Preparation control method of denitration and desulfurization activated carbon catalyst

Abstract

This invention relates to the field of activated carbon regeneration and preparation technology, and particularly to a method and system for controlling the preparation of denitrification and desulfurization activated carbon catalysts. The method includes: adding activated carbon to a prepared impregnation solution; collecting the concentration of metal ions in the impregnation solution and the microcirculation flow rate of the impregnation solution during loading in real time based on a preset time interval; obtaining the decrease rate parameter Nd of the metal ion concentration and the change rate parameter Ls of the liquid flow rate; establishing a loading uniformity classification rule based on the decrease rate parameter of the metal ion concentration and the change rate parameter of the liquid flow rate, and obtaining the loading uniformity classification result. This invention obtains the decrease rate of the metal ion concentration and the change rate of the liquid flow rate, and evaluates the loading state of the activated carbon based on these two types of dynamic parameters. The combination of the two can accurately characterize whether there are abnormal states such as excessively fast adsorption, local saturation, or mass transfer lag during the loading process from the dual dimensions of adsorption rate and liquid phase mass transfer.

Description

Technical Field

[0001] This invention relates to the field of activated carbon regeneration and preparation technology, specifically to a method for controlling the preparation of denitrification and desulfurization activated carbon catalysts. Background Technology

[0002] Activated carbon, due to its well-developed pore structure, excellent adsorption properties, and tunable surface functional groups, is widely used as a catalyst or support in flue gas denitrification (SCR / SNCR) and desulfurization (FGD) processes. By loading metal oxides onto its surface and controlling the pore size, it is possible to effectively remove SO2 and NO. x Highly efficient catalytic removal of pollutants such as [list of pollutants].

[0003] Currently, in the regeneration and preparation of activated carbon, specifically the process of controlling the loading of activated carbon into a metal-loaded solution, the metal loading is typically assessed using methods such as fixed-time impregnation or single-point concentration detection. This lack of real-time dynamic monitoring of key parameters during impregnation makes it difficult to promptly detect non-uniform adsorption, agglomeration, or localized supersaturation during the loading process. Furthermore, traditional methods often neglect the influence of the internal flow state of the impregnation solution on the mass transfer efficiency of metal ions, failing to effectively incorporate parameters characterizing the stability of the reaction environment, such as liquid flow rate. This results in an inability to respond promptly to uneven loading or initial mass transfer obstacles, reducing the overall uniformity and repeatability of the loading. Summary of the Invention

[0004] To address the above problems, this invention provides a method for controlling the preparation of denitrification and desulfurization activated carbon catalysts.

[0005] This invention employs the following technical solution: a method for controlling the preparation of a denitrification and desulfurization activated carbon catalyst, comprising:

[0006] Activated carbon is added to the prepared impregnation solution, and the activated carbon loading process is controlled according to preset control parameters, including impregnation temperature, pH value of impregnation solution and impregnation time.

[0007] Based on a preset time interval, the concentration of metal ions in the impregnation solution and the microcirculation flow rate of the impregnation solution during the loading process are collected in real time to obtain the decrease rate parameter Nd of the metal ion concentration and the change rate parameter Ls of the liquid flow rate.

[0008] Based on the decrease rate parameter of metal ion concentration and the change rate parameter of liquid flow rate, a load uniformity classification rule is established, and the load uniformity classification result is obtained.

[0009] Obtain historical production datasets of activated carbon regeneration loads, and train a pre-built load process prediction model based on the historical production datasets.

[0010] Obtain the current process control parameters and trigger different process optimization strategies based on different load uniformity classification results.

[0011] As a further description of the above technical solution: the historical production dataset includes P sets of load parameters and the process control strategies corresponding to the P sets of load parameters, where P is a positive integer greater than 0. The historical production dataset is divided into a training set and a validation set, wherein the training set is used to train the load process prediction model and the validation set is used to evaluate the generalization performance of the load process prediction model.

[0012] The process control strategy includes impregnation temperature, impregnation solution pH value, and impregnation time.

[0013] As a further description of the above technical solution: the training method of the load process prediction model includes:

[0014] During the training of the load process prediction model, a deep neural network structure based on multilayer perceptrons is adopted. The load parameters are converted into feature vectors as input, and nonlinear features in the data are extracted through multiple hidden layers. Finally, the softmax activation function is used in the output layer to generate the probability distribution of the process control strategy, and the process control strategy corresponding to the highest probability is output as the final prediction result. The training process aims to minimize the cross-entropy loss function, and an early stopping strategy is introduced to monitor the performance of the validation set. When the prediction accuracy on the validation set reaches a preset threshold, the load process prediction model is considered to have converged, and the training stops.

[0015] As a further description of the above technical solution: the loading parameters include the type of impregnation solution, the concentration of the impregnation solution, the impregnation method, and the specific surface area of ​​the activated carbon;

[0016] The impregnation solutions include Fe(NO3)3 solution, MnSO4 solution and CuCl2 solution;

[0017] The impregnation methods include the equal volume method, the ultrasonic-assisted method, the decompression method, and the microwave-assisted method.

[0018] As a further description of the above technical solution: the load uniformity grading result includes a first uniformity level, a second uniformity level, and a third uniformity level.

[0019] As a further description of the above technical solution: the method for establishing load uniformity grading rules and obtaining load uniformity grading results includes:

[0020] Preset the threshold values ​​for the rate of decrease of metal ion concentration (Nmax) and the threshold values ​​for the rate of change of liquid flow velocity (Lmax);

[0021] When the descent rate parameter Nd < Nmax and the liquid flow rate change parameter Ls < Lmax, it is determined to be the first uniformity level;

[0022] When the descent rate parameter Nd > Nmax and the liquid flow rate change parameter Ls > Lmax, it is determined to be the third level of uniformity.

[0023] The remaining cases are classified as the second level of uniformity.

[0024] As a further description of the above technical solution: the method for triggering different process optimization strategies includes:

[0025] If it is the third uniformity level, the turbidity and interface potential of the impregnation liquid in the impregnation tank are obtained. The obtained turbidity and interface potential of the impregnation liquid, as well as the real-time load parameters, are input into the pre-constructed activated carbon load result prediction model, and the activated carbon load result is output. The activated carbon load result includes two types: load meeting the standard and load not meeting the standard.

[0026] When the output load meets the standard, the real-time load parameters are input into the load process prediction model, and the process control strategy is output. The process control strategy includes impregnation temperature, impregnation solution pH value and impregnation time. The impregnation temperature, impregnation solution pH value and impregnation time in the preset control parameters are obtained to obtain the impregnation temperature adjustment amount, impregnation solution pH value adjustment amount and impregnation time adjustment amount. The impregnation temperature adjustment amount, impregnation solution pH value adjustment amount and impregnation time adjustment amount are the process optimization strategy.

[0027] When the output load is below the standard, an alarm shutdown command is generated directly.

[0028] As a further description of the above technical solution: If it is the second uniformity level, the real-time load parameters are input into the load process prediction model, and the process control strategy is output. The process control strategy includes the impregnation temperature, the pH value of the impregnation solution, and the impregnation time. The impregnation temperature, the pH value of the impregnation solution, and the impregnation time are obtained from the preset control parameters, and the impregnation temperature adjustment amount, the impregnation solution pH value adjustment amount, and the impregnation time adjustment amount are obtained. The impregnation temperature adjustment amount, the impregnation solution pH value adjustment amount, and the impregnation time adjustment amount are the process optimization strategy.

[0029] If the uniformity level is the first level, no process optimization strategy will be generated.

[0030] As a further description of the above technical solution: the training method of the activated carbon loading result prediction model includes:

[0031] Q sets of training data were collected in advance, where Q is a positive integer greater than 0. The training data included the turbidity of the impregnation solution, the interfacial potential and the real-time loading parameters, as well as the activated carbon loading results corresponding to the turbidity of the impregnation solution, the interfacial potential and the real-time loading parameters.

[0032] The activated carbon loading result prediction model was trained using training data. The turbidity of the impregnation solution, the interface potential, and the real-time loading parameters were used as inputs to the activated carbon loading result prediction model, and the activated carbon loading result was used as the output. The stochastic gradient descent method was used, and the weights and biases of the activated carbon loading result prediction model were adjusted through the backpropagation algorithm to minimize the error between the prediction result and the actual result. A loss function was set, which was the mean squared error. When the loss function value converged, the training of the activated carbon loading result prediction model was stopped, and the activated carbon loading result prediction model corresponding to the convergence of the loss function value was used as the trained activated carbon loading result prediction model.

[0033] A preparation control system for a denitrification and desulfurization activated carbon catalyst, used to implement the preparation control method of the aforementioned denitrification and desulfurization activated carbon catalyst, the system comprising:

[0034] The load control module adds activated carbon to the prepared impregnation solution and controls the activated carbon loading process according to preset control parameters, including impregnation temperature, pH value of the impregnation solution, and impregnation time.

[0035] The data acquisition module collects the concentration of metal ions in the impregnation solution and the microcirculation flow rate of the impregnation solution during the loading process in real time based on a preset time interval, and obtains the decrease rate parameter Nd of the metal ion concentration and the change rate parameter Ls of the liquid flow rate.

[0036] The uniformity judgment module establishes load uniformity classification rules based on the decrease rate parameter of metal ion concentration and the change rate parameter of liquid flow rate, and obtains the load uniformity classification results.

[0037] The model training module obtains historical production datasets of activated carbon regeneration loads and trains a pre-built load process prediction model based on these datasets.

[0038] The process optimization module obtains the current process control parameters and triggers different process optimization strategies based on different load uniformity grading results.

[0039] Beneficial effects:

[0040] In the above technical solution, the present invention provides a method for controlling the preparation of denitrification and desulfurization activated carbon catalyst. By real-time acquisition of the concentration of metal ions in the impregnation solution and the microcirculation flow rate during the impregnation process, the decrease rate of metal ion concentration and the change rate of liquid flow rate are obtained. Based on these two dynamic parameters, the loading state of activated carbon is evaluated. The decrease rate of metal ion concentration reflects the speed at which metal ions are adsorbed onto the surface of activated carbon per unit time, which is the core indicator for judging the migration and binding rate of metal components. The change rate of liquid flow rate reveals the degree of disturbance and mass transfer efficiency of the microenvironment inside the impregnation solution, which can indirectly reflect the balance of liquid-solid phase interface exchange during the loading process. The combination of the two can accurately characterize whether there are abnormal states such as excessively fast adsorption, local saturation or mass transfer lag during the loading process from the dual dimensions of adsorption rate and liquid phase mass transfer. Compared with traditional fixed-duration or single-concentration monitoring methods, this method has high timeliness and systematicity, significantly improving the ability to identify abnormal conditions during the loading process and the response efficiency, providing a reliable data basis for subsequent determination of loading uniformity level and adjustment of process strategy.

[0041] Furthermore, based on the load uniformity grading rules using Nd and Ls parameters, the load uniformity grading results are divided into three levels: first uniformity, second uniformity, and third uniformity. Differentiated process control mechanisms are set for each level. When the system identifies the third uniformity level (i.e., poor load uniformity), it automatically acquires turbidity and interface potential data to further determine whether the load effect meets the standard. If it does not meet the standard, an alarm shutdown mechanism is immediately triggered to effectively prevent the generation of low-quality products and resource waste. When the standard is met or the second uniformity level is reached, the system does not directly interrupt the process but uses a load process prediction model driven by historical production data to dynamically adjust key process parameters such as impregnation temperature, pH value, and impregnation time to improve the load balance and efficiency. This grading response mechanism has strong autonomous judgment and intervention capabilities, which not only ensures product quality but also improves the stability and automation of the entire impregnation process and enhances the intelligence of activated carbon load control. Attached Figure Description

[0042] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0043] Figure 1 This is a flowchart illustrating the preparation and control method of a denitrification and desulfurization activated carbon catalyst provided in Embodiment 1 of the present invention.

[0044] Figure 2 This is a flowchart of the steps involved in the preparation and control of a denitrification and desulfurization activated carbon catalyst provided in Example 1 of the present invention.

[0045] Figure 3 This is a flowchart of the steps of the method for triggering different process optimization strategies provided in Embodiment 1 of the present invention;

[0046] Figure 4 This is a model connection diagram of a preparation control system for a denitrification and desulfurization activated carbon catalyst provided in an embodiment of the present invention. Detailed Implementation

[0047] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0048] Example 1

[0049] Please see Figures 1-3 This invention provides a technical solution: a method for controlling the preparation of a denitrification and desulfurization activated carbon catalyst, used for the regeneration of activated carbon, i.e., controlling the loading process of the activated carbon impregnated with a metal-supported solution, comprising:

[0050] Activated carbon is added to the prepared impregnation solution, and the activated carbon loading process is controlled according to preset control parameters, including impregnation temperature, pH value of impregnation solution and impregnation time.

[0051] Based on a preset time interval, the concentration of metal ions in the impregnation solution and the microcirculation flow rate of the impregnation solution during the loading process are collected in real time to obtain the decrease rate parameter Nd of the metal ion concentration and the change rate parameter Ls of the liquid flow rate.

[0052] It should be noted that a steady or slow decrease in the concentration of metal ions indicates uniform adsorption, while a sudden drop may be due to local agglomeration and deposition. When the load is uniform, the pore size of the activated carbon decreases less, and the change rate of the liquid microcirculation velocity during the loading process is small. When the load agglomerates and causes pore blockage, the rate of change of the velocity will increase.

[0053] The formula for calculating the rate of decrease of the metal ion concentration is as follows: In the formula, Nd is the rate parameter of decrease of metal ion concentration, Δt is the time interval, and |ΔC| is the absolute value of the difference between the two metal ion concentrations detected.

[0054] The formula for calculating the liquid flow rate change parameter is as follows: In the formula, Ls is the liquid flow rate change parameter, Δt is the time interval, and |ΔV| is the absolute value of the difference between the two detected liquid flow rates.

[0055] Based on the decrease rate parameter of metal ion concentration and the change rate parameter of liquid flow rate, a load uniformity classification rule is established, and the load uniformity classification result is obtained.

[0056] The load uniformity classification results include a first uniformity level, a second uniformity level, and a third uniformity level;

[0057] The method for establishing load uniformity classification rules and obtaining load uniformity classification results includes:

[0058] Preset the threshold values ​​for the rate of decrease of metal ion concentration (Nmax) and the threshold values ​​for the rate of change of liquid flow velocity (Lmax);

[0059] It should be noted that the descent rate threshold Nmax and the liquid flow rate change threshold Lmax were determined by those skilled in the art based on a large number of experiments.

[0060] When the descent rate parameter Nd < Nmax and the liquid flow rate change parameter Ls < Lmax, it is determined to be the first uniformity level;

[0061] When the descent rate parameter Nd > Nmax and the liquid flow rate change parameter Ls > Lmax, it is determined to be the third level of uniformity.

[0062] The remaining cases are classified as the second level of uniformity.

[0063] In this embodiment, the concentration of metal ions in the impregnation solution and the microcirculation flow rate during the impregnation process are collected in real time to obtain the rate of decrease of metal ion concentration and the rate of change of liquid flow rate. The loading state of activated carbon is evaluated based on these two types of dynamic parameters. The rate of decrease of metal ion concentration reflects the speed at which metal ions are adsorbed onto the surface of activated carbon per unit time, which is the core indicator for judging the migration and binding rate of metal components. The rate of change of liquid flow rate reveals the degree of disturbance and mass transfer efficiency of the microenvironment inside the impregnation solution, which can indirectly reflect the balance of liquid-solid phase interface exchange during the loading process.

[0064] By combining the two, the abnormal states of excessively rapid adsorption, local saturation, or mass transfer lag during the loading process can be accurately characterized from both the adsorption rate and liquid phase mass transfer dimensions. This overcomes the limitations of traditional fixed-duration or single-concentration monitoring methods. This judgment method has high timeliness and systematicity, and significantly improves the ability to identify the uniformity of the load during the loading process and the response efficiency.

[0065] Obtain historical production datasets of activated carbon regeneration loads, and train a pre-built load process prediction model based on the historical production datasets.

[0066] The historical production dataset includes P sets of load parameters and the corresponding process control strategies for the P sets of load parameters, where P is a positive integer greater than 0. The historical production dataset is divided into a training set and a validation set, where the training set is used to train the load process prediction model and the validation set is used to evaluate the generalization performance of the load process prediction model.

[0067] The process control strategy includes impregnation temperature, impregnation solution pH value, and impregnation time;

[0068] The training method for the load process prediction model includes:

[0069] During the training of the load process prediction model, a deep neural network structure based on multilayer perceptron is adopted. The load parameters are converted into feature vectors as input, and nonlinear features in the data are extracted through multiple hidden layers. Finally, the softmax activation function is used in the output layer to generate the probability distribution of the process control strategy. The process control strategy corresponding to the highest probability is output as the final prediction result. The training process aims to minimize the cross-entropy loss function. At the same time, an early stopping strategy is introduced to monitor the performance of the validation set. When the prediction accuracy on the validation set reaches a preset threshold, the load process prediction model is considered to have converged, and the training stops.

[0070] It should be noted that the preset threshold for the prediction accuracy on the validation set mentioned above is determined based on specific application requirements.

[0071] The loading parameters include the type of impregnation solution, the concentration of the impregnation solution, the impregnation method, and the specific surface area of ​​the activated carbon;

[0072] The impregnation solutions include Fe(NO3)3 solution, MnSO4 solution, and CuCl2 solution;

[0073] The impregnation methods include the equal volume method, the ultrasonic-assisted method, the decompression method, and the microwave-promoted method. All of the above impregnation methods are existing technologies, and the specific methods and procedures will not be disclosed here.

[0074] The specific surface area of ​​the activated carbon was obtained by low-temperature nitrogen adsorption-desorption experiment;

[0075] Obtain the current process control parameters and trigger different process optimization strategies based on different load uniformity classification results;

[0076] The methods for triggering different process optimization strategies include:

[0077] If it is the third uniformity level, the turbidity and interface potential of the impregnation liquid in the impregnation tank are obtained. The obtained turbidity and interface potential of the impregnation liquid, as well as the real-time load parameters, are input into the pre-constructed activated carbon load result prediction model, and the activated carbon load result is output. The activated carbon load result includes two types: load meeting the standard and load not meeting the standard.

[0078] When the output load meets the target, the real-time load parameters are input into the load process prediction model, and a process control strategy is output. The process control strategy includes impregnation temperature, impregnation solution pH value, and impregnation time. The impregnation temperature, impregnation solution pH value, and impregnation time are obtained from the preset control parameters to obtain the impregnation temperature adjustment amount, impregnation solution pH value adjustment amount, and impregnation time adjustment amount. The impregnation temperature adjustment amount, impregnation solution pH value adjustment amount, and impregnation time adjustment amount are the process optimization strategy; the adjustment amount is the difference between the preset control parameters and the process control strategy.

[0079] When the output load is below the standard, an alarm shutdown command is generated directly;

[0080] If it is the second uniformity level, the real-time load parameters are input into the load process prediction model, and the process control strategy is output. The process control strategy includes impregnation temperature, impregnation solution pH value and impregnation time. The impregnation temperature, impregnation solution pH value and impregnation time in the preset control parameters are obtained to obtain the impregnation temperature adjustment amount, impregnation solution pH value adjustment amount and impregnation time adjustment amount. The impregnation temperature adjustment amount, impregnation solution pH value adjustment amount and impregnation time adjustment amount are the process optimization strategy.

[0081] If it is the first uniformity level, no process optimization strategy will be generated;

[0082] The training method for the activated carbon loading result prediction model includes:

[0083] Q sets of training data were collected in advance, where Q is a positive integer greater than 0. The training data included the turbidity of the impregnation solution, the interfacial potential and the real-time loading parameters, as well as the activated carbon loading results corresponding to the turbidity of the impregnation solution, the interfacial potential and the real-time loading parameters.

[0084] The activated carbon loading result prediction model was trained using training data. The turbidity of the impregnation solution, the interface potential, and the real-time loading parameters were used as inputs to the activated carbon loading result prediction model, and the activated carbon loading result was used as the output. The stochastic gradient descent method was used, and the weights and biases of the activated carbon loading result prediction model were adjusted through the backpropagation algorithm to minimize the error between the prediction result and the actual result. A loss function was set, which was the mean squared error. When the loss function value converged, the training of the activated carbon loading result prediction model was stopped, and the activated carbon loading result prediction model corresponding to the convergence of the loss function value was used as the trained activated carbon loading result prediction model.

[0085] In this embodiment, the load uniformity is graded into three levels based on the Nd and Ls parameters: a first uniformity level, a second uniformity level, and a third uniformity level. Differentiated process control mechanisms are set for each level. When the system identifies the third level (poor load uniformity), it automatically acquires turbidity and interface potential data to further determine whether the load effect meets the standard. If it does not meet the standard, an alarm shutdown mechanism is immediately triggered to effectively prevent the generation of low-quality products and waste of resources. When the standard is met or the second level is identified, the system does not directly interrupt the process but uses a load process prediction model driven by historical production data to dynamically adjust key process parameters such as impregnation temperature, pH value, and impregnation time to improve the load balance and efficiency. This graded response mechanism has strong autonomous judgment and intervention capabilities, which not only ensures product quality but also improves the stability and automation of the entire impregnation process and enhances the intelligence of activated carbon load control.

[0086] Example 2

[0087] Please see Figure 4 This invention provides a technical solution: a preparation control system for denitrification and desulfurization activated carbon catalyst, which is used to implement the preparation control method of the aforementioned denitrification and desulfurization activated carbon catalyst. The system includes:

[0088] The load control module adds activated carbon to the prepared impregnation solution and controls the activated carbon loading process according to preset control parameters, including impregnation temperature, pH value of the impregnation solution, and impregnation time.

[0089] The data acquisition module collects the concentration of metal ions in the impregnation solution and the microcirculation flow rate of the impregnation solution during the loading process in real time based on a preset time interval, and obtains the decrease rate parameter Nd of the metal ion concentration and the change rate parameter Ls of the liquid flow rate.

[0090] The uniformity judgment module establishes load uniformity classification rules based on the decrease rate parameter of metal ion concentration and the change rate parameter of liquid flow rate, and obtains the load uniformity classification results.

[0091] The model training module obtains historical production datasets of activated carbon regeneration loads and trains a pre-built load process prediction model based on these datasets.

[0092] The process optimization module obtains the current process control parameters and triggers different process optimization strategies based on different load uniformity grading results.

[0093] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for controlling the preparation of a denitrification and desulfurization activated carbon catalyst, characterized in that, include: Activated carbon is added to the prepared impregnation solution, and the activated carbon loading process is controlled according to preset control parameters, including impregnation temperature, pH value of impregnation solution and impregnation time. Based on a preset time interval, the concentration of metal ions in the impregnation solution and the microcirculation flow rate of the impregnation solution during the loading process are collected in real time to obtain the decrease rate parameter Nd of the metal ion concentration and the change rate parameter Ls of the liquid flow rate. Based on the decrease rate parameter of metal ion concentration and the change rate parameter of liquid flow rate, a load uniformity classification rule is established, and the load uniformity classification result is obtained. Obtain historical production datasets of activated carbon regeneration loads, and train a pre-built load process prediction model based on the historical production datasets. Obtain the current process control parameters and trigger different process optimization strategies based on different load uniformity classification results; The method for establishing load uniformity classification rules and obtaining load uniformity classification results includes: Preset the threshold values ​​for the rate of decrease of metal ion concentration (Nmax) and the threshold values ​​for the rate of change of liquid flow velocity (Lmax); When the descent rate parameter Nd < Nmax and the liquid flow rate change parameter Ls < Lmax, it is determined to be the first uniformity level; When the descent rate parameter Nd > Nmax and the liquid flow rate change parameter Ls > Lmax, it is determined to be the third level of uniformity. The remaining cases are classified as the second level of uniformity; The methods for triggering different process optimization strategies include: If it is the third uniformity level, the turbidity and interface potential of the impregnation liquid in the impregnation tank are obtained. The obtained turbidity and interface potential of the impregnation liquid, as well as the real-time load parameters, are input into the pre-constructed activated carbon load result prediction model, and the activated carbon load result is output. The activated carbon load result includes two types: load meeting the standard and load not meeting the standard. When the output load meets the standard, the real-time load parameters are input into the load process prediction model, and the process control strategy is output. The process control strategy includes impregnation temperature, impregnation solution pH value and impregnation time. The impregnation temperature, impregnation solution pH value and impregnation time in the preset control parameters are obtained to obtain the impregnation temperature adjustment amount, impregnation solution pH value adjustment amount and impregnation time adjustment amount. The impregnation temperature adjustment amount, impregnation solution pH value adjustment amount and impregnation time adjustment amount are the process optimization strategy. When the output load is below the standard, an alarm shutdown command is generated directly; If it is the second uniformity level, the real-time load parameters are input into the load process prediction model, and the process control strategy is output. The process control strategy includes impregnation temperature, impregnation solution pH value and impregnation time. The impregnation temperature, impregnation solution pH value and impregnation time in the preset control parameters are obtained to obtain the impregnation temperature adjustment amount, impregnation solution pH value adjustment amount and impregnation time adjustment amount. The impregnation temperature adjustment amount, impregnation solution pH value adjustment amount and impregnation time adjustment amount are the process optimization strategy. If the uniformity level is the first level, no process optimization strategy will be generated.

2. The method for controlling the preparation of a denitrification and desulfurization activated carbon catalyst according to claim 1, characterized in that, The historical production dataset includes P sets of load parameters and the corresponding process control strategies for the P sets of load parameters, where P is a positive integer greater than 0. The historical production dataset is divided into a training set and a validation set, where the training set is used to train the load process prediction model and the validation set is used to evaluate the generalization performance of the load process prediction model. The process control strategy includes immersion temperature, immersion solution pH value, and immersion time.

3. The method for controlling the preparation of a denitrification and desulfurization activated carbon catalyst according to claim 2, characterized in that, The training method for the load process prediction model includes: During the training of the load process prediction model, a deep neural network structure based on multilayer perceptrons is adopted. The load parameters are converted into feature vectors as input, and nonlinear features in the data are extracted through multiple hidden layers. Finally, the softmax activation function is used in the output layer to generate the probability distribution of the process control strategy, and the process control strategy corresponding to the highest probability is output as the final prediction result. The training process aims to minimize the cross-entropy loss function, and an early stopping strategy is introduced to monitor the performance of the validation set. When the prediction accuracy on the validation set reaches a preset threshold, the load process prediction model is considered to have converged, and the training stops.

4. The method for controlling the preparation of a denitrification and desulfurization activated carbon catalyst according to claim 3, characterized in that, The loading parameters include the type of impregnation solution, the concentration of the impregnation solution, the impregnation method, and the specific surface area of ​​the activated carbon; The impregnation solutions include Fe(NO3)3 solution, MnSO4 solution and CuCl2 solution; The impregnation methods include the equal volume method, the ultrasonic-assisted method, the decompression method, and the microwave-assisted method.

5. The method for controlling the preparation of a denitrification and desulfurization activated carbon catalyst according to claim 1, characterized in that, The load uniformity classification results include a first uniformity level, a second uniformity level, and a third uniformity level.

6. The method for controlling the preparation of a denitrification and desulfurization activated carbon catalyst according to claim 1, characterized in that, The training method for the activated carbon loading result prediction model includes: Q sets of training data were collected in advance, where Q is a positive integer greater than 0. The training data included the turbidity of the impregnation solution, the interfacial potential and the real-time loading parameters, as well as the activated carbon loading results corresponding to the turbidity of the impregnation solution, the interfacial potential and the real-time loading parameters. The activated carbon loading result prediction model was trained using training data. The turbidity of the impregnation solution, the interface potential, and the real-time loading parameters were used as inputs to the activated carbon loading result prediction model, and the activated carbon loading result was used as the output. The stochastic gradient descent method was used, and the weights and biases of the activated carbon loading result prediction model were adjusted through the backpropagation algorithm to minimize the error between the prediction result and the actual result. A loss function was set, which was the mean squared error. When the loss function value converged, the training of the activated carbon loading result prediction model was stopped, and the activated carbon loading result prediction model corresponding to the convergence of the loss function value was used as the trained activated carbon loading result prediction model.

7. A preparation control system for a denitrification and desulfurization activated carbon catalyst, used to implement the preparation control method for the denitrification and desulfurization activated carbon catalyst according to any one of claims 1-6, characterized in that, The system includes: The load control module adds activated carbon to the prepared impregnation solution and controls the activated carbon loading process according to preset control parameters, including impregnation temperature, pH value of the impregnation solution, and impregnation time. The data acquisition module collects the concentration of metal ions in the impregnation solution and the microcirculation flow rate of the impregnation solution during the loading process in real time based on a preset time interval, and obtains the decrease rate parameter Nd of the metal ion concentration and the change rate parameter Ls of the liquid flow rate. The uniformity judgment module establishes load uniformity classification rules based on the decrease rate parameter of metal ion concentration and the change rate parameter of liquid flow rate, and obtains the load uniformity classification results. The model training module obtains historical production datasets of activated carbon regeneration loads and trains a pre-built load process prediction model based on these datasets. The process optimization module obtains the current process control parameters and triggers different process optimization strategies based on different load uniformity grading results.

Images