Intelligent carbon potential control method and system based on multi-source data perception

Abstract

This invention relates to an intelligent carbon potential control method and system based on multi-source data sensing. The method comprehensively acquires multi-source information related to carbon potential by collecting parameter vectors corresponding to M parameter types. These parameter vectors are then rationally concatenated and processed. Through a trained time-series offset prediction model, offset mapping model, and parameter prediction model, the method can accurately predict the carbon potential at a preset target time point. Compared to traditional prediction methods relying on a single or few parameters, this method fully explores the potential relationships and time-series characteristics between different parameters, greatly improving the accuracy of carbon potential prediction and providing a reliable basis for subsequent precise control. The multi-source data sensing approach can comprehensively reflect the complex and ever-changing working conditions in actual production. The model can learn and adapt to the relationship between parameters and carbon potential under different working conditions, exhibiting strong robustness and adaptability, thus broadening the application scope of this method in different production environments.

Description

Technical Field

[0001] This invention relates to the field of carbon potential control technology, specifically to an intelligent carbon potential control method and system based on multi-source data sensing. Background Technology

[0002] With the development of heat treatment technology, carbon potential control plays a crucial role in ensuring product quality and performance. Traditional carbon potential control methods often rely on a single data source or a simple mathematical model for carbon potential regulation. For example, some methods indirectly infer carbon potential based solely on furnace temperature data. However, in actual production environments, numerous factors influence carbon potential, such as gas flow rate, composition ratio, and workpiece material properties. A single data source cannot comprehensively and accurately reflect the dynamic changes in carbon potential.

[0003] Some methods, while considering multiple parameters, fail to fully explore the inherent relationships and temporal characteristics between these parameters. When faced with complex and variable operating conditions, these traditional methods cannot predict carbon potential in a timely and accurate manner, resulting in poor carbon potential control precision and a tendency for problems such as insufficient or excessive carburizing. Insufficient carburizing will cause the workpiece surface hardness and wear resistance to fail to meet requirements, affecting product lifespan; excessive carburizing may lead to workpiece deformation and increased brittleness, similarly reducing product quality. Furthermore, due to the inability to achieve precise control, more energy and raw materials are often required to compensate for control errors, increasing production costs.

[0004] Therefore, improving the accuracy of carbon potential control has become an urgent problem to be solved. Summary of the Invention

[0005] To achieve the objectives of this invention, the technical solution adopted is as follows: an intelligent carbon potential control method based on multi-source data sensing, which includes the following steps:

[0006] S101, within a preset time period, collect parameter vectors corresponding to M parameter types respectively, where M is a positive integer;

[0007] S102, use the parameter vector with the parameter type of the first preset type as the base vector, and other parameter vectors as reference vectors;

[0008] S103, the initial parameter matrix is ​​obtained by concatenating the basic vector and each reference vector;

[0009] S104, Input the initial parameter matrix into the trained temporal offset prediction model to obtain the temporal offset corresponding to each reference vector;

[0010] S105, input the initial parameter matrix and the temporal offsets corresponding to each reference vector into the trained offset mapping model to obtain the intermediate parameter matrix;

[0011] S106, Adjust the intermediate parameter matrix according to the preset mapping function to obtain the target parameter matrix;

[0012] S107, extract the target sub-matrix from the target parameter matrix, input the target sub-matrix into the trained parameter prediction model, and obtain the target prediction vector corresponding to the preset target time point;

[0013] S108, Based on the target prediction vector, determine the predicted carbon potential corresponding to the target time point;

[0014] S109, if the predicted carbon potential at the target time point is less than the preset first carbon potential threshold, then execute the first carbon potential control process; if the predicted carbon potential at the target time point is greater than the preset second carbon potential threshold, then execute the second carbon potential control process.

[0015] Compared with the prior art, the beneficial effects of the present invention are:

[0016] 1. By collecting parameter vectors corresponding to M parameter types, comprehensive multi-source information related to carbon potential is obtained. These parameter vectors are reasonably spliced ​​and processed, and through trained time series offset prediction models, offset mapping models, and parameter prediction models, the carbon potential corresponding to the preset target time point can be accurately predicted. Compared with traditional prediction methods that rely on a single or a small number of parameters, this fully explores the potential relationships and time series characteristics between different parameters, greatly improving the accuracy of carbon potential prediction and providing a reliable basis for subsequent precise control.

[0017] 2. The multi-source data sensing method can comprehensively reflect the complex and ever-changing working conditions in actual production. The model can learn and adapt to the relationship between parameters and carbon potential under different working conditions. Whether under complex conditions such as equipment aging, environmental changes or fluctuations in raw material characteristics, it can stably and accurately predict and control carbon potential, and has strong robustness and adaptability, thus broadening the application scope of this method in different production environments. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an intelligent carbon potential control method based on multi-source data sensing in Embodiment 1 of the present invention.

[0019] Figure 2 This is a schematic diagram of the structure of an intelligent carbon potential control system based on multi-source data sensing in Embodiment 2 of the present invention. Detailed Implementation

[0020] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1:

[0022] like Figure 1 As shown, the present invention provides a technical solution: an intelligent carbon potential control method based on multi-source data sensing, characterized in that the intelligent carbon potential control method based on multi-source data sensing includes the following steps:

[0023] S101, within a preset time period, collect parameter vectors corresponding to M parameter types respectively.

[0024] Where M is a positive integer, the preset time period can include several preset time points, the time interval between adjacent preset time points is fixed at T, and a single parameter vector can include the parameter values ​​collected by the corresponding parameter type sensor at each preset time point.

[0025] S102, use the parameter vector with the parameter type of the first preset type as the base vector, and other parameter vectors as reference vectors.

[0026] The first preset type can refer to the parameter type that is directly related to the carbon potential.

[0027] Specifically, in this embodiment, there is one base vector and M-1 reference vectors.

[0028] In one specific implementation, the parameter types include at least oxygen probe signal parameters, carbon monoxide concentration signal parameters, furnace temperature signal parameters, and furnace pressure parameters;

[0029] The first preset type is oxygen probe signal parameters.

[0030] The parameter types can also include enrichment gas flow rate, furnace circulating fan current, workpiece surface infrared temperature measurement, etc. Implementers can set the parameter types according to actual needs.

[0031] The oxygen probe signal parameters are the types of parameters that are directly related to the carbon potential.

[0032] S103 is the initial parameter matrix obtained by concatenating the basic vector and each reference vector.

[0033] The initial parameter matrix is ​​obtained by concatenating the basic vector as the first row vector with other reference vectors row by row.

[0034] S104. Input the initial parameter matrix into the trained time series offset prediction model to obtain the time series offset corresponding to each reference vector.

[0035] The temporal offset is an integer by default. The architecture of the temporal offset prediction model can use convolutional modules and fully connected layers. The output of the convolutional module is connected to the input of the fully connected layer. The convolutional module is used to extract the feature information of the initial parameter matrix of the input. The fully connected layer is used to map the feature information extracted by the convolutional module into the temporal offset corresponding to each reference vector.

[0036] It should be noted that in this embodiment, the output dimension of the fully connected layer is set to be the same as the number of columns of the initial parameter matrix in order to better correspond the timing offset to the parameter type. Since the parameter vector of the first preset type is a base vector and no offset processing is performed, the timing offset corresponding to the parameter vector of the first preset type is zero by default. In this step, only the timing offsets corresponding to the other reference vectors are described.

[0037] In one specific implementation, the training process of the time-series offset prediction model includes the following steps:

[0038] Obtain the historical vectors corresponding to M parameter types respectively, and use the historical vector with parameter type of the first preset type as the first vector, and use the other historical vectors as the second vector;

[0039] The parameter sample matrix is ​​obtained by concatenating the historical vectors corresponding to the M parameter types.

[0040] For any second vector, obtain the preset initial offset of the parameter type corresponding to the second vector, use the preset initial offset as a temporary offset, and perform shift processing on the second vector according to the temporary offset to obtain the shift vector corresponding to the second vector.

[0041] A temporary matrix is ​​obtained by concatenating the first vector and the shift vector corresponding to the second vector;

[0042] Based on the preset initial offset, extract the temporary submatrix from the temporary matrix;

[0043] By inputting the temporary submatrix into the trained parameter prediction model, temporary prediction vectors corresponding to several historical time points are obtained.

[0044] Obtain the temporary true vector corresponding to each historical time point, and calculate the prediction accuracy based on the temporary prediction vector and the temporary true vector corresponding to each historical time point.

[0045] If the prediction accuracy is less than the preset accuracy threshold, the temporary offset is updated with the preset optimization algorithm until the prediction accuracy corresponding to the temporary offset meets the optimization target. The finally updated temporary offset is used as the target offset corresponding to the second vector.

[0046] Offset label data is obtained by concatenating the first preset value and the target offsets corresponding to each second vector;

[0047] Based on the parameter sample matrix, offset label data, and the preset first training loss function, the first training loss is calculated. The time series offset prediction model is then trained based on the first training loss until the first training loss converges, resulting in a trained time series offset prediction model.

[0048] The historical vector can be formed by the parameter values ​​collected by sensors of the corresponding parameter type during a historical time period.

[0049] The preset initial offset of the parameter type can be set by the implementer or it can be the offset used in the historical inference process. The purpose of introducing the preset initial offset is to ensure that when it is necessary to optimize the temporary offset, the temporary offset has a data basis, thereby minimizing the amount of calculation and time spent in optimization and improving optimization efficiency.

[0050] Specifically, when shifting the second vector according to the temporary offset, this embodiment uses zero-padding to ensure that the shift vector corresponding to the second vector is aligned with the first vector.

[0051] A temporary matrix is ​​obtained by concatenating the first vector as the first row and the shift vector corresponding to the second vector row by row.

[0052] Given the preset initial offset, the time interval covered by the shift vectors corresponding to the first vector and the second vector can be determined, that is, the effective columns in the temporary matrix can be determined, and the temporary submatrix can be extracted from the temporary matrix based on the effective columns.

[0053] The temporary submatrix is ​​input into the trained parametric prediction model, and iterative prediction is performed column by column to obtain temporary prediction vectors corresponding to several historical time points. The temporary prediction vectors are column vectors. It should be noted that, in order to ensure that the output size of the temporary submatrix and the parametric prediction model is consistent, the temporary matrix can be obtained by concatenating the first vector as the first row, the shift vector corresponding to the second vector, and the zero vectors corresponding to the other second vectors by row. Since the columns of the temporary submatrix can represent time series, iterative prediction is performed column by column. For the specific iterative prediction process, please refer to the description of the parametric prediction model.

[0054] As we know, parameters were actually collected at each historical time point. Therefore, we can directly obtain the temporary true vectors corresponding to each historical time point. Based on the temporary prediction vector and temporary true vector corresponding to each historical time point, we can calculate the sub-accuracy corresponding to each historical time point using metrics such as cosine similarity. Then, we can calculate the average of the sub-accuracy corresponding to each historical time point to obtain the prediction accuracy.

[0055] In this embodiment, the preset accuracy threshold can be 0.8. The implementer can adjust the accuracy threshold according to the actual prediction accuracy requirements. The higher the actual prediction accuracy requirements, the closer the accuracy threshold should be to 1.

[0056] In this embodiment, the preset optimization algorithm adopts the hill climbing method. After updating the temporary offset, the prediction accuracy is recalculated until the prediction accuracy corresponding to the temporary offset meets the optimization target. The finally updated temporary offset is used as the target offset corresponding to the second vector.

[0057] The first preset value can be set to 0, and the first training loss function can be the mean squared error loss function.

[0058] In one specific implementation, the optimization objective is to achieve a local maximum in the prediction accuracy corresponding to the temporary offset.

[0059] After the prediction accuracy corresponding to the temporary offset meets the optimization objective, the following is also included:

[0060] Based on the prediction accuracy corresponding to the temporary offset, determine the offset confidence level of the parameter type corresponding to the second vector.

[0061] Since this embodiment optimizes based on a preset initial offset, the optimization process can be simplified. Therefore, the hill climbing method is used for optimization, with local maxima as the optimization target. Of course, implementers can also use optimization methods such as simulated annealing algorithm, with global maxima as the optimization target.

[0062] Offset confidence can be used to indicate the time series offset during subsequent prediction processes.

[0063] Specifically, in this embodiment, the prediction accuracy is normalized, so the prediction accuracy corresponding to the temporary offset can be directly used as the offset confidence level of the corresponding parameter type.

[0064] S105, input the initial parameter matrix and the time-series offsets corresponding to each reference vector into the trained offset mapping model to obtain the intermediate parameter matrix.

[0065] The first preset value corresponding to the base vector and the temporal offset corresponding to each reference vector can be concatenated into a one-dimensional vector with the same number of columns as the initial parameter matrix. By concatenating the initial parameter matrix and the one-dimensional vector row by row, the input matrix of the trained offset mapping model can be obtained. The size of the intermediate parameter matrix is ​​the same as the size of the initial parameter matrix.

[0066] The offset mapping model can adopt a similar network structure to U-Net, the only difference being the input size of the encoder and the output size of the decoder.

[0067] Specifically, the offset mapping model is only used for offset processing to ensure the differentiability of the offset processing, which facilitates the subsequent reverse update of the time series offset. Therefore, the implementer can randomly generate training samples including matrices and offsets, determine training labels based on the shift processing, and train the offset mapping model based on the training samples and training labels, combined with the mean squared error loss function, to obtain the trained offset mapping model.

[0068] S106, Adjust the intermediate parameter matrix according to the preset mapping function to obtain the target parameter matrix.

[0069] The preset mapping function is used to extract the effective columns of the intermediate parameter matrix and ensure the differentiability of the extraction process.

[0070] In one specific implementation, the intermediate parameter matrix is ​​adjusted according to a preset mapping function to obtain the target parameter matrix, including:

[0071] Calculate the minimum value of each column in the intermediate parameter matrix, and add the minimum value of each column to the second preset value to obtain the indicator value of each column;

[0072] For any element value in the intermediate parameter matrix, the corresponding mapping value is obtained based on the element value, the indicator value of the column to which the element value belongs, and the preset mapping function;

[0073] The target parameter matrix is ​​determined based on the mapping values ​​corresponding to all element values.

[0074] The second preset value c can be a minimum value, which can be set to 10 in this embodiment. -6 .

[0075] In one specific implementation, the preset mapping function is as follows:

[0076] y=exp(-a / b j )×d ij ;

[0077] Where, d ijLet be the value of the element in the i-th row and j-th column of the intermediate parameter matrix, where i is an integer in the range [1, M], j is an integer in the range [1, K], and K is the total number of columns in the intermediate parameter matrix. j is the indicator value of column j, a is the third preset value, and exp() is the exponential function.

[0078] The third preset value is also a minimum value, but its exponent is greater than that of the second preset value. In this embodiment, the third preset value can be 10. -3 .

[0079] Specifically, if there is a zero element in the j-th column, this column can be considered not a valid column; that is, not all shift vectors have a corresponding sampled value in this column. Therefore, b... j =10 -6 -a / b j =-10 3 ,exp(-a / b j The value approaches 0, so that any element in the column is mapped to 0.

[0080] If there are no zero elements in the j-th column, the column can be considered a valid column, meaning that all shift vectors have corresponding acquisition values ​​in this column. Therefore, b j =min j +c,min j The minimum element value in column j, defaulting to b. j Much greater than the third preset value, b j It can be approximated as being similar to min j If they are the same, then -a / b j Approaching 0, exp(-a / b) j The value approaches 1, such that any element in the column maps to itself.

[0081] S107: Extract the target submatrix from the target parameter matrix, input the target submatrix into the trained parameter prediction model, and obtain the target prediction vector corresponding to the preset target time point.

[0082] Among them, the parameter prediction model can adopt a temporal prediction model, such as a recurrent neural network model, a long short-term memory network model, and a temporal convolutional network model.

[0083] In this embodiment, the parameter prediction model adopts a temporal convolutional network model. The architecture of the temporal convolutional network model will not be described in detail here. The training of the temporal convolutional network model can be directly trained based on the parameter sample matrix obtained by concatenating historical vectors. Since this embodiment improves the accuracy of the prediction process by adjusting the temporal relationship of parameter vectors, there is no need to perform additional training on the parameter prediction model, and conventional training methods can be used.

[0084] In one specific implementation, a target sub-matrix is ​​extracted from the target parameter matrix, and the target sub-matrix is ​​input into a trained parameter prediction model to obtain a target prediction vector corresponding to a preset target time point, including:

[0085] The target submatrix is ​​extracted from the target parameter matrix, where the number of columns in the target submatrix is ​​N, and N is a positive integer;

[0086] Align the left boundary of the preset sliding window with the first column of the target submatrix, and use the preset sliding window to extract the first intermediate matrix from the target submatrix. The length of the preset sliding window is L, where L is a positive integer less than N.

[0087] Input the first intermediate matrix into the trained parameter prediction model to obtain LN intermediate prediction vectors;

[0088] The prediction reliability is calculated based on the column vectors corresponding to columns L+1 to N in the target submatrix and LN intermediate prediction vectors.

[0089] If the prediction reliability is greater than or equal to the preset reliability threshold, then the right boundary of the preset sliding window is aligned with the last column of the target submatrix, and the second intermediate matrix is ​​extracted from the target submatrix using the preset sliding window;

[0090] The second intermediate matrix is ​​input into the trained parameter prediction model to obtain the target prediction vector corresponding to the target time point;

[0091] If the prediction reliability is less than the preset threshold, then the time series offset corresponding to each reference vector is optimized according to the offset confidence of each reference vector, and the process returns to step S105.

[0092] The width of the preset sliding window is M, which is the same as the number of rows in the target submatrix.

[0093] Specifically, the target sub-matrix has a size of N×M, corresponding to the target parameter vectors at the 1st to Nth preset time points. The time interval between adjacent preset time points is fixed at T. The first intermediate matrix corresponds to the target parameter vectors at the 1st to Lth preset time points. The first intermediate matrix is ​​input into the trained parameter prediction model to obtain the intermediate prediction vector at the (L+1)th preset time point. Then, the target parameter vectors at the 2nd to Lth preset time points and the intermediate prediction vector at the (L+1)th preset time point are input into the trained parameter prediction model to obtain the intermediate prediction vector at the (L+2)th preset time point. Then, the target parameter vectors at the 3rd to Lth preset time points and the intermediate prediction vector at the (L+1)th preset time point to the (L+2)th intermediate point are input into the trained parameter prediction model to obtain the intermediate prediction vector at the (L+3)th preset time point. This process is repeated iteratively to obtain LN intermediate prediction vectors.

[0094] Based on any preset time point from the (L+1)th preset time point to the Nth preset time point, the sub-reliability is calculated using the intermediate prediction vector and the column vector in the target sub-matrix corresponding to the preset time point, respectively, through measures such as cosine similarity. Then, the mean of all sub-reliability is calculated to obtain the prediction reliability.

[0095] In this embodiment, the preset threshold is set to 0.8. Similarly, the implementer can adjust the preset threshold according to the actual prediction accuracy requirements. The higher the actual prediction accuracy requirements, the closer the preset threshold should be to 1.

[0096] When the prediction reliability is greater than or equal to the preset threshold, it indicates that the time series offset is relatively reliable under this condition. There is no need to adjust the time series offset. The second intermediate matrix can be directly extracted for iterative prediction. The target time point can refer to the Rth preset time point, where R is an integer greater than N. The target prediction vector corresponding to the target time point can be obtained by iterative prediction based on the second intermediate matrix. The specific iterative prediction process will not be elaborated here.

[0097] When the prediction reliability is less than the preset threshold, it indicates that the time series offset is unreliable. This is because the time series offset is determined by using local maxima as the optimization target. Local maxima may not even meet the prediction accuracy threshold. Therefore, it is necessary to optimize the time series offset corresponding to each reference vector. The optimization method can be the reverse gradient descent method. The learning rate for updating the time series offset corresponding to the reference vector with different offset confidence is different. The higher the offset confidence, the lower the learning rate when updating the time series offset of the corresponding reference vector.

[0098] S108, Based on the target prediction vector, determine the predicted carbon potential corresponding to the target time point.

[0099] Specifically, the first row of elements in the target prediction vector is extracted to obtain the oxygen probe signal parameters corresponding to the target time point.

[0100] Based on the oxygen probe signal parameters and the prior relationship between oxygen partial pressure and carbon potential, the predicted carbon potential corresponding to the target time point can be calculated.

[0101] S109, if the predicted carbon potential at the target time point is less than the preset first carbon potential threshold, then execute the first carbon potential control process; if the predicted carbon potential at the target time point is greater than the preset second carbon potential threshold, then execute the second carbon potential control process.

[0102] The implementer can set a carbon potential range. When the predicted carbon potential corresponding to the target time point is within this range, no adjustment is required. The left boundary value of this carbon potential range can be the first carbon potential threshold, and the right boundary value can be the second carbon potential threshold.

[0103] Specifically, when the predicted carbon potential corresponding to the target time point is less than the preset first carbon potential threshold, the first carbon potential control process is executed. The first carbon potential control process may include increasing the carbon supply and reducing the opening of the exhaust valve. When the predicted carbon potential corresponding to the target time point is greater than the preset second carbon potential threshold, the second carbon potential control process is executed. The second carbon potential control process may include reducing the enrichment gas flow rate and appropriately increasing the furnace circulating fan speed.

[0104] It should be noted that the implementer can set the adjustment object corresponding to the specific carbon potential control process according to the actual situation. The specific adjustment amount for the adjustment object can be determined by existing technologies such as PID control, expert preset rules, and reinforcement learning.

[0105] This embodiment collects parameter vectors corresponding to M parameter types to comprehensively acquire multi-source information related to carbon potential. These parameter vectors are then rationally concatenated and processed. Through a trained time-series offset prediction model, offset mapping model, and parameter prediction model, the carbon potential at a preset target time point can be accurately predicted. Compared to traditional prediction methods relying on a single or few parameters, this approach fully explores the potential relationships and temporal characteristics between different parameters, significantly improving the accuracy of carbon potential prediction and providing a reliable basis for subsequent precise control. This multi-source data sensing method can comprehensively reflect the complex and ever-changing working conditions in actual production. The model can learn and adapt to the relationship between parameters and carbon potential under different working conditions, exhibiting strong robustness and adaptability, thus broadening the application scope of this method in different production environments.

[0106] Example 2:

[0107] like Figure 2As shown, the present invention provides a technical solution: an intelligent carbon potential control system based on multi-source data sensing, characterized in that the intelligent carbon potential control system based on multi-source data sensing includes:

[0108] The parameter acquisition module 201 is used to acquire parameter vectors corresponding to M parameter types within a preset time period, where M is a positive integer;

[0109] The vector classification module 202 is used to take the parameter vector with the parameter type as the first preset type as the base vector and other parameter vectors as reference vectors.

[0110] Vector concatenation module 203 is used to concatenate the base vector and each reference vector to obtain the initial parameter matrix;

[0111] The offset prediction module 204 is used to input the initial parameter matrix into the trained temporal offset prediction model to obtain the temporal offset corresponding to each reference vector.

[0112] The matrix mapping module 205 is used to input the initial parameter matrix and the temporal offsets corresponding to each reference vector into the trained offset mapping model to obtain the intermediate parameter matrix.

[0113] The matrix adjustment module 206 is used to adjust the intermediate parameter matrix according to a preset mapping function to obtain the target parameter matrix;

[0114] The parameter prediction module 207 is used to extract the target sub-matrix from the target parameter matrix, input the target sub-matrix into the trained parameter prediction model, and obtain the target prediction vector corresponding to the preset target time point.

[0115] The carbon potential prediction module 208 is used to determine the predicted carbon potential corresponding to the target time point based on the target prediction vector.

[0116] The carbon potential control module 209 is used to execute a first carbon potential control process if the predicted carbon potential at the target time point is less than a preset first carbon potential threshold, and to execute a second carbon potential control process if the predicted carbon potential at the target time point is greater than a preset second carbon potential threshold.

[0117] It should be noted that the specific limitations of the intelligent carbon potential control system based on multi-source data sensing can be found in the above-described limitations of the intelligent carbon potential control method based on multi-source data sensing, and will not be repeated here. The information interaction and execution process between the above modules are based on the same concept as the method embodiments of this invention, and their specific functions and technical effects can be found in the method embodiments section, and will not be repeated here.

[0118] The embodiments disclosed herein are preferred embodiments, but are not limited thereto. Those skilled in the art can readily grasp the spirit of the present invention based on the above embodiments and make different extensions and variations, but as long as they do not depart from the spirit of the present invention, they are all within the protection scope of the present invention.

Claims

1. A smart carbon potential control method based on multi-source data sensing, characterized in that, The intelligent carbon potential control method based on multi-source data sensing includes the following steps: S101, within a preset time period, collect parameter vectors corresponding to M parameter types respectively, where M is a positive integer; S102, use the parameter vector with the parameter type of the first preset type as the base vector, and other parameter vectors as reference vectors; S103, the initial parameter matrix is ​​obtained by concatenating the base vector and each reference vector; S104, Input the initial parameter matrix into the trained temporal offset prediction model to obtain the temporal offset corresponding to each reference vector. The training process of the temporal offset prediction model includes the following steps: Obtain the historical vectors corresponding to M parameter types respectively, and use the historical vector with parameter type of the first preset type as the first vector, and use the other historical vectors as the second vector; The parameter sample matrix is ​​obtained by concatenating the historical vectors corresponding to the M parameter types. For any second vector, obtain the preset initial offset of the parameter type corresponding to the second vector, use the preset initial offset as a temporary offset, and perform shift processing on the second vector according to the temporary offset to obtain the shift vector corresponding to the second vector. A temporary matrix is ​​obtained by concatenating the first vector and the shift vector corresponding to the second vector; Based on the preset initial offset, a temporary submatrix is ​​extracted from the temporary matrix; The temporary submatrix is ​​input into the trained parameter prediction model to obtain temporary prediction vectors corresponding to several historical time points. Obtain the temporary true vector corresponding to each historical time point, and calculate the prediction accuracy based on the temporary prediction vector and the temporary true vector corresponding to each historical time point. If the prediction accuracy is less than the preset accuracy threshold, the temporary offset is updated with the preset optimization algorithm until the prediction accuracy corresponding to the temporary offset meets the optimization target, and the finally updated temporary offset is used as the target offset corresponding to the second vector. Offset label data is obtained by concatenating the first preset value and the target offsets corresponding to each second vector; The first training loss is calculated based on the parameter sample matrix, the offset label data, and the preset first training loss function. The time series offset prediction model is trained based on the first training loss until the first training loss converges, and the trained time series offset prediction model is obtained. S105, input the initial parameter matrix and the temporal offsets corresponding to each reference vector into the trained offset mapping model to obtain the intermediate parameter matrix; S106, Adjust the intermediate parameter matrix according to the preset mapping function to obtain the target parameter matrix, including: Calculate the minimum value of each column in the intermediate parameter matrix, and add the minimum value of each column to the second preset value to obtain the indicator value of each column; For any element value in the intermediate parameter matrix, a mapped value is obtained based on the element value, the indicator value of the column to which the element value belongs, and the preset mapping function. Specifically, the preset mapping function is: y=exp(-a / (b j ))×d ij ; Where, d ij Let be the value of the element in the i-th row and j-th column of the intermediate parameter matrix, where i is an integer in the range [1, M], j is an integer in the range [1, K], and K is the total number of columns in the intermediate parameter matrix. j is the indicator value of column j, a is the third preset value, and exp() is the exponential function; The target parameter matrix is ​​determined based on the mapping values ​​corresponding to all element values. S107, extract the target sub-matrix from the target parameter matrix, input the target sub-matrix into the trained parameter prediction model, and obtain the target prediction vector corresponding to the preset target time point; S108, Based on the target prediction vector, determine the predicted carbon potential corresponding to the target time point; S109, if the predicted carbon potential corresponding to the target time point is less than the preset first carbon potential threshold, then execute the first carbon potential control process; if the predicted carbon potential corresponding to the target time point is greater than the preset second carbon potential threshold, then execute the second carbon potential control process.

2. The intelligent carbon potential control method based on multi-source data sensing according to claim 1, characterized in that, The parameter types include at least oxygen probe signal parameters, carbon monoxide concentration signal parameters, furnace temperature signal parameters, and furnace pressure parameters; The first preset type is the oxygen probe signal parameter.

3. The intelligent carbon potential control method based on multi-source data sensing according to claim 1, characterized in that, The optimization objective is to achieve a local maximum in the prediction accuracy corresponding to the temporary offset. After the prediction accuracy corresponding to the temporary offset meets the optimization objective, the following steps are also included: Based on the prediction accuracy corresponding to the temporary offset, determine the offset confidence level of the parameter type corresponding to the second vector.

4. The intelligent carbon potential control method based on multi-source data sensing according to claim 2, characterized in that, A target submatrix is ​​extracted from the target parameter matrix, and the target submatrix is ​​input into the trained parameter prediction model to obtain the target prediction vector corresponding to the preset target time point, including: A target submatrix is ​​extracted from the target parameter matrix, wherein the number of columns of the target submatrix is ​​N, and N is a positive integer; Align the left boundary of the preset sliding window with the first column of the target submatrix, and use the preset sliding window to extract the first intermediate matrix from the target submatrix, wherein the length of the preset sliding window is L, and L is a positive integer less than N; The first intermediate matrix is ​​input into the trained parameter prediction model to obtain LN intermediate prediction vectors; The prediction reliability is calculated based on the column vectors corresponding to columns L+1 to N in the target submatrix and the LN intermediate prediction vectors. If the prediction reliability is greater than or equal to a preset threshold, then the right boundary of the preset sliding window is aligned with the last column of the target sub-matrix, and the second intermediate matrix is ​​extracted from the target sub-matrix using the preset sliding window; The second intermediate matrix is ​​input into the trained parameter prediction model to obtain the target prediction vector corresponding to the target time point; If the prediction reliability is less than the preset threshold, then the time series offset corresponding to each reference vector is optimized according to the offset confidence of each reference vector, and the process returns to step S105.

5. An intelligent carbon potential control system based on multi-source data sensing, characterized in that, The intelligent carbon potential control system based on multi-source data sensing includes: The parameter acquisition module is used to acquire parameter vectors corresponding to M parameter types within a preset time period, where M is a positive integer; The vector classification module is used to take parameter vectors with parameter type as the first preset type as base vectors and other parameter vectors as reference vectors. The vector concatenation module is used to concatenate the base vector and each reference vector to obtain the initial parameter matrix; The offset prediction module is used to input the initial parameter matrix into the trained temporal offset prediction model to obtain the temporal offset corresponding to each reference vector. The training process of the temporal offset prediction model includes the following steps: Obtain the historical vectors corresponding to M parameter types respectively, and use the historical vector with parameter type of the first preset type as the first vector, and use the other historical vectors as the second vector; The parameter sample matrix is ​​obtained by concatenating the historical vectors corresponding to the M parameter types. For any second vector, obtain the preset initial offset of the parameter type corresponding to the second vector, use the preset initial offset as a temporary offset, and perform shift processing on the second vector according to the temporary offset to obtain the shift vector corresponding to the second vector. A temporary matrix is ​​obtained by concatenating the first vector and the shift vector corresponding to the second vector; Based on the preset initial offset, a temporary submatrix is ​​extracted from the temporary matrix; The temporary submatrix is ​​input into the trained parameter prediction model to obtain temporary prediction vectors corresponding to several historical time points. Obtain the temporary true vector corresponding to each historical time point, and calculate the prediction accuracy based on the temporary prediction vector and the temporary true vector corresponding to each historical time point. If the prediction accuracy is less than the preset accuracy threshold, the temporary offset is updated with the preset optimization algorithm until the prediction accuracy corresponding to the temporary offset meets the optimization target, and the finally updated temporary offset is used as the target offset corresponding to the second vector. Offset label data is obtained by concatenating the first preset value and the target offsets corresponding to each second vector; The first training loss is calculated based on the parameter sample matrix, the offset label data, and the preset first training loss function. The time series offset prediction model is trained based on the first training loss until the first training loss converges, and the trained time series offset prediction model is obtained. The matrix mapping module is used to input the initial parameter matrix and the temporal offsets corresponding to each reference vector into the trained offset mapping model to obtain the intermediate parameter matrix. A matrix adjustment module is used to adjust the intermediate parameter matrix according to a preset mapping function to obtain a target parameter matrix, including: Calculate the minimum value of each column in the intermediate parameter matrix, and add the minimum value of each column to the second preset value to obtain the indicator value of each column; For any element value in the intermediate parameter matrix, a mapped value is obtained based on the element value, the indicator value of the column to which the element value belongs, and the preset mapping function. Specifically, the preset mapping function is: y=exp(-a / (b j ))×d ij ; Where, d ij Let be the value of the element in the i-th row and j-th column of the intermediate parameter matrix, where i is an integer in the range [1, M], j is an integer in the range [1, K], and K is the total number of columns in the intermediate parameter matrix. j is the indicator value of column j, a is the third preset value, and exp() is the exponential function; The parameter prediction module is used to extract a target sub-matrix from the target parameter matrix, input the target sub-matrix into the trained parameter prediction model, and obtain the target prediction vector corresponding to the preset target time point. The carbon potential prediction module is used to determine the predicted carbon potential corresponding to the target time point based on the target prediction vector. The carbon potential control module is used to execute a first carbon potential control process if the predicted carbon potential corresponding to the target time point is less than a preset first carbon potential threshold, and to execute a second carbon potential control process if the predicted carbon potential corresponding to the target time point is greater than a preset second carbon potential threshold.

Images