441 results about "Partial differential equation" patented technology

The invention relates to a proton exchange membrane fuel cell (PEMFC) high current density performance prediction method, system, equipment and medium. The method comprises the following steps: establishing a multi-physical field coupling model which comprehensively considers complex processes such as electrochemical reaction, proton conduction, gas diffusion and heat transfer, and describing the change of each physical quantity by adopting a partial differential equation based on a basic physical law; performing grid division and numerical discretization on the proton exchange membrane fuel cell model; selecting model parameters, and verifying the model through experimental data of different working conditions; inputting actual working condition parameters into a model to predict performance, and analyzing a simulation result; using a convolutional neural network, a recurrent neural network and an auto-encoder to extract features from different types of data and fuse the features to form a comprehensive feature vector; a deep neural network prediction model is constructed, and a cross entropy loss function and an Adam optimizer are adopted for training; dropout, L1 and L2 regularization, k-fold cross validation and transfer learning are utilized to optimize the model, and the generalization ability is improved; the system, the equipment and the medium realize high current density performance prediction of the proton exchange membrane fuel cell (PEMFC) based on the method; the prediction precision is improved, the experiment cost is reduced, the internal mechanism can be deeply understood, and powerful support is provided for design optimization, operation management and fault diagnosis of the fuel cell.

The invention belongs to the technical field of turbulent flow field prediction and deep learning, and discloses an incompressible turbulent flow field prediction method based on a potential diffusion model. The method comprises the following steps: acquiring original turbulence data; processing the turbulence data; constructing a turbulence prediction model; model training; and evaluating the model and the like. The model of the technical scheme of the invention specifically comprises the following steps: designing a multi-scale Fourier auto-encoder for extracting multi-scale space and frequency domain features in a turbulence field and obtaining a global structure and a local scale structure of turbulence; a novel accelerated sampling method is proposed and introduced in the diffusion process, namely a diffusion probability model solver greatly shortens the reasoning time in a potential space and keeps high fidelity in long-time-sequence prediction; a physical constraint loss item based on a partial differential equation is introduced, and a Navier-Stokes equation (N-S) is explicitly introduced into a training process, so that the physical consistency of results is effectively improved, and errors are remarkably reduced.

The invention discloses a central air conditioner energy-saving control method based on AI self-adaptive adjustment, and relates to the technical field of intelligent control, and the method comprises the following steps: collecting environmental data, equipment operation data and energy consumption data of a central air conditioner in real time, preprocessing multi-modal data, constructing a physical constraint equation in combination with a thermodynamic law, and generating a multi-modal data set; generating a multi-target optimization control instruction according to the optimization weight, solving an optimal equipment parameter combination through a Pareto frontier algorithm, and transmitting the optimal equipment parameter combination to a central air conditioner actuator; actual data after instruction execution are collected, the deviation degree of the energy-saving efficiency and the comfort degree is calculated, physical information neural network parameters are updated through a meta-learning framework, and thermodynamic partial differential equation coefficients are adjusted. According to the method, by constructing a multi-modal physical information fusion framework and a dynamic closed-loop optimization system, the comprehensive regulation and control capability of the central air conditioner in a complex building environment is improved.

The invention relates to the technical field of ocean data interpolation filling, in particular to a space-time deficiency filling method and system based on context association and physical guidance. The method comprises the following steps: acquiring seawater dissolved oxygen data and context data; multivariable space-time dependence extraction is carried out based on the obtained seawater dissolved oxygen data and context data; gaussian noise diffusion is carried out based on the obtained seawater dissolved oxygen data; noise prediction is carried out based on double-view space-time correlation; and the prediction error is constrained based on the joint loss function. According to the method, a physical consistency constraint mechanism based on a partial differential equation is introduced in a model training process, so that model output better conforms to a physical coupling rule among variables in a marine environment. The constraint effectively inhibits non-physical fluctuation possibly occurring in the interpolation result, enhances the physical credibility and interpretability of the result, and provides a more reliable data basis for subsequent scientific analysis and process modeling.

The invention discloses a three-dimensional unstructured grid adaptive refinement method and system based on machine learning, and belongs to the field of machine learning, partial differential equation solving and computational fluid mechanics simulation. Residual errors of a fluid control equation are used as a novel error indicator and a refinement criterion, imprecise flow field data and the fluid control equation are obtained by fusing coarse grids of simulation flow through a physical information neural network, and the total residual errors of the equations are conveniently calculated by using automatic differentiation after training is completed. Therefore, the coarse grid units with relatively high residual errors can be adaptively marked and refined. By matching an h-refinement scheme, Delaunay tetrahedron subdivision is executed after vertexes are strategically inserted to maintain the quality of the refined grid. According to the method, any numerical solver can be flexibly matched to carry out grid adaptive refinement, so that various typical flow problems can be solved with high precision. The method achieves better balance between calculation precision and the number of grids, and has the advantages of being simple, convenient, high in compatibility and universality and the like.

The invention relates to the technical field of coal mine geological safety monitoring and disaster early warning, in particular to a coal mine disaster early warning method and system based on a geological model and a storage medium, and the method comprises the steps: firstly obtaining high fractal fracture information and monitoring data such as gas, hydrology and stress through chaotic wave excitation and fractal analysis; and a three-dimensional fractal grid with multi-level fracture representation is constructed. And discretizing a mechanical field and a fluid field by using a fractal partial differential equation operator, and globally solving a large-scale nonlinear equation through a quantum annealing mode or a quantum and classical combined annealing mode to form a quantum fractal coupling simulation result. And finally, integrating the grid units and monitoring data in the fuzzy hypergraph structure, carrying out fuzzy membership analysis on a multi-disaster element coupling sign, and outputting early warning information once the risk is judged to reach a threshold value. According to the invention, the precision and timeliness of coal mine geological disaster monitoring can be obviously improved.

The invention discloses a soil water and salt dynamic monitoring and quality evaluation method based on multispectral remote sensing, and relates to the field of geological monitoring, and the method comprises the steps: carrying out the scattering deviation correction and roughness correction of an observation spectrum through a satellite remote sensing image, an unmanned aerial vehicle multispectral image, a ground sample and meteorological driving data; the soil intrinsic reflectivity, the vegetation coverage component and the salt crusting component are obtained; a water-salt inversion model is established by combining thermal infrared and red edge wave band information, and multi-temporal soil volumetric moisture content and surface salinity index are inverted; by introducing meteorological conditions and vegetation dynamic characteristics, a water-salt coupling partial differential equation and a graph space-time constraint model are constructed, and continuous space-time distribution of soil water content and salinity is obtained; extracting salinity, moisture, vegetation response and soil health indexes, establishing a soil quality comprehensive evaluation model, and outputting high-risk plaques and a treatment priority sequence. The method realizes dynamic monitoring and quality grading evaluation of soil water and salt, and is suitable for saline-alkali soil treatment and ecological restoration.

The invention discloses a pipeline flow field remodeling method based on LAAF-PINN, and the method comprises the steps: collecting the flow field data of a pipeline measurement point; dividing the measuring points into monitoring points and testing points, and preprocessing the data sequence; constructing an LAAF-PINN, randomly selecting a matching point along a pipeline, and inputting a time-space sequence of the monitoring point and the matching point to obtain model output; calculating a data loss item according to the model output of the monitoring point, calculating a residual error according to the model output of the collocation point to obtain a partial differential equation loss item, and combining the two items to obtain total loss; after multiple rounds of iteration updating, a trained PINN model is obtained, a time-space sequence of a test set is input, and corresponding flow field information can be quickly and accurately reconstructed. According to the method, the LAAF-PINN is applied to pipeline hydraulic transient research, an existing flow field remodeling method is expanded, good robustness is achieved for data containing uncertain factors, and meanwhile the problem that the calculation precision is not high possibly existing in forward simulation is solved.

The invention relates to the technical field of microbial ecological information processing, and discloses a microbiome functional activity prediction method and system, and the method comprises the steps: collecting dynamic environment factor data in soil in real time through a sensor, and carrying out the preprocessing of the dynamic environment factor data; establishing a partial differential equation model based on the preprocessed data; the partial differential equation model comprises a diffusion term, a reaction term and an environmental factor coupling term; optimizing the differential equation model parameters by using a neural network, and carrying out transfer learning training on the partial differential equation model parameters based on historical data; and predicting the activity grade and function of the microorganisms on the basis of the optimized model for the functional activity of the microorganisms. The adaptability and generalization ability of the model in different soil areas are remarkably improved, and closed-loop processing from function activity numerical value prediction to function classification output is achieved.

The invention provides a topology and morphology integrated optimization design method for a thin-walled structure. The method comprises the steps that 1, a finite element model is established according to design requirements; 2, extracting a thin-wall structure morphology distribution field # imgabs0 # through two-step partial differential equation filtering and Heavide projection operation; 3, constructing a material attribute rational approximation model in combination with the thin-wall morphology distribution field # imgabs1 # and the topological density field v obtained in the step 2, thereby obtaining an elastic modulus E and mass rho in a design domain; 4, performing finite element analysis in combination with the material interpolation model, and calculating the flexibility of the structure; 5, solving to obtain the sensitivity of the structure flexibility and the quality relative to the design variables; step 6, updating the morphology design variable mu and the topology design variable v of the thin plate by adopting a moving asymptote method; and step 7, when a convergence condition is satisfied, stopping optimization iteration and outputting an optimization result. According to the method, the maximum forming angle of the thin-wall structure is controlled by adjusting the filtering radius of the anisotropic filter, and the stamping forming process is met.

A method for training a shape optimization neural network to produce an optimized point cloud defining desired shapes of materials with given properties is provided. The method comprises collecting a subject point cloud including points identified by their initial coordinates and material properties and jointly training a first neural network to iteratively modify a shape boundary by changing coordinates of a set of points in the subject point cloud to maximize an objective function and a second neural network to solve for physical fields by satisfying partial differential equations imposed by physics of the different materials of the subject point cloud having a shape produced by the changed coordinates output by the first neural network. The method also comprises outputting optimized coordinates of the set of points in the subject point cloud, produced by the trained first neural network.

The invention discloses a physical field solving method based on a Bayesian physical information extreme learning machine, and the method comprises the steps: constructing a single-layer full-connection neural network, carrying out the random initialization, and fixing the weight of an input layer; based on a partial differential equation of a physical system and boundary conditions thereof, defining a training loss item containing physical information; a physical system solving problem is converted into a linear least square problem, and a linear equation set is constructed; supposing that an output layer weight parameter obeys Gaussian prior distribution with the mean value being zero, and controlling a covariance matrix by an adjustable hyper-parameter; constructing a Gaussian likelihood function based on the observation data, and calculating posterior distribution of the output weight in combination with the prior distribution; carrying out iterative optimization on the hyper-parameter by adopting an evidence maximization method to obtain a mean value and a covariance of posterior distribution; based on posterior distribution, adopting a Monte Carlo integral method to generate prediction output of the physical system; and performing uncertainty quantization based on the variance of prediction output, and outputting a prediction value and a confidence interval thereof.

The invention discloses a method and system for solving a partial differential equation based on a KAN and MLP parallel structure, and belongs to the technical field of deep learning. The method comprises the following steps: constructing a parallel neural network comprising a KAN branch and an MLP branch, and introducing a fusion factor to carry out weighted fusion on the output of the KAN branch and the output of the MLP branch; based on a physical information neural network principle, constructing a loss function comprising an equation residual term, an initial condition term and a boundary condition term; training the parallel neural network by using an adaptive optimization algorithm, and minimizing the loss function; and inputting variables in the partial differential equation into the trained parallel neural network, and outputting a solution of the partial differential equation. By effectively combining the nonlinear expression ability of MLP and the function modeling advantage of KAN, the expression and solution ability of a multi-scale, non-stationary and complex physical field is improved; the modeling precision and convergence stability of a complex physical system, especially a partial differential equation driving problem, can be effectively improved.

The invention discloses a numerical simulation method for a semiconductor device based on spatial discretization and uncoupling iteration. The method comprises the following steps: establishing a basic equation set of the semiconductor device; spatial discretization is carried out on the non-uniform hexahedral mesh; discretizing the current equation to obtain a drift diffusion model equation set in a three-dimensional discrete format; and sequentially solving each equation in the drift diffusion model equation set by adopting a Gummel uncoupling method. According to the semiconductor device numerical simulation method based on spatial discretization and uncoupled iteration provided by the invention, a semiconductor partial differential equation is subjected to three-dimensional spatial discretization by adopting an FDM method, and a large-scale nonlinear equation set is solved by adopting a Gummel uncoupled iteration method, so that physical characteristics and internal physical images of the device are obtained; the occupied computer memory is relatively low, the method is not sensitive to an initial value, and the simulation speed is greatly improved.

The invention relates to the technical field of water conservancy projects, in particular to a water conservancy project construction progress intelligent management system. The method comprises the following steps: acquiring geometric and semantic information, a three-dimensional geological parameter distribution diagram and mechanical pose data in a BIM model, performing primary processing on the acquired data, segmenting the geometric and semantic information in multi-modal water conservancy construction data, and mapping time sequence data to a process state space to obtain cross-modal semantic alignment data; performing space-time-physical joint embedding representation on the cross-modal semantic alignment data, establishing an identification model based on a PINN physical information neural network, and introducing a partial differential equation residual term into a loss function of the model to obtain a construction progress deviation identification model; and inputting the characteristic water conservancy construction data into a construction progress deviation identification model for identification to obtain a construction progress result. The method can accurately recognize the construction progress deviation, and provides a reliable decision basis for the construction management of a water conservancy project.

The invention relates to the technical field of image analysis, and discloses an end-to-end visual tactile perception method and system based on a morphology-force field analysis model, a terminal and a storage medium, and the method comprises the steps: employing a morphology reconstruction module to achieve the analysis of a micron-order contact surface, employing a force field analysis model to achieve the precise analysis of a force field in each direction, and taking the two results as a data set, inputting and mapping the image into contact morphology, normal and shear force distribution through an end-to-end network, integrating the high efficiency of data driving and the reliability of physical consistency, and finally introducing partial differential equation residual error sum into a loss function to carry out optimization. And the morphology and the force field meet the set physical consistency on the whole. According to the method, micron-sized morphology analysis and multi-axial force synchronous estimation are realized, the real-time robust performance under a high-frequency dynamic task is realized, the cost of tactile perception is reduced, and meanwhile, the accuracy of a tactile perception result is improved.

The invention discloses a rapid heat transfer simulation method and device based on a neural network, and relates to the technical field of physical simulation. The method comprises the steps that a hybrid neural network model is trained, the model learns operator mapping from an input function to a temperature or heat flow field, and meanwhile physical constraints such as a heat conduction partial differential equation are coded into a loss function; for a new simulation task, single forward inference is carried out by using the operator mapping, and an initial prediction result is rapidly generated; then, according to physical constraints of coding, calculating a physical residual error of initial prediction, and when the residual error exceeds a preset threshold value, executing a small amount of optimization iteration by taking the prediction as an initial value to carry out rapid local correction; the problems that a traditional numerical method is long in calculation time and an existing neural network method is insufficient in physical fidelity are solved, and high efficiency and high precision of heat transfer simulation are achieved.

The invention discloses a security boundary consistency control method for a multi-agent spatio-temporal dynamic system. The method specifically comprises the following technical steps: firstly, constructing a leader-follower spatio-temporal dynamic mathematical model of the multi-agent system based on a partial differential equation; then establishing an abnormal working condition mathematical model containing an actuator fault and a network spoofing attack; deducing to obtain an error dynamic system equation by defining a system consistency control target; constructing an energy function by using a Lyapunov stability theory, and deducing to obtain a system stability criterion; finally, a security boundary consistency control method is designed, and multi-agent space-time dynamic consistency control is achieved. The method effectively solves the problem of high cost caused by arrangement of actuators in a whole spatial domain in a traditional scheme and the problem that a multi-agent system is easily interfered by network attacks in a communication process, and has remarkable application value in engineering practice.

The invention relates to the technical field of underground engineering safety monitoring and disaster simulation, in particular to a digital twinborn deduction platform for underground engineering disaster chain evolution simulation, which comprises a multi-disaster coupling numerical simulation module used for acquiring multi-source data including design drawing data, geological data and engineering monitoring data, and based on the multi-source data, establishing an initial stress field model by using a numerical calculation method combining a finite element and a finite difference, and carrying out leakage-settlement-structure failure multi-disaster coupling numerical simulation. According to the method, the partial differential equation for describing the evolution law of the underground engineering is used as a constraint term to be embedded into the deduction model, and online identification and dynamic extrapolation of parameters are executed by fusing real-time monitoring data, so that the accuracy of disaster evolution simulation under complex working conditions is remarkably improved; and the problem that the traditional numerical calculation method is difficult to meet the real-time requirement of digital twinning is solved.

The invention provides a flow field solving method based on a time sequence physical information neural network, and belongs to a partial differential equation calculation method. The method comprises the following steps: firstly, determining a flow field motion law and a flow field computational domain based on a fluid motion control equation, configuring data points for training and performing data fusion operation; secondly, constructing a physical constraint residual equation, taking the training data set as input data of a neural network, calculating a data residual term, weighting a calculation result of the physical constraint residual equation and a calculation result of the data residual term, and constructing a data-physical mixed loss function; and finally, constructing a time sequence-based physical information neural network, performing training optimization on parameters of the time sequence-based physical information neural network, completing training and obtaining simulated flow field features. By introducing the time sequence module, the accuracy of flow field dynamic simulation based on the time sequence physical information neural network is improved, the requirement for the model solving speed can be met, and the feasibility and accuracy of the model are guaranteed.

The invention discloses a multi-modal physical constraint energy storage battery health state cross-domain evaluation method and system, and the method comprises the steps: constructing an RC equivalent circuit model of an energy storage battery, and constructing an electrothermal coupling model of the energy storage battery based on the RC equivalent circuit model and a thermodynamic equation; constructing a neural network model by taking the electrothermal coupling model as a constraint condition, constructing a three-dimensional temperature field in the battery based on a loss function of a partial differential equation in the neural network model, and training the neural network model; constructing a battery segment capacity attenuation model according to the three-dimensional temperature field, and optimizing the battery segment capacity attenuation model through the trained neural network model; and further adjusting the optimized battery segment capacity attenuation model according to the chemical composition of the energy storage battery to obtain a battery segment capacity attenuation model adaptive to the chemical composition of the current energy storage battery, and predicting the health state of the energy storage battery. According to the invention, high-precision and low-cost cross-domain battery health assessment can be realized.

The invention relates to a basin flood simulation optimization method based on a physical information neural network, and aims to construct a basin flood acceleration solution mode based on PINNs, and ensure the continuity and the physical characteristics of mass conservation during model operation, and the method comprises the following steps: simulating basin flood by using an urban flood space-time simulation model; obtaining a data set required for constructing the physical neural network; physical rules are integrated and simplified into limiting conditions, so that integration of a neural network structure is facilitated; establishing a partial differential equation (PDE) used for describing river flood routing, and taking the partial differential equation as a loss function of the physical information neural network model; constructing and training a PINNs drainage basin flood model constructed on the basis of a CNN (Convolutional Neural Network); and establishing a model evaluation index and performing evaluation. The method has the beneficial effect that the speed and precision of basin flood simulation are improved.

The invention relates to the technical field of engineering mechanical structure optimization, in particular to an excavator structure topological optimization method and system based on real-time stress field reconstruction. The method comprises the following steps of: designing finite element analysis software, modeling and simulating in the finite element analysis software to obtain virtual data, designing a Kalman filter to carry out error correction on the virtual data and actual test data, training the Kalman filter by utilizing the corrected data to obtain prediction of real-time stress field distribution, and further, establishing a topological optimization model of an excavator structure, according to the method, material attributes are coded into a weight matrix in a neural network, and a stress constraint partial differential equation is used as a constraint condition, so that high-precision reconstruction of the stress field of the movable arm of the excavator under an actual working condition is realized, and an accurate boundary condition is provided for topological optimization.

The invention discloses a three-dimensional physical field real-time prediction method and system based on geometric deep learning and physical constraint, and relates to the technical field of computer-aided engineering and artificial intelligence. The method comprises the following steps: directly extracting native boundary representation data (B-Rep) of a three-dimensional model from a computer aided design system; constructing a heterogeneous dual graph taking a parameterized curved surface as a graph node, skipping finite element grid division, and aggregating local and global topological features by using a graph neural network; in combination with a physical information driving mechanism, a partial differential equation (PDE) residual error is introduced as a loss function for constraint training, and generalization prediction of a novel geometric structure is realized; and finally, the physical field state quantity is predicted through direct regression and is rendered in real time. An incremental reasoning mechanism based on a local topology subgraph is adopted, millisecond-level physical field real-time feedback under design modification is achieved, and the method is suitable for scheme rapid screening and trend prediction in the initial stage of design.

The invention discloses a water quality antibiotic traceability method and system based on data analysis, and the method comprises the steps: firstly constructing a water system network into a topological graph, building a physical information-graph attention network as a pollutant propagation prediction model, and enabling the model to integrate a pollutant convective diffusion partial differential equation as a physical constraint into a loss function, the physical fidelity prediction of the spatial and temporal distribution of the antibiotic concentration in the water body is realized; secondly, constructing a pollution source tracing problem as a reverse reinforcement learning task, and reversely learning a pollution source reward function capable of optimally explaining an observation result from real downstream monitoring data by adopting an antagonism reverse reinforcement learning framework and taking the propagation prediction model as a dynamic environment through antagonism training of a generator and a discriminator; and finally, generating a probability distribution diagram of the upstream candidate pollution source position according to the reward function, and performing accurate positioning in combination with a hydrodynamic constraint condition, thereby remarkably improving the accuracy, robustness and calculation efficiency of traceability.

The invention provides a switch cabinet contact temperature rise wireless sensing method and system based on surface acoustic wave resonance, and relates to the technical field of power system switch cabinet detection, and the method comprises the steps: calculating the real-time offset of a fundamental frequency and a third harmonic relative to a reference frequency, and obtaining the fundamental frequency offset and the third harmonic offset; in combination with the fundamental frequency offset and the equivalent acoustic impedance variable quantity of the oxide film, surface acoustic wave propagation path deformation caused by thermal expansion of the contact material is compensated by using a thermal-acoustic coupling partial differential equation, and the fundamental frequency net offset after temperature decoupling is obtained; sparse reconstruction is carried out on the temperature field of the contact area of the contact by adopting a compressed sensing algorithm, and three-dimensional temperature gradient distribution of the contact area of the contact is output; and inputting the three-dimensional temperature gradient distribution and historical time sequence data thereof into an intrinsic mode decomposition model for processing, and generating a dynamic self-adaptive early warning threshold value. The system has the advantages that passive, wireless, high-precision and full-space temperature rise monitoring is achieved, and intelligent operation and maintenance decision making of the switch cabinet is directly driven.

The invention discloses a rock and ore microscopic image splicing method and system based on deep learning, and relates to the field of image processing and the technical field of microscopes, and the method comprises the steps: obtaining a local rock and ore microscopic image of a rock and ore slice, carrying out the preprocessing of the local rock and ore microscopic image, and carrying out the overlapping region coarse registration of the preprocessed local rock and ore microscopic image through a phase correlation method; based on an improved image feature detection model, basic features and description features in the local rock and ore microscopic image after coarse registration are extracted, and local image features of the local rock and ore microscopic image are obtained; performing feature matching on the local image features of the two groups of local rock and mineral microscopic images by using an image feature matching model to obtain a matching corresponding relation of the local image features; and based on an image fusion algorithm of a homography matrix and a partial differential equation, splicing and optimizing the local rock and ore microscopic images in combination with a matching corresponding relation of local image features, and generating a large-view-field rock and ore microscopic image. According to the method, the complex transformation between the images can be better processed.

The invention relates to the technical field of intelligent monitoring, and discloses an intelligent monitoring method and system for forging production. The method comprises the following steps: carrying out temperature data acquisition and polar coordinate conversion on the wind power flange forge piece to obtain polar coordinate temperature field data; inputting the polar coordinate temperature field data into an annular thermal field gradient decomposition network for feature extraction to obtain a temperature field feature vector; performing phase change critical temperature sensitivity enhancement on the temperature field feature vector to obtain a phase change sensitivity enhanced feature vector; on the basis of heat conduction partial differential equation constraints, submerged space reconstruction and interpolation are carried out on the phase change sensitivity enhancement feature vectors, and full-field temperature distribution data are obtained; and generating temperature field anomaly score and anomaly type probability distribution information according to the full-field temperature distribution data. The monitoring problem that the surface temperature is normal but the dangerous temperature gradient exists in the depth direction is effectively solved, and the production qualification rate of wind power flange forgings is increased.

The invention belongs to the technical field of aero-engine performance testing, particularly relates to a physical constraint-based aero-engine state prediction model construction method and system, and aims to solve the problems of high calculation complexity and poor data quality of an existing physical information model. The method comprises the following steps: acquiring historical operation data of the aero-engine and a physical constraint rule set of engineering simplification; predicting performance parameters by adopting a deep learning model; constructing a total loss function formed by weighting a data loss item and a physical loss item to train the model; wherein the physical loss item is generated based on the deviation degree of the predicted performance parameter and the engineering simplified physical constraint rule set, and is used for replacing the complex partial differential equation constraint. According to the method, the engineering simplified physical rule is introduced, so that the calculation overhead of model training is remarkably reduced, the model is effectively guided to learn the characteristics conforming to the physical rule, and the accuracy and generalization ability of the prediction model are remarkably improved under the condition of limited data.

The invention discloses a GNN-PINN fusion dynamic error enhancement current transformer error prediction method, and the method comprises the following steps: firstly, building a sliding window steady-state recognition method based on a kurtosis-entropy collaborative criterion, and achieving the precise segmentation of current features under a complex working condition through multi-dimensional statistic fusion; secondly, a dynamic error simulation injection mechanism is designed, a current data enhancement strategy guided by a physical rule is constructed, and the working condition coverage of training data is effectively improved; on the basis, a GNN-PINN-based current transformer error prediction model is provided, and deep coupling of an error mechanism and data characteristics is realized through embedding a physical connection relation of equipment through a topological graph and combining partial differential equation constraints, so that a current transformer error is obtained. According to the current transformer error prediction method provided by the invention, the key problems of difficulty in steady-state feature extraction, scarcity of data samples, lack of physical constraints and the like in current transformer error prediction during online calibration of the transformer substation current transformer can be solved.