CN-121980757-A - Rapid prediction method for thermochemical energy storage carrier particle thermo-mechanical coupling process

CN121980757ACN 121980757 ACN121980757 ACN 121980757ACN-121980757-A

Abstract

The invention discloses a rapid prediction method of a particle thermal-force coupling process, which is characterized by scanning carrier particles based on a micro-focus computed tomography technology to generate a three-dimensional digital model, importing the three-dimensional digital model into a multi-physical-field simulation platform to establish a transient full-order model, generating training data sets and test data sets under different working conditions and corresponding temperature fields and stress fields based on the transient full-order model, decomposing and calculating the training data sets by utilizing an intrinsic orthogonal decomposition POD (point-of-sale) reduced-order model to obtain POD basis functions and spectrum coefficients, establishing a mapping model from an input working condition to an output spectrum coefficient based on a BP (back propagation) neural network, inputting the target working condition into the mapping model to output the POD spectrum coefficient, and reconstructing full-field physical quantity distribution at any moment under the target working condition by linear combination based on a calculation result and the POD spectrum coefficient. The method solves the problems of low calculation efficiency and high resource consumption of the traditional numerical simulation.

Inventors

DENG YAJUN
LIU ZHENMING
BIAN RUIHAO
ZHAO YIMING
ZHANG WEI

Assignees

北京石油化工学院

Dates

Publication Date: 20260505
Application Date: 20251223

Claims (9)

1. A method for rapidly predicting a thermochemical energy-storage energy-carrier particle thermo-mechanical coupling process, the method comprising: step 1, scanning energy carrier particles based on a micro-focus computer tomography technology to generate a three-dimensional digital model reflecting the real geometric morphology of the energy carrier particles; step 2, the three-dimensional digital model obtained in the step 1 is imported into a multi-physical-field simulation platform, and a transient full-order model for describing the thermal-force coupling process of the energy carrier particles under solar irradiation is established; step 3, generating training data sets and test data sets under different working conditions and corresponding temperature fields and stress fields based on the transient full-order model established in the step 2; Step 4, performing decomposition calculation on the training data set by utilizing an intrinsic orthogonal decomposition POD reduced order model to obtain a POD base function and a spectral coefficient; Step 5, based on the BP neural network learning the mapping relation between different working conditions and the spectrum coefficients, establishing a mapping model from the input working conditions to the output spectrum coefficients; Step 6, inputting target working conditions into the mapping model in the step 5, and outputting predicted POD spectrum coefficients after dimension reduction; step 7, reconstructing full-field physical quantity distribution at any moment under the target working condition through linear combination based on the calculated result in the step 4 and the POD spectrum coefficient output in the step 6; And 8, verifying the prediction accuracy by using the test data set generated in the step 3.
2. The method for rapidly predicting the thermal-mechanical coupling process of thermochemical energy-storage carrier particles according to claim 1, wherein in step 1, specifically, a microfocus computer tomography technology is adopted to perform high-resolution scanning on carrier particles with complex porous structures, so as to obtain a plurality of two-dimensional gray slice images inside the carrier particles; the method comprises the steps of preprocessing an acquired slice image by utilizing three-dimensional image processing and reconstruction software, and specifically comprises filtering denoising processing, threshold segmentation processing, morphological operation and flaw correction, wherein the filtering denoising processing is used for improving the signal to noise ratio; and finally, generating a three-dimensional digital model capable of reflecting the real geometric morphology of the energy carrier particles.
3. The method for rapid prediction of thermochemical energy-storage energy carrier particle thermo-mechanical coupling process according to claim 1, wherein in step 2, unstructured meshing is performed on the three-dimensional digital model and local encryption is performed in areas with large temperature gradients; then applying boundary conditions conforming to actual working conditions, including applying convection-radiation composite heat exchange boundary on the surface and restricting displacement of the bottom of the carrier particles; wherein, the heat conduction equation of the carrier particle is: ; Wherein ρ s is the density of the carrier particles, C p,s is the specific heat capacity of the carrier particles, T s is the temperature of the carrier particles, lambda s is the thermal conductivity of the carrier particles, T is the time, and subscript s is solid; for the gas fraction, the relationship between gas density and temperature is: ; Wherein ρ g is the gas density, ρ 0 is the gas density at the reference temperature, β is the thermal expansion coefficient of the gas, Δt is the temperature difference between the actual temperature and the reference temperature T 0 , and subscript g is the gas; the gas continuity equation is expressed as: ; Wherein u g is the gas velocity; The gas momentum equation is expressed as: ; wherein eta is dynamic viscosity, p is gas pressure; Gravitational acceleration; the gas energy equation is expressed as: ; Wherein C p,g is the specific heat capacity of the gas, T g is the temperature of the gas, and lambda g is the thermal conductivity of the gas.
4. A method for rapid prediction of thermochemical energy-storage energy-carrier particle thermo-mechanical coupling process according to claim 1, wherein in step 3, the different working conditions P include incident solar energy intensity I, ambient temperature The surface convective heat transfer coefficient h and the initial properties of the particulate material; Generating k groups of combinations { p 1 ,p 2 ,…,p k } with space filling property and representative working conditions; For each set of working conditions p i , i is from 1 to k, the transient full-order model established in the step 2 is operated, and complete transient coupling calculation is executed; after the calculation is completed, outputting full-field data snapshots including a temperature field and a stress field according to a fixed time interval; dividing all snapshot data into a training data set and a test data set according to a set proportion, wherein the training data set is used for constructing a POD basis function and a training neural network, and the test data set is used for verifying model prediction accuracy.
5. The method for rapid prediction of thermochemical energy-storage energy-carrier particle thermo-mechanical coupling process according to claim 1, wherein the process of step 4 is specifically: First, a global time average field is calculated for all S snapshot data from a training dataset, wherein the average field of the temperature field The method comprises the following steps: ; wherein: is a space coordinate representing a position in a three-dimensional space; is the j-th time point in the snapshot sequence; The average field Characterizing dominant static features of the system; subsequently, the pulse field of each snapshot data is calculated I.e. deviating from the average field Is expressed as: ; Pulse field Key information of dynamic evolution of the system is obtained; all S pulse fields Arranged in columns to form a snapshot matrix Wherein M is the total number of space grid nodes; Is a real number domain; The objective of the eigen-orthogonal decomposition POD reduced order model is to find a group of optimal orthogonal basis, and after linear reconstruction is performed based on the first N basis vectors, the mean square error between the obtained prediction field and the original field is the smallest, in particular to a snapshot matrix Singular value decomposition is performed, expressed as: ; Wherein, the Is of the column vector of (2) For POD space modes, namely POD basis functions, the POD space modes and the POD basis functions form an optimal orthogonal coordinate system for describing the system dynamics; the system is a diagonal matrix, diagonal elements of the diagonal matrix are called singular values, squares of the singular values are characteristic values, and energy or variance contribution carried by corresponding POD spatial modes is represented; the column vector of (1) contains time coefficient information of a spatial mode; intercepting the front N modes, wherein the interception order N is determined by a preset energy duty ratio and is expressed as: ; wherein: Characterizing the energy contribution of an ith POD spatial modality for the modality; representing a preset energy duty ratio threshold value for determining the cutoff order N; Whereby truncated POD basis functions are obtained Pulse field The approximation is expressed as: ; wherein the coefficient is Called POD spectral coefficients, which constitute the original field, i.e. the coordinates of the original known sample data in the low-dimensional space spanned by the POD basis functions By pulsing the field Projection to the POD basis functions is calculated, expressed as: ; wherein: representing an inner product over a computational domain; is the kth space grid node; the weight of the kth node; performing this operation on all S snapshot data to obtain POD spectral coefficient time series of corresponding temperature field Expressed as: ; the superscript T denotes a transpose; the same operation is performed on the stress field, resulting in the POD basis functions and spectral coefficients of the stress field.
6. The method for rapid prediction of thermochemical energy-storage energy carrier particle thermo-mechanical coupling process of claim 1, wherein in step 5, the BP neural network adopts a multi-layer feed-forward design comprising an input layer, a hidden layer and an output layer; Each set of working conditions I is the total number of working conditions from 1 to k, k is taken as an input sample of the BP neural network, and the corresponding output label is the working conditions All spectral coefficients calculated below And finally forming training data pair ; Performing standardized processing on all input and output data so as to accelerate training convergence; Using a supervised learning mode with mean square error as a loss function Expressed as: ; wherein: For training sample number; the true spectrum coefficient vector is the ith training sample; Predicting the spectral coefficient vector for the neural network of the ith training sample; Represents an L2 norm; Calculation of loss function using back propagation algorithm Regarding the gradient of BP neural network weight, and adopting an adaptive learning rate optimizer to update the weight; After training is completed, a mapping model from the input working condition to the output spectrum coefficient is obtained Expressed as: 。
7. the method for rapid prediction of a thermochemical energy-storage energy-carrier particle thermo-mechanical coupling process according to claim 5, wherein in step 7, the average field of the temperature field calculated in step 4 is used Average field of stress field And truncated POD basis functions And (3) reconstructing full-field physical quantity distribution at any moment under a target working condition through linear combination by combining the POD spectrum coefficients output in the step (6), wherein: Reconstructed temperature field Expressed as: ; wherein: the ith POD spectrum coefficient predicted for the BP neural network; An ith POD basis function for the temperature field; reconstructed stress field Expressed as: ; wherein: Is the i-th POD basis function of the stress field.
8. An electronic device comprising a memory and a processor, characterized in that the memory has stored therein a computer program, the processor being arranged to run the computer program to perform the method of any of claims 1 to 7.
9. A computer storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the method of any one of claims 1 to 7.

Description

Rapid prediction method for thermochemical energy storage carrier particle thermo-mechanical coupling process Technical Field The invention relates to the technical field of thermochemical energy storage systems, in particular to a rapid prediction method of a thermochemical energy storage carrier particle thermal-force coupling process. Background In a solar high-temperature thermochemical energy storage system, the temperature of the energy carrier particles is changed severely in the circulation process, complex heat-force load is easy to generate, and the internal heat-force coupling behavior of the energy carrier particles directly influences the energy storage efficiency and the structural stability of the particles. Therefore, the realization of the rapid and accurate prediction of the thermal-force multi-physical field coupling behavior of the three-dimensional energy carrier particles in the circulation process becomes the key for optimizing the particle design and improving the system performance. At present, the field mainly depends on the traditional numerical methods, such as a finite volume method, a finite element method and the like, and performs full-order simulation on the processes of heat conduction, stress evolution and the like in particles. Although the method can provide accurate simulation results, when a three-dimensional particle model, nonlinear material behaviors and transient coupling processes are processed, the method has the advantages of high consumption of computing resources and low solving speed, and is difficult to meet the real-time requirements of rapid evaluation and multi-parameter optimization in engineering. To improve computational efficiency, a reduced order modeling approach based on eigen-orthogonal decomposition (Proper Orthogonal Decomposition, POD) was introduced. The POD can extract a low-dimensional characteristic basis function from full-order simulation data of a high-temperature thermal-mechanical coupling process, and characterization of carrier particle behaviors is achieved. However, in the traditional POD framework, the solution of the spectral coefficient still needs to construct and solve the reduced differential equation on line, the deducing process is complex, and the adaptability to the thermal-force process with strong nonlinearity and large parameter variation is insufficient, so that the application of the POD framework in quick prediction is limited. Disclosure of Invention The invention aims to provide a quick prediction method for a thermochemical energy storage carrier particle thermal-force coupling process, which is used for establishing end-to-end mapping from working condition parameters to POD spectrum coefficients by combining intrinsic orthogonal decomposition POD with BP neural network, so that the problems of low calculation efficiency, high resource consumption and insufficient adaptability to complex working conditions of the traditional numerical simulation calculation method, which still need to solve differential equations in real time, are effectively solved. The invention aims at realizing the following technical scheme: A method for rapid prediction of thermochemical energy-storage energy-carrier particle thermo-mechanical coupling processes, the method comprising: step 1, scanning energy carrier particles based on a micro-focus computer tomography technology to generate a three-dimensional digital model reflecting the real geometric morphology of the energy carrier particles; step 2, the three-dimensional digital model obtained in the step 1 is imported into a multi-physical-field simulation platform, and a transient full-order model for describing the thermal-force coupling process of the energy carrier particles under solar irradiation is established; step 3, generating training data sets and test data sets under different working conditions and corresponding temperature fields and stress fields based on the transient full-order model established in the step 2; Step 4, performing decomposition calculation on the training data set by utilizing an intrinsic orthogonal decomposition POD reduced order model to obtain a POD base function and a spectral coefficient; Step 5, based on the BP neural network learning the mapping relation between different working conditions and the spectrum coefficients, establishing a mapping model from the input working conditions to the output spectrum coefficients; Step 6, inputting target working conditions into the mapping model in the step 5, and outputting predicted POD spectrum coefficients after dimension reduction; step 7, reconstructing full-field physical quantity distribution at any moment under the target working condition through linear combination based on the calculated result in the step 4 and the POD spectrum coefficient output in the step 6; And 8, verifying the prediction accuracy by using the test data set generated in the step 3. According to the technical scheme provi