map of the k-th layer. It is the ReLU activation function. Let be the conditional mapping weight matrix of the k-th layer. z For conditional coding vectors, H k The height of the feature map at layer k is given by the given information. W k The width of the feature map at layer k is given by the given value. Size is H k × W k A matrix of all ones is used to expand the conditional vector into a two-dimensional feature map.

[0010] In one embodiment, the encoder comprises multiple convolutional blocks that extract spatial features and embed conditional vectors by progressively increasing dimensionality; the bottleneck layer serves as the core of feature fusion, increasing the number of channels to strengthen the binding between the conditional vectors and spatial features; the decoder comprises multiple convolutional blocks that add skip connections on top of the encoder convolutional blocks to achieve dimensionality reduction of high-dimensional features while preserving the detailed information of the encoder, and each layer embeds conditional vectors to provide a spatial feature map of matching dimension for the output layer; each convolutional block consists of multiple convolutional layers using a same-padding pattern; the output module uses point convolutional layers with a linear activation function.

[0011] In one embodiment, the conditional feature layer and the hydrodynamic feature map are fused by splicing. After fusion, channel compression is performed through a point convolutional layer to achieve deep coupling between multi-factor conditional information and spatial feature information, as shown below: In the formula, This is the output feature map after fusing the conditional features at the k-th layer. This is the original spatial feature map of the k-th layer.

[0012] In one embodiment, the MHCE-Net model is optimized based on the mean squared error loss function during the training phase, using the Adam optimizer, and combining an early stopping mechanism with an adaptive learning rate decay strategy to improve the model's generalization ability.

[0013] According to a second aspect of the present disclosure, a smart prediction system for hydrodynamic effects based on a multi-factor, full-level conditional embedding network is provided, comprising: The input vector construction and preprocessing module obtains multiple key factor parameters that affect hydrodynamic effects, constructs multi-factor conditional input vectors, and preprocesses these conditional input vectors to obtain standardized conditional input vectors. The network construction module builds a multi-factor full-level conditional embedding network MHCE-Net, which includes a conditional input module, a conditional encoding module, a conditional mapping module, an encoder, a bottleneck layer, a decoder, and an output module. The feature enhancement module inputs the standardized conditional input vector into the conditional input module of MHCE-Net, and then enhances the conditional input vector through the conditional encoding module to obtain the enhanced conditional feature vector. The mapping module maps the enhanced conditional feature vectors to conditional feature layers that match the spatial and channel dimensions of the feature maps at each level of the encoder, bottleneck layer, and decoder through the conditional mapping module. The embedding and fusion module embeds the mapped conditional feature layer into the corresponding layers of the encoder, bottleneck layer, and decoder, and fuses it with the convolutional feature maps of each layer. The prediction module extracts and reconstructs the fused features layer by layer, trains the MHCE-Net model, and obtains the prediction results of the hydrodynamic effect spatial field. The intelligent prediction module performs inverse normalization on the prediction results to obtain the spatial distribution field of the actual physical quantities, thereby realizing rapid intelligent prediction of hydrodynamic effects.

[0014] According to a third aspect of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the memory, wherein the processor executes the program to implement the intelligent prediction method for hydrodynamic effects based on a multi-factor full-level conditional embedding network.

[0015] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the intelligent prediction method for hydrodynamic effects based on a multi-factor full-level conditional embedding network.

[0016] The advantages of the above technical solutions adopted in this invention compared with the prior art are as follows: 1. The MHCE-Net model constructed in this invention features an innovative full-level conditional embedding mechanism. It embeds multi-factor conditional feature layers into the encoder, bottleneck layer, and decoder layers, achieving full-level deep fusion of multi-factor conditional information and hydrodynamic spatial features. This effectively solves the technical problems of existing models that simply splice conditional vectors at the input layer and do not fully utilize multi-factor coupling information, thus significantly improving the prediction accuracy of hydrodynamic effects.

[0017] 2. This invention enhances the features of low-dimensional multi-factor conditional vectors through a conditional encoding module, and achieves precise dimensional matching between the conditional vectors and the feature maps of each layer of the network by combining the conditional mapping module. This fully explores the deep value of multi-factor coupled information, solves the industry problem of the difficulty in adapting low-dimensional conditional information to high-dimensional spatial features, and further optimizes the model's prediction performance.

[0018] 3. This invention adopts an end-to-end deep learning architecture, which eliminates the need for iterative calculations based on complex hydrodynamic equations, significantly reducing computational complexity and improving prediction efficiency. It can quickly output prediction results of the hydrodynamic effect spatial field, meeting the actual needs of real-time decision-making and rapid control in engineering sites. It can also effectively reduce the requirements and costs of hardware equipment, and has strong engineering applicability. Attached Figure Description

[0019] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application.

[0020] Figure 1 This is a diagram of the MHCE-Net model architecture in the embodiment; Figure 2 The example shows the scatter plots of the MHCE-Net model fitted to each dataset. Detailed Implementation

[0021] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0022] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0023] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0024] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and systems according to various embodiments of this disclosure. It should be noted that each block in a flowchart or block diagram may represent a module, segment, or portion of code, which may include one or more executable instructions for implementing the logical functions specified in the various embodiments. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutively represented blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, may be implemented using a dedicated hardware-based system that performs the specified functions or operations, or using a combination of dedicated hardware and computer instructions.

[0025] Example 1: This embodiment provides an intelligent prediction method for hydrodynamic effects based on a multi-factor, full-level conditional embedding network, including the following steps: S1. Obtain the parameters of multiple key factors affecting hydrodynamic effects, construct a multi-factor conditional input vector, preprocess the conditional input vector to obtain a standardized conditional input vector; In this embodiment, the deformation field of the interception net of the water-related project is selected as the research object of hydrodynamic effect, and the key factors affecting the hydrodynamic effect are determined to be flow velocity, water level, and blockage rate. Four flow velocity levels (0.1 m / s, 0.5 m / s, 1.0 m / s, and 3.0 m / s), two water level levels (7.03 m and 9.06 m), and two blockage rate levels (0 and 0.25) are selected. Based on these, 12 sets of working conditions are designed for numerical simulation, and the simulation results are extracted as the ground truth dataset. This dataset is divided into a training set (2 / 3), a validation set, and a test set (1 / 6 each). Each subset is processed using the StandardScaler method to obtain the standardized conditional input vector.

[0026] S2. Construct a multi-factor full-level conditional embedding network MHCE-Net, which includes a conditional input module, a conditional encoding module, a conditional mapping module, an encoder, a bottleneck layer, a decoder, and an output module; like Figure 1As shown, the conditional encoding module has 128-dimensional and 256-dimensional fully connected layers and uses the ReLU activation function to extract and enhance features from the conditional input vector, realizing the mapping from the low-dimensional conditional input vector to the high-dimensional conditional feature vector. The conditional mapping module first maps the enhanced conditional feature vector to a dimension consistent with the number of channels of the target feature map, then expands it to a dimension consistent with the spatial dimension of the target feature map through repeated vector operations, and finally obtains a conditional feature map that perfectly matches the dimension of the target feature map through a reshaping operation.

[0027] The encoder consists of two convolutional layers with 64 and 128 channels respectively. It extracts spatial features by progressively increasing the dimensionality and embedding conditional vectors. The bottleneck layer, as the core of feature fusion, increases the channel count to 256 dimensions, further strengthening the binding relationship between the conditional vectors and spatial features. The decoder also consists of two convolutional layers with 128 and 64 channels respectively. It adds skip connections to the encoder's convolutional layers to achieve dimensionality reduction of high-dimensional features while preserving the encoder's detailed information. Each layer embeds conditional vectors, and after dimensionality adjustment, provides a 64-dimensional spatial feature map for the output layer. Each convolutional layer consists of two 3×3 convolutional layers, all using the same-padding pattern. The output module has a 1×1 convolutional layer using a linear activation function.

[0028] S3. Input the standardized conditional input vector into the conditional input module of MHCE-Net, and perform feature enhancement on the conditional input vector through the conditional encoding module to obtain the enhanced conditional feature vector; S4. The enhanced conditional feature vectors are mapped to conditional feature layers that match the spatial and channel dimensions of the feature maps at each level of the encoder, bottleneck layer, and decoder through the conditional mapping module. S5. The mapped conditional feature layer is embedded into the corresponding layers of the encoder, bottleneck layer, and decoder, and fused with the convolutional feature maps of each layer. After fusion, channel compression is performed through a 1×1 convolutional layer to achieve deep coupling between multi-factor conditional information and spatial feature information. S6. The fused features are extracted and reconstructed layer by layer, and the MHCE-Net model is trained. The mean squared error loss function is used for optimization during the training phase. The Adam optimizer is used, and the model's generalization ability is improved by combining early stopping mechanism and adaptive learning rate decay strategy to obtain the prediction results of hydrodynamic effect space field. S7. The prediction results are inversely normalized to obtain the deformation field of the actual water-related engineering interception net, thus realizing rapid and intelligent prediction of hydrodynamic effects.

[0029] To verify the effectiveness and reliability of the method of this invention, the prediction performance of the Multi-Factor Full-Level Conditional Embedding Network (MHCE-Net) was evaluated on the training set, validation set, and test set. The results are as follows: Figure 2As shown. The coefficient of determination R for each dataset. 2 All values ​​are above 0.98, with sample points closely distributed around the diagonal y=x, showing no obvious systematic bias or outliers. The high-density regions are concentrated around the diagonal, indicating that the model's predictions under different operating conditions are highly consistent with the true values. The small difference between the performance metrics of the training and test sets indicates that the model constructed in this invention has not experienced overfitting, possesses good generalization ability, and can meet the accuracy and reliability requirements for hydrodynamic effect prediction in engineering scenarios.

[0030] Example 2: This embodiment provides an intelligent prediction system for hydrodynamic effects based on a multi-factor, full-level conditional embedding network, including: The input vector construction and preprocessing module obtains multiple key factor parameters that affect hydrodynamic effects, constructs multi-factor conditional input vectors, and preprocesses these conditional input vectors to obtain standardized conditional input vectors. The network construction module builds a multi-factor full-level conditional embedding network MHCE-Net, which includes a conditional input module, a conditional encoding module, a conditional mapping module, an encoder, a bottleneck layer, a decoder, and an output module. The feature enhancement module inputs the standardized conditional input vector into the conditional input module of MHCE-Net, and then enhances the conditional input vector through the conditional encoding module to obtain the enhanced conditional feature vector. The mapping module maps the enhanced conditional feature vectors to conditional feature layers that match the spatial and channel dimensions of the feature maps at each level of the encoder, bottleneck layer, and decoder through the conditional mapping module. The embedding and fusion module embeds the mapped conditional feature layer into the corresponding layers of the encoder, bottleneck layer, and decoder, and fuses it with the convolutional feature maps of each layer. The prediction module extracts and reconstructs the fused features layer by layer, trains the MHCE-Net model, and obtains the prediction results of the hydrodynamic effect spatial field. The intelligent prediction module performs inverse normalization on the prediction results to obtain the spatial distribution field of the actual physical quantities, thereby realizing rapid intelligent prediction of hydrodynamic effects.

[0031] The above modules can be deployed on the same device or distributed devices; the division of modules is only a functional logic description and does not limit the specific physical boundaries or implementation order.

[0032] Example 3: An electronic device is provided for running the aforementioned "Intelligent Prediction Method for Hydrodynamic Effects Based on Multi-Factor Full-Level Conditional Embedding Network". The electronic device includes a processor, a memory, and optional communication interfaces / display devices / input devices, etc.; the memory stores a computer program that can run on the processor, and when the processor executes the program, it implements steps S1 to S7 of the method described in Embodiment 1, specifically including but not limited to: S1. Obtain the parameters of multiple key factors affecting hydrodynamic effects, construct a multi-factor conditional input vector, preprocess the conditional input vector to obtain a standardized conditional input vector; S2. Construct a multi-factor full-level conditional embedding network MHCE-Net, which includes a conditional input module, a conditional encoding module, a conditional mapping module, an encoder, a bottleneck layer, a decoder, and an output module; S3. Input the standardized conditional input vector into the conditional input module of MHCE-Net, and perform feature enhancement on the conditional input vector through the conditional encoding module to obtain the enhanced conditional feature vector; S4. The enhanced conditional feature vectors are mapped to conditional feature layers that match the spatial and channel dimensions of the feature maps at each level of the encoder, bottleneck layer, and decoder through the conditional mapping module. S5. Embed the mapped conditional feature layer into the corresponding layers of the encoder, bottleneck layer, and decoder, and fuse it with the convolutional feature maps of each layer; S6. Extract and reconstruct the fused features layer by layer, train the MHCE-Net model, and obtain the prediction results of the hydrodynamic effect spatial field. S7. Perform inverse normalization on the prediction results to obtain the actual physical quantity spatial distribution field, thereby realizing rapid and intelligent prediction of hydrodynamic effects.

[0033] The electronic device hardware can be one of a server, personal computer, workstation, industrial controller, edge computing device, or mobile terminal; the processor can be a general-purpose CPU, GPU, NPU, FPGA, or a combination thereof; the memory can be RAM, ROM, flash memory, or disk array. The device can interact with local / remote data storage (acquiring observation data and outputting inversion results) through a communication interface. The above hardware configuration does not constitute a limitation of the present invention.

[0034] Example 4: A computer-readable storage medium storing a computer program, which, when run on a processor of an electronic device, causes the program to execute the method steps S1 to S7 described in Embodiment 1; the storage medium may be a disk, optical disk, flash memory, solid-state drive, read-only memory, random access memory, or any combination of the above media.

[0035] Those skilled in the art will understand that the modules or steps described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, which can then be stored in a storage device for execution by a computer device. Alternatively, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. This disclosure is not limited to any particular combination of hardware and software.

[0036] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0037] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A smart prediction method for hydrodynamic effects based on multi-factor, full-level conditional embedding networks, characterized in that, Includes the following steps: We obtain parameters of multiple key factors affecting hydrodynamic effects, construct a multi-factor conditional input vector, and preprocess the conditional input vector to obtain a standardized conditional input vector. A multi-factor full-level conditional embedding network MHCE-Net is constructed, which includes a conditional input module, a conditional encoding module, a conditional mapping module, an encoder, a bottleneck layer, a decoder, and an output module. The standardized conditional input vector is input into the conditional input module of MHCE-Net, and the conditional encoding module performs feature enhancement on the conditional input vector to obtain the enhanced conditional feature vector. The enhanced conditional feature vectors are mapped to conditional feature layers that match the spatial and channel dimensions of the feature maps at each level of the encoder, bottleneck layer, and decoder through the conditional mapping module. The mapped conditional feature layer is embedded into the corresponding layers of the encoder, bottleneck layer, and decoder, and fused with the convolutional feature maps of each layer. The fused features are extracted and reconstructed layer by layer, and the MHCE-Net model is trained to obtain the prediction results of the hydrodynamic effect spatial field. The prediction results are inversely normalized to obtain the spatial distribution field of the actual physical quantities, thereby enabling rapid and intelligent prediction of hydrodynamic effects.

2. The intelligent prediction method for hydrodynamic effects based on a multi-factor, full-level conditional embedding network according to claim 1, characterized in that, The preprocessing includes dataset partitioning and standardization; the standardization maps key factor parameters to the same numerical range; the key factors include flow rate, water level, and media blockage status.

3. The intelligent prediction method for hydrodynamic effects based on a multi-factor, full-level conditional embedding network according to claim 1, characterized in that, The conditional coding module includes multiple fully connected layers. These fully connected layers employ non-linear activation functions to extract and enhance features from the conditional input vector, mapping the low-dimensional conditional input vector to a high-dimensional conditional feature vector; and enhancing the conditional features. e The method of obtaining it is: In the formula, W 1. W 2. b 1. b 2 represents the parameters of the fully connected layer; c Input for multi-factor conditions; , This is the ReLU activation function.

4. The intelligent prediction method for hydrodynamic effects based on a multi-factor, full-level conditional embedding network according to claim 1, characterized in that, The mapping process of the conditional mapping module includes: first, mapping the enhanced conditional feature vector to a dimension consistent with the number of channels in the target feature map through a fully connected layer; then, expanding it to a dimension consistent with the target feature map space through repeated vector operations; and finally, obtaining a conditional feature map that perfectly matches the dimension of the target feature map through a reshaping operation. Specifically: In the formula, E (k) This is the conditional feature map of the k-th layer. It is the ReLU activation function. Let be the conditional mapping weight matrix of the k-th layer. z For conditional coding vectors, H k The height of the feature map at layer k is given by the given information. W k The width of the feature map at layer k is given by the given value. Size is H k × W k A matrix of all ones is used to expand the conditional vector into a two-dimensional feature map.

5. The intelligent prediction method for hydrodynamic effects based on a multi-factor, full-level conditional embedding network according to claim 1, characterized in that, The encoder contains multiple layers of convolutional blocks, which extract spatial features by progressively increasing the dimensionality and embedding them into the working condition vector; the bottleneck layer serves as the core of feature fusion, increasing the number of channels to strengthen the binding between the condition vector and the spatial features; The decoder contains multiple layers of convolutional blocks, with skip connections added on top of the encoder's convolutional blocks to achieve dimensionality reduction of high-dimensional features while preserving the encoder's detailed information. Each layer embeds a conditional vector to provide a spatial feature map of the matching dimension for the output layer. Each convolutional block consists of multiple convolutional layers and uses a same-padding pattern. The output module uses point convolutional layers with a linear activation function.

6. The intelligent prediction method for hydrodynamic effects based on a multi-factor, full-level conditional embedding network according to claim 1, characterized in that, The conditional feature layer and the hydrodynamic feature map are fused by concatenation. After fusion, channel compression is performed through a point convolutional layer to achieve deep coupling between multi-factor conditional information and spatial feature information, as shown below: In the formula, This is the output feature map after fusing the conditional features at the k-th layer. This is the original spatial feature map of the k-th layer.

7. The intelligent prediction method for hydrodynamic effects based on a multi-factor, full-level conditional embedding network according to claim 1, characterized in that, The MHCE-Net model is optimized during the training phase based on the mean squared error loss function, using the Adam optimizer, and combining an early stopping mechanism with an adaptive learning rate decay strategy to improve the model's generalization ability.

8. A smart prediction system for hydrodynamic effects based on a multi-factor, full-level conditional embedding network, characterized in that, include: The input vector construction and preprocessing module obtains multiple key factor parameters that affect hydrodynamic effects, constructs multi-factor conditional input vectors, and preprocesses these conditional input vectors to obtain standardized conditional input vectors. The network construction module builds a multi-factor full-level conditional embedding network MHCE-Net, which includes a conditional input module, a conditional encoding module, a conditional mapping module, an encoder, a bottleneck layer, a decoder, and an output module. The feature enhancement module inputs the standardized conditional input vector into the conditional input module of MHCE-Net, and then enhances the conditional input vector through the conditional encoding module to obtain the enhanced conditional feature vector. The mapping module maps the enhanced conditional feature vectors to conditional feature layers that match the spatial and channel dimensions of the feature maps at each level of the encoder, bottleneck layer, and decoder through the conditional mapping module. The embedding and fusion module embeds the mapped conditional feature layer into the corresponding layers of the encoder, bottleneck layer, and decoder, and fuses it with the convolutional feature maps of each layer. The prediction module extracts and reconstructs the fused features layer by layer, trains the MHCE-Net model, and obtains the prediction results of the hydrodynamic effect spatial field. The intelligent prediction module performs inverse normalization on the prediction results to obtain the spatial distribution field of the actual physical quantities, thereby realizing rapid intelligent prediction of hydrodynamic effects.

9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and running thereon, characterized in that, When the processor executes the program, it implements the intelligent prediction method for hydrodynamic effects based on multi-factor full-level conditional embedding networks as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the intelligent prediction method for hydrodynamic effects based on a multi-factor full-level conditional embedding network as described in any one of claims 1-7.

Images