Desulfurization parameter optimization method and device, storage medium and program product

Abstract

The invention relates to the field of flue gas desulfurization, and discloses a desulfurization parameter optimization method and equipment, a storage medium and a program product. The desulfurization parameter optimization method provided by the embodiment of the invention comprises the following steps: constructing a prediction model set according to preprocessed historical desulfurization data; calculating the selected probability of each prediction model in the prediction model set according to the real-time desulfurization data so as to generate model probability distribution, and selecting a prediction model used for calculating a target optimization function of the production line in the prediction model set; and based on the real-time desulfurization data and the selected prediction model, constructing an input population corresponding to the target optimization function, iteratively calculating the fitness of the input population, selecting the optimal input parameter of the target optimization function, determining the material consumption of the production line, and adjusting the material parameter of the input production line according to the calculated material consumption. According to the method provided by the embodiment of the invention, the energy consumption and material consumption cost of sintering CFB desulfurization are reduced, and the economic benefits of enterprises are improved.

Description

Technical Field

[0001] This invention relates to the field of flue gas desulfurization, and in particular to a method, equipment, storage medium, and program product for optimizing desulfurization parameters. Background Technology

[0002] SO2 emissions from the sintering process account for approximately 40% to 60% of the annual emissions from steel enterprises. Controlling SO2 emissions during sintering is not only a key focus of SO2 pollution control in steel enterprises but also crucial for achieving SO2 emission reduction targets and plans within the steel industry. Semi-dry desulfurization technology offers advantages such as low investment, small footprint, low water consumption, minimal equipment corrosion, dry byproducts, no wastewater generation, and simple process, effectively overcoming some of the problems and shortcomings of wet desulfurization. Therefore, semi-dry desulfurization has gradually become the dominant approach for sintering flue gas desulfurization.

[0003] Currently, circulating fluidized bed (CFB) desulfurization systems are commonly used to desulfurize flue gas generated during steel production. In semi-dry flue gas desulfurization processes, the flue gas still contains SO2 after denitrification. The SO2-containing flue gas enters the bottom of the reaction tower through the flue and then enters the absorption reaction zone. The SO2 in the flue gas will fully mix and react with the absorbent and atomized water to remove the SO2 from the flue gas.

[0004] Existing desulfurization methods suffer from drawbacks such as dynamic response lag and flow field uniformity issues. Because semi-dry desulfurization involves a strongly coupled process of flow-mass transfer-reaction, there is a time delay in measuring key parameters (such as SO2 concentration and bed pressure drop) in the desulfurization system. This makes it difficult for traditional PID control used in the desulfurization system to adjust accurately in real time, easily leading to excessive SO2 levels in the flue gas or over-dosing of absorbent. Furthermore, uneven airflow distribution can easily occur within the desulfurization tower during the desulfurization process, resulting in localized material accumulation or poor fluidization. This necessitates multiple adjustments to the gas delivery rate, increasing the difficulty and cost of engineering implementation. Summary of the Invention

[0005] This invention provides a method, equipment, storage medium, and program product for optimizing desulfurization parameters to solve at least some of the above-mentioned problems.

[0006] In a first aspect, embodiments of the present invention provide a method for optimizing desulfurization parameters, comprising: acquiring historical desulfurization data of a production line; preprocessing the historical desulfurization data; and constructing a set of prediction models based on the preprocessed historical desulfurization data; acquiring real-time desulfurization data of the production line; calculating the selection probability of each prediction model in the set of prediction models based on the real-time desulfurization data to generate a model probability distribution; selecting a prediction model from the set of prediction models to calculate the target optimization function of the production line based on the model probability distribution; constructing an input population corresponding to the target optimization function based on the real-time desulfurization data and the selected prediction model to iteratively calculate the fitness of the input population; selecting the optimal input parameters of the target optimization function based on the fitness of the input population to determine the material consumption of the production line; and adjusting the material parameters input to the production line based on the calculated material consumption of the production line.

[0007] The desulfurization parameter optimization method provided in this invention can dynamically adjust the amount of materials used for desulfurization based on actual desulfurization data in a dynamic desulfurization scenario, so as to avoid waste of materials used for desulfurization. While ensuring the desulfurization effect and pollutant emission compliance, it reduces the energy and material consumption costs of sintering CFB desulfurization, improves the economic benefits of enterprises, and achieves green cost-reducing production.

[0008] Optionally, the prediction model set is constructed based on the following steps: S1: Obtain the first process time from multiple desulfurization inlets to the reaction center and the second process time from the reaction center to the desulfurization outlet of the flue gas generated by the production line; S2: Based on the first process time, the second process time, and pre-processed historical desulfurization data, define the input vector and output result of the prediction model set corresponding to the first process time and the second process time; S3: Divide the input vector and output result into a training set and a test set; S4: Repeatedly train multiple prediction models based on the training set, and calculate the root mean square error of each prediction model through the test set until a preset number of training iterations is reached; S5: Determine whether the root mean square error of each prediction model is greater than a preset error threshold. If the root mean square error is greater than or equal to the preset error threshold, the prediction model is removed; if the root mean square error is less than the preset error threshold, the prediction model is retained; S6: Construct a prediction model set based on all retained prediction models.

[0009] Optionally, the desulfurization parameter optimization method further includes: after training the first prediction model, calculating the optimized convergence time of the first prediction model from the start of training to model convergence; if the second process time is less than the optimized convergence time, updating the second process time and the corresponding first process time, such that the updated second process time is greater than or equal to the optimized convergence time; and re-executing steps S2 to S6 based on the updated first process time and second process time to construct a set of prediction models.

[0010] Optionally, the input vector is The output result is Where y is + + The instantaneous measured value of sulfur dioxide concentration at the desulfurization outlet at a given time. For instantaneous measurement time, For the first process time, For the second process time, For the prediction model to be trained, for The flow rate of quicklime at any given time. for Blast furnace gas consumption at any given time. for The process water flow rate at any given time, for Instantaneous power consumption at any given moment After that Other input features for historical desulfurization data.

[0011] Optionally, the step of calculating the selection probability of each prediction model in the prediction model set includes: based on historical desulfurization data, obtaining the first outlet flue gas temperature, the second outlet flue gas temperature, the instantaneous outlet flue gas temperature, and the instantaneous inlet flue gas temperature of the flue gas heater on the production line when heating the flue gas, and calculating the average temperature measurement value; inputting the first outlet flue gas temperature, the second outlet flue gas temperature, the instantaneous outlet flue gas temperature, and the instantaneous inlet flue gas temperature into the prediction model set, and calculating the comprehensive deviation value between the predicted temperature value output by each prediction model and the average temperature measurement value; and calculating the selection probability of each prediction model based on the comprehensive deviation value.

[0012] Alternatively, the overall deviation value is calculated based on the following formula:

[0013] in, This is the overall deviation value. The root mean square error of the prediction model, The weighting coefficient for evaluating reaction temperature deviation. The weighting coefficient for evaluating SO2 concentration deviation. This is a measure of average temperature. To measure the inlet SO2 concentration, To predict the average inlet SO2 concentration for the model, the average temperature measure is calculated based on the following formula:

[0014] in, This is a measure of average temperature. The temperature of the first outlet flue gas. The second outlet flue gas temperature, The instantaneous temperature of the flue gas at the outlet. The instantaneous temperature of the inlet flue gas.

[0015] Optionally, the objective function is:

[0016]

[0017] in, Optimize the function for the objective. This is the amount of quicklime to input. To determine the amount of gas to be input into the blast furnace. This is the input industrial water consumption. Cost per unit of quicklime Cost per unit of blast furnace gas Cost per unit of industrial water The unit price of electricity This refers to the sulfur dioxide emission standard concentration value. It is a constant. All parameters are measured and known parameters of real-time desulfurization data, excluding the input quicklime quantity, blast furnace gas quantity, and industrial water quantity. The function is a piecewise function. The instantaneous power consumption predicted by the prediction model, This refers to the predicted sulfur dioxide concentration at the export site, as determined by the forecasting model.

[0018] Optionally, the desulfurization parameter optimization method further includes: calculating the optimal input parameters of the objective optimization function under preset constraints; wherein the constraints are: determining whether the calcium-sulfur ratio parameter of each calculated model input vector is within the pre-acquired calcium-sulfur ratio threshold range; if it is not within the calcium-sulfur ratio threshold range, the corresponding model input vector is removed; otherwise, it is retained.

[0019] Optionally, the calcium-to-sulfur ratio threshold range is:

[0020] in, This refers to the mass fraction of effective calcium element in the desulfurizing agent. This represents the molar mass of calcium. This refers to the dry flue gas volume under standard conditions. To achieve the target desulfurization efficiency, This is the lower limit of the calcium-to-sulfur ratio. This represents the upper limit of the calcium-to-sulfur ratio.

[0021] Optionally, the desulfurization parameter optimization method further includes: determining whether the optimal input parameters meet the preset convergence conditions; if not, repeating the step of iteratively calculating the fitness of the input population using the objective optimization function until the convergence conditions are met, and then obtaining the adjusted optimal input parameters.

[0022] Optionally, the convergence condition is that the current iteration number reaches the preset maximum iteration number.

[0023] Optionally, both historical and real-time desulfurization data include the SO2 concentration at the desulfurization inlet of the production line, the instantaneous SO2 concentration at the desulfurization outlet, the flow rate of combustion air, process water, and quicklime used for desulfurization, the oxygen concentration at the desulfurization outlet, the instantaneous temperature of the flue gas at the desulfurization inlet, the first outlet flue gas temperature at the flue gas heater outlet, the instantaneous outlet flue gas temperature at the desulfurization outlet, the outlet pressure of the desulfurization tower, the second outlet flue gas temperature of the desulfurization tower, the inlet pressure of the desulfurization tower, the inlet flue gas temperature of the desulfurization tower, the internal desulfurization pressure of the desulfurization tower, and the instantaneous power consumption during desulfurization.

[0024] In a second aspect, embodiments of the present invention provide a desulfurization parameter optimization device, comprising: a processor and a memory, wherein the memory stores instructions; the processor invokes the instructions in the memory to cause the processor to execute the desulfurization parameter optimization method of any of the foregoing embodiments of the first aspect of the present invention.

[0025] The processor of the desulfurization parameter optimization device provided in this embodiment of the invention executes the desulfurization parameter optimization method of any of the foregoing embodiments of the first aspect of the invention by calling instructions in the memory. In dynamic desulfurization scenarios, it can dynamically adjust the amount of materials used for desulfurization according to the actual desulfurization data on site, so as to avoid waste of materials used for desulfurization. While ensuring the desulfurization effect and pollutant emission compliance, it reduces the energy and material consumption costs of sintering CFB desulfurization, improves the economic benefits of enterprises, and achieves green cost reduction production.

[0026] Thirdly, embodiments of the present invention provide a computer-readable storage medium storing instructions that, when executed by a processor, implement the desulfurization parameter optimization method of any of the foregoing embodiments of the first aspect of the present invention.

[0027] The instructions stored in the computer-readable storage medium provided in the embodiments of the present invention can be called by a processor and executed by the desulfurization parameter optimization method of any of the foregoing embodiments of the first aspect of the present invention. This enables the dynamic adjustment of the amount of materials used for desulfurization based on the actual desulfurization data on site in dynamic desulfurization scenarios, so as to avoid waste of materials used for desulfurization. While ensuring the desulfurization effect and pollutant emission compliance, it reduces the energy and material consumption costs of sintering CFB desulfurization, improves the economic benefits of enterprises, and achieves green cost-reducing production.

[0028] Fourthly, embodiments of the present invention provide a computer program product, which includes a computer program that, when executed by a processor, implements the desulfurization parameter optimization method of any of the foregoing embodiments of the first aspect of the present invention.

[0029] When the computer program in the computer program product provided in the embodiments of the present invention is executed by the processor, it can implement the desulfurization parameter optimization method of any of the foregoing embodiments of the first aspect of the present invention. This enables the computer program product to dynamically adjust the amount of material used for desulfurization based on the actual desulfurization data on site in a dynamic desulfurization scenario, so as to avoid waste of materials used for desulfurization. While ensuring the desulfurization effect and pollutant emission compliance, it reduces the energy and material consumption costs of sintering CFB desulfurization, improves the economic benefits of enterprises, and achieves green cost-reducing production. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0031] Figure 1 This is a flowchart of one embodiment of the desulfurization parameter optimization method of the present invention; Figure 2 This is a flowchart of step S110 in one embodiment of the desulfurization parameter optimization method of the present invention; Figure 3 This is a flowchart of step S114 in one embodiment of the desulfurization parameter optimization method of the present invention; Figure 4 This is a flowchart of step S120 in one embodiment of the desulfurization parameter optimization method of the present invention; Figure 5 This is a flowchart illustrating the construction of a prediction model set and the updating of the first and second process times in one embodiment of the desulfurization parameter optimization method of the present invention. Figure 6 This is a flowchart illustrating the initialization algorithm population in one embodiment of the desulfurization parameter optimization method of the present invention. Figure 7 This is a flowchart illustrating the calculation of optimal input parameters using a target optimization function in one embodiment of the desulfurization parameter optimization method of the present invention. Figure 8 This is a general flowchart of an embodiment of the desulfurization parameter optimization method of the present invention; Figure 9 This is a structural block diagram of one embodiment of the desulfurization parameter optimization equipment of the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0033] It should be noted that all directional indications in the embodiments of the present invention, such as up, down, left, right, front, back, etc., are only used to explain the relative positional relationship and movement of the components in a specific posture as shown in the attached figure. If the specific posture changes, the directional indication will also change accordingly.

[0034] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0035] For ease of understanding, the desulfurization parameter optimization method of this invention will be described below, such as... Figure 1 As shown, the desulfurization parameter optimization method in this embodiment of the invention includes steps S110 to S150.

[0036] In step S110, historical desulfurization data of the production line is acquired, the historical desulfurization data is preprocessed, and a set of prediction models is constructed based on the preprocessed historical desulfurization data.

[0037] In this embodiment, historical desulfurization data includes the SO2 concentration at the desulfurization inlet and the instantaneous SO2 concentration at the desulfurization outlet of the production line's CEMS (Continuous Emission Monitoring System) over a period of time (e.g., more than two months), the flow rate of combustion air, process water, and quicklime used for desulfurization, the oxygen concentration at the desulfurization outlet, the instantaneous temperature of the inlet flue gas at the desulfurization inlet, the first outlet flue gas temperature at the outlet of the flue gas heater, the instantaneous outlet flue gas temperature at the desulfurization outlet, the outlet pressure of the desulfurization tower, the second outlet flue gas temperature of the desulfurization tower, the inlet pressure of the desulfurization tower, the inlet flue gas temperature of the desulfurization tower, the internal desulfurization pressure of the desulfurization tower, and the instantaneous power consumption during desulfurization.

[0038] After obtaining historical desulfurization data, the data underwent data cleaning and preprocessing. During the data preprocessing process, the instantaneous measurements of SO2 concentration in the desulfurization inlet flue gas of the CEMS, instantaneous measurements of SO2 concentration in the desulfurization outlet flue gas of the CEMS, combustion air flow rate, process water flow rate, desulfurization quicklime flow rate, desulfurization outlet SO2, instantaneous measurements of temperature of the CEMS inlet flue gas, temperature of the flue gas outlet of the flue gas heater, instantaneous measurements of temperature of the CEMS outlet flue gas, outlet pressure of the desulfurization tower, temperature of the desulfurization tower outlet flue gas, inlet pressure of the desulfurization tower, inlet flue gas temperature of the desulfurization tower, internal pressure of the desulfurization tower, and instantaneous power consumption were synchronized with a unified timestamp format to ensure the time correspondence of data at different times.

[0039] Next, the data was cleaned and missing value handling was performed. Differentiated strategies were adopted for missing values ​​based on their range. For small-range, scattered missing values ​​(such as single points or a small number of consecutive missing values), forward imputation or backward imputation methods were preferred. This fully utilized the continuity of the data in time or series or the correlation of adjacent observations, using the nearest known valid value before or after the missing point to fill the gap, thus ensuring the coherence of the local data.

[0040] For large, continuous missing data segments, simple imputation may lead to significant bias. Therefore, linear interpolation or imputation based on the mean of data from adjacent time points is used. For data with a certain trend, linear interpolation is used to generate smooth transition values ​​within the missing interval. For data with relatively stable fluctuations, imputation based on the mean of data from adjacent time points is used. This method estimates the missing value by calculating the average of data within the adjacent windows before and after the missing period, thereby reducing the impact of local data loss on the overall analysis.

[0041] After cleaning and pretreatment, outlier detection and processing are performed on historical desulfurization data to ensure data quality and robustness of analysis results.

[0042] In outlier detection, the 3σ criterion is applied: for data features that conform to or are approximately normally distributed, the mean μ and standard deviation σ are calculated to eliminate or correct all extreme data points that fall outside the interval [μ - 3σ, μ + 3σ], in order to identify outliers that significantly deviate from the central tendency of the data.

[0043] For detected outliers, based on the CFB desulfurization mechanism and data properties, strategies are adopted to either eliminate them or correct them using reasonable estimates (such as the median, upper and lower cutoff values).

[0044] After outlier detection and processing, the processed data is standardized. Specifically, Z-Score standardization is used to convert the original historical desulfurization data into a standard normal distribution with a mean of 0 and a standard deviation of 1. The formula is as follows: .

[0045] in, The mean of all data. The standard deviation represents the average value. This method is suitable for data with an approximately normal distribution or where normalization based on the average fluctuation (standard deviation) is required. It effectively handles data with outliers. Standardization ensures that all features are on comparable orders of magnitude, improving the efficiency and effectiveness of subsequent modeling.

[0046] Finally, a training set is established based on historical desulfurization data after cleaning and pretreatment to train the SO2 concentration prediction model at the desulfurization outlet. The specific steps for constructing the SO2 concentration prediction model are as follows.

[0047] like Figure 2 As shown, in some optional embodiments, the step of constructing the prediction model set in step S110 includes steps S111 to S116.

[0048] In step S111, the first process time of the flue gas generated by the production line from multiple desulfurization inlets to the reaction center and the second process time from the reaction center of the production line to the desulfurization outlet are obtained.

[0049] In step S112, based on the first process time, the second process time, and the pre-processed historical desulfurization data, the input vector and output result of the prediction model set corresponding to the first process time and the second process time are defined.

[0050] In step S113, the input vector and output result are divided into training set and test set.

[0051] In step S114, multiple prediction models are repeatedly trained based on the training set, and the root mean square error of each prediction model is calculated using the test set until the preset number of training iterations is reached.

[0052] In step S115, it is determined whether the root mean square error of each prediction model is greater than a preset error threshold. If the root mean square error is greater than or equal to the preset error threshold, the prediction model is removed. If the root mean square error is less than the preset error threshold, the prediction model is retained.

[0053] In step S116, a set of prediction models is constructed based on all the retained prediction models.

[0054] In this embodiment, the input vector is The output result is Where y is + + The instantaneous measured value of sulfur dioxide concentration at the desulfurization outlet at a given time. For instantaneous measurement time, For the first process time, For the second process time, For the prediction model to be trained, for The flow rate of quicklime at any given time. for Blast furnace gas consumption at any given time. for The process water flow rate at any given time, for Instantaneous power consumption at any given moment After that Other input features for historical desulfurization data.

[0055] First, initialize the first process time of the flue gas generated in the production line from the CEMS desulfurization inlet to the CEMS reaction center. And the second process time from the CEMS reaction center to the CEMS desulfurization outlet. .

[0056] The instantaneous measurement value of SO2 concentration in flue gas at the CEMS desulfurization inlet ( (Time), combustion air flow ( (Time), process water flow rate ( (Time), desulfurization quicklime flow rate ( (Time), instantaneous measurement of CEMS inlet flue gas temperature ( At any given moment, the instantaneous measurement value of oxygen content in the flue gas at the inlet of the CEMS environmental protection system ( At any given time, the outlet flue gas temperature of the flue gas heater ( (Time), instantaneous measurement of flue gas temperature at the CEMS outlet ( At any given time, the outlet flue gas temperature of the desulfurization tower ( (Time), blast furnace gas consumption (unit: Nm3 / h) (Time), process water flow rate ( (Time), desulfurization quicklime flow rate ( The instantaneous measured value of SO2 concentration in the flue gas at the CEMS desulfurization outlet is obtained by taking the time as input. (Time), instantaneous power consumption ( (Time) as output.

[0057] A training set is established based on historical desulfurization data after data cleaning and preprocessing. A prediction model for SO2 concentration at the CFB desulfurization outlet is trained, and the root mean square error is judged using a test set. When the root mean square error is less than a preset error threshold, the prediction model is updated to the model in the prediction model set.

[0058] Specifically, in this embodiment, the prediction model can be constructed using a BP neural network for relationship fitting: a detailed description of the neural network prediction process for SO2 concentration and power consumption after sintering desulfurization includes the following key formulas and steps: In network design, the number of inter-layer nodes n corresponds to the standardized features. Dimension. Input vector The number of nodes in each layer was determined through experimental optimization after two layers of hidden layer propagation. The hidden layers used the ReLU activation function.

[0059] To accelerate gradient updates and convergence, the output of the first hidden layer is calculated using the following formula: , This is the output of the hidden layer of layer 1. This is the weight vector of the first hidden layer. This is the bias of the first hidden layer. For the above network input.

[0060] The output of the second hidden layer is calculated using the following formula: , This is the output of the hidden layer of layer 2. This is the weight vector of the second hidden layer. The bias of the second hidden layer, This is the output of the first layer network.

[0061] The output layer is a two-node linear activation layer, outputting a combined vector of predicted SO2 concentration and instantaneous power consumption. : , The weight vector of the output layer. This represents the bias of the output layer. The training objective is achieved through the mean squared error loss function: ; Where N is the sample size. Output the target value. These are predicted values.

[0062] The training process is based on the backpropagation algorithm: forward propagation calculates the predicted values ​​layer by layer, and gradient backpropagation uses the chain rule to calculate the partial derivatives of the loss with respect to the parameters. Parameter updates use the Adam optimizer to adjust the weights. ; in, This is the weight matrix. For learning rate, Let be the error loss function. Overfitting is prevented through an early stopping mechanism, and hyperparameters are optimized using grid search.

[0063] Applying a branch-weighted improved loss function to high-concentration samples enhances the prediction accuracy of high-risk points. The model evaluation uses the root mean square error (RMSE) metric. ; Predictive models that meet the root mean square error (RMSE) requirement are selected and retained (for example, the threshold for the RMSE requirement can be 0.1 or 0.05). Based on all the retained predictive models, a set of predictive models is constructed.

[0064] Furthermore, in step S114, when repeatedly training multiple prediction models based on the training set, the optimized convergence time from the start of training to model convergence is first calculated based on the first prediction model obtained from training, and then the first process time and the second process time are updated based on the optimized convergence time.

[0065] like Figure 3 As shown, specifically, step S114 includes steps S1141 to S1143.

[0066] In step S1141, after the first prediction model is trained, the optimized convergence time of the first prediction model from the start of training to model convergence is calculated.

[0067] In step S1142, if the second process time is less than the optimized convergence time, the second process time and the corresponding first process time are updated so that the updated second process time is greater than or equal to the optimized convergence time.

[0068] In step S1143, based on the updated first process time and second process time, steps S112 to S116 are re-executed to construct a set of prediction models.

[0069] like Figure 5 As shown in this embodiment, during the model training process, based on the first prediction model constructed initially, the optimized convergence time from the start of training to model convergence is calculated. .

[0070] Based on the optimized convergence time Determine the time of the second process used to build the subsequent prediction model. Is it less than the optimization convergence time? If the second process time Less than the optimal convergence time Then update the second process time. and the corresponding first process time Until the second process time Greater than or equal to the optimization convergence time .

[0071] Choose again , Then, train the model set and calculate... The threshold, based on the initial model calculation, is statistically determined by... Optimize the objective for the target: ; The convergence time of the optimization algorithm based on the objective of calculating SO2 concentration to meet environmental protection requirements is optimized. Select the one that satisfies And it conforms to the desulfurization reaction law , The prediction model is then retrained using the above process, and the model set is continuously updated until the model network set is completed after a preset time period.

[0072] During the training of the prediction model, the input is recorded. This represents the input vector corresponding to each of the above input features. ,in Indicates the flow rate of quicklime ( time), Indicates blast furnace gas consumption ( time), Indicates process water flow rate ( time), Instantaneous power consumption ( (Time), the rest represent other input features respectively.

[0073]

[0074] in, This represents the relationship between the trained prediction models. Input as features The output result at that time.

[0075]

[0076] That is, the power consumption corresponding to this input. That is, the instantaneous measurement value of SO2 concentration in the flue gas at the CEMS desulfurization outlet corresponding to this input. time).

[0077] For models where the error RMSE is below the RMSE threshold, calculate the sample average temperature. and the average inlet SO2 concentration of the sample Add model information , Indicates time taking , The relationship between the prediction model and time. This represents the root mean square error (RMSE) of the model.

[0078] Repeat the above training steps to construct multiple prediction models for predicting SO2 concentration at the CFB desulfurization outlet and power consumption of the desulfurization process. Select and retain prediction models that meet the root mean square error requirement. Based on all retained prediction models, construct a prediction model set.

[0079] In step S120, real-time desulfurization data of the production line is acquired, and the selection probability of each prediction model in the prediction model set is calculated based on the real-time desulfurization data to generate a model probability distribution.

[0080] The real-time desulfurization data includes the instantaneous SO2 concentration at the desulfurization inlet and outlet of the production line, the flow rate of combustion air, process water, and quicklime used for desulfurization, the oxygen concentration at the desulfurization outlet, the instantaneous temperature of the flue gas at the desulfurization inlet, the first outlet flue gas temperature at the flue gas heater outlet, the instantaneous outlet flue gas temperature at the desulfurization outlet, the outlet pressure of the desulfurization tower, the second outlet flue gas temperature of the desulfurization tower, the inlet pressure of the desulfurization tower, the inlet flue gas temperature of the desulfurization tower, the internal desulfurization pressure of the desulfurization tower, and the instantaneous power consumption during desulfurization.

[0081] In this embodiment, as Figure 6 As shown, after the SO2 prediction model set is trained, the monitoring and control optimization decision calculation for pollutants begins, and real-time desulfurization data is collected: based on the SO2 concentration at the desulfurization inlet ( (Time), combustion air flow rate used for desulfurization ( (Time), process water flow rate ( (Time), desulfurization quicklime flow rate ( At any given moment, the instantaneous temperature of the flue gas at the desulfurization inlet ( (moment), instantaneous measurement value of oxygen content in flue gas at the CEMS inlet of the environmental protection system ( At any given moment, the first outlet flue gas temperature at the flue gas heater outlet ( At any given moment, the instantaneous temperature of the flue gas exiting the CEMS outlet ( At any given time, the outlet flue gas temperature of the desulfurization tower ( (Time), blast furnace gas consumption ( (Time), process water flow rate ( (Time), desulfurization quicklime flow rate ( Data such as time (or time) are used as input.

[0082] Instantaneous measurement of SO2 concentration in the flue gas from the desulfurization outlet of CEMS ( (Time), instantaneous power consumption ( (Time) as output. like Figure 4 As shown, further, the step of calculating the selection probability of each prediction model in the prediction model set in step S120 includes steps S121 to S123.

[0083] In step S121, based on historical desulfurization data, the first outlet flue gas temperature of the flue gas heater on the production line when heating the flue gas, the second outlet flue gas temperature of the desulfurization tower, the instantaneous outlet flue gas temperature, and the instantaneous inlet flue gas temperature are obtained, and the average temperature measurement value is calculated.

[0084] In step S122, the first outlet flue gas temperature, the second outlet flue gas temperature, the instantaneous outlet flue gas temperature, and the instantaneous inlet flue gas temperature are input into the prediction model set, and the comprehensive deviation between the predicted temperature value output by each prediction model and the average temperature measurement value is calculated.

[0085] In step S123, the probability of each prediction model being selected is calculated based on the comprehensive deviation value.

[0086] like Figure 6 As shown, specifically, the set of SO2 prediction models is traversed and calculated. The combined deviations of each model are used to determine the probability distribution of model selection. The combined deviation value is calculated based on the following formula:

[0087] in, This is the overall deviation value. The root mean square error of the prediction model, The weighting coefficient for evaluating reaction temperature deviation. The weighting coefficient for evaluating SO2 concentration deviation. This is a measure of average temperature. To measure the inlet SO2 concentration, To predict the average inlet SO2 concentration for the model, the average temperature measure is calculated based on the following formula:

[0088] in, This is a measure of average temperature. The temperature of the first outlet flue gas. The second outlet flue gas temperature, The instantaneous temperature of the flue gas at the outlet. The instantaneous temperature of the inlet flue gas.

[0089] After calculating the comprehensive deviation values, the selection probability of each prediction model is then calculated. For the i-th model, its selection probability is... for: ; in, This represents the overall deviation value of the i-th model. represents the overall deviation value of the j-th model, and m represents the number of models in the prediction model set.

[0090] In step S130, based on the model probability distribution, a prediction model for calculating the target optimization function of the production line is selected from the set of prediction models.

[0091] To ensure the generalization rationality of the overall algorithm, this method randomly selects a prediction model based on the probability distribution formed by the selection probabilities of each model, and then performs subsequent optimization analysis and calculation based on the selected model.

[0092] In this embodiment, the SO2 concentration at the desulfurization inlet ( (Time), combustion air flow rate used for desulfurization ( (Time), process water flow rate ( (Time), desulfurization quicklime flow rate ( At any given moment, the instantaneous temperature of the flue gas at the desulfurization inlet ( (moment), instantaneous measurement value of oxygen content in flue gas at the CEMS inlet of the environmental protection system ( At any given moment, the first outlet flue gas temperature at the flue gas heater outlet ( At any given moment, the instantaneous temperature of the flue gas exiting the CEMS outlet ( At any given time, the outlet flue gas temperature of the desulfurization tower ( (Time), blast furnace gas consumption ( (Time), process water flow rate ( (Time), desulfurization quicklime flow rate ( Data such as time (or time) are used as input.

[0093] Network relationships of the selected prediction model: ,according to , Select measured data and record the current time. Take the instantaneous measurement value of SO2 concentration in the flue gas at the CEMS desulfurization inlet. (Time), combustion air flow ( (Time), process water flow rate ( (Time), desulfurization quicklime flow rate ( (Time), instantaneous measurement of flue gas temperature at the CEMS inlet ( At any given moment, the instantaneous measurement value of oxygen content in the flue gas at the CEMS environmental protection system inlet ( At any given moment, the first outlet flue gas temperature of the flue gas heater ( At any given moment, the instantaneous temperature of the flue gas exiting the CEMS outlet ( At any given time, the outlet flue gas temperature of the desulfurization tower ( time).

[0094] Record the flow rate of desulfurization quicklime ( (Time) is Blast furnace gas consumption (unit: Nm3 / h) (Time) is The process water flow rate is Construct an optimization algorithm input individual and according to Input to construct the corresponding pollutant prediction model network For any have In the selected prediction model This indicates that the characteristics are consistent, and at this state time, These are measured, known values. Let them be... .

[0095] Therefore, the objective optimization function is:

[0096]

[0097] in, Optimize the function for the objective. This is the amount of quicklime to input. To determine the amount of gas to be input into the blast furnace. This is the input industrial water consumption. Cost per unit of quicklime Cost per unit of blast furnace gas Cost per unit of industrial water The unit price of electricity This refers to the sulfur dioxide emission standard concentration value. It is a constant. All parameters are measured and known parameters of real-time desulfurization data, excluding the input quicklime quantity, blast furnace gas quantity, and industrial water quantity. The function is a piecewise function. The instantaneous power consumption predicted by the prediction model, This refers to the predicted sulfur dioxide concentration at the export site, as determined by the forecasting model.

[0098] Furthermore, based on the aforementioned objective optimization function, and under preset constraints, the optimal input parameters of the objective optimization function are calculated.

[0099] The constraint is as follows: determine whether the calcium-sulfur ratio parameter of each model input vector is within the pre-acquired calcium-sulfur ratio threshold range. If it is not within the calcium-sulfur ratio threshold range, the corresponding model input vector is removed; otherwise, it is retained.

[0100] Specifically, the calcium-to-sulfur ratio threshold range is:

[0101] in, This refers to the mass fraction of effective calcium element in the desulfurizing agent. This represents the molar mass of calcium. This refers to the dry flue gas volume under standard conditions. To achieve the target desulfurization efficiency, This is the lower limit of the calcium-to-sulfur ratio. This represents the upper limit of the calcium-to-sulfur ratio.

[0102] The constraint is the SO2 concentration of pollutants that meets the emission standards. The optimization objective of this application is to achieve the optimal comprehensive cost that minimizes energy and material consumption while meeting the constraints. The constraints are converted into soft constraints to construct the objective function.

[0103] In step S140, based on real-time desulfurization data and the selected prediction model, an input population corresponding to the objective optimization function is constructed to iteratively calculate the fitness of the input population.

[0104] In this embodiment, as Figure 7 One feasible algorithm for the optimization calculation shown is the particle swarm optimization algorithm, and the algorithm flow is as follows: randomly select Individual composition in relation to the entire population, the position of each individual .

[0105] The optimal position of particle i at step k. This represents the historical position of particle i that best satisfies the fitness function up to this iteration step. This represents the historical position of the most satisfied fitness function that all individuals in the entire population have experienced up to this step.

[0106] To avoid getting trapped in local optima and to maximize the algorithm's convergence speed while ensuring search capability, this application employs a particle swarm optimization algorithm with a compression factor, and dynamically adjusts the learning factor. The speed update is as follows: ; ; ; in, The inertia coefficient, , r1 and r2 are learning factors, and are random numbers between [0,1].

[0107] Furthermore, the compression factor can control the convergence of the particle swarm optimization algorithm, giving particles the opportunity to search different regions in the space, improving the algorithm's fitness, obtaining high-quality particles, and greatly improving the convergence speed and accuracy of the particle swarm optimization algorithm. This represents the individual cognitive coefficient, which relates to the algorithm's ability to escape local limitations and perform global search. The social cognition coefficient is related to the convergence speed of the algorithm. It draws inspiration from simulated annealing and is dynamically adjusted. , : ; ; in, , for , The adjustment factor, where k is the current iteration number. This represents the total number of iterations.

[0108] The fitness of the particle changes dynamically with the number of iterations, effectively improving its global search capability and enhancing particle convergence. Through population updates and iterations, the fitness is gradually optimized, and the optimal particle is selected after reaching convergence or the upper limit of the number of iterations. , , recorded as , .

[0109] Through the above steps, the fitness of the input population is iteratively calculated based on the objective optimization function.

[0110] In some optional embodiments, after calculating the optimal input parameters, it is determined whether the optimal input parameters meet the preset convergence condition. If not, the step of iteratively calculating the fitness of the input population with the objective optimization function is repeated until the convergence condition is met, and then the adjusted optimal input parameters are obtained.

[0111] The convergence condition is that the current iteration number reaches the preset maximum iteration number.

[0112] In step S150, based on the fitness of the input population, the optimal input parameters of the objective optimization function are selected to determine the material usage of the production line, and the material parameters input to the production line are adjusted according to the calculated material usage of the production line.

[0113] In this embodiment, the parameters of the corresponding feeding equipment are adjusted according to the determined material consumption of the production line, such as the demand for quicklime, blast furnace gas, and industrial water.

[0114] For example, based on the characteristics set for different quicklime feeders, the required quicklime consumption can be converted into the corresponding amount of material that the quicklime feeder needs to convey. The opening of the blast furnace gas valve can be adjusted according to the blast furnace gas demand, and the opening of the industrial water valve can be adjusted according to the industrial water demand. Specifically, DCS / PLC control can be combined to achieve control of the quicklime feeder and valves, forming a closed-loop control system.

[0115] like Figure 8 As shown in Figure 8, this application uses the method steps described in Figure 8, specifically using Python to write the algorithm service, Vue as the front-end framework, and .NET architecture as the back-end business framework. Taking the production and flue gas treatment data of a steel plant as an example, the overall method application is verified. The specific steps are as follows: Deploy a data acquisition system to collect PLC data and instrument data such as on-site sintering machine production and flue gas treatment through an IoT platform or gateway. The data includes all the features required by the prediction model in the method of the above embodiments of this application.

[0116] Historical desulfurization data is collected, cleaned, and preprocessed to form training samples for the SO2 prediction model. The model is then trained using the appropriate training method, forming a model set. The algorithm service interfaces with the ERP system to obtain quicklime cost and blast furnace gas cost data, determining the comprehensive energy-material consumption optimization objective function. Real-time data collected by the acquisition system is used to select from the model set, and the optimization model calculates the optimal energy-material consumption cost for quicklime and blast furnace gas usage to meet SO2 emission requirements. Through conversion relationships, based on the characteristics of different quicklime feeders, the quicklime material consumption demand is converted into quicklime feeder settings, providing a basis for quicklime feeder control. Blast furnace gas demand is converted into blast furnace gas valve opening, providing a basis for blast furnace gas valve control; similarly, industrial water demand is converted into industrial water valve opening, providing a basis for industrial water valve control. PID closed-loop control (setpoint = optimized output value) is employed, with feedforward compensation for fluctuations in flue gas oxygen content.

[0117] This method ensures the desulfurization effect of steel plant flue gas and meets pollutant emission standards, while effectively saving the comprehensive cost of energy and material consumption in sintering CFB desulfurization, thereby generating significant economic benefits and realizing green cost reduction production for steel enterprises.

[0118] The desulfurization parameter optimization method provided in this embodiment of the invention includes: acquiring historical desulfurization data of the production line, preprocessing the historical desulfurization data, and constructing a set of prediction models based on the preprocessed historical desulfurization data; acquiring real-time desulfurization data of the production line, calculating the selection probability of each prediction model in the prediction model set based on the real-time desulfurization data to generate a model probability distribution; selecting a prediction model from the prediction model set to calculate the target optimization function of the production line based on the model probability distribution; constructing an input population corresponding to the target optimization function based on the real-time desulfurization data and the selected prediction model, and iteratively calculating the fitness of the input population; selecting the optimal input parameters of the target optimization function based on the fitness of the input population to determine the material consumption of the production line, and adjusting the material parameters input to the production line based on the calculated material consumption of the production line.

[0119] The desulfurization parameter optimization method provided in this invention can dynamically adjust the amount of materials used for desulfurization based on actual desulfurization data in a dynamic desulfurization scenario, so as to avoid waste of materials used for desulfurization. While ensuring the desulfurization effect and pollutant emission compliance, it reduces the energy and material consumption costs of sintering CFB desulfurization, improves the economic benefits of enterprises, and achieves green cost-reducing production.

[0120] In addition to the above method embodiments, the present invention also provides, for example, Figure 9 The desulfurization parameter optimization device shown includes a processor 201 and a memory 202, wherein the memory 202 stores instructions.

[0121] The processor 201 can call instructions in the memory 202 to execute the desulfurization parameter optimization method of any of the above embodiments of the present invention.

[0122] The desulfurization parameter optimization method provided in the above embodiments of the present invention includes: acquiring historical desulfurization data of the production line, preprocessing the historical desulfurization data, and constructing a set of prediction models based on the preprocessed historical desulfurization data; acquiring real-time desulfurization data of the production line, calculating the selection probability of each prediction model in the prediction model set based on the real-time desulfurization data to generate a model probability distribution; selecting a prediction model from the prediction model set to calculate the target optimization function of the production line based on the model probability distribution; constructing an input population corresponding to the target optimization function based on the real-time desulfurization data and the selected prediction model, and iteratively calculating the fitness of the input population; selecting the optimal input parameters of the target optimization function based on the fitness of the input population to determine the material consumption of the production line, and adjusting the material parameters input to the production line based on the calculated material consumption of the production line.

[0123] The desulfurization parameter optimization equipment provided in this embodiment of the invention can dynamically adjust the amount of materials used for desulfurization based on the actual desulfurization data on site in dynamic desulfurization scenarios, thereby avoiding waste of materials used for desulfurization. While ensuring the desulfurization effect and pollutant emission compliance, it reduces the energy and material consumption costs of sintering CFB desulfurization, improves the economic benefits of enterprises, and achieves green cost-reducing production.

[0124] Furthermore, the desulfurization parameter optimization device provided in this embodiment of the invention may also include a communication interface 203 and a bus 204, with the processor 201, memory 202 and communication interface 203 electrically connected via the bus 204.

[0125] The memory 202 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 203 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 204 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 9 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0126] Processor 201 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 201 or by instructions in software form. The processor 201 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 202. The processor 201 reads the information in memory 202 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0127] This invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the above-described desulfurization parameter optimization method.

[0128] The computer-readable storage medium provided in this embodiment of the invention stores data and computer-executable instructions for the above-described desulfurization parameter optimization method. The desulfurization parameter optimization method includes: acquiring historical desulfurization data from the production line; preprocessing the historical desulfurization data; and constructing a set of prediction models based on the preprocessed historical desulfurization data; acquiring real-time desulfurization data from the production line; calculating the selection probability of each prediction model in the prediction model set based on the real-time desulfurization data to generate a model probability distribution; selecting a prediction model from the prediction model set to calculate the target optimization function of the production line based on the model probability distribution; constructing an input population corresponding to the target optimization function based on the real-time desulfurization data and the selected prediction model; iteratively calculating the fitness of the input population; and selecting the optimal input parameters of the target optimization function based on the fitness of the input population to determine the material usage of the production line, and adjusting the material parameters input to the production line based on the calculated material usage of the production line.

[0129] The computer-readable storage medium provided in this embodiment of the invention, by implementing the above method, can dynamically adjust the amount of materials used for desulfurization based on the actual desulfurization data on site in a dynamic desulfurization scenario, so as to avoid waste of materials used for desulfurization. While ensuring the desulfurization effect and pollutant emission compliance, it reduces the energy and material consumption costs of sintering CFB desulfurization, improves the economic benefits of enterprises, and achieves green cost-reducing production.

[0130] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: Step S110: Obtain historical desulfurization data from the production line, preprocess the historical desulfurization data, and construct a set of prediction models based on the preprocessed historical desulfurization data.

[0131] Step S120: Obtain real-time desulfurization data from the production line, and calculate the selection probability of each prediction model in the prediction model set based on the real-time desulfurization data to generate a model probability distribution.

[0132] Step S130: Based on the model probability distribution, select a prediction model from the set of prediction models to calculate the objective optimization function of the production line.

[0133] Step S140: Based on real-time desulfurization data and the selected prediction model, construct an input population corresponding to the objective optimization function, and iteratively calculate the fitness of the input population.

[0134] The computer program product provided in this invention, by implementing the above method, can dynamically adjust the amount of materials used for desulfurization based on actual desulfurization data in a dynamic desulfurization scenario, so as to avoid waste of materials used for desulfurization. While ensuring the desulfurization effect and pollutant emission compliance, it reduces the energy and material consumption costs of sintering CFB desulfurization, improves the economic benefits of enterprises, and achieves green cost-reducing production.

[0135] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0136] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, 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 cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0137] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing desulfurization parameters, characterized in that, include: Historical desulfurization data of the production line is acquired, the historical desulfurization data is preprocessed, and a set of prediction models is constructed based on the preprocessed historical desulfurization data. Acquire real-time desulfurization data from the production line, and calculate the selection probability of each prediction model in the prediction model set based on the real-time desulfurization data to generate a model probability distribution. Based on the model probability distribution, the prediction model used to calculate the target optimization function of the production line is selected from the set of prediction models; Based on the real-time desulfurization data and the selected prediction model, an input population corresponding to the objective optimization function is constructed to iteratively calculate the fitness of the input population. Based on the fitness of the input population, the optimal input parameters of the objective optimization function are selected to determine the material usage of the production line, and the material parameters input to the production line are adjusted according to the calculated material usage of the production line.

2. The desulfurization parameter optimization method according to claim 1, characterized in that, The set of prediction models is constructed based on the following steps: S1: Obtain the first process time of the flue gas generated by the production line from multiple desulfurization inlets to the reaction center and the second process time from the reaction center of the production line to the desulfurization outlet; S2: Based on the first process time, the second process time, and the preprocessed historical desulfurization data, define the input vector and output result of the prediction model set corresponding to the first process time and the second process time; S3: Divide the input vector and the output result into a training set and a test set; S4: Repeatedly train multiple prediction models based on the training set, and calculate the root mean square error of each prediction model using the test set, until a preset number of training iterations is reached; S5: Determine whether the root mean square error of each prediction model is greater than a preset error threshold. If the root mean square error is greater than or equal to the preset error threshold, then the prediction model is removed. If the root mean square error is less than the preset error threshold, then the prediction model is retained. S6: Based on all the retained prediction models, construct the prediction model set.

3. The desulfurization parameter optimization method according to claim 2, characterized in that, The method further includes: After the first prediction model is obtained through training, the optimized convergence time of the first prediction model from the start of training to model convergence is calculated. If the second process time is less than the optimized convergence time, then update the second process time and the corresponding first process time, so that the updated second process time is greater than or equal to the optimized convergence time; Based on the updated first process time and second process time, steps S2 to S6 are re-executed to construct the prediction model set.

4. The desulfurization parameter optimization method according to claim 2, characterized in that, The input vector is The output result is ; Where y is + + The instantaneous measured value of sulfur dioxide concentration at the desulfurization outlet at a given time. For instantaneous measurement time, The time for the first process. The time for the second process, The prediction model to be trained. for The flow rate of quicklime at any given time. for Blast furnace gas consumption at any given time. for The process water flow rate at any given time, for Instantaneous power consumption at any given moment After that Other input features for the historical desulfurization data.

5. The desulfurization parameter optimization method according to claim 1, characterized in that, The steps for calculating the selection probability of each prediction model in the set of prediction models include: Based on the historical desulfurization data, the first outlet flue gas temperature, the second outlet flue gas temperature of the desulfurization tower, the instantaneous outlet flue gas temperature, and the instantaneous inlet flue gas temperature of the flue gas heater on the production line are obtained, and the average temperature measurement value is calculated. The first outlet flue gas temperature, the second outlet flue gas temperature, the instantaneous outlet flue gas temperature, and the instantaneous inlet flue gas temperature are input into the prediction model set, and the comprehensive deviation value between the predicted temperature value output by each prediction model and the average temperature measurement value is calculated. Based on the comprehensive deviation value, the probability of selection for each prediction model is calculated.

6. The desulfurization parameter optimization method according to claim 5, characterized in that, The overall deviation value is calculated based on the following formula: in, The comprehensive deviation value is... The root mean square error of the prediction model is... The weighting coefficient for evaluating reaction temperature deviation. The weighting coefficient for evaluating SO2 concentration deviation. The average temperature measurement value. To measure the inlet SO2 concentration, The average inlet SO2 concentration for the prediction model is calculated based on the following formula: in, The average temperature measurement value. The temperature of the first outlet flue gas. The second outlet flue gas temperature, The instantaneous temperature of the outlet flue gas, The instantaneous temperature of the inlet flue gas is denoted as .

7. The desulfurization parameter optimization method according to claim 1, characterized in that, The objective optimization function is: in, Let the objective optimization function be... This is the amount of quicklime to input. To determine the amount of gas to be input into the blast furnace. This is the input industrial water consumption. Cost per unit of quicklime Cost per unit of blast furnace gas Cost per unit of industrial water The unit price of electricity This refers to the sulfur dioxide emission standard concentration value. It is a constant. All parameters are measured and known parameters of the real-time desulfurization data, excluding the input quicklime quantity, blast furnace gas quantity, and industrial water quantity. The function is a piecewise function. The instantaneous power consumption predicted by the prediction model. The predicted sulfur dioxide concentration at the export site is given by the prediction model.

8. The desulfurization parameter optimization method according to claim 7, characterized in that, The method further includes: Under preset constraints, calculate the optimal input parameters of the objective optimization function; The constraints are as follows: Determine whether the calcium-sulfur ratio parameter of each model input vector is within the pre-acquired calcium-sulfur ratio threshold range. If it is not within the calcium-sulfur ratio threshold range, then the corresponding model input vector is discarded; otherwise, it is retained.

9. The desulfurization parameter optimization method according to claim 8, characterized in that, The calcium-to-sulfur ratio threshold range is: in, This refers to the mass fraction of effective calcium element in the desulfurizing agent. This represents the molar mass of calcium. This refers to the dry flue gas volume under standard conditions. To achieve the target desulfurization efficiency, This is the lower limit of the calcium-to-sulfur ratio. This represents the upper limit of the calcium-to-sulfur ratio.

10. The desulfurization parameter optimization method according to claim 1, characterized in that, The method further includes: Determine whether the optimal input parameters meet the preset convergence condition. If not, repeat the step of iteratively calculating the fitness of the input population using the objective optimization function until the convergence condition is met, and then obtain the adjusted optimal input parameters.

11. The desulfurization parameter optimization method according to claim 10, characterized in that, The convergence condition is that the current iteration number reaches the preset maximum iteration number.

12. The desulfurization parameter optimization method according to claim 1, characterized in that, Both the historical desulfurization data and the real-time desulfurization data include the SO2 concentration at the desulfurization inlet of the production line, the instantaneous SO2 concentration at the desulfurization outlet, the flow rate of combustion air, process water, and quicklime used for desulfurization, the oxygen concentration at the desulfurization outlet, the instantaneous temperature of the flue gas at the desulfurization inlet, the first outlet flue gas temperature at the outlet of the flue gas heater, the instantaneous outlet flue gas temperature at the desulfurization outlet, the outlet pressure of the desulfurization tower, the second outlet flue gas temperature of the desulfurization tower, the inlet pressure of the desulfurization tower, the inlet flue gas temperature of the desulfurization tower, the internal desulfurization pressure of the desulfurization tower, and the instantaneous power consumption during desulfurization.

13. A desulfurization parameter optimization device, characterized in that, The desulfurization parameter optimization device includes a processor and a memory, wherein the memory stores instructions; The processor invokes the instructions in the memory to cause the desulfurization parameter optimization device to implement the desulfurization parameter optimization method as described in any one of claims 1 to 12.

14. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the desulfurization parameter optimization method as described in any one of claims 1 to 12.

15. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the desulfurization parameter optimization method as described in any one of claims 1 to 12.

Images