PINN-based ammonia production operation regulation and control method and system

Abstract

The invention relates to the technical field of renewable energy driven ammonia production, and discloses a PINN-based ammonia production operation regulation and control method and system, and the regulation and control method comprises the steps: modeling different units of an ammonia synthesis system, and coupling an ammonia synthesis neural network model and an interstage heat exchange model to establish an ammonia synthesis simulation model; a raw material feeding and mixing model, a multi-stage compression process model, an inlet and outlet heat exchange model, an ammonia synthesis simulation model, a cooling process model and an ammonia separation process HDMR model are coupled according to the ammonia synthesis system connection relation to form an ammonia synthesis circulation ring; optimizing variables influencing variable load operation are selected, the leveling ammonia cost serves as a target function, optimization variable solving is carried out on the basis of an ammonia synthesis circulation ring through an optimization algorithm under the target production load, and the optimization variables are used for regulating and controlling an ammonia synthesis system. The method is suitable for the condition of variable load regulation and control of the ammonia synthesis circle driven by renewable energy sources, the calculation cost is lower, the precision is higher, and regulation and control of the green ammonia production process are facilitated so as to enhance the system economy.

Description

Technical Field

[0001] The present invention relates to the technical field of ammonia production driven by renewable energy, and particularly relates to a method and system for regulating ammonia production operation based on PINN. Background Art

[0002] The ammonia synthesis system is an industrial device used to convert nitrogen and hydrogen into ammonia under specific conditions. As Figure 1 shown, it includes a raw material intake unit, a multi-stage compression unit, an ammonia synthesis reaction unit, a cooling unit, and a flash separation unit. The raw material intake unit is used to mix the required hydrogen and nitrogen in the required proportion. The multi-stage compression unit is used to pressurize the mixed hydrogen and nitrogen. The ammonia synthesis reaction unit conducts the reaction of hydrogen and nitrogen through an ammonia synthesis reactor. The cooling unit is used to cool the reactants in the ammonia synthesis reaction unit. The flash separation unit is used to separate the generated ammonia from the unreacted reactants. The obtained liquid ammonia product enters a storage tank or a refining unit, while the unreacted reactants are recycled back to the ammonia synthesis reactor to improve the utilization rate of the raw material gas. The traditional ammonia synthesis system mainly obtains raw materials by using fossil energy (coal, natural gas, etc.) to produce hydrogen and air separation to produce nitrogen, and synthesizes ammonia under the action of a catalyst in the ammonia synthesis reactor. However, chemical plants using fossil energy as raw materials will emit a large amount of carbon dioxide during the hydrogen production process, exacerbating the greenhouse effect. In order to reduce carbon dioxide emissions and alleviate global warming and other negative impacts, renewable new energy such as wind energy, solar energy, hydropower, and photovoltaic can be used to generate electricity to provide energy supply for electrolytic water hydrogen production and ammonia synthesis, so as to reduce the use of fossil energy.

[0003] The "green electricity - green hydrogen - green ammonia" system driven by renewable energy is an important technical path for energy transformation. Different from the ammonia synthesis system based on traditional fossil energy, the supply of renewable energy is random, intermittent, and volatile, resulting in the inability of the new ammonia synthesis system production technology to operate according to the production technology requirements of "long - cycle, stable, and full - load" of the traditional fossil energy process. That is, the ammonia synthesis production process based on traditional fossil energy (mainly natural gas and coal) generally operates at a stable load state within a relatively long time period, and there are basically no frequent production load switching scenarios. However, when coupling wind power, photovoltaic power, etc. into the energy supply of the ammonia synthesis system, due to the inevitable high - frequency and wide - amplitude volatility of renewable energy such as wind and light, in order to ensure the stable progress of the production process, it is necessary to frequently adjust the load of the ammonia synthesis system to adapt to the production raw materials and ensure the green attribute of the production system. At present, the production load switching of the ammonia synthesis system basically relies on manual experience. If the ammonia synthesis system driven by renewable energy completely relies on manual experience to achieve frequent load switching, this will sharply increase the manual labor intensity. At the same time, the judgment of process control parameters by manual experience is somewhat single and rough to a certain extent, and cannot meet the economic requirements of the system under frequent multi - working - condition load adjustments. Summary of the Invention

[0004] The purpose of the present invention is to overcome the problems of high labor intensity, relatively single and poor - precision control parameters existing in the production load switching of the existing ammonia synthesis system relying on manual experience, and provide an ammonia production operation control method and system based on PINN (Physics - Informed Neural Network).

[0005] In order to achieve the above - mentioned invention purpose, the present invention provides the following technical solutions:

[0006] An ammonia production operation control method based on PINN includes the following steps:

[0007] Model different units of the ammonia synthesis system, including establishing a raw material feed mixing model, a multi - stage compression process model, an inlet - outlet heat exchange model, an inter - stage heat exchange model, and a cooling process model. Based on PINN, establish an ammonia synthesis neural network model for the reaction process of the ammonia synthesis reactor; use a high - dimensional response surface model to establish a demixing process HDMR model for the flash demixing process;

[0008] Based on the structure of the ammonia synthesis reactor, couple the ammonia synthesis neural network model with the inter - stage heat exchange model to establish an ammonia synthesis simulation model; according to the connection relationship of the ammonia synthesis system, couple the raw material feed mixing model, the multi - stage compression process model, the inlet - outlet heat exchange model, the ammonia synthesis simulation model, the cooling process model, and the demixing process HDMR model to form an ammonia synthesis cycle loop;

[0009] Select the optimization variables that affect the variable load operation, and perform simulations based on the ammonia synthesis recycle loop at the target production load. Using the levelized ammonia cost as the objective function, an optimization algorithm is employed to solve for the optimization variables. The ammonia synthesis system is regulated according to the optimization variables obtained at the target production load.

[0010] In the above technical solution, the regulation method includes modeling different operation units, forming the ammonia synthesis recycle loop and optimizing the variable load operation of the ammonia synthesis recycle loop. By modeling different operation units of the ammonia synthesis system, a raw material feed mixing model, a multistage compression process model, an inlet and outlet heat exchange model, an ammonia synthesis neural network model, an inter-stage heat exchange model, a cooling process model, and a partial ammonia separation process HDMR model are respectively established. Then, they are coupled according to the connection relationship of the ammonia synthesis system to form the ammonia synthesis recycle loop. Then, the optimization variables output by optimizing the variable load operation of the ammonia synthesis recycle loop are determined. Using the levelized ammonia cost as the objective function, simulation calculations are performed based on the ammonia synthesis recycle loop at the target production load using an optimization algorithm to obtain the optimal optimization variables at the target production load. The control values of the optimization variables at the target production load are used to regulate the variable load of the ammonia synthesis system. The present invention establishes a deep learning architecture that integrates the physical and chemical mechanisms of the process based on PINN, which can achieve high-precision simulation of complex physical and chemical processes, be used for modeling the reaction process of the ammonia synthesis reactor, form the ammonia synthesis recycle loop, conduct simulation and optimization calculations of the ammonia synthesis system, realize system energy consumption optimization, and support load adjustment control decisions. The present invention solves the problems of large labor intensity, relatively single regulation parameters, and poor accuracy existing in the production load switching of the existing ammonia synthesis system relying on manual experience.

[0011] Through the above technical solution, the method of the present invention has the advantages of high calculation accuracy, low cost, low human labor intensity, and good system economy, and is applicable to the variable load production regulation of the green ammonia production process driven by renewable energy; it not only ensures the accuracy of process simulation and simulation but also reduces the model calculation complexity, reduces the solution cost, and ensures accuracy; the coupling of unit operation process modeling based on machine learning and system integration optimization based on mathematical programming is easy to solve and the results are highly available.

[0012] As a preferred embodiment of the present invention, the exergy conservation equation of the multistage compression process model is as follows:

[0013]

[0014] T out = f(T in , P in , P out )

[0015] Q = f(Cp, T in , T out , F)

[0016] Where Power is the power consumption during the compression process, T and P represent the temperature and pressure of the stream respectively, Q is the heat, Cp is the isobaric specific heat capacity of the stream, and F is the stream flow rate.

[0017] As a more preferred embodiment of the present invention, the specific calculation formula of Power is:

[0018]

[0019] T out The specific calculation formula is:

[0020]

[0021] The specific calculation formula of Q is:

[0022]

[0023] Among them, is the compressor efficiency, γ is the polytropic exponent, Mw and Cp are the molecular weight and specific heat capacity of the gas mixture respectively, and the subscript l represents the compressor.

[0024] As a preferred embodiment of the present invention, the exergy conservation equation of the inlet and outlet heat exchange model or the inter-stage heat exchange model of the ammonia synthesis reactor is as follows:

[0025] T out = f(T t in , T s in , ε)

[0026] Among them, T out is the temperature of the outlet stream, ε is the efficiency parameter reflecting the heat exchanger configuration, which is determined by the heat exchanger configuration, T s in is the shell-side feed temperature, and T t in is the tube-side feed temperature.

[0027] As a more preferred embodiment of the present invention, the tube-side outlet temperature of the inter-stage heat exchange model of the ammonia synthesis reactor is determined by the following equation:

[0028]

[0029] Among them, is the tube-side outlet temperature, is the inlet temperature of the ammonia synthesis reactor, and the subscript k represents the split at the inlet of the ammonia synthesis reactor, and RHX represents the inter-stage cooler after the corresponding catalyst bed.

[0030] As a preferred embodiment of the present invention, the specific implementation method for establishing the ammonia synthesis neural network model is as follows:

[0031] Based on the kinetics and thermodynamics of the ammonia synthesis catalytic reaction process and the physicochemical characteristics of the catalyst bed in the ammonia synthesis reactor, a nonlinear partial differential mechanism equation is established. The nonlinear partial differential mechanism equation is coupled with the physics-informed neural network PINN, and after training to meet the accuracy requirements, the ammonia synthesis neural network model is obtained.

[0032] As a more preferred embodiment of the present invention, the nonlinear partial differential mechanism equation is as follows:

[0033]

[0034] where i represents the reactant, b represents the number of catalyst layers, v is the stoichiometric coefficient, X is the conversion rate of the corresponding reactant, V is the volume of the catalyst bed, is the reaction rate, is the molar flow rate of reactant i entering catalyst bed b; T represents the temperature of the reaction mixture, ΔH is the heat of reaction, m b represents the total mass flow rate of the reaction mixture, F b,i is the mass flow rate of component i in b, Mw i is the molecular weight of component i, and Cp represents the specific heat capacity of the reaction mixture.

[0035] As a more preferred embodiment of the present invention, when training the ammonia synthesis neural network model, the sobol sampling technique is used to collect multiple input variables in the target data space to form sample data X. The sample data X is input into the ammonia synthesis neural network model to obtain output variables, forming the target value Y. The sample data X→Y is divided into a training set and a test set, which are respectively used to train and test the ammonia synthesis neural network model. During the training process, the loss function of the neural network and the residual of the nonlinear partial differential mechanism equation are optimized until the accuracy meets the requirements. The input variables include the molar flow rate, molar composition fraction, bed temperature, bed pressure, and catalyst bed volume of the inlet logistics of the catalyst bed, and the output variables include the hydrogen conversion rate and the outlet temperature of the catalyst bed logistics.

[0036] As a preferred embodiment of the present invention, the calculation formula of the cooling process model is as follows:

[0037]

[0038] where, represents the ammonia concentration in the gas phase flow, P (atm) and T (K) respectively represent the operating pressure and temperature of the flash separation unit, Q cool is the energy consumption of the ammonia refrigeration system, is the discharge temperature, is the inlet temperature, is the output flow rate, Mw n is the molecular weight.

[0039] As a preferred embodiment of the present invention, the optimization variables include: the hydrogen-nitrogen ratio at the reactor inlet, the inert gas content, the outlet pressure of the multistage compressor, the recycle material flow rate, the purge material flow rate, and the terminal temperature of the ammonia cooling process.

[0040] As a preferred embodiment of the present invention, the constraint conditions include: the energy conservation equation for the compression process, the energy conservation equation for the heat exchange process, the mass conservation equation for the reaction process, the energy conservation equation for the reaction process, the energy conservation equation for the separation process, the mass conservation equation for the separation process, and the process stream connection equation.

[0041] As a preferred embodiment of the present invention, the levelized ammonia cost includes capital expenditure, operating expenditure, by-product revenue, and total ammonia production. The calculation formula for the levelized ammonia cost is as follows:

[0042]

[0043] where LCOA represents the levelized ammonia cost of the product, CAPEX j and OPEX j respectively represent the capital cost and operating cost of unit operation j, is the electrolyzer oxygen product flow rate, is the ammonia product flow rate at the outlet of the flash separation, is the molecular weight of ammonia, ΔHr is the annual operating time, ACCR is the depreciation rate, and ir and ny are the interest rate and plant life respectively.

[0044] Another aspect of the present invention provides an ammonia production operation regulation system based on PINN, and the system includes:

[0045] A modeling module for modeling different units of the ammonia synthesis system, including establishing a raw material feed mixing model, a multistage compression process model, an inlet and outlet heat exchange model, an inter-stage heat exchange model, a cooling process model, and establishing an ammonia synthesis neural network model for the reaction process of the ammonia synthesis reactor based on PINN; using a high-dimensional response surface model to establish a HDMR model for the ammonia separation process in the flash ammonia separation process;

[0046] An ammonia synthesis recycle loop module for coupling the ammonia synthesis neural network model and the inter-stage heat exchange model based on the structure of the ammonia synthesis reactor to establish an ammonia synthesis simulation model; coupling the raw material feed mixing model, the multistage compression process model, the inlet and outlet heat exchange model, the ammonia synthesis simulation model, the cooling process model, and the HDMR model for the ammonia separation process according to the connection relationship of the ammonia synthesis system to form an ammonia synthesis recycle loop;

[0047] An optimization calculation module, configured to select optimization variables affecting variable load operation, simulate based on the ammonia synthesis recycle loop at a target production load, use the levelized ammonia cost as an objective function, and solve the optimization variables using an optimization algorithm;

[0048] A regulation module, configured to regulate the ammonia synthesis system according to the optimization variables at the target production load obtained by the solution.

[0049] The present invention also provides an electronic device, including at least one processor, and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned method for regulating ammonia production operation based on PINN.

[0050] Compared with the prior art, the beneficial effects of the present invention are:

[0051] 1. The present invention provides a method for regulating ammonia production operation based on PINN, which establishes a deep learning architecture integrating process physical and chemical mechanisms based on PINN, can achieve high-precision simulation of complex physical and chemical processes, is used for modeling the reaction process of an ammonia synthesis reactor, forms an ammonia synthesis recycle loop, conducts simulation and optimization calculation of the ammonia synthesis system, realizes system energy consumption optimization, and is used to support the variable load operation optimization of a green ammonia production system.

[0052] 2. The present invention is applicable to the situation of variable load regulation of an ammonia synthesis loop driven by renewable energy. Applying PINN to the fidelity modeling of an ammonia synthesis catalytic bed layer and supporting the variable load regulation optimization of the synthesis loop has lower calculation costs and higher precision, which is beneficial to the regulation of the green ammonia production process to enhance system economy. Description of the Drawings

[0053] Figure 1 It is a schematic structural diagram of an ammonia synthesis system;

[0054] Figure 2 It is a flowchart of the method for regulating ammonia production operation based on PINN of the present invention

[0055] Figure 3 It is a schematic structural diagram of an ammonia synthesis reactor in Embodiment 1

[0056] Figure 4 It is a modeling strategy diagram of the ammonia synthesis reactor in Embodiment 1;

[0057] Figure 5 It is a modeling diagram of the flash separation unit in Embodiment 1;

[0058] Figure 6 It is the optimization calculation results of the hydrogen-nitrogen ratio, inert gas content, and multi-stage compression outlet pressure under different production loads in Embodiment 2

[0059] Figure 7 It is the optimal calculation result of the ammonia separation temperature under different production loads in Example 2. Specific embodiments

[0060] In order to more clearly describe the invention purpose, technical solution and technical effect advantages in the specific embodiments of the present invention, the following will combine the accompanying drawings of the present invention to elaborate on the solutions in the specific embodiments in detail. The specific technical solutions involved in the following specific embodiments are only for clearly and completely describing the innovative technical solutions of the present invention. They are only a part of the specific implementation solutions that the present invention can adopt, not all embodiments, and should not be construed as a limitation on the innovative solutions of the present invention. Any solution adopting the same inventive concept of the present invention should be included in the protection scope of the present invention.

[0061] Secondly, the relevant descriptions of the accompanying drawings in the specific embodiments of the present invention are only for facilitating those skilled in the art to understand the solutions of the present invention. Some details shown in the drawings are for clearly presenting the technical solutions. It should not be considered that all technical features in the drawings must be incorporated into the specific embodiments, nor can the detailed features in the drawings be regarded as additional limitations on the innovative technical solutions of the present invention. The components in each embodiment described and shown in the drawings can be combined and arranged in different configurations, and these changes in combination and arrangement should be regarded as a part of all embodiments of the innovative solutions of the present invention and be included in the scope to be protected by the present invention.

[0062] It should be noted that, without special instructions, in the description of the specific embodiments of the present invention, the expression terms indicating the orientation or positional relationship such as "upper", "lower", "left", "right", "center", "inner", "outer", etc. are all based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship when the invention product / device / device is usually used and placed. These terms of orientation or positional relationship are only for facilitating the description of the solutions of the present invention or simplifying the description in the specific embodiments, so as to enable those skilled in the art to quickly understand the solutions, rather than indicating or implying that a specific device / component / element must have a specific orientation or be constructed and operated in a specific positional relationship. Therefore, it cannot be understood as a limitation on the present invention.

[0063] In addition, when terms such as "horizontal", "vertical", "hanging", etc. appear, it does not mean that the corresponding device / component / element is required to be absolutely horizontal, vertical or hanging, but it can be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and it does not mean that the structure must be completely horizontal, but it can be slightly inclined. Or, it can be simply understood that the corresponding device / component / element is arranged in a specific direction such as "horizontal", "vertical", "hanging", etc., and can have an error / deviation of ±10% relative to the corresponding direction setting, more preferably an error / deviation within ±8%, more preferably an error / deviation within ±6%, more preferably an error / deviation within ±5%, more preferably an error / deviation within ±4%. As long as the corresponding device / component / element is within the error / deviation range, it can still achieve its function in the solution of the present invention.

[0064] In addition, when expressions such as "first", "second", "third", etc. appear in the terms, they are only used to distinguish the description of the same or similar components, and should not be understood as emphasizing or implying the relative importance of specific components.

[0065] In addition, in the description of the embodiments of the present invention, "several", "multiple", "a number of" represent at least 2. It can be any situation such as 3, 4, 5, 6, 7, 8, 9, etc., and even can be a situation exceeding 9.

[0066] In addition, in the description of the technical solution of the present invention, unless otherwise clearly specified / defined / restricted, when terms such as "set", "installed", "connected", "linked" appear, they should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection. It can be connection means commonly used in the art such as welding, riveting, bolting, threaded connection, etc. This kind of connection can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components.

[0067] Embodiment 1

[0068] In this embodiment, the ammonia synthesis system is as Figure 1As shown in the figure, the production process flow is as follows: raw material hydrogen and nitrogen are mixed in a certain proportion in the raw material inlet unit, compressed by a multi-stage compressor when passing through the multi-stage compression unit, and heat-exchanged with the outlet reaction stream after compression; before entering the ammonia synthesis reactor in the ammonia synthesis reaction unit, the mixed gas flow is divided into a cold jet stream and two cold extraction streams, and then enters the ammonia synthesis reactor. After a series of material and energy conversions, the generated ammonia is cooled in the cooling unit, and then enters the flash separation unit for ammonia separation. The liquid-phase stream is then stored or sent to the downstream refining unit. Part of the gas-phase stream is recycled to the inlet of the ammonia synthesis reactor for heat exchange, and the other part is vented. When using renewable energy to drive the ammonia synthesis system, the renewable energy supply is random, intermittent, and volatile, and the ammonia synthesis system needs to be frequently adjusted in load to adapt to the production raw materials. However, at present, when switching the production load of the ammonia synthesis system, it basically relies on manual experience, resulting in a large labor intensity, relatively single control parameters and poor accuracy, and being unable to meet the economic requirements of the system under frequent multi-condition load adjustments.

[0069] To solve the above problems, this embodiment provides a PINN-based ammonia production operation control method, as Figure 2 follows:

[0070] S1. Model different units of the ammonia synthesis system, including establishing a raw material feed mixing model, a multi-stage compression process model, an inlet and outlet heat exchange model, an inter-stage heat exchange model, and a cooling process model, and establishing an ammonia synthesis neural network model for the reaction process of the ammonia synthesis reactor based on PINN; establish a HDMR model for the ammonia separation process using a high-dimensional response surface model for the flash ammonia separation process;

[0071] First, model different operating units of the ammonia synthesis system. In the ammonia synthesis system of this embodiment, hydrogen is generated by alkaline water electrolysis, nitrogen is sent by an air separation unit, and hydrogen and nitrogen are mixed in a certain proportion. Based on this, the exergy conservation equation for the raw material gas mixing and feeding process is established as follows:

[0072]

[0073] In the formula, m mix is the total mass flow rate of the mixed gas, F i is the molar flow rate of the i-th component; Mw i is the molecular weight of the i-th component, T mix is the temperature of the mixed gas stream; m i is the mass flow rate of the i-th component, T i is the temperature of the i-th component.

[0074] Ammonia synthesis is carried out under high-pressure conditions of 15 MPa to 30 MPa, under which the activity of the catalyst can be maintained. Therefore, a multistage compressor is installed in the ammonia synthesis system to boost the raw material gas flow before the ammonia synthesis reactor. Usually, the same compression ratio is set for each compression stage, and an interstage cooler is used to remove the heat generated during the compression process of the previous stage. In this way, the energy consumption of the compression process can be reduced. The molar ratio r of hydrogen to nitrogen is an important operating parameter for ammonia synthesis, especially in the case of green ammonia production. In this study, the molar ratio of hydrogen to nitrogen was set as the process control variable to be optimized. Assume that the compression of the raw material gas is divided into C stages. The inlet of the first-stage compressor is calculated by the compression equation. For the last-stage compressor, its inlet includes, in addition to the outlet of the previous stage, a recycle gas stream with unreacted hydrogen and nitrogen. The compression equation is as follows:

[0075]

[0076] where is the flow rate of the compression input, is the flow rate of hydrogen, is the flow rate output from the previous-stage compressor, F recycle is the recycle flow rate, r is the molar ratio of hydrogen to nitrogen, l is the label of the compressor, l = 1 represents the first-stage compressor, L is the number of compressors, and l = L represents the last stage of the compressor. The energy balance of the compressor determines the energy consumption of the compression process and the required cooling load. Based on this, the exergy conservation of the multistage compression process model is calculated by the following formula:

[0077] Power = f(P in , P out , η, γ)

[0078] T out = f(T in , P in , P out )

[0079] Q = f(Cp, T in , T out , F)

[0080] In the formula, Power is the power consumption of the compression process, T and P represent the temperature and pressure of the stream respectively, Q is the heat, Cp is the isobaric heat capacity of the stream, and F is the flow rate of the stream. More specifically, Power is used to calculate the power requirement of the compressor to compress the gas stream with a volume flow rate of V in from the pressure P in to the pressure P out . The specific calculation formula of Power is:

[0081]

[0082] where T out is used to estimate the temperature of the outlet stream, T out The specific calculation formula is:

[0083]

[0084] And Q is used to calculate the required cooling load. The specific calculation formula for Q is:

[0085]

[0086] Where is the compressor efficiency, γ is the polytropic exponent. Mw and Cp are the molecular weight and specific heat capacity of the gas mixture respectively. The mass balance at the inlet and outlet of each compressor, the connection between the intermediate-stage compressors, and the component mass balance are omitted here.

[0087] The mixed gas stream after multi-stage compression is mixed with the recycle gas stream with unreacted hydrogen and nitrogen at the last-stage compressor. For this process, an inlet and outlet heat exchange model is established; the inter-stage cooler between the catalyst beds exchanges heat with the gas stream, and an inter-stage heat exchange model is established for this. The inter-stage cooler between the catalyst beds can be equivalent to a heat exchanger. Therefore, the outlet heat exchange model and the inter-stage heat exchange model adopt the same exergy conservation equation. The exergy conservation equation is as follows:

[0088] T out = f(T t in , T s in , ε)

[0089] Where ε is the efficiency parameter reflecting the heat exchanger configuration, which is determined by the heat exchanger configuration. T s in is the shell-side feed temperature, and T t in is the tube-side feed temperature.

[0090] The ammonia synthesis reactor is the core of the ammonia synthesis system, and the catalyst bed is the heart of the ammonia synthesis reactor. The ammonia synthesis reactor used in this embodiment consists of three catalyst beds, two inter-stage coolers (RHX, also known as intermediate coolers), and a mixer, all of which are installed in a pressure shell. Its schematic diagram is as Figure 3As shown. Before entering the reactor, the feed mixed gas stream is divided into three sub-streams: two quench streams and one cold surge stream. One of the quench streams is directly sent to the second intercooler (RHX = 2) between the second catalyst bed and the third catalyst bed for heat exchange, while the other quench stream first flows along the shell of the ammonia synthesis reactor to cool the shell and then enters the first intercooler (RHX = 1) between the first catalyst bed and the second catalyst bed. Heat exchange from the catalyst beds and the outlet product stream to these streams is ignored during the simulation. After heat exchange with the reaction product, the two quench streams converge with the cold surge stream in a mixer, and then the mixture enters the first catalyst bed. The flow rate control of the cold surge stream and the quench streams is crucial for ensuring the temperature control of the system. The mass balance equations for the fluid split are as follows:

[0091]

[0092] where is the total inflow mass flow rate of component i, and F k,i is the mass flow rate of component i in the reactor inlet split stream k.

[0093] The ammonia synthesis reaction is an exothermic reaction that occurs on the catalyst surface. During this process, hydrogen and nitrogen are consumed to produce ammonia. The mathematical modeling of the ammonia synthesis reactor includes three parts: the ammonia synthesis neural network model of the catalyst bed, the intercooling heat exchange model of the intercooler, and the mixer model.

[0094] Based on the kinetics and thermodynamics mechanisms of the ammonia synthesis catalytic reaction process and the physical and chemical characteristics of the catalyst bed in the ammonia synthesis reactor, a nonlinear partial differential mechanism equation is established. The nonlinear partial differential mechanism equation is coupled with the physics-informed neural network PINN. After training to meet the accuracy requirements, the ammonia synthesis neural network model is obtained. More specifically, as Figure 4 , the ammonia synthesis neural network model is trained until the accuracy meets the requirements. The training uses experimental / engineering data to establish the ammonia synthesis neural network model. The sobol sampling technique is used to collect sample data X in the target data space. The sample data X is input into the linear partial differential mechanism equation of the ammonia synthesis neural network model to obtain the target value Y. The sample data X→Y is divided into a training set and a test set, which are used to train and test the neural network model respectively. If the accuracy does not meet the requirements, the data volume is further increased and training continues until the accuracy meets the requirements. Industrial data and mechanism data are integrated for model training and testing, and the model prediction accuracy is higher than 95%.

[0095] The modeling of the catalytic reaction process in the catalyst bed is achieved using PINN, which incorporates the physical laws described by differential equations into its loss function to guide the learning process to obtain solutions that are more in line with the basic physical laws. The input variables include the molar flow rate, molar composition fraction, bed temperature, bed pressure, and catalyst bed volume of the inlet stream of the catalyst bed. The output variable Y includes the hydrogen conversion rate and the outlet temperature of the catalyst bed stream. The nonlinear partial differential mechanism equation is embedded in the training process of the neural network together with experimental / engineering data, enabling the model to not only rely on data but also be constrained by physical laws. During the model training process, the loss function of the neural network is optimized with the residuals of the ammonia synthesis reaction kinetics and thermodynamics partial differential equations to ensure accurate prediction of the target ammonia synthesis catalyst bed on the premise that the model satisfies the physical and chemical laws. During the catalyst bed modeling process, a radial flow catalyst bed is considered and assumed to be under isobaric and adiabatic conditions. The nonlinear partial differential mechanism equation in the catalyst bed is as follows:

[0096]

[0097]

[0098] where i represents the reactant, b represents the number of catalyst beds, v is the stoichiometric coefficient, X is the conversion rate of the corresponding reactant, V is the volume of the catalyst bed, is the reaction rate, is the molar flow rate of reactant i entering catalyst bed b; T represents the temperature of the reaction mixture, ΔH is the heat of reaction, m b represents the total mass flow rate of the reaction mixture, and Cp represents the specific heat capacity of the reaction mixture. Existing conventional methods are used for the calculation of the reaction rate, heat of reaction, and specific heat capacity of the reaction mixture. The calculation formula of is:

[0099]

[0100] where k 2 is the reverse reaction constant, K is the reaction equilibrium constant, is the fugacity of nitrogen, and α is a constant, which is taken as 0.5 in this embodiment.

[0101] During the training process, the loss function of PINN consists of data loss and physical loss. The data loss is based on actual engineering data, and the physical loss is based on the nonlinear partial differential mechanism equation, with the requirement that the residual < 10%.

[0102] The inter-stage cooler between catalyst beds can be equivalent to a heat exchanger, where the hot fluid flows in the shell side and the cold fluid flows in the tube side. The heat transfer process occurs in a combined way of cross-flow and counter-flow. Assuming the heat transfer process is adiabatic, no phase change occurs in the fluid, and the temperature change of the fluid only occurs axially. For the convenience of calculation, the thermophysical properties of the gas are considered constant. In addition, it is also assumed that no chemical reaction occurs outside the catalyst bed. The exergy conservation equation of the inter-stage heat transfer model involves the efficiency parameter ε. For the inter-stage cooler, the outlet temperature of the tube side is determined by the following equation:

[0103]

[0104] where is the outlet temperature of the tube side, is the temperature of the reactor inlet stream. The subscript k represents the split at the inlet of the ammonia synthesis reactor, and RHX represents the inter-stage cooler after the corresponding catalyst bed. The constant ε reflects the effectiveness of the inter-stage cooler, which is independent of the change in fluid temperature and is determined by the structure of the inter-stage cooler. The calculation formula of ε RHX is as follows:

[0105]

[0106] In the formula is the specific heat capacity, and NTU RHX represents the number of heat transfer units.

[0107]

[0108]

[0109] In the formula and are the mass flow rates of the cold fluid and the hot fluid respectively, Cp cold and CP hot are their specific heat capacities respectively; U is the overall heat transfer coefficient of the heat exchanger, and A RHX is the heat transfer area of the heat exchanger.

[0110] In the ammonia synthesis reactor, the energy balance equation of the inter-stage cooler is as follows:

[0111]

[0112] In the formula, RHX represents the inter-stage cooler.

[0113] According to the energy balance formula of the inter-stage cooler, the outlet temperature of the shell side can be derived from it.

[0114] It is assumed that the mixing of hydrogen and nitrogen in the mixer is ideal and instantaneous. In the ammonia synthesis reactor, isobaric and adiabatic conditions are assumed. Therefore, pressure drop and heat of mixing are neglected. Since no conversion of substances occurs and the flow rate and composition remain unchanged, the mass balance is omitted here. The energy balance equation of the mixer is as follows:

[0115]

[0116] In the practice of ammonia synthesis production, in the cooling unit, the generated ammonia is usually separated from the reaction products (a mixture of ammonia, hydrogen, and nitrogen) by refrigeration condensation. The ammonia content in the gas phase of the reaction products after cooling plays a key role in determining the production capacity of the ammonia synthesis reactor. A key strategy to increase ammonia production is to reduce the ammonia content in the inlet gas of the ammonia synthesis reactor, which makes the ammonia concentration at the inlet of the ammonia synthesis reactor an important operating parameter that needs to be controlled. The ammonia content at the inlet of the ammonia synthesis reactor is determined by the gas-phase outlet of the flash separation unit, and the latter depends on the outlet temperature and pressure of the ammonia refrigeration system. This indirect relationship is represented by the Larson-Black empirical formula as follows:

[0117]

[0118] where, represents the ammonia concentration in the gas-phase stream, and P (atm) and T (K) represent the operating pressure and temperature of the flash separation unit respectively. Here, the pressure is determined by the multistage compression system, and the temperature is determined by the ammonia refrigeration system. The energy consumption of the ammonia refrigeration system can be calculated by the following equation:

[0119]

[0120] Thus, a cooling process model is established. It should be noted that gas compression and ammonia refrigeration are considered to be the two most energy-consuming links in the ammonia production system. They lay the foundation for subsequent ammonia separation and determine the ammonia production.

[0121] Flash separation is a key link in the ammonia synthesis cycle, used to separate ammonia from the unreacted reactants in the reaction products. The ammonia product stream separated by the flash separation unit is then sent to the downstream refining unit or storage, while the unreacted reactants are recycled back to the ammonia synthesis reactor. The operating temperature and pressure of flash separation are of great significance for the synthesis cycle in terms of energy consumption and ammonia production. Establishing a high-fidelity and low-cost computational model of the flash separation unit is crucial for system optimization. Given that the principle-based flash separation process model involves iterative calculations and verification of gas-liquid equilibrium at specific temperatures and pressures, which will pose significant obstacles to the subsequent mathematical optimization process, alternative modeling techniques are adopted to study the flash separation mechanism. As Figure 5As shown, the cooled reaction product is fed into a two-stage flash separation unit. The gas-phase outlet stream mainly containing unreacted hydrogen and nitrogen is recycled back to the ammonia synthesis reactor to improve system efficiency, while the liquid-phase stream containing product ammonia is sent to a storage tank or further refined. The modeling of flash separation is completed in two steps. First, data sampling is carried out based on the physical and chemical mechanisms of the ammonia separation process. A high-fidelity ammonia flash separation simulation model is developed using a third-party authoritative software package, which is Aspen Plus in this study. The key operating variables (including the inlet molar flow rates, temperature, and pressure of reaction stream components H 2 , N 2 , NH 3 , Ar) are set as input variables, and the process variables (covering the molar flow rates of components H 2 , N 2 , NH 3 , Ar in the gas-phase and liquid-phase outlet streams of the second-stage separation stage) are selected as output variables (state variables) for surrogate model development. The possible operating ranges of these six input variables are determined based on system analysis and engineering experience to form a six-dimensional cubic mathematical space. To improve model fidelity, as Figure 6 shown, the Sobol sequence is applied to generate sufficient X points in the predetermined hexagonal cubic space because of its ease of use and good space-filling ability. Then the sampled X data set is used as the input to the simulation model to obtain the corresponding Y values. Next is the second step of surrogate modeling. The sampled data [X→Y] is divided into two groups in a certain proportion (usually 8:2), namely the training set and the test set. The training set is used for model development, and the test set is used for verification. Using HDMR, an iterative procedure is applied to find the most suitable fitting order of the polynomial. The information of the gas-phase outlet stream in the first-stage flash separation stage is obtained through mass balance:

[0122]

[0123] where is the flow rate of component i in the gas-phase outlet stream after the first-stage flash separation, is the flow rate of component i in the feed entering the first-stage flash separation, is the flow rate of component i in the gas-phase outlet stream after the second-stage flash separation, is the flow rate of component i in the liquid-phase outlet stream after the second-stage flash separation. According to the above method, the information of the gas-phase outlet stream in the second-stage flash separation stage is obtained.

[0124] Step S2: Based on the structure of the ammonia synthesis reactor, couple the established ammonia synthesis neural network model with the inter-stage heat exchange model to establish an ammonia synthesis simulation model; couple the raw material feed mixing model, multi-stage compression process model, inlet and outlet heat exchange model, ammonia synthesis simulation model, cooling process model, and ammonia separation process HDMR model according to the connection relationship of the ammonia synthesis system to form an ammonia synthesis recycle loop.

[0125] The simulation of the ammonia synthesis recycle loop is to couple and connect the models established for different units of the ammonia synthesis system. According to the connection relationship of the ammonia synthesis system, successively couple the raw material feed mixing model, multi-stage compression process model, inlet and outlet heat exchange model, ammonia synthesis simulation model, cooling process model, and ammonia separation process HDMR model, and recycle the flash separation gas phase outlet stream back to the inlet of the last-stage compressor of the multi-stage compression process. Process stream connection equations are added between different models to form a system of equations for the ammonia synthesis recycle loop, so as to realize the simulation calculation of the ammonia synthesis loop under the subsequent iterative calculation strategy framework. The process stream connection equations connect the systems of different operating units of the ammonia synthesis system, and the output parameters of the previous unit are the input parameters of the connected unit, such as temperature, pressure, and flow rate.

[0126] Step S3: Select the optimization variables affected by variable load operation, perform simulation based on the ammonia synthesis recycle loop under the production load, use the levelized ammonia cost as the objective function, and adopt an optimization algorithm to solve the optimization variables; regulate the ammonia synthesis system according to the optimization variables under the production load.

[0127] In this step, it is necessary to perform multi-steady-state variable load integrated optimization of the ammonia synthesis recycle loop. Taking the levelized ammonia cost of the overall system economy as the objective function and the system process description equations as the constraints, establish a nonlinear mathematical programming model, solve and calculate the variable load operation optimization of the ammonia synthesis loop to achieve system energy consumption optimization and support load adjustment control decisions.

[0128] According to the structure and connection of the ammonia synthesis system, determine the optimization variables affected under variable load operation. The optimization variables include: hydrogen-nitrogen ratio at the reactor inlet, inert gas content, multi-stage compressor outlet pressure, recycle material flow rate, vent material flow rate, terminal temperature of the ammonia cooling process, etc.

[0129] There are various optimization algorithms for parameter optimization, such as genetic algorithm, particle swarm optimization algorithm, gradient descent method, Bayesian optimization algorithm, etc.

[0130] During the optimization solution process, the constraints include: energy conservation equation for the compression process, energy conservation equation for the heat exchange process, mass conservation equation for the reaction process, energy conservation equation for the reaction process, energy conservation equation for the separation process, mass conservation equation for the separation process, and process stream connection equation.

[0131] In this work, the levelized cost of ammonia (LCOA) is adopted as the objective function of the model. LCOA is calculated by the annual average cost method and is used to evaluate the operating costs of different operation strategies, aiming to identify potential improvement measures to reduce costs. The levelized cost of ammonia includes capital expenditure, operating expenditure, by-product revenue, and total ammonia production. The calculation formula for the levelized cost of ammonia is as follows:

[0132]

[0133] where LCOA represents the levelized cost of ammonia of the product, CAPEX j and OPEX j represent the capital cost and operating cost of unit operation j, respectively, covering photovoltaic (PV), wind turbine (WT), electrolyzer (ele), storage tank (S), compressor (cps), heat exchanger (HEX), reactor (R), and separator (sps). is the molecular weight of ammonia, is the oxygen product flow rate of the electrolyzer, is the ammonia product flow rate at the outlet of the secondary flash separation. ΔHr is the annual operating time, ACCR is the depreciation rate, and ir and ny are the interest rate and plant life, respectively. In the calculation, the operating expenditure, by-product revenue, and ammonia production are defined on an annual basis, while the capital expenditure involves the total investment cost of equipment and catalysts during the project life cycle. Therefore, the depreciation rate is calculated according to the interest rate and the project life cycle.

[0134] Example 2

[0135] This example provides a method for regulating ammonia production operations based on PINN. Using the scheme of Example 1, taking a 100,000-ton-per-year ammonia synthesis plant in the southwest as an example, the optimization algorithm uses the CONOPT4 algorithm in GAMS software for optimization. CONOPT4 is an efficient solver for solving nonlinear programming problems and belongs to a local solver, mainly used to find local optimal solutions and is suitable for large-scale nonlinear models.

[0136] Through simulation calculations, the hydrogen-nitrogen ratio, inert gas content, final outlet pressure of multistage compression, and ammonia separation temperature required to maintain the normal operation of the reactor while minimizing production costs during different production load switches are obtained, as shown in Figure 6 , Figure 7 . During the production process, when a load change is required, the optimal optimization variables at different loads are determined according to the present invention and are used to regulate the ammonia synthesis system.

[0137] Example 3

[0138] This example provides a system for regulating ammonia production operations based on PINN to implement the method of Example 1. The system includes:

[0139] A modeling module for modeling different units of an ammonia synthesis system, including establishing a raw material feed mixing model, a multi-stage compression process model, an inlet and outlet heat exchange model, an inter-stage heat exchange model, and a cooling process model, and establishing an ammonia synthesis neural network model for the reaction process of the ammonia synthesis reactor based on PINN; using a high-dimensional response surface model to establish a HDMR model for the ammonia separation process in the flash ammonia separation process;

[0140] An ammonia synthesis recycle loop module for coupling the ammonia synthesis neural network model with the inter-stage heat exchange model based on the structure of the ammonia synthesis reactor to establish an ammonia synthesis simulation model; coupling the raw material feed mixing model, the multi-stage compression process model, the inlet and outlet heat exchange model, the ammonia synthesis simulation model, the cooling process model, and the HDMR model for the ammonia separation process according to the connection relationship of the ammonia synthesis system to form an ammonia synthesis recycle loop;

[0141] An optimization calculation module for selecting optimization variables affecting variable load operation, simulating based on the ammonia synthesis recycle loop under the target production load, using the ammonia cost per unit of production as the objective function, and using an optimization algorithm to solve for the optimization variables;

[0142] A regulation module for regulating the ammonia synthesis system according to the optimization variables obtained under the target production load.

[0143] The system or module etc. illustrated in the above embodiments can be specifically implemented by a computer chip or entity, or by a product with a certain function. For the convenience of description, when describing the above device, various modules are described separately according to their functions. Of course, when implementing the present application, the functions of each module can be implemented in the same or multiple software and / or hardware, or the modules implementing the same function can be realized by a combination of multiple sub-modules, etc.

[0144] This embodiment also provides an electronic device, including at least one processor, a memory communicatively connected to at least one processor, and at least one input / output interface communicatively connected to at least one processor; the memory stores instructions executable by at least one processor, and the instructions are executed by at least one processor to enable at least one processor to execute a method for regulating ammonia production operations based on PINN in the foregoing embodiments. The input / output interface may include a display, a keyboard, a mouse, and a USB interface for inputting and outputting data.

[0145] The electronic device may be an electronic device for a client, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute a method for regulating ammonia production operations based on PINN in Embodiment 1.

[0146] Those skilled in the art can understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including the above method embodiments; and the foregoing storage medium includes: various media such as removable storage devices, read-only memory (ROM), magnetic disks, or optical discs that can store program codes.

[0147] When the above integrated unit of the present invention is implemented in the form of a software functional unit and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods of the various embodiments of the present invention. And the foregoing storage medium includes: various media such as removable storage devices, ROM, magnetic disks, or optical discs that can store program codes.

[0148] For those skilled in the art, when understanding the solutions described in the specific embodiments of the present invention, they can refer to the conventional technical manuals in the art. At the same time, for the places where the above terms appear, they can make appropriate understandings or adjustments referentially. Without creative efforts, the same or similar technical solution implementation situations can be deduced.

[0149] The above embodiments only describe the basic principles, main features, and / or advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and the description in the invention content part of the specification are only the principles or specific cases of the present invention. Without departing from the essence of the innovative idea of the present invention, there are various changes and improvements to the innovative solutions of the present invention, and these changes and improvements all fall within the scope of protection required by the present invention.

Claims

1. A PINN-based ammonia production operation control method, characterized in that: The following steps are involved: Modeling of different units of the ammonia synthesis system, including the establishment of raw material feed mixing model, multi-stage compression process model, inlet and outlet heat exchange model, inter-stage heat exchange model, cooling process model, and the establishment of an ammonia synthesis neural network model based on PINN for the reaction process of the ammonia synthesis reactor; and the establishment of an HDMR model of the ammonia separation process using a high-dimensional response surface model for the flash ammonia separation process; Based on the structure of the ammonia synthesis reactor, the ammonia synthesis neural network model is coupled with the inter-stage heat exchange model to establish an ammonia synthesis simulation model; according to the connection relationship of the ammonia synthesis system, the raw material feed mixing model, the multi-stage compression process model, the inlet and outlet heat exchange model, the ammonia synthesis simulation model, the cooling process model and the HDMR model of the ammonia separation process are coupled to form an ammonia synthesis cycle; The optimization variables that affect the variable load operation are selected, and a simulation is performed based on the ammonia synthesis cycle at the target production load. The levelized ammonia cost is used as the objective function, and an optimization algorithm is used to solve the optimization variables; the ammonia synthesis system is regulated based on the optimized variables at the target production load obtained by solving the optimization variables.

2. The PINN-based ammonia production operation control method according to claim 1, characterized in that: The energy and mass conservation equations of the multi-stage compression process model are as follows: T out =f(T in ,P in ,P out ) Q=f(Cp,T in ,T out ,F) Where Power is the power consumption of the compression process, T and P represent the stream temperature and pressure respectively. is the compressor efficiency, γ is the polytropic index, Q is the heat, Cp is the isobaric specific heat capacity of the stream, and F is the stream flow rate.

3. The PINN-based ammonia production operation control method according to claim 1, characterized in that: The energy and mass conservation equations of the outlet heat exchange model or the inter-stage heat exchange model are as follows: T out =f(T t in ,T s in ,ε) Where T out is the outlet flow temperature, ε is the efficiency parameter of the heat exchanger configuration, which is determined by the heat exchanger configuration, T s in is the shell side feed temperature, T t in is the tube feed temperature.

4. The PINN-based ammonia production operation control method according to claim 1, characterized in that: The specific implementation method of establishing the ammonia synthesis neural network model is: According to the kinetic and thermodynamic mechanism of the ammonia synthesis catalytic reaction process and the physicochemical characteristics of the catalyst bed in the ammonia synthesis reactor, a nonlinear partial differential mechanism equation is established. The nonlinear partial differential mechanism equation is coupled with the physical information neural network PINN. After training to meet the accuracy requirements, the ammonia synthesis neural network model is obtained.

5. The PINN-based ammonia production operation control method according to claim 4, characterized in that: The nonlinear partial differential mechanism equation is as follows: Where i represents the reactant, b represents the number of catalyst layers, ν is the stoichiometric number, X is the conversion rate of the corresponding reactant, and V is the volume of the catalyst bed. is the reaction rate, is the molar flow rate of reactant i entering catalyst bed b; T represents the temperature of the reaction mixture, ΔH is the heat of reaction, and m b represents the total mass flow rate of the reaction mixture, F b,i is the mass flow rate of component i in b, Mw i is the molecular weight of component i, and Cp represents the specific heat capacity of the reaction mixture.

6. The PINN-based ammonia production operation control method according to claim 4, characterized in that: When training the ammonia synthesis neural network model, a sobol sampling technique is used to collect multiple input variables in the target data space to form sample data X, and the sample data X is input into the ammonia synthesis neural network model to obtain output variables to form a target value Y. The sample data X→Y is divided into a training set and a test set, which are used to train and test the ammonia synthesis neural network model respectively. During the training process, the loss function of the neural network and the residual of the nonlinear partial differential mechanism equation are optimized until the accuracy meets the requirements. The input variables include the catalyst bed inlet logistics molar flow rate, molar composition fraction, bed temperature, bed pressure, and catalyst bed volume. The output variables include hydrogen conversion rate and catalyst bed logistics outlet temperature.

7. The PINN-based ammonia production operation control method according to any one of claims 1 to 6, characterized in that: The optimization variables include: reactor inlet hydrogen-nitrogen ratio, inert gas content, multi-stage compressor outlet pressure, circulating material flow rate, venting material flow rate, and ammonia cooling process terminal temperature.

8. The PINN-based ammonia production operation control method according to any one of claims 1 to 6, characterized in that: The constraints include: energy conservation equation for compression process, energy conservation equation for heat exchange process, mass conservation equation for reaction process, energy conservation equation for reaction process, energy conservation equation for separation process, mass conservation equation for separation process, and process stream connection equation.

9. The PINN-based ammonia production operation control method according to any one of claims 1 to 6, characterized in that: The levelized ammonia cost includes capital expenditures, operating expenditures, by-product revenues and total ammonia production.

10. An ammonia production operation control system based on PINN, characterized in that: The system comprises: Modeling module, used to model different units of the ammonia synthesis system, including the establishment of raw material feed mixing model, multi-stage compression process model, inlet and outlet heat exchange model, inter-stage heat exchange model, cooling process model, and the establishment of an ammonia synthesis neural network model based on PINN for the reaction process of the ammonia synthesis reactor; and the establishment of an HDMR model of the ammonia separation process using a high-dimensional response surface model for the flash ammonia separation process; An ammonia synthesis cycle module is used to couple the ammonia synthesis neural network model with the interstage heat exchange model based on the ammonia synthesis reactor structure to establish an ammonia synthesis simulation model; according to the connection relationship of the ammonia synthesis system, the raw material feed mixing model, the multi-stage compression process model, the inlet and outlet heat exchange model, the ammonia synthesis simulation model, the cooling process model and the HDMR model of the ammonia separation process are coupled to form an ammonia synthesis cycle; An optimization calculation module is used to select optimization variables that affect variable load operation, simulate based on the ammonia synthesis cycle under the target production load, take the levelized ammonia cost as the objective function, and use an optimization algorithm to solve the optimization variables; The control module is used to control the ammonia synthesis system according to the optimized variables under the target production load obtained by solving.

Images