Ammonia cracking hydrogen production integrated reaction device

Abstract

The invention provides an ammonia cracking hydrogen production integrated reaction device and a control method thereof, and aims to realize real-time monitoring and automatic regeneration treatment of a catalyst state. The device comprises a reactor, a catalyst bed, a heating system and a spectrum online detection module, and the poisoning depth and the recommended heating temperature of the catalyst are accurately predicted by collecting Raman spectrum data and utilizing deep learning model analysis. The system can automatically adjust heating and gas flow, realizes online regeneration of the catalyst, improves the efficiency of hydrogen production by ammonia cracking and prolongs the service life of the catalyst. According to the scheme, the ammonia cracking hydrogen production process is optimized through real-time monitoring, accurate control and automatic regeneration, and the method has the advantages of being high in automation degree, good in safety, environmentally friendly and the like and has great significance in promoting development of the hydrogen energy technology.

Description

Technical Field

[0001] The invention relates to the field of hydrogen production by cracking ammonia, and in particular to an integrated reaction device for hydrogen production by cracking ammonia. Background Art

[0002] As the global demand for clean energy grows, hydrogen has attracted widespread attention as an efficient and clean energy carrier. Ammonia cracking hydrogen production technology has become a research hotspot due to its high efficiency and safety. At present, ammonia cracking hydrogen production technology mainly relies on the use of high-efficiency catalysts, among which ruthenium-based, nickel-based and iron-based catalysts are widely studied due to their excellent cracking performance.

[0003] However, the catalyst is susceptible to poisoning during long-term use, resulting in a decrease in activity, which affects the efficiency of the entire hydrogen production process. Therefore, how to monitor the state of the catalyst in real time and regenerate it in time has become one of the key technologies to improve the efficiency of ammonia cracking hydrogen production. Existing technologies mainly rely on offline analysis and regular maintenance. These methods are not only time-consuming and labor-intensive, but also difficult to achieve real-time monitoring and precise control. To address this problem, this scheme proposes an integrated reaction device for ammonia cracking hydrogen production, which integrates an online spectral detection module and an intelligent control module, which can monitor the poisoning state of the catalyst in real time and automatically adjust the operating conditions to achieve in-situ regeneration of the catalyst.

[0004] Traditional methods require that the catalyst be removed from the reactor and then tested offline using laboratory-scale analytical techniques. This method not only interrupts the hydrogen production process, but also fails to provide real-time data, making it impossible to monitor and adjust the catalyst status in a timely manner. Although Raman spectroscopy is currently considered a potential online monitoring tool because it can provide molecular vibration information, in actual applications, the Raman spectroscopy signal itself is weak, and the use of mesh catalysts further weakens the signal, making it difficult to directly use conventional characteristic peak methods for analysis. Therefore, the development of an online monitoring technology that can overcome the effects of weak signals and catalyst morphology is of great significance for improving the efficiency of the ammonia cracking hydrogen production process and the accuracy of catalyst management. Summary of the invention

[0005] In order to solve the above-mentioned problem, the present invention provides an integrated reaction device for producing hydrogen by cracking ammonia, comprising a reactor, a catalyst bed (4), a heating system (2) and a spectrum online detection module (3).

[0006] The reactor comprises a reactor shell (1), an ammonia inlet (6), a heater inlet, an optical inlet and a mixed gas outlet; a catalyst bed (4) is arranged in the reactor shell, and a three-dimensional mesh catalyst is arranged in the catalyst bed; the catalyst is used to crack the ammonia into hydrogen and nitrogen under heating; the reaction equation is:

[0007] 2NH3→3H2+N2;

[0008] The catalyst is one or more of a ruthenium-based catalyst, a nickel-based catalyst or an iron-based catalyst;

[0009] The top of the reactor shell is connected to a mixed gas outlet (7), and the bottom of the reactor shell is connected to a waste outlet (5);

[0010] The bottom of the side of the reactor shell is connected to an ammonia inlet (6), which is connected to the inside of the reactor shell and then to an ammonia disperser for dispersing ammonia into the catalyst bed;

[0011] A heating system (2) and a spectrum online detection module (3) are arranged outside the reactor shell. The heating system (2) enters the reactor shell through a heater inlet and heats the catalyst bed. The spectrum online detection module (3) enters the catalyst bed through an optical inlet. The spectrum online detection module (3) emits laser light toward the catalyst bed and collects spectrum data to analyze the catalyst poisoning condition.

[0012] The reaction device also includes a main control computer, which obtains data collected by the spectrum online detection module and analyzes it using the spectrum analysis module to obtain the catalyst poisoning situation;

[0013] The main control machine controls the heating module to heat and regenerate the catalyst according to the poisoning condition of the catalyst.

[0014] The main control machine is connected to the spectrum analysis module, the spectrum online detection module, the catalytic heating control module, the regeneration heating control module and the gas flow control module;

[0015] The spectrum online detection module includes a light emitter and a light collector;

[0016] The catalytic heating control module is used to control the catalyst temperature during catalytic cracking of ammonia, the regeneration heating control module is used to adjust the heating temperature of the catalyst bed during catalyst heating regeneration, and the gas flow control module is used to control the ammonia inlet flow and the mixed gas outlet flow.

[0017] The catalyst bed is cylindrical, the inner surface of which is a smooth light reflecting surface, and the catalyst is arranged inside through a three-dimensional grid structure; a gas outlet (42) is arranged on the top of the catalyst bed, a gas inlet (41) is arranged on the bottom of the catalyst bed, and a light emitter (32) and a light collector (31) are installed on the side of the catalyst bed, the light emitter is used to emit a detection laser toward the catalyst inside the catalyst bed, and the light collector is used to collect Raman scattered light generated by the detection laser on the surface of the catalyst; the spectral online detection module sends the Raman spectrum data collected by the light collector to the main control computer.

[0018] The main control computer sends the Raman spectrum data to the spectrum analysis module, and the spectrum analysis module processes the Raman spectrum data collected by the spectrum online detection module to determine the poisoning of the catalyst. The processing includes:

[0019] a. Data preprocessing: Normalize the Raman spectral data to ensure that the data is on the same scale, use a smoothing algorithm to reduce noise, and make the spectral curve smoother;

[0020] b. Input model: The processed spectral features are directly input into the deep learning model as a whole to obtain the catalyst poisoning depth and recommended heating temperature.

[0021] The structure of the deep learning model is:

[0022] Input Layer:

[0023] The size of the input layer is determined according to the number of wavelength points of the spectral data; if the spectral data has n wavelength points, the size of the input layer is n;

[0024] First convolutional layer:

[0025] Number of convolution kernels: 32 convolution kernels are selected to capture different features;

[0026] Convolution kernel size: The convolution kernel size is selected as 5×1 to capture local features;

[0027] Activation function: Use ReLU activation function to increase nonlinearity;

[0028] First pooling layer:

[0029] Pooling type: Use maximum pooling to reduce the spatial size of features, and choose a pooling kernel of 2×1 size;

[0030] Second convolutional layer:

[0031] Number of convolution kernels: Increase the number of convolution kernels to 64 to further abstract features;

[0032] Convolution kernel size: Select the convolution kernel size as 3×1;

[0033] Activation function: Use ReLU activation function;

[0034] Second pooling layer:

[0035] Pooling type: Use maximum pooling to reduce the spatial size of features, and choose a pooling kernel of 2×1 size;

[0036] Flattened layer:

[0037] Flatten the multi-dimensional convolutional layer output into one dimension so that it can be input into the fully connected layer;

[0038] First fully connected layer

[0039] Number of neurons: 128 neurons were selected;

[0040] Activation function: Use ReLU activation function;

[0041] Second fully connected layer

[0042] Number of neurons: 64 neurons were selected;

[0043] Activation function: Use ReLU activation function;

[0044] Branch 1 is used to predict poisoning depth:

[0045] Branch 1 fully connected layer: connected from the shared bottom layer network for further processing of poisoning depth features;

[0046] Branch 1 output layer: a single neuron, using a linear activation function, outputs a continuous value of the poisoning depth;

[0047] Branch 2 is used to predict the heating temperature:

[0048] Branch 2 fully connected layer: connected from the shared bottom layer network for further processing of poisoning depth features;

[0049] Branch 2 output layer: a single neuron, using a linear activation function, outputs a continuous value of the poisoning depth;

[0050] Loss function: Combine two losses using a weighted sum:

[0051] L=λ1·L depth +λ2·L temperature ;

[0052] Where L depth is the loss of poisoning depth, L temperature is the loss of the recommended heating temperature, while λ1 and λ2 are weights used to balance the importance of the two tasks.

[0053] An integrated reaction control method for ammonia cracking and hydrogen production, using the integrated reaction device for ammonia cracking and hydrogen production, comprises the following steps:

[0054] Step 1: Data collection, start the spectrum online detection module (3), emit detection laser to the catalyst bed through the light emitter (32), collect Raman scattered light through the light collector (31), and send the collected Raman spectrum data to the main control computer;

[0055] Step 2: Data preprocessing: In the main control computer, the collected Raman spectrum data is normalized to ensure that the data is on the same scale, and a smoothing algorithm is used to reduce noise and make the spectrum curve smoother.

[0056] Step 3: Deep learning model analysis: input the preprocessed spectral data into the deep learning model, which analyzes the data and predicts the poisoning depth of the catalyst and the recommended heating temperature;

[0057] Step 4: Poisoning depth assessment: Based on the output of the deep learning model, the catalyst poisoning depth is assessed and the degree of poisoning is determined;

[0058] Step 5: Heating temperature adjustment: According to the heating temperature recommended by the deep learning model, the heating temperature of the catalyst bed is adjusted through the regeneration heating control module to promote the regeneration and recovery of the catalyst activity;

[0059] Step 6: Feedback control: during the catalyst heating regeneration process, continuously collect Raman spectrum data and repeat steps 2 to 5 to monitor the change of the catalyst state until the catalyst activity is restored to a predetermined level;

[0060] Step 7: Operation adjustment: According to the poisoning depth of the catalyst and the heating regeneration effect, the main control machine automatically adjusts the parameters of the catalytic heating control module and the gas flow control module to optimize the ammonia catalytic cracking process.

[0061] Before step 1, the method further includes the following steps:

[0062] Step 1: System startup, initializing the integrated reaction device for ammonia cracking and hydrogen production, including starting the gas supply system, heating system, spectrum online detection module and main control computer;

[0063] Step 2: System inspection: perform a comprehensive functional and safety check on the reactor to ensure that all sensors, actuators and control modules are working properly and there are no leaks or other potential safety issues;

[0064] Step 3: preheating and stabilization, starting the heating system (2) to preheat the catalyst bed (4) to achieve an optimal catalytic cracking temperature and keep the system stable;

[0065] Step 4: Initial gas flow setting: setting the initial ammonia flow rate through the gas flow control module to ensure that ammonia is input into the reactor at an appropriate rate;

[0066] Step 5: spectral system calibration, calibrating the light emitter (32) and light collector (31) in the spectral online detection module (3) to ensure the accuracy and repeatability of Raman spectral data;

[0067] Step 6: Data acquisition. After confirming that the system is stable and calibrated, start continuously collecting Raman spectral data of the catalyst bed for subsequent analysis and monitoring.

[0068] The beneficial effects of the present invention are:

[0069] This solution integrates the Raman spectroscopy online detection module to achieve real-time poisoning detection without removing the catalyst, avoiding the limitations of traditional offline detection methods. The spectral acquisition system designed specifically for mesh catalysts effectively captures the spectral information on the catalyst surface and can perform effective analysis even when the signal is weak. In order to address the problem of weak Raman spectroscopy signals, this solution directly inputs the Raman spectrum as a whole into the deep learning model to directly predict the poisoning situation and activation temperature, thereby improving detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0071] Attached Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0072] Attached Figure 2 It is a schematic structural diagram of the catalyst bed of the present invention;

[0073] Attached Figure 3 Schematic diagram of the architecture of the detection module of the present invention. DETAILED DESCRIPTION

[0074] Embodiment 1:

[0075] See also Figures 1 to 3 The present invention provides an integrated reaction device for producing hydrogen by cracking ammonia, comprising a reactor, a catalyst bed (4), a heating system (2) and a spectrum online detection module (3).

[0076] The reactor comprises a reactor shell (1), an ammonia inlet (6), a heater inlet, an optical inlet and a mixed gas outlet; a catalyst bed (4) is arranged in the reactor shell, and a three-dimensional mesh catalyst is arranged in the catalyst bed; the catalyst is used to crack the ammonia into hydrogen and nitrogen under heating; the reaction equation is:

[0077] 2NH3→3H2+N2;

[0078] The catalyst is one or more of a ruthenium-based catalyst, a nickel-based catalyst or an iron-based catalyst;

[0079] The top of the reactor shell is connected to a mixed gas outlet (7), and the bottom of the reactor shell is connected to a waste outlet (5);

[0080] The bottom of the side of the reactor shell is connected to an ammonia inlet (6), which is connected to the inside of the reactor shell and then to an ammonia disperser for dispersing ammonia into the catalyst bed;

[0081] A heating system (2) and a spectrum online detection module (3) are arranged outside the reactor shell. The heating system (2) enters the reactor shell through a heater inlet and heats the catalyst bed. The spectrum online detection module (3) enters the catalyst bed through an optical inlet. The spectrum online detection module (3) emits laser light toward the catalyst bed and collects spectrum data to analyze the catalyst poisoning condition.

[0082] The reaction device also includes a main control computer, which obtains data collected by the spectrum online detection module and analyzes it using the spectrum analysis module to obtain the catalyst poisoning situation;

[0083] The main control machine controls the heating module to heat and regenerate the catalyst according to the poisoning condition of the catalyst.

[0084] The main control machine is connected to the spectrum analysis module, the spectrum online detection module, the catalytic heating control module, the regeneration heating control module and the gas flow control module;

[0085] The spectrum online detection module includes a light emitter and a light collector;

[0086] The catalytic heating control module is used to control the catalyst temperature during catalytic cracking of ammonia, the regeneration heating control module is used to adjust the heating temperature of the catalyst bed during catalyst heating regeneration, and the gas flow control module is used to control the ammonia inlet flow and the mixed gas outlet flow.

[0087] The catalyst bed is cylindrical, the inner surface of which is a smooth light reflecting surface, and the catalyst is arranged inside through a three-dimensional grid structure; a gas outlet (42) is arranged on the top of the catalyst bed, a gas inlet (41) is arranged on the bottom of the catalyst bed, and a light emitter (32) and a light collector (31) are installed on the side of the catalyst bed, the light emitter is used to emit a detection laser toward the catalyst inside the catalyst bed, and the light collector is used to collect Raman scattered light generated by the detection laser on the surface of the catalyst; the spectral online detection module sends the Raman spectrum data collected by the light collector to the main control computer.

[0088] The main control computer sends the Raman spectrum data to the spectrum analysis module, and the spectrum analysis module processes the Raman spectrum data collected by the spectrum online detection module to determine the poisoning of the catalyst. The processing includes:

[0089] a. Data preprocessing: Normalize the Raman spectral data to ensure that the data is on the same scale, use a smoothing algorithm to reduce noise, and make the spectral curve smoother;

[0090] b. Input model: The processed spectral features are directly input into the deep learning model as a whole to obtain the catalyst poisoning depth and recommended heating temperature.

[0091] The structure of the deep learning model is:

[0092] Input Layer:

[0093] The size of the input layer is determined according to the number of wavelength points of the spectral data; if the spectral data has n wavelength points, the size of the input layer is n;

[0094] First convolutional layer:

[0095] Number of convolution kernels: 32 convolution kernels are selected to capture different features;

[0096] Convolution kernel size: The convolution kernel size is selected as 5×1 to capture local features;

[0097] Activation function: Use ReLU activation function to increase nonlinearity;

[0098] First pooling layer:

[0099] Pooling type: Use maximum pooling to reduce the spatial size of features, and choose a pooling kernel of 2×1 size;

[0100] Second convolutional layer:

[0101] Number of convolution kernels: Increase the number of convolution kernels to 64 to further abstract features;

[0102] Convolution kernel size: Select the convolution kernel size as 3×1;

[0103] Activation function: Use ReLU activation function;

[0104] Second pooling layer:

[0105] Pooling type: Use maximum pooling to reduce the spatial size of features, and choose a pooling kernel of 2×1 size;

[0106] Flattened layer:

[0107] Flatten the multi-dimensional convolutional layer output into one dimension so that it can be input into the fully connected layer;

[0108] First fully connected layer

[0109] Number of neurons: 128 neurons were selected;

[0110] Activation function: Use ReLU activation function;

[0111] Second fully connected layer

[0112] Number of neurons: 64 neurons were selected;

[0113] Activation function: Use ReLU activation function;

[0114] Branch 1 is used to predict poisoning depth:

[0115] Branch 1 fully connected layer: connected from the shared bottom layer network for further processing of poisoning depth features;

[0116] Branch 1 output layer: a single neuron, using a linear activation function, outputs a continuous value of the poisoning depth;

[0117] Branch 2 is used to predict the heating temperature:

[0118] Branch 2 fully connected layer: connected from the shared bottom layer network for further processing of poisoning depth features;

[0119] Branch 2 output layer: a single neuron, using a linear activation function, outputs a continuous value of the poisoning depth;

[0120] Loss function: Combine two losses using a weighted sum:

[0121] L=λ1·L depth +λ2·L temperature ;

[0122] Where L depth is the loss of poisoning depth, L temperature is the loss of the recommended heating temperature, while λ1 and λ2 are weights used to balance the importance of the two tasks.

[0123] Embodiment 2:

[0124] An integrated reaction control method for ammonia cracking and hydrogen production, using the integrated reaction device for ammonia cracking and hydrogen production, comprises the following steps:

[0125] Step 1: Data collection, start the spectrum online detection module (3), emit detection laser to the catalyst bed through the light emitter (32), collect Raman scattered light through the light collector (31), and send the collected Raman spectrum data to the main control computer;

[0126] Step 2: Data preprocessing: In the main control computer, the collected Raman spectrum data is normalized to ensure that the data is on the same scale, and a smoothing algorithm is used to reduce noise and make the spectrum curve smoother.

[0127] Step 3: Deep learning model analysis: input the preprocessed spectral data into the deep learning model, which analyzes the data and predicts the poisoning depth of the catalyst and the recommended heating temperature;

[0128] Step 4: Poisoning depth assessment: Based on the output of the deep learning model, the catalyst poisoning depth is assessed and the degree of poisoning is determined;

[0129] Step 5: Heating temperature adjustment: According to the heating temperature recommended by the deep learning model, the heating temperature of the catalyst bed is adjusted through the regeneration heating control module to promote the regeneration and recovery of the catalyst activity;

[0130] Step 6: Feedback control: during the catalyst heating regeneration process, continuously collect Raman spectrum data and repeat steps 2 to 5 to monitor the change of the catalyst state until the catalyst activity is restored to a predetermined level;

[0131] Step 7: Operation adjustment: According to the poisoning depth of the catalyst and the heating regeneration effect, the main control machine automatically adjusts the parameters of the catalytic heating control module and the gas flow control module to optimize the ammonia catalytic cracking process.

[0132] Before step 1, the method further includes the following steps:

[0133] Step 1: System startup, initializing the integrated reaction device for ammonia cracking and hydrogen production, including starting the gas supply system, heating system, spectrum online detection module and main control computer;

[0134] Step 2: System inspection: perform a comprehensive functional and safety check on the reactor to ensure that all sensors, actuators and control modules are working properly and there are no leaks or other potential safety issues;

[0135] Step 3: preheating and stabilization, starting the heating system (2) to preheat the catalyst bed (4) to achieve an optimal catalytic cracking temperature and keep the system stable;

[0136] Step 4: Initial gas flow setting: setting the initial ammonia flow rate through the gas flow control module to ensure that ammonia is input into the reactor at an appropriate rate;

[0137] Step 5: spectral system calibration, calibrating the light emitter (32) and light collector (31) in the spectral online detection module (3) to ensure the accuracy and repeatability of Raman spectral data;

[0138] Step 6: Data acquisition. After confirming that the system is stable and calibrated, start continuously collecting Raman spectral data of the catalyst bed for subsequent analysis and monitoring.

[0139] In this embodiment, the light emitted by the light emitter (32) is a monochromatic laser, and the wavelength range used is between visible light and near infrared light. When these light beams irradiate the catalyst on the catalyst bed (4), the following scattering phenomena will occur:

[0140] Rayleigh Scattering: This is the most common scattering process when light interacts with matter. In Rayleigh scattering, the wavelength of the scattered light is the same as the wavelength of the incident light, so it does not generate new frequencies. Since the size of the catalyst particles is much smaller than the wavelength of light, the intensity of the Rayleigh scattered light is inversely proportional to the fourth power of the wavelength, resulting in a stronger scattering of short wavelength light.

[0141] Raman scattering: Unlike Rayleigh scattering, Raman scattering involves the interaction of photons with the vibration or rotational energy levels of matter, resulting in the frequency of the scattered light being different from the frequency of the incident light. This frequency shift provides information about the molecular vibration energy levels, so Raman scattering is an inelastic scattering process.

[0142] In the integrated reaction device for hydrogen production by ammonia cracking, the inner surface of the catalyst bed (4) is designed to be a smooth light reflecting surface to enhance the collection efficiency of scattered light. When the light emitted by the light emitter (32) irradiates the catalyst surface, Rayleigh scattering and Raman scattering occur simultaneously. Since the intensity of the Rayleigh scattering light is much greater than the intensity of the Raman scattering light, the Rayleigh scattering light is the main component in the collected scattered light, while the Raman scattering light is the weak signal that we need to analyze.

[0143] Since the inner wall of the catalyst bed is a reflective structure, the scattered light is reflected multiple times between the inner walls of the bed, which increases the path length of the light, so that more Raman scattered light has the opportunity to interact with the catalyst particles and be scattered again. In this way, even if the Raman scattered light is very weak when it is initially emitted, the multiple reflections on the inner wall can increase the probability of it being detected by the light collector (31).

[0144] Finally, a portion of the scattered light, including Rayleigh scattered light and Raman scattered light, will be collected by a light collector (31). Since the intensity of Rayleigh scattered light is much greater than that of Raman scattered light, it is usually necessary to use a filter or other optical device to suppress the Rayleigh scattered light so that the Raman scattered light signal can be detected more clearly. The collected Raman scattered light is then sent to a spectral analysis module. By analyzing the frequency shift and intensity of the Raman scattered light, the chemical state and structural information of the catalyst can be obtained, and then the poisoning of the catalyst can be evaluated.

[0145] Thus far, the description of the above-described embodiments is provided for the purpose of illustration and description. It is not intended to be exhaustive or to limit the present disclosure. The individual elements or features of a specific embodiment are generally not subject to the limitations of the specific embodiment, but when applicable, they can be interchanged and used for selected embodiments even if not specifically shown or described. In many respects, the same element or feature also can be changed. This variation is not considered to be a departure from the present disclosure, and all such modifications are intended to be included in the scope of the present disclosure.

[0146] Example embodiments are provided so that the present disclosure will be thorough and the scope will be fully conveyed to those skilled in the art. In order to thoroughly understand the embodiments of the present disclosure, numerous details are set forth, such as examples of specific parts, devices, and methods. Obviously, for those skilled in the art, it is not necessary to use specific details, the example embodiments can be implemented in many different forms, and neither should be construed as limiting the scope of the present disclosure. In certain example embodiments, well-known processes, well-known device structures, and well-known technologies are not described in detail.

[0147] Here, specialized vocabulary is used only for the purpose of describing specific example embodiments, and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "a" and "the" used herein may be intended to include plural forms as well. The terms "including" and "having" are inclusive, and therefore specify the presence of the claimed features, wholes, steps, operations, elements and / or components, but do not exclude the presence or additional presence of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof. Unless the order of execution is clearly indicated, the method steps, processing and operations described herein are not interpreted as necessarily needing to be performed in the specific order discussed and shown. It should also be understood that additional or optional steps may be adopted.

Claims

1. An integrated reaction device for producing hydrogen by cracking ammonia, comprising a reactor, a catalyst bed (4), a heating system (2) and a spectrum online detection module (3), characterized in that: The reactor comprises a reactor shell (1), an ammonia inlet (6), a heater inlet, an optical inlet and a mixed gas outlet; a catalyst bed (4) is arranged in the reactor shell, and a three-dimensional mesh catalyst is arranged in the catalyst bed; the catalyst is used to crack the ammonia into hydrogen and nitrogen under heating; the reaction equation is: 2NH3→3H2+N2; The catalyst is one or more of a ruthenium-based catalyst, a nickel-based catalyst or an iron-based catalyst; The top of the reactor shell is connected to a mixed gas outlet (7), and the bottom of the reactor shell is connected to a waste outlet (5); the bottom of the side of the reactor shell is connected to an ammonia inlet (6), and the ammonia inlet is connected to the inside of the reactor shell and then connected to an ammonia disperser for dispersing ammonia into the catalyst bed; A heating system (2) and a spectrum online detection module (3) are arranged outside the reactor shell. The heating system (2) enters the reactor shell through a heater inlet and heats the catalyst bed. The spectrum online detection module (3) enters the catalyst bed through an optical inlet. The spectrum online detection module (3) emits laser light toward the catalyst bed and collects spectrum data to analyze the catalyst poisoning condition.

2. The integrated reaction device for producing hydrogen by cracking ammonia according to claim 1, characterized in that: The reaction device also includes a main control computer, which obtains data collected by the spectrum online detection module and analyzes it using the spectrum analysis module to obtain the catalyst poisoning situation; The main control machine controls the heating module to heat and regenerate the catalyst according to the poisoning condition of the catalyst.

3. The integrated reaction device for producing hydrogen by cracking ammonia according to claim 2, characterized in that: The main control machine is connected to the spectrum analysis module, the spectrum online detection module, the catalytic heating control module, the regeneration heating control module and the gas flow control module; The spectrum online detection module includes a light emitter and a light collector; The catalytic heating control module is used to control the catalyst temperature during catalytic cracking of ammonia, the regeneration heating control module is used to adjust the heating temperature of the catalyst bed during catalyst heating regeneration, and the gas flow control module is used to control the ammonia inlet flow and the mixed gas outlet flow.

4. The integrated reaction device for producing hydrogen by cracking ammonia according to claim 3, characterized in that: The catalyst bed is cylindrical, the inner surface of which is a smooth light reflecting surface, and the catalyst is arranged inside through a three-dimensional grid structure; a gas outlet (42) is arranged on the top of the catalyst bed, a gas inlet (41) is arranged on the bottom of the catalyst bed, and a light emitter (32) and a light collector (31) are installed on the side of the catalyst bed, the light emitter is used to emit a detection laser toward the catalyst inside the catalyst bed, and the light collector is used to collect Raman scattered light generated by the detection laser on the surface of the catalyst; the spectral online detection module sends the Raman spectrum data collected by the light collector to the main control computer.

5. The integrated reaction device for producing hydrogen by cracking ammonia according to claim 4, characterized in that: The main control computer sends the Raman spectrum data to the spectrum analysis module, and the spectrum analysis module processes the Raman spectrum data collected by the spectrum online detection module to determine the poisoning of the catalyst. The processing includes: a. Data preprocessing: Normalize the Raman spectral data to ensure that the data is on the same scale, use a smoothing algorithm to reduce noise, and make the spectral curve smoother; b. Input model: The processed spectral features are directly input into the deep learning model as a whole to obtain the catalyst poisoning depth and recommended heating temperature.

6. The integrated reaction device for producing hydrogen by cracking ammonia according to claim 5, characterized in that: The structure of the deep learning model is: Input Layer: The size of the input layer is determined according to the number of wavelength points of the spectral data; if the spectral data has n wavelength points, the size of the input layer is n; First convolutional layer: Number of convolution kernels: 32 convolution kernels are selected to capture different features; Convolution kernel size: The convolution kernel size is selected as 5×1 to capture local features; Activation function: Use ReLU activation function to increase nonlinearity; First pooling layer: Pooling type: Use maximum pooling to reduce the spatial size of features, and select a pooling kernel of 2×1 size; Second convolutional layer: Number of convolution kernels: Increase the number of convolution kernels to 64 to further abstract features; Convolution kernel size: Select the convolution kernel size as 3×1; Activation function: Use ReLU activation function; Second pooling layer: Pooling type: Use maximum pooling to reduce the spatial size of features, and choose a pooling kernel of 2×1; Flattening layer: Flatten the multi-dimensional convolutional layer output into one dimension so that it can be input into the fully connected layer; First fully connected layer Number of neurons: 128 neurons were selected; Activation function: Use ReLU activation function; Second fully connected layer Number of neurons: 64 neurons were selected; Activation function: Use ReLU activation function; Branch 1 is used to predict poisoning depth: Branch 1 fully connected layer: connected from the shared bottom layer network for further processing of poisoning depth features; Branch 1 output layer: a single neuron, using a linear activation function, outputs a continuous value of the poisoning depth; branch 2 is used to predict the heating temperature: Branch 2 fully connected layer: connected from the shared underlying network for further processing of poisoning depth features; Branch 2 output layer: a single neuron, using a linear activation function, outputs a continuous value of the poisoning depth; Loss function: uses a weighted sum to combine two losses: L=λ1·L depth +λ2·L temperature ; Where L depth is the loss of poisoning depth, L temperature is the loss of the recommended heating temperature, while λ1 and λ2 are weights used to balance the importance of the two tasks.

7. A method for controlling an integrated reaction of ammonia cracking to produce hydrogen, using an integrated reaction device of ammonia cracking to produce hydrogen according to any one of claims 1 to 6, characterized in that The steps include: Step 1: Data collection, start the spectrum online detection module (3), emit detection laser to the catalyst bed through the light emitter (32), collect Raman scattered light through the light collector (31), and send the collected Raman spectrum data to the main control computer; Step 2: Data preprocessing: In the main control computer, the collected Raman spectrum data is normalized to ensure that the data is on the same scale, and a smoothing algorithm is used to reduce noise and make the spectrum curve smoother. Step 3: Deep learning model analysis: input the preprocessed spectral data into the deep learning model, which analyzes the data and predicts the poisoning depth of the catalyst and the recommended heating temperature; Step 4: Poisoning depth assessment: Based on the output of the deep learning model, the catalyst poisoning depth is assessed and the degree of poisoning is determined; Step 5: Heating temperature adjustment: According to the heating temperature recommended by the deep learning model, the heating temperature of the catalyst bed is adjusted through the regeneration heating control module to promote the regeneration and recovery of the catalyst activity; Step 6: Feedback control: during the catalyst heating and regeneration process, continuously collect Raman spectrum data and repeat steps 2 to 5 to monitor the change of the catalyst state until the catalyst activity is restored to a predetermined level; Step 7: Operation adjustment: According to the poisoning depth of the catalyst and the heating regeneration effect, the main control machine automatically adjusts the parameters of the catalytic heating control module and the gas flow control module to optimize the ammonia catalytic cracking process.

8. The integrated reaction control method for ammonia cracking to produce hydrogen according to claim 7, characterized in that Before step 1, the method further includes the following steps: Step 1: System startup, initializing the integrated reaction device for ammonia cracking and hydrogen production, including starting the gas supply system, heating system, spectrum online detection module and main control computer; Step 2: System inspection: perform a comprehensive functional and safety check on the reactor to ensure that all sensors, actuators and control modules are working properly and there are no leaks or other potential safety issues; Step 3: preheating and stabilization, starting the heating system (2) to preheat the catalyst bed (4) to achieve an optimal catalytic cracking temperature and maintain system stability; Step 4: Initial gas flow setting: setting the initial ammonia flow rate through the gas flow control module to ensure that ammonia is input into the reactor at an appropriate rate; Step 5: spectral system calibration, calibrating the light emitter (32) and light collector (31) in the spectral online detection module (3) to ensure the accuracy and repeatability of Raman spectral data; Step 6: Data acquisition. After confirming that the system is stable and calibrated, start continuously collecting Raman spectral data of the catalyst bed for subsequent analysis and monitoring.

Images