A Hydraulic Design Data Modeling Method and System for Submersible Well Pumps

Abstract

The present invention discloses a method for hydraulic design data modeling of a submersible well pump. The method includes five steps: sensing and acquisition, hydraulic data enhancement, hydraulic performance prediction, intelligent material matching, and modeling output. The present invention relates to the technical field of pump hydraulic modeling analysis, specifically referring to a method and system for hydraulic design data modeling of a submersible well pump. By using a physics-informed adversarial generative network combined with a spatio-temporal Kriging interpolation algorithm, three-dimensional flow field data is enhanced; a multi-modal graph neural network integrating physical control losses is constructed to achieve dynamic prediction of performance indicators such as pump body efficiency, head, and net positive suction head; based on a corrosion-wear coupling model and a reinforcement learning algorithm, the lifespan, cost, and processability of candidate materials are evaluated, and the optimal combination is intelligently recommended, providing systematic support for the design of high-efficiency, low-cost, and long-life submersible well pumps.

Description

Technical Field

[0001] The present invention relates to the technical field of pump hydraulic modeling analysis, and particularly to a method and system for hydraulic design data modeling of a submersible well pump. Background Art

[0002] The method for hydraulic design data modeling of a submersible well pump refers to the process of quantitatively modeling and performance prediction of the internal flow field distribution and hydraulic performance indexes of a submersible well pump by using basic data such as pump body structure parameters, operating conditions, and fluid characteristics through theoretical calculations, numerical simulations, or data-driven algorithms. The role of this method is to provide a quantitative basis for pump body structure optimization, operating state evaluation, and material selection, and it is one of the core technical means to ensure the efficient, safe, and long-term operation of a submersible well pump.

[0003] However, in the existing methods for hydraulic design data modeling of submersible well pumps, there are technical problems that the existing methods usually divide hydraulic simulation, performance analysis, and material selection into multiple links, with inconsistent data interfaces and poor physical consistency, making it difficult to support rapid iterative design; in the existing hydraulic data enhancement process, there are technical problems that current conventional data simulation methods face large computational overheads and low sampling densities, while traditional data-driven interpolation methods are difficult to balance local accuracy and physical rationality; in the existing hydraulic performance prediction process, there are technical problems that most traditional machine learning methods rely on large-scale data regression fitting, making it difficult to introduce physical law constraints, and they have weak processing capabilities for heterogeneous information (such as materials, structures, operating conditions) and poor generalization; in the existing intelligent material matching process, there are technical problems that existing material selection is mostly based on empirical methods or static scoring methods, unable to meet the trade-off between life changes and performance under different flow fields and environmental conditions, resulting in unclear optimal material selection strategies and poor adaptability to operating conditions. Summary of the Invention

[0004] In view of the above situation, to overcome the defects of the existing technology, the present invention provides a method and system for hydraulic design data modeling of submersible well pumps. In the existing methods for hydraulic design data modeling of submersible well pumps, there are technical problems that the existing methods usually divide hydraulic simulation, performance analysis, and material selection into multiple links, with inconsistent data interfaces and poor physical consistency, making it difficult to support rapid iterative design. This solution creatively adopts an integrated optimization method for hydraulic design data modeling of submersible well pumps that combines hydraulic data enhancement, hydraulic performance prediction, and intelligent material matching, realizing the integrated linkage of hydraulic performance modeling and material matching decision-making, effectively improving the overall efficiency and scientificity of data-driven design; in view of the technical problems that in the existing hydraulic data enhancement process, the current conventional data simulation methods face large computational overhead and low sampling density, and traditional data-driven interpolation methods are difficult to balance local accuracy and physical rationality, this solution creatively adopts a method that combines a physics-informed generative adversarial network and improved spatio-temporal Kriging interpolation for hydraulic data enhancement, realizing high-resolution, physically consistent three-dimensional flow field reconstruction under low-sample and high-complexity flow field conditions, effectively improving the modeling accuracy and simulation restoration degree; in view of the technical problems that in the existing hydraulic performance prediction process, most traditional machine learning methods rely on large-scale data regression fitting, are difficult to introduce physical law constraints, and have weak processing capabilities for heterogeneous information (such as materials, structures, operating conditions) and poor generalization ability, this solution creatively adopts a multi-modal graph neural network combined with physical control embedding for hydraulic performance prediction, breaking through the limitations of traditional models in processing unstructured parameters, material properties, and flow field data, and realizing accurate prediction of pump body performance (efficiency, head, net positive suction head) and structural optimization feedback under complex operating conditions; in view of the technical problems that in the existing intelligent material matching process, most existing material selection methods are based on empirical methods or static scoring methods, unable to meet the trade-off between life changes and performance under different flow fields and environmental conditions, resulting in unclear optimal material selection strategies and poor operating condition adaptability, this solution creatively adopts a depth reinforcement learning method specialized for wear and corrosion modeling for material performance evaluation and matching, realizing optimal material recommendation under multiple objectives and multiple operating conditions, and significantly improving the adaptability and service reliability of submersible well pumps in high-corrosion and high-wear environments.

[0005] The technical solution adopted by the present invention is as follows: A method for hydraulic design data modeling of a submersible well pump provided by the present invention, the method comprising the following steps:

[0006] Step S1: Sensing and acquisition;

[0007] Step S2: Hydraulic data enhancement;

[0008] Step S3: Hydraulic performance prediction;

[0009] Step S4: Intelligent material matching;

[0010] Step S5: Hydrodynamic modeling of submersible well pumps.

[0011] Further, in step S1, the sensing and acquisition is used to collect multi-dimensional raw data required for hydrodynamic design data modeling of submersible well pumps. Specifically, data is collected by deploying an Internet of Things sensor array to obtain a raw data set for hydrodynamic design of submersible well pumps.

[0012] The raw data set for hydrodynamic design of submersible well pumps includes vibration sensing data, pressure sensing data, temperature sensing data, flow rate sensing data, sediment content sensing data, and pH value sensing data.

[0013] Further, in step S2, the hydrodynamic data enhancement is used to perform enhancement and optimization processing on the raw data. Specifically, based on the raw data set for hydrodynamic design of submersible well pumps, a method combining a physics-informed generative adversarial network and improved spatio-temporal Kriging interpolation is adopted to perform hydrodynamic data enhancement to obtain enhanced three-dimensional flow field hydrodynamic data, including the following steps:

[0014] Step S21: Construct a standard generative adversarial network, which is used as the basic network architecture for hydrodynamic physics constraint data enhancement. Specifically, a standard generative adversarial network including a generator and a discriminator is constructed for generative adversarial training.

[0015] The generator is used to output a predicted generated value of the flow field based on the raw data set for hydrodynamic design of submersible well pumps.

[0016] The discriminator specifically distinguishes the predicted generated value of the flow field from the real fluid mechanics simulation data through a three-dimensional convolutional neural network.

[0017] The predicted generated value of the flow field is used to represent the hydrodynamic velocity vector and pressure distribution.

[0018] Step S22: Embedding of physical loss. Specifically, a physical constraint generator loss function is constructed. Based on the standard adversarial loss, a physical constraint calculation function for mass conservation and momentum conservation is introduced to obtain a physical loss function, and the physical loss function is used to optimize the generator parameters.

[0019] Step S23: Construct an improved spatio-temporal Kriging interpolation algorithm, which is used to perform local prediction generation on data at other positions of submersible well pumps where sensor arrays are not arranged. Specifically, by combining the spatial position and time information of the water body, covariance information is fitted and the model hyperparameters are optimized through maximum likelihood estimation to establish a spatio-temporal composite kernel function, and an improved spatio-temporal Kriging interpolation model is constructed. Through the improved spatio-temporal Kriging interpolation model, the collected historical data and real-time data are fused to perform transient flow field prediction to obtain local transient flow field prediction data.

[0020] Step S24: Interpolation prediction enhancement, which is used to enhance the authenticity of the local transient flow field prediction data. Specifically, on the basis of the improved spatio-temporal Kriging interpolation algorithm, vibration acceleration is introduced as a dynamic perturbation weight term, and according to the dynamic perturbation weight term, interpolation weighting is performed on the local transient flow field prediction data to realize the simulation and optimization of the unsteady perturbation in the interpolation process, and interpolation prediction enhanced data is obtained;

[0021] Step S25: Hydraulic data enhancement. Specifically, through the construction of the standard generative adversarial network and the embedding of the physical loss, a physical information generative adversarial network is constructed to obtain physically constrained generated flow field data. Through the construction of the improved spatio-temporal Kriging interpolation algorithm and the interpolation prediction enhancement, spatio-temporal interpolation prediction is performed to obtain spatio-temporal interpolation enhanced flow field data. By fusing the physically constrained generated flow field data and the spatio-temporal interpolation enhanced flow field data, enhanced three-dimensional flow field hydraulic data is obtained.

[0022] Further, in step S3, the hydraulic performance prediction is used to dynamically predict the hydraulic performance of the submersible pump for well use. Specifically, based on the enhanced three-dimensional flow field hydraulic data, a multi-modal graph neural network combined with physical control embedding is used to perform hydraulic performance prediction to obtain hydraulic performance prediction reference data, including the following steps:

[0023] Step S31: Construct a flow field prediction subnet. Specifically, a standard deep neural network including a hidden layer and an output layer is constructed, and the physical loss function is introduced as a loss term to construct a flow field prediction subnet;

[0024] Step S32: Improvement of the physical control loss. Specifically, in combination with the physical loss function, a joint loss function is constructed to obtain a physical control joint loss function, and based on the physical control joint loss function, the training and optimization of the flow field prediction subnet are carried out;

[0025] The calculation formula of the physical control joint loss function is:

[0026] L F =a1·L BCE +a2·L phy ;

[0027] In the formula, L F is the physical control joint loss function, a1 is the cross-entropy loss weight, L BCE is the cross-entropy loss function of the flow field prediction, a2 is the physical control loss weight, and L phy is the physical loss function;

[0028] Step S33: Construct a multi-modal graph neural subnet. Specifically, construct a heterogeneous graph of the pump body and combine it with a standard graph attention neural network to construct a multi-modal graph neural subnet, obtaining a pump condition perception graph neural network;

[0029] The heterogeneous graph of the pump body includes a pump body node set and a pump body edge set;

[0030] The pump body node set includes geometric nodes, pump material nodes, and pump condition nodes;

[0031] The pump body edge set includes material-geometric connection edges and condition-structure interaction edges;

[0032] The material-geometric connection edges are used to represent the strength and corrosion resistance of the pump body material; the condition-structure interaction edges are used to represent the stress on the impeller by the rotational speed;

[0033] Step S34: Integrate the hydraulic performance prediction output. Specifically, construct a max pooling layer, map the output hidden state of the pump condition perception graph neural network to obtain a global feature vector, and map the global feature vector to the prediction output of the hydraulic performance prediction index;

[0034] Step S35: Hydraulic performance prediction. Specifically, through the construction of the flow field prediction subnet, the improvement of the physical control loss, the construction of the multi-modal graph neural subnet, and the integration of the hydraulic performance prediction output, perform flow field prediction through the flow field prediction subnet, perform performance prediction of the submersible well pump through the multi-modal graph neural subnet, train a hydraulic performance prediction model, and use the hydraulic performance prediction model based on the enhanced three-dimensional flow field hydraulic data to perform hydraulic performance prediction, obtaining hydraulic performance prediction reference data.

[0035] Further, in step S4, the intelligent material matching is used to construct a recommended optimal material optimization plan for the submersible well pump. Specifically, based on the enhanced three-dimensional flow field hydraulic data and the hydraulic performance prediction reference data, adopt a depth reinforcement learning method specialized for wear and corrosion modeling to perform material performance evaluation and matching, obtaining recommended data for the optimal material combination, including the following steps:

[0036] Step S41: Corrosion modeling. Specifically, based on the enhanced three-dimensional flow field hydraulic data, construct a corrosion modeling equation to predict the corrosion thickness of the submersible well pump material, obtaining corrosion prediction reference data;

[0037] The corrosion modeling equation includes a corrosion medium concentration term, a pH value term, and a material diffusion characteristic term, and performs corrosion depth integral calculation based on Fick's second law of diffusion to obtain the corrosion depth value;

[0038] Step S42: Wear modeling. Specifically, adopt a material wear modeling method that combines the particle erosion mechanism to construct an improved particle equation for the flow field pressure distribution, conduct wear prediction, and obtain a wear prediction reference value;

[0039] The improved particle equation for the flow field pressure distribution conducts comprehensive wear modeling by combining force, sliding, material hardness, and particle size effects;

[0040] Step S43: Coupled modeling specialization. Specifically, based on the corrosion prediction reference data and the wear prediction reference value, conduct unified loss modeling, and through constructing an overall model of the pump material with adaptive adjustment, conduct modeling calculation of the corrosion and wear loss rate to obtain a total loss prediction reference value;

[0041] Step S44: Construct a reinforcement learning architecture. Specifically, construct a standard deep Q-network as the learning framework for the material recommendation strategy, and through the definition of the state space and the action space, construct the reinforcement learning architecture, and achieve policy iteration update through the greedy strategy;

[0042] Step S45: Improve the reward construction. Specifically, conduct reinforcement reward improvement by combining material life, processing cost, and processing feasibility to obtain an improved reward function, which is used for intelligent material matching reinforcement learning;

[0043] Step S46: Intelligent material matching. Specifically, through the constructed reinforcement learning architecture and the improved reward construction, conduct reinforcement learning for intelligent material matching. In the training stage, the reinforcement learning agent conducts policy learning in the virtual environment, utilizes the target network and the main network switching mechanism until stable convergence, and in the recommendation stage, through maximizing the Q value and Pareto non-dominated sorting, screen out the optimal material combination in the three-objective space of performance-cost-processability to obtain the optimal material combination recommendation data.

[0044] Further, in step S5, the hydraulic modeling of the submersible well pump is used to generate a digital twin model of the hydraulic data of the submersible well pump. Specifically, combine the hydraulic performance prediction reference data and the optimal material combination recommendation data, and conduct data integration to obtain the parametric comprehensive reference data for the hydraulic modeling of the submersible well pump;

[0045] The parametric comprehensive reference data for the hydraulic modeling of the submersible well pump specifically includes operating condition parameters, structural parameters, hydraulic performance parameters, material matching parameters, and model meta-information parameters.

[0046] An intelligent relay protection hidden danger detection system provided by the present invention includes a sensing and acquisition module, a hydraulic data enhancement module, a hydraulic performance prediction module, an intelligent material matching module, and a hydraulic modeling module for submersible well pumps;

[0047] The said sensing and acquisition module is used for sensing and acquisition. Through sensing and acquisition, the original data set of the hydraulic design of the submersible well pump is obtained, and the original data set of the hydraulic design of the submersible well pump is sent to the hydraulic data enhancement module;

[0048] The said hydraulic data enhancement module is used for hydraulic data enhancement. Through hydraulic data enhancement, the enhanced three-dimensional flow field hydraulic data is obtained, and the enhanced three-dimensional flow field hydraulic data is sent to the hydraulic performance prediction module and the intelligent material matching module;

[0049] The said hydraulic performance prediction module is used for hydraulic performance prediction. Through hydraulic performance prediction, the reference data for hydraulic performance prediction is obtained, and the reference data for hydraulic performance prediction is sent to the intelligent material matching module and the hydraulic modeling module of the submersible well pump;

[0050] The said intelligent material matching module is used for intelligent material matching. Through intelligent material matching, the recommended data of the optimal material combination is obtained, and the recommended data of the optimal material combination is sent to the hydraulic modeling module of the submersible well pump;

[0051] The said hydraulic modeling module of the submersible well pump is used for the hydraulic modeling of the submersible well pump. Through the hydraulic modeling of the submersible well pump, the parametric comprehensive reference data for the hydraulic modeling of the submersible well pump is obtained.

[0052] The beneficial effects achieved by the present invention using the above solution are as follows:

[0053] (1) Aiming at the technical problems existing in the existing hydraulic design data modeling method of submersible well pumps, where the existing methods usually divide hydraulic simulation, performance analysis, and material selection into multiple links, with inconsistent data interfaces and poor physical consistency, making it difficult to support rapid iterative design, this solution creatively adopts an integrated optimization method for the hydraulic design data modeling of submersible well pumps that combines hydraulic data enhancement, hydraulic performance prediction, and intelligent material matching, realizing the integrated linkage of hydraulic performance modeling and material matching decision-making, and effectively improving the overall efficiency and scientificity of data-driven design;

[0054] (2) Aiming at the technical problems existing in the existing hydraulic data enhancement process, where the current conventional data simulation methods face problems such as large computational overhead and low sampling density, and the traditional data-driven interpolation methods are difficult to balance local accuracy and physical rationality, this solution creatively adopts a method that combines a physics-informed adversarial generative network and improved spatio-temporal Kriging interpolation for hydraulic data enhancement, realizing the high-resolution and physically consistent three-dimensional flow field reconstruction under low-sample and high-complex flow field conditions, and effectively improving the modeling accuracy and simulation restoration degree;

[0055] (3) In view of the technical problems existing in the existing hydraulic performance prediction process, such as most traditional machine learning methods relying on large-scale data regression fitting, being difficult to introduce physical law constraints, having weak processing capabilities for heterogeneous information (such as materials, structures, working conditions), and poor generalization, this solution creatively uses a multi-modal graph neural network combined with physical control embedding for hydraulic performance prediction, breaking through the limitations of traditional models in processing unstructured parameters, material properties, and flow field data, and achieving accurate prediction of pump body performance (efficiency, head, suction specific speed) under complex working conditions and feedback for structural optimization;

[0056] (4) In view of the technical problems existing in the existing intelligent material matching process, such as most existing material selections being based on empirical methods or static scoring methods, being unable to meet the trade-off between life changes and performance under different flow fields and environmental conditions, resulting in an unclear optimal material selection strategy and poor working condition adaptability, this solution creatively uses a depth reinforcement learning method specialized for wear and corrosion modeling for material performance evaluation and matching, achieving optimal material recommendation under multiple objectives and multiple working conditions, and significantly improving the adaptability and service reliability of submersible well pumps in high-corrosion and high-wear environments. Brief Description of the Drawings

[0057] Figure 1 It is a schematic flow chart of a hydraulic design data modeling method for a submersible well pump provided by the present invention;

[0058] Figure 2 It is a schematic diagram of a hydraulic design data modeling system for a submersible well pump provided by the present invention;

[0059] Figure 3 It is a schematic flow chart of hydraulic data enhancement in step S2;

[0060] Figure 4 It is a schematic flow chart of hydraulic performance prediction in step S3;

[0061] Figure 5 It is a schematic flow chart of intelligent material matching in step S4.

[0062] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention. Detailed Embodiments

[0063] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0064] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the drawings. These are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention.

[0065] Example 1. Refer to Figure 1 , a method for modeling hydraulic design data of a submersible pump for wells provided by the present invention, the method comprising the following steps:

[0066] Step S1: Sensing and acquisition;

[0067] Step S2: Hydraulic data enhancement;

[0068] Step S3: Hydraulic performance prediction;

[0069] Step S4: Intelligent material matching;

[0070] Step S5: Hydraulic modeling of the submersible pump for wells.

[0071] By performing the above operations, in view of the technical problems existing in the existing method for modeling hydraulic design data of a submersible pump for wells, where the existing method usually divides hydraulic simulation, performance analysis, and material selection into multiple links, with inconsistent data interfaces and poor physical consistency, making it difficult to support rapid iterative design, this solution creatively adopts an integrated optimization method for modeling hydraulic design data of a submersible pump for wells that combines hydraulic data enhancement, hydraulic performance prediction, and intelligent material matching, achieving an integrated linkage between hydraulic performance modeling and material matching decision-making, and effectively improving the overall efficiency and scientificity of data-driven design.

[0072] Example 2. Refer to Figure 1 and Figure 2 , in Step S1, the sensing and acquisition is used to collect multi-dimensional original data required for modeling hydraulic design data of a submersible pump for wells. Specifically, by deploying an Internet of Things sensor array, data acquisition is performed to obtain an original dataset of hydraulic design of a submersible pump for wells;

[0073] The deployment of the Internet of Things sensor array is specifically arranged on the impeller shaft, diffuser outlet, and seal cavity of the submersible pump for wells;

[0074] The data acquisition includes historical data acquisition and real-time data acquisition;

[0075] The original data set for the hydraulic design of the submersible pump for wells includes vibration sensing data, pressure sensing data, temperature sensing data, flow rate sensing data, sediment content sensing data, and pH value sensing data.

[0076] Example 3, refer to Figure 1 and Figure 3 , based on the above example, in step S2, the hydraulic data enhancement is used to perform enhancement and optimization processing on the original data. Specifically, according to the original data set for the hydraulic design of the submersible pump for wells, a method combining a physics-informed generative adversarial network and improved spatio-temporal Kriging interpolation is adopted to perform hydraulic data enhancement, and enhanced three-dimensional flow field hydraulic data is obtained, including the following steps:

[0077] Step S21: Construct a standard generative adversarial network, which is used as the basic network architecture for hydraulic physics constraint data enhancement. Specifically, by constructing a standard generative adversarial network including a generator and a discriminator, generative adversarial training is performed;

[0078] The generator is used to output the predicted generated value of the flow field according to the original data set for the hydraulic design of the submersible pump for wells;

[0079] The discriminator specifically distinguishes the predicted generated value of the flow field from the real fluid mechanics simulation data through a three-dimensional convolutional neural network;

[0080] The predicted generated value of the flow field is used to represent the hydraulic velocity vector and pressure distribution;

[0081] Step S22: Physical loss embedding, specifically, construct a physical constraint generator loss function. On the basis of the standard adversarial loss, introduce the physical constraint calculation functions of mass conservation and momentum conservation to obtain the physical loss function, and use the physical loss function to optimize the generator parameters;

[0082] The calculation formula of the physical loss function is:

[0083] ;

[0084] In the formula, L phy is the physical loss function, is the weight of the mass conservation term, as a whole is the mass conservation constraint, where, is the divergence vector of the velocity field, is the vector differential operator, which is used to represent the local fluid volume conservation, is the weight of the momentum conservation term, as a whole is the momentum conservation constraint, where, as a whole is the reciprocal of the velocity field with respect to time, which is used to represent the unsteady characteristics of the velocity changing with time, and u is the velocity field vector. is the convective derivative term, and v is the dynamic viscosity coefficient. is the fluid velocity diffusion vector. is the fluid density value. is the pressure gradient vector, ||·|| 2 is the sum of squared residuals operator;

[0085] Step S23: Construct an improved spatio-temporal Kriging interpolation algorithm for locally predicting and generating data at other positions of the submersible pump in the well where the sensor array is not arranged. Specifically, by combining the spatial position and time information of the water body, fitting the covariance information and optimizing the model hyperparameters through maximum likelihood estimation, establishing a spatio-temporal composite kernel function, constructing an improved spatio-temporal Kriging interpolation model, and through the improved spatio-temporal Kriging interpolation model, fusing the collected historical data and real-time data to perform transient flow field prediction and obtain local transient flow field prediction data;

[0086] The calculation formula of the improved spatio-temporal Kriging interpolation algorithm is:

[0087] ;

[0088] where k(·) is the Kriging interpolation kernel function for outputting the interpolation result, s i is the spatial coordinate of the first sampling point i, s j is the spatial coordinate of the second sampling point j, t i is the sampling time of the first sampling point i, t j is the sampling time of the second sampling point j, i is the index of the first sampling point, j is the index of the second sampling point, is the signal variance value, which is used to represent the overall intensity of the flow field fluctuation, l s is the spatial scale parameter, l t is the time scale parameter, is the noise control function for controlling whether to add a noise term, which takes the value of 0 when i≠j and 1 when i = j, is the noise variance parameter;

[0089] Step S24: Interpolation prediction enhancement is used to enhance the authenticity of the local transient flow field prediction data. Specifically, on the basis of the improved spatio-temporal Kriging interpolation algorithm, vibration acceleration is introduced as a dynamic perturbation weight term, and according to the dynamic perturbation weight term, the local transient flow field prediction data is interpolated and weighted to realize the non-steady perturbation simulation and optimization of the interpolation process and obtain interpolation prediction enhancement data;

[0090] The calculation formula of the dynamic perturbation weight term is:

[0091] ;

[0092] where w i (t) is the dynamic disturbance weight value of the first sampling point i at time t, is the target predicted spatial coordinate, is the target predicted time point, is the vibration sensitivity coefficient, is the vibration acceleration vector;

[0093] Step S25: Hydraulic data enhancement. Specifically, through the construction of the standard generative adversarial network and the embedding of the physical loss, a physical information generative adversarial network is constructed to obtain physically constrained generated flow field data. And through the construction of the improved spatio-temporal Kriging interpolation algorithm and the interpolation prediction enhancement, spatio-temporal interpolation prediction is performed to obtain spatio-temporal interpolation enhanced flow field data. And by fusing the physically constrained generated flow field data and the spatio-temporal interpolation enhanced flow field data, enhanced three-dimensional flow field hydraulic data is obtained.

[0094] By performing the above operations, in view of the problems that in the existing hydraulic data enhancement process, the current conventional data simulation methods face large computational overhead and low sampling density, and the traditional data-driven interpolation methods are difficult to balance local accuracy and physical rationality, this solution creatively adopts a method combining a physical information adversarial generation network and an improved spatio-temporal Kriging interpolation to perform hydraulic data enhancement, realizing high-resolution and physically consistent three-dimensional flow field reconstruction under low-sample and high-complexity flow field conditions, and effectively improving the modeling accuracy and simulation restoration degree.

[0095] Example 4, refer to Figure 1 、 Figure 2 and Figure 4 , based on the above example, in step S3, the hydraulic performance prediction is used to dynamically predict the hydraulic performance of the submersible pump for well use. Specifically, based on the enhanced three-dimensional flow field hydraulic data, a multi-modal graph neural network combined with physical control embedding is used to perform hydraulic performance prediction to obtain hydraulic performance prediction reference data, including the following steps:

[0096] Step S31: Construct a flow field prediction subnet. Specifically, a standard deep neural network including a hidden layer and an output layer is constructed, and the physical loss function is introduced as a loss term to construct a flow field prediction subnet;

[0097] Step S32: Improve the physical control loss. Specifically, in combination with the physical loss function, a joint loss function is constructed to obtain a physical control joint loss function, and based on the physical control joint loss function, the training optimization of the flow field prediction subnet is performed;

[0098] The calculation formula of the physical control joint loss function is:

[0099] L F = a1·L BCE + a2·L phy ;

[0100] Wherein, L F is the physical control combined loss function, a1 is the cross-entropy loss weight, and L BCE is the cross-entropy loss function for flow field prediction, a2 is the physical control loss weight, and L phy is the physical loss function;

[0101] Step S33: Construct a multi-modal graph neural subnet, specifically by constructing a pump body heterogeneous graph and combining a standard graph attention neural network to construct a multi-modal graph neural subnet to obtain a pump operating condition perception graph neural network;

[0102] The pump body heterogeneous graph includes a pump body node set and a pump body edge set;

[0103] The pump body node set includes geometric nodes, pump material nodes, and pump operating condition nodes;

[0104] The pump body edge set includes material-geometry connection edges and operating condition-structure interaction edges;

[0105] The material-geometry connection edge is used to represent the strength and corrosion resistance of the pump body material; the operating condition-structure interaction edge is used to represent the influence of rotational speed on the impeller stress;

[0106] Step S34: Integrate the hydraulic performance prediction output, specifically by constructing a max pooling layer, mapping the output hidden state of the pump operating condition perception graph neural network to obtain a global feature vector, and mapping the global feature vector to the prediction output of the hydraulic performance prediction index;

[0107] The calculation formula for the prediction output of the hydraulic performance prediction index is:

[0108] ;

[0109] Wherein, The whole is the prediction output of the hydraulic performance prediction index, is the efficiency of the submersible borehole pump, H is the head, is the net positive suction head, and W o is the output weight, h global is the global feature vector, and b o is the output bias term;

[0110] Step S35: Hydraulic performance prediction. Specifically, through the construction of the flow field prediction subnet, the improvement of physical control loss, the construction of the multi-modal graph neural subnet, and the integration of hydraulic performance prediction output, the flow field is predicted through the flow field prediction subnet, the performance of the submersible pump for well use is predicted through the multi-modal graph neural subnet, the hydraulic performance prediction model is trained to obtain the hydraulic performance prediction model, and based on the enhanced three-dimensional flow field hydraulic data, the hydraulic performance prediction model is used to perform hydraulic performance prediction to obtain the hydraulic performance prediction reference data.

[0111] By performing the above operations, in the existing hydraulic performance prediction process, there are technical problems that most traditional machine learning methods rely on large-scale data regression fitting, it is difficult to introduce physical law constraints, and the processing ability for heterogeneous information (such as materials, structures, working conditions) is weak and the generalization ability is poor. This solution creatively uses a multi-modal graph neural network combined with physical control embedding for hydraulic performance prediction, breaking through the limitations of traditional models in processing unstructured parameters, material properties, and flow field data, and realizing accurate prediction of pump body performance (efficiency, head, net positive suction head) and structural optimization feedback under complex working conditions.

[0112] Example 5, refer to Figure 1 、 Figure 2 and Figure 5 Based on the above example, in step S4, the intelligent material matching is used to construct a recommendation for the optimal material optimization plan for the submersible pump for well use. Specifically, based on the enhanced three-dimensional flow field hydraulic data and the hydraulic performance prediction reference data, a depth reinforcement learning method specialized for wear and corrosion modeling is used to perform material performance evaluation and matching to obtain the recommended data of the optimal material combination, including the following steps:

[0113] Step S41: Corrosion modeling. Specifically, based on the enhanced three-dimensional flow field hydraulic data, a corrosion modeling equation is constructed to predict the corrosion thickness of the materials of the submersible pump for well use to obtain the corrosion prediction reference data;

[0114] The corrosion modeling equation includes a corrosion medium concentration term, a pH value term, and a material diffusion characteristic term, and the corrosion depth integral calculation is performed based on Fick's second law of diffusion to obtain the corrosion depth value;

[0115] The calculation formula of the corrosion modeling equation is:

[0116] ;

[0117] In the formula, The overall is the change rate parameter of the corrosion medium concentration, C is the corrosion medium concentration, D is the corrosion ion diffusion coefficient, is the concentration diffusion Laplacian operator, The overall corrosion intensity comprehensive rate term, where k c is the corrosion reaction rate constant, pH is the acidity and alkalinity, C Cl- It is the chloride ion concentration, which is the main corrosion factor of the corrosive medium;

[0118] The calculation formula of the corrosion depth integral is:

[0119] ;

[0120] Where, is the corrosion depth value, which is used as a reference data for corrosion prediction, T is the total prediction time, M mat is the molar mass of the pump material, is the pump material density;

[0121] Step S42: Wear modeling, specifically, using a material wear modeling method combined with a particle erosion mechanism to construct an improved flow field pressure distribution particle equation, perform wear prediction, and obtain a wear prediction reference value;

[0122] The improved flow field pressure distribution particle equation combines force, sliding, material hardness and particle size effects to perform comprehensive wear modeling, and the calculation formula is:

[0123] ;

[0124] Where V wear is the wear volume prediction value, which is used as a reference value for wear prediction, K wear is the pump material wear coefficient, F n is the normal force parameter, L is the relative sliding distance during the wear process, and H v is the hardness parameter of the pump material, The whole is the particle distribution correction factor, where d sand is the average diameter of wear and erosion particles;

[0125] Step S43: coupling modeling specialization, specifically, performing unified loss modeling based on the corrosion prediction reference data and the wear prediction reference value, and constructing an adaptively adjusted overall pump material model to perform corrosion wear loss rate modeling calculation to obtain a total loss prediction reference value;

[0126] The calculation formula of the adaptively adjusted pump material overall model is:

[0127] ;

[0128] Where, The overall value is the total material loss rate, which is used as a reference value for total loss prediction and as the basis for calculating material life bonus items. is the corrosion-dominated weight factor, and its specific calculation formula is , where is the environmental correction coefficient of the submersible pump for wells, C sand is the sand concentration in the liquid, is the corrosion depth value, V wear is the predicted value of the wear volume;

[0129] Step S44: Construct a reinforcement learning architecture, specifically by constructing a standard deep Q-network as the learning framework for the material recommendation strategy, and through the definition of the state space and the action space, construct the reinforcement learning architecture, and achieve policy iteration update through the greedy policy;

[0130] The calculation formula of the state space is:

[0131] ;

[0132] In the formula, s t is the state space parameter set, H v is the hardness parameter of the pump material, is the material yield strength parameter, pH is the pH value, is the temperature parameter, C Cl- is the chloride ion concentration, d sand is the average diameter of the wear and erosion particles, C mat is the unit price of the pump material, P proc is the processing cost of the pump material;

[0133] The definition of the action space is specifically defined through a predefined set of candidate materials, and the candidate material combinations include 304 stainless steel, 316L stainless steel, duplex stainless steel 2205, WC-10CoHVOF-coated duplex stainless steel 2205, titanium nitride-coated 316L stainless steel, and high-silicon-coated high-chromium cast iron;

[0134] Step S45: Improve the reward structure, specifically by combining the material life, processing cost, and processing feasibility to improve the reinforcement reward, obtain the improved reward function, and use it for intelligent material matching reinforcement learning;

[0135] The calculation formula of the improved reward function is:

[0136] ;

[0137] In the formula, R is the improved reward function, w1 is the weight of the material life reward term, The whole is the material life reward term, where, L max is the maximum reference life of the material, L(s t ) is the predicted life of the candidate material, and its specific calculation formula is , where is the maximum material loss, as a whole is the total material loss rate value, w2 is the weight of the cost reward item, as a whole is the cost reward item, C mat is the unit price of the pump material, C budget is the upper limit of the cost budget, w3 is the weight of the processing feasibility reward item, as a whole is the processing feasibility reward item, P proc is the processing cost of the pump material, P max is the maximum processing complexity parameter;

[0138] Step S46: Intelligent material matching, specifically, through the constructed reinforcement learning architecture and the improved reward construction, perform reinforcement learning for intelligent material matching. In the training stage, perform policy learning through a reinforcement learning agent in a virtual environment, utilize the target network and the main network switching mechanism until stable convergence, and in the recommendation stage, through maximizing the Q value and Pareto non-dominated sorting, screen out the optimal material combination in the three-objective space of performance-cost-processability, and obtain the optimal material combination recommendation data.

[0139] By performing the above operations, in view of the technical problems existing in the existing intelligent material matching process, where most of the existing material selection is based on empirical methods or static scoring methods, and it is impossible to meet the trade-off between life change and performance under different flow fields and environmental conditions, resulting in an unclear optimal material selection strategy and poor working condition adaptability, this solution creatively adopts a depth reinforcement learning method specialized for wear and corrosion modeling to evaluate and match material performance, realizes optimal material recommendation under multiple objectives and multiple working conditions, and significantly improves the adaptability and service reliability of submersible well pumps in high-corrosion and high-wear environments.

[0140] Example six, refer to Figure 1 and Figure 2 , based on the above example, in step S5, the hydraulic modeling of the submersible well pump is used to generate a digital twin model of the submersible well pump hydraulic data, specifically, by combining the hydraulic performance prediction reference data and the optimal material combination recommendation data, and performing data integration to obtain the parametric submersible well pump hydraulic modeling comprehensive reference data;

[0141] The parametric submersible well pump hydraulic modeling comprehensive reference data specifically includes working condition parameters, structural parameters, hydraulic performance parameters, material matching parameters, and model meta-information parameters;

[0142] Table 1 is a detailed data content reference information table for the comprehensive reference data of the parametric submersible pump for wells in hydraulic modeling. As shown in the table, the operating conditions parameters include the rated flow rate and rotational speed; the structural parameters include the impeller outlet diameter, number of blades, blade wrap angle parameter, impeller shaft length, impeller shaft diameter, impeller hub diameter, seal cavity length, and diffuser outlet area; the hydraulic performance parameters include the three-dimensional flow field reconstruction data and the hydraulic performance prediction result data; the material matching parameters include the material combination result, mechanical property parameters, corrosion and wear prediction life, and economic index parameters; the model meta-information parameters include the standard generative adversarial network parameters, multi-modal graph neural network model parameters, and reinforcement learning process scoring value.

[0143] Table 1 Detailed data content reference information table for the comprehensive reference data of the parametric submersible pump for wells in hydraulic modeling

[0144]

[0145] Example Seven, refer to Figure 1 and Figure 2 This embodiment is based on the above embodiments. An intelligent hidden danger detection system for relay protection provided by the present invention includes a sensing and acquisition module, a hydraulic data enhancement module, a hydraulic performance prediction module, an intelligent material matching module, and a submersible pump for wells hydraulic modeling module;

[0146] The sensing and acquisition module is used for sensing and acquisition. Through sensing and acquisition, the original data set of the hydraulic design of the submersible pump for wells is obtained, and the original data set of the hydraulic design of the submersible pump for wells is sent to the hydraulic data enhancement module;

[0147] The hydraulic data enhancement module is used for hydraulic data enhancement. Through hydraulic data enhancement, the enhanced three-dimensional flow field hydraulic data is obtained, and the enhanced three-dimensional flow field hydraulic data is sent to the hydraulic performance prediction module and the intelligent material matching module;

[0148] The hydraulic performance prediction module is used for hydraulic performance prediction. Through hydraulic performance prediction, the hydraulic performance prediction reference data is obtained, and the hydraulic performance prediction reference data is sent to the intelligent material matching module and the submersible pump for wells hydraulic modeling module;

[0149] The intelligent material matching module is used for intelligent material matching. Through intelligent material matching, the recommended data of the optimal material combination is obtained, and the recommended data of the optimal material combination is sent to the submersible pump for wells hydraulic modeling module;

[0150] The submersible pump for wells hydraulic modeling module is used for the hydraulic modeling of the submersible pump for wells. Through the hydraulic modeling of the submersible pump for wells, the comprehensive reference data of the parametric submersible pump for wells in hydraulic modeling is obtained.

[0151] It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0152] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention.

[0153] The above description of the present invention and its embodiments is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and, without departing from the gist of the present invention, design similar structural modes and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. A hydraulic design data modeling method for submersible pumps used in wells, characterized in that: The method includes the following steps: Step S1: Sensing acquisition to obtain the original data set for the hydraulic design of submersible pumps for wells; Step S2: Hydraulic data enhancement. Using the method of improving the physical information adversarial generation network combined with spatio-temporal Kriging interpolation to perform hydraulic data enhancement to obtain enhanced three-dimensional flow field hydraulic data, including the following steps: Step S21: Construct a standard generative adversarial network; Step S22: Embed physical losses. Based on the standard adversarial loss, introduce physical constraint calculation functions for mass conservation and momentum conservation; Step S23: Construct an improved spatio-temporal Kriging interpolation algorithm; Step S24: Interpolation prediction enhancement. Based on the improved spatio-temporal Kriging interpolation algorithm, introduce vibration acceleration as a dynamic perturbation weight term; Step S25: Hydraulic data enhancement; Step S3: Hydraulic performance prediction. Using a multi-modal graph neural network combined with physical control embedding to perform hydraulic performance prediction to obtain reference data for hydraulic performance prediction, including the following steps: Step S31: Construct a flow field prediction subnet; Step S32: Improve physical control losses. Combine physical loss functions to construct a joint loss function; Step S33: Construct a multi-modal graph neural subnet; Step S34: Integrate the output of hydraulic performance prediction; Step S35: Hydraulic performance prediction; Step S4: Intelligent material matching. Using a deep reinforcement learning method specialized for wear and corrosion modeling to perform material performance evaluation and matching to obtain recommended data for the optimal material combination, including the following steps: Step S41: Corrosion modeling; Step S42: Wear modeling; Step S43: Specialize coupling modeling; Step S44: Construct a reinforcement learning architecture; Step S45: Improve reward construction. Improve the reinforcement reward by combining material life, processing cost, and processing feasibility; Step S46: Intelligent material matching; Step S5: Hydraulic modeling of submersible pumps for wells to obtain parametric comprehensive reference data for the hydraulic modeling of submersible pumps for wells.

2. A hydraulic design data modeling method for a submersible pump for wells according to claim 1, characterized in that: In step S1, the sensing acquisition is used to collect multi-dimensional original data required for the hydraulic design data modeling of submersible pumps for wells. Specifically, data is collected by deploying an Internet of Things sensor array to obtain the original data set for the hydraulic design of submersible pumps for wells; The original data set for the hydraulic design of submersible pumps for wells includes vibration sensing data, pressure sensing data, temperature sensing data, flow sensing data, sediment concentration sensing data, and pH value sensing data.

3. A hydraulic design data modeling method for a submersible well pump according to claim 2, characterized in that: In step S2, the hydraulic data enhancement is used to perform enhancement and optimization processing on the original data. Specifically, based on the original data set for the hydraulic design of submersible pumps for wells, the method of improving the physical information adversarial generation network combined with spatio-temporal Kriging interpolation is used to perform hydraulic data enhancement to obtain enhanced three-dimensional flow field hydraulic data, including the following steps: Step S21: Construct a standard generative adversarial network, which is used as the basic network architecture for hydraulic physical constraint data enhancement. Specifically, a standard generative adversarial network including a generator and a discriminator is constructed for generative adversarial training; The generator is used to output the generated value of the flow field prediction based on the original data set for the hydraulic design of submersible pumps for wells; The discriminator specifically uses a three-dimensional convolutional neural network to distinguish between the predicted generated values of the flow field and the real hydrodynamic simulation data; The predicted generated values of the flow field are used to represent the hydraulic velocity vector and pressure distribution; Step S22: Physical loss embedding, specifically constructing a physical constraint generator loss function. On the basis of the standard adversarial loss, introducing physical constraint calculation functions for mass conservation and momentum conservation to obtain a physical loss function, and using the physical loss function to optimize the generator parameters; Step S23: Construct an improved spatio-temporal Kriging interpolation algorithm for local prediction generation of data at other positions of the submersible pump for wells where sensor arrays are not arranged. Specifically, combining the spatial position and time information of the water body, fitting covariance information and optimizing the model hyperparameters through maximum likelihood estimation, establishing a spatio-temporal composite kernel function, constructing an improved spatio-temporal Kriging interpolation model, and through the improved spatio-temporal Kriging interpolation model, fusing the collected historical data and real-time data to perform transient flow field prediction to obtain local transient flow field prediction data; Step S24: Interpolation prediction enhancement, used to enhance the authenticity of the local transient flow field prediction data. Specifically, on the basis of the improved spatio-temporal Kriging interpolation algorithm, introducing vibration acceleration as a dynamic perturbation weight term, and weighting the local transient flow field prediction data according to the dynamic perturbation weight term to achieve non-steady perturbation simulation and optimization of the interpolation process to obtain interpolation prediction enhanced data; Step S25: Hydraulic data enhancement, specifically constructing a physical information generation adversarial network through the constructed standard generative adversarial network and the physical loss embedding to obtain physically constrained generated flow field data, and performing spatio-temporal interpolation prediction through the constructed improved spatio-temporal Kriging interpolation algorithm and the interpolation prediction enhancement to obtain spatio-temporal interpolation enhanced flow field data, and fusing the physically constrained generated flow field data and the spatio-temporal interpolation enhanced flow field data to obtain enhanced three-dimensional flow field hydraulic data.

4. A hydraulic design data modeling method for a submersible pump for wells according to claim 3, characterized in that: In step S3, the hydraulic performance prediction is used to dynamically predict the hydraulic performance of the submersible pump for wells. Specifically, based on the enhanced three-dimensional flow field hydraulic data, using a multi-modal graph neural network combined with physical control embedding to perform hydraulic performance prediction to obtain hydraulic performance prediction reference data, including the following steps: Step S31: Construct a flow field prediction subnet, specifically constructing a standard deep neural network including a hidden layer and an output layer, and introducing the physical loss function as a loss term to construct a flow field prediction subnet; Step S32: Physical control loss improvement, specifically combining the physical loss function to construct a joint loss function to obtain a physical control joint loss function, and training and optimizing the flow field prediction subnet according to the physical control joint loss function; The calculation formula of the physical control joint loss function is: L F = a1·L BCE + a2·L phy ; where L F is the physical control combined loss function, a1 is the cross-entropy loss weight, L BCE is the cross-entropy loss function for flow field prediction, a2 is the physical control loss weight, L phy is the physical loss function; Step S33: Construct a multi-modal graph neural subnet, specifically constructing a multi-modal graph neural subnet by constructing a heterogeneous graph of the pump body and combining a standard graph attention neural network to obtain a pump condition perception graph neural network; The heterogeneous graph of the pump body includes a pump body node set and a pump body edge set; The pump body node set includes geometric nodes, pump material nodes, and pump operating condition nodes; The pump body edge set includes material-geometry connection edges and operating condition-structure interaction edges; The material-geometry connection edges are used to represent the strength and corrosion resistance of the pump body material; the operating condition-structure interaction edges are used to represent the stress on the impeller due to the rotational speed; Step S34: Hydraulic performance prediction output integration, specifically by constructing a max pooling layer, mapping a global feature vector based on the output hidden state of the pump operating condition-aware graph neural network, and mapping the global feature vector to the prediction output of the hydraulic performance prediction index; Step S35: Hydraulic performance prediction, specifically by constructing the flow field prediction subnet, improving the physical control loss, constructing the multi-modal graph neural subnet, and integrating the hydraulic performance prediction output. The flow field prediction subnet is used to predict the flow field, the multi-modal graph neural subnet is used to predict the performance of the submersible well pump, the hydraulic performance prediction model is trained to obtain the hydraulic performance prediction model, and the hydraulic performance prediction reference data is obtained by using the hydraulic performance prediction model based on the enhanced three-dimensional flow field hydraulic data; 5. A hydraulic design data modeling method for a submersible well pump according to claim 4, characterized in that: In step S4, the intelligent material matching is used to construct an optimal recommendation for the submersible well pump material optimization plan, specifically by using a depth reinforcement learning method specialized for wear and corrosion modeling based on the enhanced three-dimensional flow field hydraulic data and the hydraulic performance prediction reference data to evaluate and match the material properties, and obtain the optimal material combination recommendation data, including the following steps: Step S41: Corrosion modeling, specifically by constructing a corrosion modeling equation based on the enhanced three-dimensional flow field hydraulic data to predict the corrosion thickness of the submersible well pump material and obtain the corrosion prediction reference data; The corrosion modeling equation includes a corrosion medium concentration term, a pH value term, and a material diffusion characteristic term, and the corrosion depth value is obtained by integrating the corrosion depth according to Fick's second law of diffusion; Step S42: Wear modeling, specifically by using a material wear modeling method combined with the particle erosion mechanism to construct an improved flow field pressure distribution particle equation to predict wear and obtain the wear prediction reference value; The improved flow field pressure distribution particle equation combines force, sliding, material hardness, and particle size effects for comprehensive wear modeling; Step S43: Coupling modeling specialization, specifically by performing unified loss modeling based on the corrosion prediction reference data and the wear prediction reference value, and constructing an adaptive adjustment pump material overall model to calculate the corrosion and wear loss rate and obtain the total loss prediction reference value; Step S44: Constructing a reinforcement learning architecture, specifically by constructing a standard deep Q network as the material recommendation strategy learning framework, and constructing the reinforcement learning architecture through state space definition and action space definition, and realizing policy iteration update through the greedy strategy; Step S45: Improve the reward structure, specifically by combining material lifespan, processing cost, and processing feasibility to perform enhanced reward improvement, obtaining an improved reward function, which is used for intelligent material matching reinforcement learning; Step S46: Intelligent material matching, specifically by using the constructed reinforcement learning architecture and the improved reward structure to perform reinforcement learning for intelligent material matching. During the training phase, the reinforcement learning agent performs policy learning in a virtual environment, utilizes the target network and main network switching mechanism until stable convergence, and during the recommendation phase, through maximizing the Q-value and Pareto non-dominated sorting, filters out the optimal material combination in the three-objective space of performance-cost-processability, obtaining optimal material combination recommendation data.

6. A hydraulic design data modeling method for a submersible well pump according to claim 5, characterized in that: In step S5, the hydraulic modeling of the submersible pump for wells is used to generate a digital twin model of the hydraulic data of the submersible pump for wells. Specifically, by combining the hydraulic performance prediction reference data and the optimal material combination recommendation data, and performing data integration, parametric comprehensive reference data for the hydraulic modeling of the submersible pump for wells is obtained; The parametric comprehensive reference data for the hydraulic modeling of the submersible pump for wells specifically includes operating condition parameters, structural parameters, hydraulic performance parameters, material matching parameters, and model metadata parameters.

7. A submersible well pump hydraulic design data modeling system for implementing a submersible well pump hydraulic design data modeling method as described in any one of claims 1-6, characterized in that: It includes a sensing and acquisition module, a hydraulic data enhancement module, a hydraulic performance prediction module, an intelligent material matching module, and a hydraulic modeling module for the submersible pump for wells.

8. A hydraulic design data modeling system for submersible well pumps according to claim 7, characterized in that: The sensing and acquisition module is used for sensing and acquisition. Through sensing and acquisition, the original dataset of the hydraulic design of the submersible pump for wells is obtained, and the original dataset of the hydraulic design of the submersible pump for wells is sent to the hydraulic data enhancement module; The hydraulic data enhancement module is used for hydraulic data enhancement. Through hydraulic data enhancement, enhanced three-dimensional flow field hydraulic data is obtained, and the enhanced three-dimensional flow field hydraulic data is sent to the hydraulic performance prediction module and the intelligent material matching module; The hydraulic performance prediction module is used for hydraulic performance prediction. Through hydraulic performance prediction, hydraulic performance prediction reference data is obtained, and the hydraulic performance prediction reference data is sent to the intelligent material matching module and the hydraulic modeling module for the submersible pump for wells; The intelligent material matching module is used for intelligent material matching. Through intelligent material matching, optimal material combination recommendation data is obtained, and the optimal material combination recommendation data is sent to the hydraulic modeling module for the submersible pump for wells; The hydraulic modeling module for the submersible pump for wells is used for hydraulic modeling of the submersible pump for wells. Through hydraulic modeling of the submersible pump for wells, parametric comprehensive reference data for the hydraulic modeling of the submersible pump for wells is obtained.

Images