Intelligent control method for chemical industry park based on digital twinning

Abstract

The invention relates to the technical field of intelligent management and control, and discloses a digital twinning-based intelligent management and control method for a chemical industrial park, which comprises the following steps of: 1, constructing a multi-physical field particle model of the chemical industrial park by using dissipative particle dynamics, dispersing the characteristics of a fluid field, a thermal field and a chemical field in the park into a particle system, and defining the physical attribute of each particle to obtain a particle system; constructing an interaction relationship among the particles, and determining the state and trajectory of the particles by utilizing the motion equation of the particles and combining the interaction relationship; 2, extracting macroscopic variables in the particle model, and inputting the macroscopic variables as core data of the digital twinborn body; a chemical industry park multi-physics field particle model is constructed through dissipative particle dynamics, macroscopic variables are extracted, refined modeling of dynamic changes of a fluid field, a thermal field and a chemical field is achieved, and the effects of accurately describing multi-physics field coupling dynamic behaviors and improving modeling precision are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent management and control, and specifically to an intelligent management and control method for a chemical park based on digital twins. Background Art

[0002] Chemical parks are an important part of modern industry. The operation process involves complex coupling processes of multiple physical fields such as fluid transportation, heat conduction, and chemical reactions. The coupling process is highly nonlinear, dynamic, and cross-scale, which determines the safety, energy efficiency, and environmental performance of the park's operation. Traditional chemical park management and control methods usually rely on empirical models or simple static monitoring methods, which cannot meet the needs of real-time, high-precision modeling and dynamic control of complex chemical systems.

[0003] In the prior art, the intelligent management method for chemical parks has the following problems:

[0004] The complex interactions among fluid, thermal and chemical fields in chemical parks require high-precision modeling and coupling description. However, traditional modeling methods are mostly based on continuum theory, which can only provide limited descriptions of multi-physical fields at a macroscopic scale, but cannot effectively capture the impact of microscopic particle interactions on system behavior, resulting in reduced modeling accuracy and an inability to accurately reflect the dynamic changes of the system.

[0005] In the coupled description of velocity field, temperature field and chemical concentration field, existing methods usually adopt separate single-field modeling strategies, which cannot achieve global unified modeling of multiple physical fields. The coupled model lacks integrity in dealing with the complex interactions between fluid dynamics, heat conduction and chemical reactions, and cannot fully reflect the dynamic behavior of the actual operation of the chemical park.

[0006] The operating environment of chemical parks is complex and changeable, and the uncertainty of parameters will significantly affect the accuracy of the model. Traditional methods are usually unable to dynamically correct model parameters, resulting in the inability of the twin to adapt to changes in actual operating conditions, limiting its application in real-time monitoring and optimization.

[0007] The complex systems of chemical parks usually require a high-computational solution process. Traditional methods show low efficiency in real-time simulation and feedback, and cannot meet the needs of real-time management and control of the park. In addition, the calculation process fails to effectively utilize parallelization technology, further limiting the real-time dynamic adjustment and process optimization capabilities of the chemical park.

[0008] Therefore, those skilled in the art provide an intelligent management and control method for chemical parks based on digital twins to solve the above-mentioned problems. Summary of the invention

[0009] In view of the shortcomings of the prior art, the present invention provides an intelligent management and control method for a chemical park based on digital twins to solve the problems raised in the above background technology.

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: A chemical park intelligent management and control method based on digital twins, comprising:

[0011] Step 1: Use dissipative particle dynamics to construct a multi-physics field particle model of the chemical park, discretize the characteristics of the fluid field, thermal field and chemical field in the park into a particle system, define the physical properties of each particle, construct the interaction relationship between particles, and use the particle motion equation combined with the interaction relationship to determine the state and trajectory of the particle;

[0012] Step 2: Extract macro variables in the particle model and use them as the core data input of the digital twin. Establish an energy functional model based on fluid viscosity dissipation, thermal energy change, chemical reaction rate and external heat source intensity, and define the model calculation domain and boundary conditions.

[0013] Step 3: Construct a multi-physics coupling model of the chemical park through macro variables and energy functional models, establish the optimization problem in the variational form of energy functionals, clarify the constraints involved in the coupling process, and combine boundary conditions to form a complete multi-physics coupling description to reflect the actual operation status and process change process of the chemical park;

[0014] Step 4: Discretize the multi-physics field coupling model, divide the calculation domain into multiple groups of finite elements using the finite element method, discretize the field variables through linear basis functions, transform the continuous coupling model into a set of algebraic equations on discrete nodes, and solve the discretized coupling model in combination with numerical solution methods to obtain the specific distribution data of the fluid field, thermal field and chemical field of the chemical park;

[0015] Step 5: Combine the discretized calculation results with the real-time data collected by the chemical park sensors, use multi-layer Bayesian inference to correct the model parameters, establish a mapping relationship between sensor data and twin prediction values, introduce parameter uncertainty models and observation noise, and dynamically update the posterior distribution of model parameters to reflect the real-time operation status of the chemical park;

[0016] Step 6. Based on the digital twin after parameter correction, calculate the distribution of fluid field, thermal field and chemical field in real time, combine the calculation results to generate the operation data of the chemical park, and feed it back to the park management and control system to guide the adjustment and optimization of the chemical process and realize intelligent dynamic management and control of the chemical park.

[0017] Preferably, the interaction between particles in step 1 includes:

[0018] Conservative forces are used to describe the attraction or repulsion between particles and are expressed as:

[0019]

[0020] in, is the conservative force between particles i and j, A is the strength of the conservative force, w C (r ij ) is the weight function, r ij is the distance between particle i and particle j, is the unit direction vector from particle i to particle j;

[0021] Dissipative force is used to describe the loss of kinetic energy between particles and is expressed as:

[0022]

[0023] in, is the dissipation force between particles i and j, y is the dissipation coefficient, w D (r ij ) is the weight function of the dissipative force, r ij is the distance between particle i and particle j, is the unit direction vector from particle i to particle j, v ij is the relative velocity of particles i and j;

[0024] Random forces are used to simulate microscopic thermal fluctuations and are expressed as:

[0025]

[0026] in, is the random force between particles i and j, o is the random force intensity, w R (r ij ) is the weight function of the random force, r ij is the distance between particle i and particle j, is the unit direction vector from particle i to particle j, I ij is the random noise factor between particles.

[0027] Preferably, the macro variables extracted in step 2 include:

[0028] The velocity field is expressed as:

[0029]

[0030] The temperature field is expressed as:

[0031]

[0032] The chemical concentration field is expressed as:

[0033]

[0034] Where v(x) represents the macroscopic velocity field at position x, T(x) represents the macroscopic temperature field at position x, C(x) represents the macroscopic chemical concentration field at position x, and m i is the mass of the ith particle, T i is the temperature of the ith particle, C i is the chemical concentration of the ith particle, v i is the velocity vector of the ith particle, r i is the position vector of the ith particle,

[0035] W(xr i ) is the kernel function, which describes the influence weight of the i-th particle at position x.

[0036] Preferably, the energy functional model established in step 3 is:

[0037]

[0038] Where E(u) represents the total energy functional of the multi-physics field, m represents the computational domain of the chemical park, v represents the velocity field, #v represents the gradient of the velocity field, u represents the fluid viscosity, T represents the temperature field, #T represents the gradient of the temperature field, k represents the thermal conductivity, Q represents the intensity of the heat source, and C i represents the concentration field of the chemical substance, i represents the number of the chemical substance, R i (C i ) represents the chemical reaction rate, and dm represents the volume element in the computational domain m.

[0039] Preferably, the optimization condition of the multi-physics field coupling model based on energy functional in step 3 is:

[0040] <NE(u),a-h> ≥0,

[0041] Where NE(u) is the variation of the energy functional, a is the test function, h is the solution, and the optimization problem is subject to the boundary conditions of fluid inlet velocity, wall temperature, and chemical concentration.

[0042] Preferably, in step 4, the computational domain m is discretized into multiple groups of finite elements using the finite element method, and the field variables are represented by the following discretization formula:

[0043] Among them, u h represents the finite element approximate solution of the field variables, φ i is the basis function, u i The discretized model is solved by iterative method to obtain the specific distribution of fluid field, thermal field and chemical field.

[0044] Preferably, in step 5, multi-layer Bayesian inference is used to calibrate the model parameters, and the calibration parameters include fluid viscosity u, thermal conductivity k, chemical reaction rate coefficient R i and the intensity of the external heat source Q, the posterior distribution is updated by the following formula: p(Z|D)∝p(D|Z)p(Z),

[0045] Where Z = (u, k, Q, R i ) are model parameters, D is the observed data,

[0046] p(D|Z) is the likelihood function, the mapping relationship between D and the twin prediction value is described by the observation noise model, and p(Z) is the parameter prior distribution.

[0047] Preferably, the observation data in step 5 includes the distribution information of the fluid field, the thermal field and the chemical field, and the observation model is described by the following formula: D = F (u) + ∈,

[0048] Among them, F(u) is the predicted value of the twin model, ∈ is the measurement noise that obeys the normal distribution, and D is the observed data.

[0049] Preferably, in step 6, GPU parallel computing is used to accelerate real-time computing, and the distribution data of the flow velocity field, temperature field and chemical concentration field predicted by the twin model are fed back to the chemical park management and control system.

[0050] Preferably, the calculation results of the twin are used to dynamically adjust the operating parameters of the chemical park, including the temperature setting value of the reactor, the feed flow rate and the cooling medium flow rate.

[0051] The present invention provides an intelligent management and control method for a chemical park based on digital twins, which has the following beneficial effects:

[0052] 1. The present invention constructs a multi-physical field particle model of a chemical park through dissipative particle dynamics and extracts macroscopic variables, thereby achieving refined modeling of the dynamic changes of fluid fields, thermal fields and chemical fields, and accurately describing the dynamic behavior of multi-physical field coupling and improving the modeling accuracy.

[0053] 2. The present invention constructs a multi-physical field coupling model based on variational optimization of energy functional and combines boundary condition constraints to achieve a global optimization description of the velocity field, temperature field and chemical concentration field, thereby achieving the effect of comprehensively expressing the interaction between multiple physical fields and enhancing the integrity of the coupling model.

[0054] 3. The present invention uses multi-layer Bayesian inference combined with real-time observation data to dynamically correct model parameters, thereby realizing the uncertainty quantification and real-time updating of a series of parameters such as fluid viscosity, thermal conductivity, and chemical reaction rate in the model, thereby improving the adaptive ability of the digital twin and accurately reflecting the actual operating status.

[0055] 4. The present invention accelerates the solution of the discretized multi-physics field model and realizes real-time feedback through GPU parallel computing, realizes the rapid calculation and dynamic adjustment of the flow field, thermal field and chemical field distribution in the chemical park, and improves the real-time simulation capability and optimizes the operating efficiency of the chemical process. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0057] In order to make the technical personnel in the technical field understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in combination with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a partial embodiment of the present invention, not a complete embodiment. Based on the embodiment of the present invention, other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0058] The present invention is described in detail below in conjunction with the accompanying drawings:

[0059] Example:

[0060] Please see attached Figure 1 , an embodiment of the present invention provides an intelligent management and control method for a chemical park based on digital twins, comprising:

[0061] Step 1: Use dissipative particle dynamics to construct a multi-physics field particle model of the chemical park, discretize the characteristics of the fluid field, thermal field and chemical field in the park into a particle system, define the physical properties of each particle, construct the interaction relationship between particles, and use the particle motion equation combined with the interaction relationship to determine the state and trajectory of the particle;

[0062] Step 2: Extract macro variables in the particle model and use them as the core data input of the digital twin. Establish an energy functional model based on fluid viscosity dissipation, thermal energy change, chemical reaction rate and external heat source intensity, and define the model calculation domain and boundary conditions.

[0063] Step 3: Construct a multi-physics coupling model of the chemical park through macro variables and energy functional models, establish the optimization problem in the variational form of energy functionals, clarify the constraints involved in the coupling process, and combine boundary conditions to form a complete multi-physics coupling description to reflect the actual operation status and process change process of the chemical park;

[0064] Step 4: Discretize the multi-physics field coupling model, divide the calculation domain into multiple groups of finite elements using the finite element method, discretize the field variables through linear basis functions, transform the continuous coupling model into a set of algebraic equations on discrete nodes, and solve the discretized coupling model in combination with numerical solution methods to obtain the specific distribution data of the fluid field, thermal field and chemical field of the chemical park;

[0065] Step 5: Combine the discretized calculation results with the real-time data collected by the chemical park sensors, use multi-layer Bayesian inference to correct the model parameters, establish a mapping relationship between sensor data and twin prediction values, introduce parameter uncertainty models and observation noise, and dynamically update the posterior distribution of model parameters to reflect the real-time operation status of the chemical park;

[0066] Step 6. Based on the digital twin after parameter correction, calculate the distribution of fluid field, thermal field and chemical field in real time, combine the calculation results to generate the operation data of the chemical park, and feed it back to the park management and control system to guide the adjustment and optimization of the chemical process and realize intelligent dynamic management and control of the chemical park.

[0067] Benefits of step 1: By discretizing complex multi-physics fields into particle systems, particle dynamics can accurately capture the microscopic interactions between fluid fields, thermal fields, and chemical fields. Constructing a model of the interaction between conservative forces, dissipative forces, and random forces between particles enables the model to reflect nonlinear dynamic behaviors at the microscopic level. Compared with traditional macroscopic modeling methods, particle dynamics can describe the evolution of physical fields at a finer granularity and improve the accuracy of the model;

[0068] Benefits of step 2: Extracting macro variables such as velocity field, temperature field, and chemical concentration field as the core data input of the digital twin helps to achieve global modeling of the system. An energy functional model is established based on fluid viscosity dissipation, thermal energy change, and chemical reaction rate, so that the model can uniformly describe the coupling relationship of multiple physical fields. Define the calculation domain and boundary conditions to provide a basis for subsequent optimization problems, ensure the complete physical meaning of modeling, and enhance the ability to describe the synergy of multiple physical fields;

[0069] Benefits of step 3: Constructing the optimization problem through the variational form of the energy functional ensures that the interaction between multiple physical fields can be solved in a globally optimal way. Clarifying the constraints in the coupling process, including boundary conditions such as flow rate, temperature and concentration, and enhancing the physical constraints of the model. The coupling model can fully reflect the dynamic operation status and process change process of the chemical park, providing basic data support for subsequent analysis;

[0070] Benefits of step 4: The finite element method is used to divide the computational domain into finite elements. The field variables are discretized through linear basis functions, which can effectively transform the complex continuous coupling model into a set of algebraic equations. The discretization process combined with the numerical solution method can effectively improve the computational efficiency and solution stability of the model. The distribution data after the solution includes the precise numerical description of the fluid field, thermal field and chemical field, providing detailed operation data for the subsequent optimization decision of the park;

[0071] Benefits of step 5: Through multi-layer Bayesian inference, the uncertainty quantification and real-time correction mechanism of parameters are introduced, so that the model can dynamically adjust key parameters such as fluid viscosity, thermal conductivity, and chemical reaction rate. Establishing a mapping relationship between sensor data and twin prediction values, and introducing observation noise can significantly improve the adaptability of the model and ensure that the twin always reflects the real-time operating status of the chemical park. Dynamically updating the posterior distribution of parameters can effectively deal with uncertainty problems in complex park operating environments and improve the adaptive ability of digital twins;

[0072] Benefits of step 6: Based on the corrected digital twin, the distribution of fluid field, thermal field and chemical field is calculated in real time, which can provide dynamic operation data support for the park management and control system. The feedback mechanism combined with the real-time calculation results can effectively guide the adjustment and optimization of chemical processes to ensure the efficiency and safety of park operation. It can realize intelligent dynamic control of chemical parks, improve the level of automated management of parks, and meet the requirements of energy saving, environmental protection and efficient production.

[0073] The interactions between particles in step 1 include:

[0074] Conservative forces are used to describe the attraction or repulsion between particles and are expressed as:

[0075]

[0076] in, is the conservative force between particles i and j, A is the strength of the conservative force, w C (r ij ) is the weight function, r ij is the distance between particle i and particle j, is the unit direction vector from particle i to particle j;

[0077] Dissipative force is used to describe the loss of kinetic energy between particles and is expressed as:

[0078]

[0079] in, is the dissipation force between particles i and j, y is the dissipation coefficient, w D (r ij ) is the weight function of the dissipative force, r ijis the distance between particle i and particle j, is the unit direction vector from particle i to particle j, v ij is the relative velocity between particles i and j;

[0080] Random forces are used to simulate microscopic thermal fluctuations and are expressed as:

[0081]

[0082] in, is the random force between particles i and j, o is the random force intensity, w R (r ij ) is the weight function of the random force, r ij is the distance between particle i and particle j, is the unit direction vector from particle i to particle j, I ij is the random noise factor between particles.

[0083] Conservative forces are used to describe the attraction or repulsion relationship between particles, ensuring that the particle system has a stable microstructure. In the multi-physics field modeling of chemical parks, conservative forces reflect the basic attraction or repulsion relationship between particles due to the interaction between substances, which helps to simulate the microscopic dynamic characteristics of fluids, thermal fields and chemical fields.

[0084] Dissipative force is used to describe the kinetic energy loss between particles. It makes the system stable by simulating the energy dissipation mechanism during particle motion. In the multi-physics field modeling of chemical parks, dissipative force reflects the resistance effect inside the fluid through viscous dissipation, providing important support for the real dynamic characteristics of the fluid field.

[0085] Random forces are used to simulate microscopic thermal fluctuations and capture random dynamic effects in particle systems, making the system closer to the fluctuation characteristics of real physical fields. In the modeling of chemical parks, random forces are used to reflect the impact of random effects at the microscopic scale on the behavior of macroscopic systems, which is especially important in describing thermal fields and chemical reactions.

[0086] The macro variables extracted in step 2 include:

[0087] The velocity field is expressed as:

[0088]

[0089] The temperature field is expressed as:

[0090]

[0091] The chemical concentration field is expressed as:

[0092]

[0093] Where v(x) represents the macroscopic velocity field at position x, T(x) represents the macroscopic temperature field at position x, C(x) represents the macroscopic chemical concentration field at position x, and m i is the mass of the ith particle, T i is the temperature of the ith particle, C i is the chemical concentration of the ith particle, v i is the velocity vector of the ith particle, r i is the position vector of the ith particle,

[0094] W(xr i ) is the kernel function, which describes the influence weight of the i-th particle at position x.

[0095] The advantage of velocity field extraction formula is that it maps the local velocity distribution of discrete particle system into a continuous macro velocity field, which can fully reflect the dynamic behavior of fluid in chemical park. i ) weights are assigned according to the distance between particle i and target position x to ensure smoothness and accuracy of the velocity field. The macroscopic velocity field is an important parameter for describing the motion characteristics of fluids and provides basic data support for subsequent energy functional modeling.

[0096] The benefits of the temperature field extraction formula: using the temperature T of each particle in the particle model i The calculated macroscopic temperature field can fully reflect the distribution characteristics of heat conduction in the chemical park. The weighting of the kernel function makes the temperature field change continuously and smoothly in the calculation domain, effectively avoiding the spatial inhomogeneity of discrete particle data. The temperature field is crucial to the operating status monitoring of chemical equipment and provides key input for safety management and process optimization.

[0097] Benefits of chemical concentration field extraction formula: Extracting chemical concentration field can accurately describe the distribution characteristics of reactants and products in the chemical park in space. By introducing kernel function weights, the concentration field is not smooth and can accurately capture the local characteristics of chemical distribution. The concentration field is an important parameter for describing the chemical reaction process and dynamic evolution, and provides accurate input for the chemical reaction rate term in the energy functional.

[0098] The energy functional model established in step 3 is:

[0099]

[0100] Where E(u) represents the total energy functional of the multi-physics field, m represents the computational domain of the chemical park, v represents the velocity field, #v represents the gradient of the velocity field, u represents the fluid viscosity, T represents the temperature field, #T represents the gradient of the temperature field, k represents the thermal conductivity, Q represents the intensity of the heat source, and C i represents the concentration field of the chemical substance, i represents the number of the chemical substance, Ri (C i ) represents the chemical reaction rate, dm represents the volume element in the computational domain m;

[0101] The optimization conditions of the multi-physics coupling model based on energy functional in step 3 are:

[0102] <NE(u),a-h> ≥0,

[0103] Where NE(u) is the variation of the energy functional, a is the test function, h is the solution, and the optimization problem is subject to the boundary conditions of fluid inlet velocity, wall temperature, and chemical concentration.

[0104] Benefits of building energy functional models:

[0105] Through the energy functional model, fluid dynamics, heat conduction and chemical reaction processes are integrated into a unified mathematical framework, so that the dynamic behavior of multiple physical fields can be optimized and solved simultaneously. Compared with the single-field separation modeling method, the coupled model can more comprehensively reflect the actual physical processes of the chemical park.

[0106] Each item in the energy functional describes the contribution of fluid shear deformation, thermal energy change, external heat source and chemical reaction rate to the total energy of the system, ensuring that the physical meaning of the model is clear and fully expressed.

[0107] The form of energy functional can intuitively reflect the energy exchange and coupling relationship between various physical fields, provide a mathematical basis for subsequent variational optimization solutions, and transform complex multi-field problems into solvable mathematical problems.

[0108] Benefits of energy functional based optimization:

[0109] Through the optimization conditions in variational form, the optimal solution of the energy functional satisfies the principle of global minimum energy, so that the solution conforms to the laws of physics and can achieve overall optimization in the computational domain.

[0110] The boundary conditions of fluid inlet velocity, wall temperature and chemical concentration are explicitly introduced into the optimization problem, so that the solution of the model can meet the constraints in the actual operation of the chemical park, further improving the accuracy and applicability of the model.

[0111] The variational operation in the optimization conditions transforms the multi-physics field coupling problem into a mathematical problem that can be solved iteratively, and can effectively handle the nonlinear coupling dynamics of fluid flow, heat conduction and chemical reactions in the park.

[0112] In summary, the construction of the energy functional model in step 3 and the coupled solution based on optimization conditions have the following key advantages:

[0113] The process description of fluid dynamics, heat conduction and chemical reactions is integrated through the energy functional framework, so that the interactions between different physical fields can be accurately modeled and optimized.

[0114] The variational optimization conditions of the energy functional ensure the global optimality of the solution, and the actual operating boundary conditions are introduced to enhance the physical accuracy of the solution.

[0115] By considering the dynamic constraints of the chemical park, such as flow rate, temperature and concentration, the model can accurately reflect the actual operating status within the park and provide input data for subsequent steps.

[0116] In step 4, the finite element method is used to discretize the computational domain m into multiple groups of finite elements, and the field variables are expressed by the following discretization formula:

[0117] Among them, u h represents the finite element approximate solution of the field variables, φ i is the basis function, u i The discretized model is solved by iterative method to obtain the specific distribution of fluid field, thermal field and chemical field.

[0118] In step 5, multi-layer Bayesian inference is used to calibrate the model parameters, including fluid viscosity u, thermal conductivity k, chemical reaction rate coefficient R i and the intensity of the external heat source Q, the posterior distribution is updated by the following formula: p(Z|D)∝p(D|Z)p(Z),

[0119] Where Z = (u, k, Q, R i ) are model parameters, D is the observed data,

[0120] p(D|Z) is the likelihood function, the mapping relationship between D and the twin prediction value is described by the observation noise model, and p(Z) is the parameter prior distribution;

[0121] The observation data in step 5 includes the distribution information of fluid field, thermal field and chemical field. The observation model is described by the following formula: D = F (u) + ∈,

[0122] Among them, F(u) is the predicted value of the twin model, ∈ is the measurement noise that obeys the normal distribution, and D is the observed data.

[0123] In step 6, GPU parallel computing is used to accelerate real-time computing, and the distribution data of the velocity field, temperature field, and chemical concentration field predicted by the twin model are fed back to the chemical park management and control system;

[0124] The calculation results of the twin are used to dynamically adjust the operating parameters of the chemical park, including the temperature set point of the reactor, the feed flow rate and the cooling medium flow rate.

[0125] The computational domain m is divided into multiple groups of finite elements through the finite element method. The discretization of the field variables by the basis function transforms the complex continuous multi-physics field model into a discrete set of algebraic equations, effectively reducing the difficulty of solving. The selection of basis functions and the flexibility of unit division allow the use of adaptive high-resolution grids in different areas, which helps to improve the solution accuracy in key areas. The discretization model can obtain the specific distribution of fluid fields, thermal fields and chemical fields in the chemical park through iterative solution methods, providing basic data for the digital twin to reflect the dynamic operation of the park in real time.

[0126] Multi-layer Bayesian inference combines observation data and prior knowledge to dynamically update the posterior distribution and correct key parameters, so that the twin can adapt to changes in the park's operating environment in real time. The introduction of the observation noise model can quantify and deal with random errors and environmental disturbances in sensor data, improve the robustness of the digital twin, and more accurately reflect the actual operating status. The mapping relationship between the observation data and the twin's predicted value is integrated through the construction of the likelihood function to ensure the reliability and physical consistency of the parameter correction results.

[0127] Through GPU parallel computing, the solution process of the discretized multi-physics field model is efficiently accelerated, and the distribution data of the real-time velocity field, temperature field and chemical concentration field can be quickly generated to meet the needs of real-time feedback and dynamic adjustment of the chemical park. The twin calculation results can be fed back to the management and control system in real time to guide the dynamic adjustment of the park's operating parameters. Through real-time calculation and adjustment, the twin can optimize the multi-objective operating efficiency of the park, including improving product yield, reducing energy consumption and reducing emissions, etc., to meet the requirements of modern chemical parks for energy conservation, environmental protection and safety.

[0128] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A chemical park intelligent management and control method based on digital twins, characterized in that: include: Step 1: Use dissipative particle dynamics to construct a multi-physics field particle model of the chemical park, discretize the characteristics of the fluid field, thermal field and chemical field in the park into a particle system, define the physical properties of each particle, construct the interaction relationship between particles, and use the particle motion equation combined with the interaction relationship to determine the state and trajectory of the particle; Step 2: Extract macro variables in the particle model and use them as the core data input of the digital twin. Establish an energy functional model based on fluid viscosity dissipation, thermal energy change, chemical reaction rate and external heat source intensity, and define the model calculation domain and boundary conditions. Step 3: Construct a multi-physics coupling model of the chemical park through macro variables and energy functional models, establish the optimization problem in the variational form of energy functionals, clarify the constraints involved in the coupling process, and combine boundary conditions to form a complete multi-physics coupling description to reflect the actual operation status and process change process of the chemical park; Step 4: Discretize the multi-physics field coupling model, divide the calculation domain into multiple groups of finite elements using the finite element method, discretize the field variables through linear basis functions, transform the continuous coupling model into a set of algebraic equations on discrete nodes, and solve the discretized coupling model in combination with numerical solution methods to obtain the specific distribution data of the fluid field, thermal field and chemical field of the chemical park; Step 5: Combine the discretized calculation results with the real-time data collected by the chemical park sensors, use multi-layer Bayesian inference to correct the model parameters, establish a mapping relationship between sensor data and twin prediction values, introduce parameter uncertainty models and observation noise, and dynamically update the posterior distribution of model parameters to reflect the real-time operation status of the chemical park; Step 6. Based on the digital twin after parameter correction, calculate the distribution of fluid field, thermal field and chemical field in real time, combine the calculation results to generate the operation data of the chemical park, and feed it back to the park management and control system to guide the adjustment and optimization of the chemical process and realize intelligent dynamic management and control of the chemical park.

2. According to claim 1, a chemical park intelligent management and control method based on digital twins is characterized in that: The interaction between particles in step 1 includes: Conservative forces are used to describe the attraction or repulsion between particles and are expressed as: in, is the conservative force between particles i and j, A is the strength of the conservative force, w C (r ij ) is the weight function, r ij is the distance between particle i and particle j, is the unit direction vector from particle i to particle j; Dissipative force is used to describe the loss of kinetic energy between particles and is expressed as: in, is the dissipation force between particles i and j, y is the dissipation coefficient, w D (r ij ) is the weight function of the dissipative force, r ij is the distance between particle i and particle j, is the unit direction vector from particle i to particle j, v ij is the relative velocity between particles i and j; Random forces are used to simulate microscopic thermal fluctuations and are expressed as: in, is the random force between particles i and j, o is the random force intensity, w R (r ij ) is the weight function of the random force, r ij is the distance between particle i and particle j, is the unit direction vector from particle i to particle j, I ij is the random noise factor between particles.

3. According to claim 1, a chemical park intelligent management and control method based on digital twins is characterized in that: The macro variables extracted in step 2 include: The velocity field is expressed as: The temperature field is expressed as: The chemical concentration field is expressed as: Where v(x) represents the macroscopic velocity field at position x, T(x) represents the macroscopic temperature field at position x, C(x) represents the macroscopic chemical concentration field at position x, and m i is the mass of the ith particle, T i is the temperature of the ith particle, C i is the chemical concentration of the ith particle, v i is the velocity vector of the ith particle, r i is the position vector of the ith particle, W(xr i ) is the kernel function, which describes the influence weight of the i-th particle at position x.

4. According to claim 1, a chemical park intelligent management and control method based on digital twins is characterized in that: The energy functional model established in step 3 is: Where E(u) represents the total energy functional of the multi-physics field, m represents the computational domain of the chemical park, v represents the velocity field, #v represents the gradient of the velocity field, u represents the fluid viscosity, T represents the temperature field, #T represents the gradient of the temperature field, k represents the thermal conductivity, Q represents the intensity of the heat source, and C i represents the concentration field of the chemical substance, i represents the number of the chemical substance, R i (C i ) represents the chemical reaction rate, and dm represents the volume element in the computational domain m.

5. According to claim 4, a chemical park intelligent management and control method based on digital twins is characterized in that: The optimization condition of the multi-physics field coupling model based on energy functional in step 3 is: <NE(u),a-h> ≥0, Where NE(u) is the variation of the energy functional, a is the test function, h is the solution, and the optimization problem is subject to the boundary conditions of fluid inlet velocity, wall temperature, and chemical concentration.

6. The intelligent management and control method of a chemical park based on digital twin according to claim 1 is characterized in that: In step 4, the computational domain m is discretized into multiple groups of finite elements using the finite element method, and the field variables are expressed by the following discretization formula: Among them, u h represents the finite element approximate solution of the field variables, φ i is the basis function, u i The discretized model is solved by iterative method to obtain the specific distribution of fluid field, thermal field and chemical field.

7. The intelligent management and control method of a chemical park based on digital twin according to claim 6 is characterized in that: In step 5, multi-layer Bayesian inference is used to calibrate the model parameters, and the calibration parameters include fluid viscosity u, thermal conductivity k, chemical reaction rate coefficient R i and the intensity of the external heat source Q, the posterior distribution is updated by the following formula: p(Z|D)∝p(D|Z)p(Z), Where Z = (u, k, Q, R i ) are model parameters, D is the observed data, p(D|Z) is the likelihood function. The mapping relationship between D and the twin prediction value is described by the observation noise model. p(Z) is the parameter prior distribution.

8. The intelligent management and control method of a chemical park based on digital twin according to claim 7 is characterized in that: The observation data in step 5 includes the distribution information of fluid field, thermal field and chemical field, and the observation model is described by the following formula: D = F (u) + ∈, Among them, F(u) is the predicted value of the twin model, ∈ is the measurement noise that obeys the normal distribution, and D is the observed data.

9. The intelligent management and control method of a chemical park based on digital twin according to claim 1 is characterized in that: In step 6, GPU parallel computing is used to accelerate real-time computing, and the distribution data of the flow velocity field, temperature field and chemical concentration field predicted by the twin model are fed back to the chemical park management and control system.

10. The intelligent management and control method of a chemical park based on digital twin according to claim 9 is characterized in that: The calculation results of the twin are used to dynamically adjust the operating parameters of the chemical park, including the temperature set point of the reactor, the feed flow rate and the cooling medium flow rate.

Images