Pretreatment analysis and optimization method, device, medium and equipment for pellet quality anomaly

Abstract

The application discloses a method, device, medium and equipment for pre-analysis and optimization of abnormal pellet quality, and relates to the technical field of pellet production. The method comprises the following steps: obtaining pellet production data of a current time node and a historical time node respectively; determining historical time sequence data of different pellet process key parameters according to the pellet production data of the current time node and the historical time node; inputting the historical time sequence data of the different pellet process key parameters into corresponding index prediction models for index prediction to obtain predicted values of different pellet quality indexes, wherein the predicted values of the different pellet quality indexes comprise a total iron predicted value, a silicon dioxide predicted value, a compressive strength predicted value and a drum strength predicted value; and if it is determined that the pellet quality is abnormal according to the predicted values of the different pellet quality indexes, an operation suggestion matrix is constructed based on operation parameters to optimize the pellet quality. The application can realize early prediction and optimization of the pellet quality.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of pellet production, in particular to a pellet quality abnormality early analysis and optimization method, device, medium and equipment. BACKGROUND

[0002] The pellet process is an important link in steel production. With the development of low-carbon smelting technology, the application of high-quality pellets has become the key to achieving the low-carbon goal. Blast furnace ironmaking relies heavily on pellets, and ensuring pellet quality not only affects smooth production, but also relates to resource utilization and environmental protection goals. Compared with the sintering process, shaft furnace pellet smelting has low energy consumption, less emission and small equipment footprint, and is suitable for small-scale production. Pellets have good permeability, which not only helps to improve fuel efficiency, reduce coke consumption and reduce carbon emissions, but also helps to optimize airflow distribution and improve hot metal yield and quality.

[0003] Currently, the pellet quality is usually detected and optimized by manual operation. However, the existing method cannot meet the real-time monitoring and adjustment requirements of pellet quality, and the detection is lagging, often leading to abnormal batches being discovered only after entering the next process, so that the quality problems of the pellets cannot be discovered and corrected in time, thereby affecting the production efficiency and quality stability of the pellets and causing production losses. SUMMARY

[0004] Therefore, the application provides a pellet quality abnormality early analysis and optimization method, device, medium and equipment, which can realize early prediction and optimization of pellet quality and solve the problem of lagging detection and correction of pellet quality.

[0005] According to a first aspect of the application, a pellet quality abnormality early analysis and optimization method is provided, which comprises:

[0006] Obtain pellet production data of a current time node and a historical time node, respectively;

[0007] According to the pellet production data of the current time node and the pellet production data of the historical time node, historical time series data corresponding to different pellet process key parameters are determined, respectively;

[0008] The historical time series data corresponding to the different pellet process key parameters are input into the corresponding index prediction model for index prediction to obtain predicted values of different pellet quality indexes, wherein the predicted values of the different pellet quality indexes include predicted values of total iron, predicted values of silicon dioxide, predicted values of compressive strength and predicted values of drum strength;

[0009] If it is determined that the pellet quality is abnormal according to the predicted values of the different pellet quality indexes, an operation suggestion matrix is constructed based on adjustable operation parameters;

[0010] based on the operation suggestion matrix, optimizing the pellet quality.

[0011] According to a second aspect of the present application, a pellet quality anomaly early analysis and optimization device is provided, which comprises:

[0012] an acquisition unit, configured to acquire pellet production data of a current time node and historical time nodes respectively;

[0013] a determination unit, configured to determine historical time sequence data corresponding to different pellet process key parameters respectively according to the pellet production data of the current time node and the pellet production data of the historical time nodes;

[0014] a prediction unit, configured to input the historical time sequence data corresponding to the different pellet process key parameters into corresponding index prediction models respectively for index prediction to obtain predicted values of different pellet quality indexes, wherein the predicted values of the different pellet quality indexes comprise a total iron predicted value, a silicon dioxide predicted value, a compressive strength predicted value and a drum strength predicted value;

[0015] a construction unit, configured to, if it is determined that a pellet quality anomaly exists according to the predicted values of the different pellet quality indexes, construct an operation suggestion matrix based on adjustable operation parameters;

[0016] an optimization unit, configured to optimize the pellet quality based on the operation suggestion matrix.

[0017] According to a third aspect of the present application, a storage medium having a computer program stored thereon is provided, and the program is executed by a processor to implement the above-mentioned pellet quality anomaly early analysis and optimization method.

[0018] According to a fourth aspect of the present application, an electronic device is provided, which comprises a storage medium, a processor and a computer program stored on the storage medium and executable on the processor, and the processor implements the above-mentioned pellet quality anomaly early analysis and optimization method when executing the program.

[0019] By the technical scheme, the balling quality abnormality early analysis and optimization method, device, medium and equipment provided by the application can determine the historical time sequence data corresponding to different balling process key parameters according to the balling production data of the current time node and the balling production data of the historical time node, respectively, and input the historical time sequence data corresponding to different balling process key parameters into the corresponding index prediction model for index prediction to obtain the prediction values of different balling quality indexes. If it is determined that the balling quality is abnormal according to the prediction values of different balling quality indexes, an operation suggestion matrix is constructed based on adjustable operation parameters, and finally the balling quality is optimized based on the operation suggestion matrix. As can be seen, the application can discover the balling quality problem in time by using the index prediction model to predict the prediction values of different balling quality indexes, thereby solving the problem of detection lag in the prior art. At the same time, the application can optimize the balling quality by constructing the operation suggestion matrix, thereby being able to solve the problems in the balling generation process in time, ensuring the stability of the balling quality, improving the production efficiency of the balling, and avoiding production loss.

[0020] The above description is only a summary of the technical scheme of the application. In order to more clearly understand the technical means of the application, the application can be implemented according to the content of the specification, and in order to make the above and other purposes, features and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described. BRIEF DESCRIPTION OF DRAWINGS

[0021] The drawings described herein are used to provide further understanding of the application, and form a part of the application. The schematic embodiments of the application and the description thereof are used to explain the application, and do not constitute an improper limitation on the application. In the drawings:

[0022] Figure 1 A flowchart of a balling quality abnormality early analysis and optimization method provided by an embodiment of the application is shown;

[0023] Figure 2 A flowchart of an index prediction model training method provided by an embodiment of the application is shown;

[0024] Figure 3 A structure diagram of a balling quality abnormality early analysis and optimization device provided by an embodiment of the application is shown. DETAILED DESCRIPTION

[0025] In the following, the application will be described in detail with reference to the drawings and in combination with the embodiments. It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0026] The existing manual operation mode cannot meet the real-time monitoring and adjustment requirements of the pellet quality, the detection is lagging, and the abnormal batch is often found after entering the next process, so that the quality problem of the pellet cannot be found and corrected in time, affecting the production efficiency and quality stability of the pellet, and causing production loss.

[0027] To solve the above problems, the embodiment of the application provides a pellet quality abnormality front-end analysis and optimization method, as shown in the figure, which comprises the following steps: Figure 1

[0028] Step 10, respectively acquiring the pellet production data of the current time node and the historical time node.

[0029] The pellet production data comprises a plurality of parameter data, such as stockyard raw material data, mixing ratio data, green ball quality data, shaft furnace operation data, etc., the stockyard raw material data specifically comprises raw material chemical components TFe, FeO, CaO, SiO2, MgO, Al2O3, S, P, etc., the green ball quality data specifically comprises green ball compression strength and green ball drop strength, and the shaft furnace operation data specifically comprises 60 shaft furnace parameter data, such as shaft furnace roasting temperature, roasting zone temperature, steam drum pressure, air guide wall inlet temperature, total gas cumulative amount, drying bed temperature, smoke hood temperature, air flow, air main pressure, combustion chamber pressure, combustion chamber temperature, cold air pressure, cold air flow, gas flow, cooling zone temperature, air guide wall outlet temperature, etc.

[0030] The embodiment of the application is mainly applicable to the abnormality front-end analysis and optimization scene of the pellet quality. The execution subject of the embodiment of the application is a device or equipment capable of performing abnormality front-end analysis and optimization on the pellet quality, such as being arranged on the server side.

[0031] In the actual production scene of the factory, the full-process data of the pellet production is monitored and collected in real time by using the field-deployed data acquisition system, so that the pellet production data of different time nodes can be acquired. Since the embodiment of the application is to perform front-end analysis on the pellet quality, that is, to make an early prediction, the pellet production data of the current time node and the historical time node need to be acquired, and the pellet quality of the next time node is predicted based on the above data. For example, the current time node is 11:00 am, and when the data acquisition is performed, the pellet production data of 11:00 am needs to be acquired, and the pellet production data before 11:00 am, i.e., the pellet production data of 10:45, 10:30, 10:15 and 10:00, also needs to be acquired.

[0032] Step 20, respectively determining the historical time sequence data corresponding to different pellet process key parameters according to the pellet production data of the current time node and the pellet production data of the historical time node. ​

[0033] wherein the different pellet process key parameters include a first pellet process key parameter, a second pellet process key parameter, a third pellet process key parameter and a fourth pellet process key parameter, the first pellet process key parameter is a parameter associated with the total iron (TFe) prediction value among all parameters involved in the pellet production process, the second pellet process key parameter is a parameter associated with the silicon dioxide (SiO2) prediction value among all parameters involved in the pellet production process, the third pellet process key parameter is a parameter associated with the pellet compression strength prediction value among all parameters involved in the pellet production process, and the fourth pellet process key parameter is a parameter associated with the pellet drum strength prediction value among all parameters involved in the pellet production process. In addition, the historical time series data includes the key parameter data of the current time node and the key parameter data of the historical time node.

[0034] For the embodiment of the present application, the historical time series data corresponding to the first pellet process key parameter, the second pellet process key parameter, the third pellet process key parameter and the fourth pellet process key parameter are determined respectively according to the pellet production data of the current time node and the pellet production data of the historical time node.

[0035] For example, the first pellet process key parameter (a parameter associated with the total iron (TFe) prediction value among all parameters involved in the pellet production process) includes raw material chemical composition TFe ratio, shaft furnace roasting temperature, roasting zone temperature, cold air flow and coal gas flow. The historical time series data of the first pellet process key parameter is determined according to the TFe ratio, the shaft furnace roasting temperature, the roasting zone temperature, the cold air flow and the coal gas flow of the current time node and the historical time node. Similarly, the historical time series data corresponding to the second pellet process key parameter, the third pellet process key parameter and the fourth pellet process key parameter can be determined respectively.

[0036] Step 30, input the historical time series data corresponding to the different pellet process key parameters into the corresponding index prediction model respectively to perform index prediction, and obtain the prediction values of different pellet quality indexes.

[0037] wherein the prediction values of different pellet quality indexes include total iron prediction value, silicon dioxide prediction value, compression strength prediction value and drum strength prediction value.

[0038] For the embodiment of the application, the historical time series data corresponding to the first balling process key parameter is input into a preset total iron prediction model for prediction to obtain a total iron prediction value of the pellet; the historical time series data corresponding to the second balling process key parameter is input into a preset silicon dioxide prediction model for prediction to obtain a silicon dioxide prediction value of the pellet; the historical time series data corresponding to the third balling process key parameter is input into a preset pellet compression strength prediction model for prediction to obtain a compression strength prediction value of the pellet; and the historical time series data corresponding to the fourth balling process key parameter is input into a preset pellet drum strength prediction model for prediction to obtain a drum strength prediction value of the pellet. Thus, through the above index prediction models, the total iron prediction value, the silicon dioxide prediction value, the pellet compression strength prediction value and the pellet drum strength prediction value at the current time node can be predicted respectively.

[0039] The preset total iron prediction model, the preset silicon dioxide prediction model, the preset pellet compression strength prediction model and the preset pellet drum strength prediction model can be Adaboost models, XGboost models or other prediction models, and the embodiment of the application does not make specific limitation thereto.

[0040] In order to further improve the prediction accuracy of the prediction values of the pellet quality indexes, the application embodiment will correct the predicted pellet quality index values. Based on this, the method further comprises: determining the prediction errors of different pellet quality indexes of the previous time node corresponding to the current time node; and compensating the prediction values of different pellet quality indexes of the current time node based on the prediction errors to obtain the compensated prediction values of different pellet quality indexes.

[0041] Since the prediction values and the actual collected values of different pellet quality indexes of the previous time node are known, the prediction errors of different pellet quality indexes of the previous time node can be calculated. Then, the prediction values of different pellet quality indexes of the current time node are compensated through the prediction errors of the previous time node, i.e. the compensated prediction values of different pellet quality indexes are equal to the sum of the prediction errors of the previous time node and the prediction values of different pellet quality indexes of the current time node.

[0042] Specifically, the total iron prediction value of the current time node is compensated according to the total iron prediction error of the previous time node to obtain a compensated total iron prediction value. At the same time, the silicon dioxide prediction value of the current time node is compensated according to the silicon dioxide prediction error of the previous time node to obtain a compensated silicon dioxide prediction value. Similarly, the compressive strength prediction value of the current time node is compensated according to the compressive strength prediction error of the previous time node to obtain a compensated compressive strength prediction value, and the drum strength prediction value of the current time node is compensated according to the drum strength prediction error of the previous time node to obtain a compensated drum strength prediction value.

[0043] The embodiment of the present application adopts a real-time correction mechanism, that is, the prediction error of the previous time is used for real-time monitoring and correction, which is used as a correction factor for the current time prediction value to dynamically correct the prediction result, thereby improving the accuracy and stability of the index prediction.

[0044] Step 40, if the pellet quality is abnormal according to the prediction values of different pellet quality indexes, an operation suggestion matrix is constructed based on adjustable operation parameters.

[0045] The adjustable operation parameters include a cold wind flow parameter, a cold wind pressure parameter, a south combustion chamber pressure parameter and a north combustion chamber pressure parameter.

[0046] For the embodiment of the present application, if the pellet quality is abnormal according to the compensated prediction values of different pellet quality indexes, an operation suggestion matrix is constructed based on adjustable operation parameters. In the analysis and optimization of the pellet quality, if any one of the total iron prediction value, the silicon dioxide prediction value, the compressive strength prediction value and the drum strength prediction value exceeds the corresponding preset index range, it is determined that the pellet quality is abnormal; the cold wind flow parameter, the cold wind pressure parameter, the south combustion chamber pressure parameter and the north combustion chamber pressure parameter are used as the operation parameters, and an operation suggestion matrix is constructed according to the operation range and adjustment step length of the operation parameters.

[0047] The preset index range can be set according to actual business requirements.

[0048] Specifically, the index determination standard when the pellet quality is abnormal can be set according to expert experience, for example, the compressive strength prediction value lower than 2400N or the drum strength prediction value lower than 90% is used as the condition to trigger the pellet quality abnormality. The abnormality determination standards of the silicon dioxide prediction value and the total iron prediction value are set according to the differences of raw materials and the actual production situation on site.

[0049] After determining the abnormality of the pellet quality according to the total iron prediction value, the silicon dioxide prediction value, the compressive strength prediction value and the drum strength prediction value, an operation suggestion matrix is constructed according to the step length and the operation range of the cold air flow parameter, the cold air pressure parameter, the south combustion chamber pressure parameter and the north combustion chamber pressure parameter, and the operation suggestion matrix records a plurality of operation suggestions, and each operation suggestion includes the correction suggestion value of the cold air flow parameter, the cold air pressure parameter, the south combustion chamber pressure parameter and the north combustion chamber pressure parameter.

[0050] For example, the compressive strength prediction value of the next time node output by the preset pellet compressive strength prediction model is 2383N, and since it is less than 2400N, it is determined that the pellet quality is abnormal, at which time the adjustment step length of the cold air flow parameter is determined to be 500m 3 , the adjustment step length of the cold air pressure parameter is 0.25kp, the adjustment step length of the south combustion chamber pressure parameter is 0.5kp, the adjustment step length of the north combustion chamber pressure parameter is 0.5kp, the operation range of the cold air flow parameter is ±1000m 3 , the operation range of the cold air pressure parameter is ±0.5kp, the operation range of the south combustion chamber pressure parameter is ±1kp, and the operation range of the north combustion chamber pressure parameter is ±1kp. Then, according to the adjustment step length and the operation range of the cold air flow parameter, the cold air pressure parameter, the south combustion chamber pressure parameter and the north combustion chamber pressure parameter, a plurality of operation suggestions about different correction suggestion values of the cold air flow parameter, the cold air pressure parameter, the south combustion chamber pressure parameter and the north combustion chamber pressure parameter are generated under the condition that other parameters remain unchanged, so as to generate the operation suggestion matrix.

[0051] Further, in order to ensure the determination accuracy of the operation parameter, the embodiment of the present application designs a method for determining the operation parameter from a plurality of parameters. Specifically, first, a plurality of common parameters in the first pellet process key parameter, the second pellet process key parameter, the third pellet process key parameter and the fourth pellet process key parameter are determined. Then, the recursive feature elimination (RFE) algorithm is used to screen out key parameters that have a greater impact on the pellet quality from the plurality of common parameters, which are used as operation parameters. Specifically, sample data sets of the plurality of common parameters and different pellet quality indicators are collected, and a parameter evaluation model is constructed for each pellet quality indicator. The parameter evaluation model can be a random forest model. For each parameter evaluation model, the parameter with the lowest weight is excluded each time, and the process is repeated until a preset number of key parameters are left, which are output. Finally, the operation parameter is determined according to the key parameters output by each parameter evaluation model, for example, the intersection of the key parameters output by each parameter evaluation model can be taken. In this way, the operation parameter can be accurately determined, and by adjusting the operation parameter, the pellet quality can be optimized.

[0052] Step 50, optimizing the pellet quality based on the operation suggestion matrix.

[0053] For the embodiment of the present application, after the operation suggestion matrix is constructed, the pellet quality is optimized based on the operation suggestion matrix. In the specific optimization, the index prediction model is used to predict the optimization result of each operation suggestion in the operation suggestion matrix; and according to the optimization result of each operation suggestion, a front operation optimization suggestion is output to optimize the pellet quality.

[0054] Specifically, the correction suggestion value of the cold air flow parameter, the cold air pressure parameter, the south combustion chamber pressure parameter and the north combustion chamber pressure parameter in each operation suggestion of the operation suggestion matrix is taken as the parameter value of the current time node, and other parameters do not change, and then the index prediction model is used to predict the total iron prediction value, the silicon dioxide prediction value, the compressive strength prediction value and the drum strength prediction value corresponding to each operation suggestion, so as to obtain the optimization result corresponding to each operation suggestion. If the total iron prediction value, the silicon dioxide prediction value, the compressive strength prediction value and the drum strength prediction value in the optimization result all develop in a good direction and improve the pellet quality, it is indicated that the operation suggestion can be adopted, and thus the operation suggestion which can optimize the pellet quality index and improve the pellet quality is output. If multiple operation suggestions can optimize the pellet quality index and improve the pellet quality, the operation suggestions can be sorted according to the size of the optimization effect, and the target operation suggestion in the preset ranking range is output according to the sorting result, such as the first 20 operation suggestions.

[0055] The pellet quality abnormality front analysis and optimization method provided by the embodiment of the present application is based on process data (stockyard raw materials, green ball quality and shaft furnace running parameters, etc.) for prediction and optimization, dynamically analyzes the pellet quality in combination with expert experience and data analysis, has strong adaptability and good interpretability. In addition, the embodiment of the present application does not need to add detection equipment, and can accurately identify abnormal pellet quality through existing production line data, has low cost, strong generalization ability of index prediction and high real-time performance. Further, the embodiment of the present application optimizes the pellet quality based on the index prediction model and provides operation suggestions, which can help the on-site operator to make quick decisions, improve production stability and reduce cost.

[0056] Further, the embodiment of the present application further provides a training method of an index prediction model, as shown in Figure 2 , which comprises the following steps.

[0057] Step 60, obtaining sample pellet production data, and determining sample historical time sequence data corresponding to each pellet process parameter according to the sample pellet production data.

[0058] The sample pellet production data includes various pellet process parameter data at different time nodes, and the various pellet process parameter data includes stockyard raw material data, mixing ratio data, green ball quality data, shaft furnace operation data, etc. The stockyard raw material data specifically includes chemical components such as TFe, FeO, CaO, SiO2, MgO, Al2O3, S, P, etc. The green ball quality data specifically includes green ball compressive strength and green ball drop strength. The shaft furnace operation data specifically includes 60 shaft furnace parameter data such as shaft furnace roasting temperature, roasting zone temperature, steam drum pressure, air guide wall inlet temperature, total gas cumulative amount, drying bed temperature, smoke hood temperature, air flow, air main pressure, combustion chamber pressure, combustion chamber temperature, cold air pressure, cold air flow, gas flow, cooling zone temperature, air guide wall outlet temperature, etc.

[0059] For the embodiment of the present application, the original pellet production data and the original index data are monitored and collected in real time through the field-deployed data collection system, so as to obtain various pellet process parameter data (sample pellet production data) and index data at different time nodes. Then, according to the various pellet process parameter data at different time nodes, sample historical time sequence data corresponding to various pellet process parameters is generated. The index data specifically includes total iron content data, total iron content data, pellet compressive strength and drum strength at different time nodes.

[0060] For the embodiment of the present application, in order to ensure data quality when obtaining sample pellet production data and index data, the original pellet production data and the original index data collected above need to be data cleaned, so as to obtain sample data. Based on this, the method comprises: obtaining original pellet production data; performing abnormality detection and missing filling on the original pellet production data respectively, to obtain the sample pellet production data. When performing missing filling, different missing filling strategies are adopted for different missing proportions of the original pellet production data.

[0061] Specifically, after obtaining the original balling production data and the original index data, in order to ensure the data quality, the embodiment of the present application adopts a multi-dimensional outlier detection mechanism. Specifically, first, the intelligent recognition ability of the isolation forest algorithm is combined to remove outliers, so that the embodiment of the present application has high accuracy when processing high-dimensional data, then the box plot technology is applied to identify and remove outliers, and then further combined with the domain knowledge of process experts, from the dual perspectives of data characteristics and production process, trigger, accurately locate the root cause of abnormal data. In addition to being able to detect outliers for balling production data, the embodiment of the present application can also fill in missing data. When supplementing the missing data, the embodiment of the present application can adopt different missing data filling strategies according to different data missing ratios. For example, for a certain balling process parameter data in the original balling production data, if the data missing ratio is below 5%, according to the normal distribution or skew distribution of the data, the mean value or mode is used for filling; if the data missing ratio is between 5% and 25%, a machine learning method is used for filling; if the data missing ratio is above 25%, the parameter data is discarded. At the same time, the time frequency of each parameter data is unified in the form of time point, so as to complete the data cleaning of the original balling production data and the original index data.

[0062] Step 70, based on the sample historical time series data corresponding to each balling process parameter, performing correlation analysis on each balling process parameter to determine balling process key parameters associated with different balling quality indicators respectively.

[0063] Among them, the balling process key parameters associated with different balling quality indicators include first balling process key parameters, second balling process key parameters, third balling process key parameters and fourth balling process key parameters, the first balling process key parameters are parameters associated with total iron (TFe) prediction value in each balling process parameter, the second balling process key parameters are parameters associated with silicon dioxide (SiO2) prediction value in each balling process parameter, the third balling process key parameters are parameters associated with balling compression strength prediction value in each balling process parameter, and the fourth balling process key parameters are parameters associated with balling drum strength prediction value in each balling process parameter.

[0064] For the embodiment of the present application, after obtaining the sample historical time series data corresponding to each pellet process parameter, the key parameters related to the pellet quality are obtained through correlation analysis based on the sample historical time series data corresponding to each pellet process parameter. Specifically, based on the sample historical time series data corresponding to each pellet process parameter, the above data is analyzed in depth to ensure that comprehensive and accurate production process information is obtained, and the above data covers each link of pellet production. On this basis, combined with the field operation principle, the pellet process state quality parameters that the field operators pay attention to are extracted, which are usually the key parameters in the production process.

[0065] Further, Pearson, Spearman, random forest and information entropy algorithm are used to perform correlation analysis from the linear and nonlinear angles to deeply mine the key pellet process parameters closely related to the pellet quality. The correlation analysis of the embodiment of the present application not only focuses on the traditional process parameters, but also introduces a plurality of process state parameters such as cold air flow, cold air pressure, drying bed temperature, combustion chamber pressure and the like. These parameters have an important influence on the final quality of the pellet, because they directly participate in the regulation and control of the key links such as the thermodynamic process, chemical reaction and gas flow of the shaft furnace. By analyzing the relationship between the above parameters and the pellet quality indicators, it can be identified which parameters have strong correlation with the pellet quality, so as to help the operators quickly identify potential quality fluctuations and abnormal factors in the production process.

[0066] Step 80, determining a sample training set according to the sample historical time series data corresponding to the different pellet process key parameters associated with the different pellet quality indicators.

[0067] For the embodiment of the present application, the sample historical time series data corresponding to the different pellet process key parameters associated with the different pellet quality indicators is divided into a test set and a sample training set. Specifically, the sample historical time series data corresponding to the first pellet process key parameter is divided into a first test set and a first sample training set, the sample historical time series data corresponding to the second pellet process key parameter is divided into a second test set and a second sample training set, the sample historical time series data corresponding to the third pellet process key parameter is divided into a third test set and a third sample training set, and the sample historical time series data corresponding to the fourth pellet process key parameter is divided into a fourth test set and a fourth sample training set.

[0068] Step 90, training different indicator prediction models based on the sample training set.

[0069] The different index prediction models include a preset total iron prediction model, a preset silicon dioxide prediction model, a preset pellet compression strength prediction model, and a preset pellet drum strength prediction model. The preset total iron prediction model, the preset silicon dioxide prediction model, the preset pellet compression strength prediction model, and the preset pellet drum strength prediction model can be an Adaboost model or an XGboost model, or other prediction models, which are not limited in the embodiments of the present application.

[0070] For the embodiments of the present application, the preset total iron prediction model, the preset silicon dioxide prediction model, the preset pellet compression strength prediction model, and the preset pellet drum strength prediction model are trained based on the first sample training set, the second sample training set, the third sample training set, and the fourth sample training set, respectively. After the training is completed, the performance of the prediction models can be verified on a test set by comparing the difference between the predicted value and the true value.

[0071] Further, as a specific implementation of the method shown in Figure 1 and Figure 2 The embodiments of the present application provide a pellet quality abnormality pre-analysis and optimization device, as shown in Figure 3 The device includes an acquisition unit 101, a determination unit 102, a prediction unit 103, a construction unit 104, and an optimization unit 105.

[0072] The acquisition unit 101 can be configured to acquire pellet production data at a current time node and historical time nodes, respectively.

[0073] The determination unit 102 can be configured to determine historical time series data corresponding to different pellet process key parameters according to the pellet production data at the current time node and the pellet production data at the historical time nodes, respectively.

[0074] The prediction unit 103 can be configured to input the historical time series data corresponding to the different pellet process key parameters into corresponding index prediction models for index prediction, to obtain predicted values of different pellet quality indexes, wherein the predicted values of the different pellet quality indexes include total iron prediction values, silicon dioxide prediction values, compression strength prediction values, and drum strength prediction values.

[0075] The construction unit 104 can be configured to construct an operation suggestion matrix based on adjustable operation parameters, if it is determined that there is a pellet quality abnormality according to the predicted values of the different pellet quality indexes.

[0076] The optimization unit 105 can be configured to optimize the pellet quality based on the operation suggestion matrix.

[0077] In some embodiments, when the different pellet process key parameters include a first pellet process key parameter, a second pellet process key parameter, a third pellet process key parameter, and a fourth pellet process key parameter, the determining unit 102 can be specifically configured to determine historical time series data corresponding to the first pellet process key parameter, the second pellet process key parameter, the third pellet process key parameter, and the fourth pellet process key parameter, respectively, according to the pellet production data at the current time node and the pellet production data at the historical time node.

[0078] In some embodiments, the prediction unit 103 can be specifically configured to input the historical time series data corresponding to the first pellet process key parameter into a preset total iron prediction model for prediction to obtain a total iron prediction value of the pellet; input the historical time series data corresponding to the second pellet process key parameter into a preset silicon dioxide prediction model for prediction to obtain a silicon dioxide prediction value of the pellet; input the historical time series data corresponding to the third pellet process key parameter into a preset pellet compressive strength prediction model for prediction to obtain a compressive strength prediction value of the pellet; and input the historical time series data corresponding to the fourth pellet process key parameter into a preset pellet drum strength prediction model for prediction to obtain a drum strength prediction value of the pellet.

[0079] In some embodiments, the device further includes a compensation unit.

[0080] The compensation unit is configured to determine a prediction error of different pellet quality indicators of a previous time node corresponding to the current time node; and compensate a prediction value of the different pellet quality indicators of the current time node based on the prediction error to obtain a compensated prediction value of the different pellet quality indicators.

[0081] The construction unit 104 can be specifically configured to construct an operation suggestion matrix based on adjustable operation parameters if it is determined that the pellet quality is abnormal according to the compensated prediction value of the different pellet quality indicators.

[0082] In some embodiments, the construction unit 104 can be specifically configured to determine that the pellet quality is abnormal if any one of the total iron prediction value, the silicon dioxide prediction value, the compressive strength prediction value, and the drum strength prediction value exceeds a corresponding preset index range; and take a cold air flow parameter, a cold air pressure parameter, a south combustion chamber pressure parameter, and a north combustion chamber pressure parameter as the operation parameters, and construct an operation suggestion matrix according to an operation range and an adjustment step length of the operation parameters.

[0083] In some embodiments, the optimization unit 105 can be specifically configured to predict the optimization result of each operation suggestion in the operation suggestion matrix by using the index prediction model; and output a pre-operation optimization suggestion according to the optimization result of each operation suggestion, so as to optimize the pellet quality.

[0084] In some embodiments, the device further comprises a construction unit.

[0085] The construction unit can be configured to obtain sample pellet production data; determine sample historical time series data corresponding to each pellet process parameter according to the sample pellet production data; perform correlation analysis on the each pellet process parameter based on the sample historical time series data corresponding to the each pellet process parameter, and determine pellet process key parameters associated with the different pellet quality indexes respectively; determine a sample training set according to sample historical time series data corresponding to the pellet process key parameters associated with the different pellet quality indexes; and train different index prediction models based on the sample training set respectively.

[0086] The construction unit can be specifically configured to obtain original pellet production data; perform abnormality detection and missing data filling on the original pellet production data respectively to obtain the sample pellet production data, wherein, when performing the missing data filling, different missing data filling strategies are adopted for different missing proportions of the original pellet production data.

[0087] It should be noted that other corresponding descriptions of the various functional units involved in the pellet quality abnormality pre-analysis and optimization device provided in this embodiment can be referred to the corresponding descriptions in Figure 1 and Figure 2 , which will not be described here in detail.

[0088] Based on the above method as shown in Figure 1 and Figure 2 , correspondingly, the present embodiment also provides a storage medium having a computer program stored thereon, which is executed by a processor to implement the above pellet quality abnormality pre-analysis and optimization method as shown in Figure 1 and Figure 2 .

[0089] Based on such understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.), and includes a plurality of instructions for causing an electronic device (which can be a personal computer, a server, or a network device, etc.) to execute the method of each implementation scenario of the present application.

[0090] Based on the above method as shown in Figure 1 and Figure 2 , and Figure 3In order to achieve the above-mentioned purposes, the electronic device provided by the embodiment of the present application can be a personal computer, a tablet computer, a server, or other network devices, etc., which comprises a storage medium and a processor; the storage medium is used for storing a computer program; and the processor is used for executing the computer program to realize the above-mentioned Figure 1 and Figure 2 The balling quality abnormality pre-analysis and optimization method shown in the figure.

[0091] Optionally, the entity device can further comprise a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface can comprise a display, an input unit such as a keyboard, etc. The optional user interface can further comprise a USB interface, a card reader interface, etc. The network interface can optionally comprise a standard wired interface, a wireless interface (such as a WI-FI interface), etc.

[0092] Those skilled in the art can understand that the above-mentioned structure of the entity device provided by the embodiment does not constitute a limitation on the entity device, and can comprise more or fewer components, or combine certain components, or different component arrangements.

[0093] The storage medium can further comprise an operating system and a network communication module. The operating system is a program for managing hardware and software resources of the entity device, and supports the running of an information processing program and other software and / or programs. The network communication module is used for realizing the communication between components in the storage medium, and the communication with other hardware and software in the information processing entity device.

[0094] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary general hardware platforms, or by hardware.

[0095] The embodiment of the present application can predict the prediction values of different balling quality indexes by adopting the index prediction model, can discover the balling quality problems in time, and thus solves the problem of detection lag in the prior art. At the same time, the embodiment of the present application can optimize the balling quality by constructing the operation suggestion matrix, can solve the problems in the balling generation process in time, ensures the stability of the balling quality, improves the production efficiency of the balling, and avoids causing production loss.

[0096] Those skilled in the art can understand that the modules or flows in the drawings are not necessarily required for implementing the present application. Those skilled in the art can understand that the modules in the devices in the implementation scenarios can be distributed in the devices in the implementation scenarios according to the description of the implementation scenarios, or can be changed to be located in one or more devices different from the implementation scenarios. The modules in the above implementation scenarios can be combined into one module, or can be further split into multiple sub-modules.

[0097] The above application numbers are only for description, and do not represent the advantages and disadvantages of the implementation scenarios. The above disclosure is only some specific implementation scenarios of the present application, but the present application is not limited thereto, and any variations that can be thought of by those skilled in the art should fall within the protection scope of the present application.

Claims

1. A method for analyzing and optimizing the quality of a balling process, characterized in that, The method comprises the following steps: obtaining the pellet production data of the current time node and the historical time node respectively; determining the historical time series data corresponding to different pellet process key parameters respectively according to the pellet production data of the current time node and the pellet production data of the historical time node; inputting the historical time series data corresponding to the different pellet process key parameters into the corresponding index prediction model respectively for index prediction to obtain the predicted values of different pellet quality indexes, wherein the predicted values of different pellet quality indexes include the predicted value of total iron, the predicted value of silicon dioxide, the predicted value of compressive strength and the predicted value of tumbler strength; if it is determined that the pellet quality is abnormal according to the predicted values of different pellet quality indexes, constructing an operation suggestion matrix based on adjustable operation parameters, determining multiple common parameters in the first pellet process key parameter, the second pellet process key parameter, the third pellet process key parameter and the fourth pellet process key parameter when determining the adjustable operation parameters; collecting a sample data set of the multiple common parameters and different pellet quality indexes, constructing a parameter evaluation model for each pellet quality index; for each parameter evaluation model, excluding the parameter with the lowest weight each time, repeating the process until a preset number of key parameters are left; taking the intersection of the key parameters output by each parameter evaluation model to obtain the adjustable operation parameters; optimizing the pellet quality based on the operation suggestion matrix.

2. The method of claim 1, wherein, When the different pellet process key parameters include the first pellet process key parameter, the second pellet process key parameter, the third pellet process key parameter and the fourth pellet process key parameter, the method of determining the historical time series data corresponding to different pellet process key parameters respectively according to the pellet production data of the current time node and the pellet production data of the historical time node comprises the following steps: determining the historical time series data corresponding to the first pellet process key parameter, the second pellet process key parameter, the third pellet process key parameter and the fourth pellet process key parameter respectively according to the pellet production data of the current time node and the pellet production data of the historical time node; the method of inputting the historical time series data corresponding to the different pellet process key parameters into the corresponding index prediction model respectively for index prediction to obtain the predicted values of different pellet quality indexes comprises the following steps: inputting the historical time series data corresponding to the first pellet process key parameter into a preset total iron prediction model for prediction to obtain the predicted value of total iron of the pellet; inputting the historical time series data corresponding to the second pellet process key parameter into a preset silicon dioxide prediction model for prediction to obtain the predicted value of silicon dioxide of the pellet; inputting the historical time series data corresponding to the third pellet process key parameter into a preset pellet compressive strength prediction model for prediction to obtain the predicted value of compressive strength of the pellet; inputting the historical time series data corresponding to the fourth pellet process key parameter into a preset pellet tumbler strength prediction model for prediction to obtain the predicted value of tumbler strength of the pellet.

3. The method of claim 1, wherein, After the historical time series data corresponding to the different pellet process key parameters are respectively input into the corresponding index prediction model for index prediction to obtain the predicted values of different pellet quality indexes, the method further comprises: determining the prediction error of the different pellet quality indexes of the previous time node corresponding to the current time node; compensating the predicted values of the different pellet quality indexes of the current time node based on the prediction error to obtain the compensated predicted values of the different pellet quality indexes; if it is determined that the pellet quality is abnormal according to the predicted values of the different pellet quality indexes, then an operation suggestion matrix is constructed based on adjustable operation parameters, comprising: if it is determined that the pellet quality is abnormal according to the compensated predicted values of the different pellet quality indexes, then an operation suggestion matrix is constructed based on adjustable operation parameters.

4. The method of claim 1, wherein, if any one of the total iron predicted value, the silicon dioxide predicted value, the compressive strength predicted value and the drum strength predicted value exceeds the corresponding preset index range, it is determined that the pellet quality is abnormal; the cold air flow parameter, the cold air pressure parameter, the south combustion chamber pressure parameter and the north combustion chamber pressure parameter are taken as the operation parameters, and an operation suggestion matrix is constructed according to the operation range and adjustment step length of the operation parameters. the optimization of the pellet quality based on the operation suggestion matrix comprises:

5. The method of claim 1, wherein, the index prediction model is used to predict the optimization result of each operation suggestion in the operation suggestion matrix; according to the optimization result of each operation suggestion, a pre-operation optimization suggestion is output to optimize the pellet quality. Before the historical time series data corresponding to the different pellet process key parameters are respectively input into the corresponding index prediction model for index prediction to obtain the predicted values of different pellet quality indexes, the method further comprises:

6. The method according to any one of claims 1 to 5, characterized in that, obtaining sample pellet production data; determining sample historical time series data corresponding to each pellet process parameter according to the sample pellet production data; based on the sample historical time series data corresponding to each pellet process parameter, the correlation analysis of the sample historical time series data corresponding to each pellet process parameter is performed to determine the pellet process key parameters associated with the different pellet quality indexes; determining a sample training set according to the sample historical time series data corresponding to the pellet process key parameters associated with the different pellet quality indexes; training different index prediction models based on the sample training set. the obtaining of the sample pellet production data comprises:

7. The method of claim 6, wherein, obtaining original pellet production data; performing abnormality detection and missing data filling on the original pellet production data respectively to obtain the sample pellet production data, wherein different missing data filling strategies are adopted for different missing proportions of the original pellet production data when performing missing data filling. comprises:

8. A device for analyzing and optimizing the quality of a pellet before an abnormality occurs, characterized by, an acquisition unit, configured to acquire pellet production data of a current time node and historical time nodes respectively; ​ A determination unit is configured to determine historical time series data corresponding to different pellet process key parameters respectively according to the current time node pellet production data and the historical time node pellet production data; A prediction unit is configured to input the historical time series data corresponding to the different pellet process key parameters into corresponding index prediction models respectively for index prediction to obtain predicted values of different pellet quality indexes, wherein the predicted values of the different pellet quality indexes include total iron predicted values, silicon dioxide predicted values, compressive strength predicted values and drum strength predicted values; A construction unit is configured to, if it is determined that the pellet quality is abnormal according to the predicted values of the different pellet quality indexes, construct an operation suggestion matrix based on adjustable operation parameters, determine multiple common parameters in a first pellet process key parameter, a second pellet process key parameter, a third pellet process key parameter and a fourth pellet process key parameter when the adjustable operation parameters are determined, collect a sample data set of the multiple common parameters and different pellet quality indexes, construct a parameter evaluation model for each pellet quality index, exclude the parameter with the lowest weight each time for each parameter evaluation model, repeat the process until a preset number of key parameters are left, and obtain the adjustable operation parameters by taking the intersection of the key parameters output by each parameter evaluation model; An optimization unit is configured to optimize the pellet quality based on the operation suggestion matrix.

9. A storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the method in any one of claims 1 to 7.

10. An electronic device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, The processor executes the computer program to implement the method in any one of claims 1 to 7.

Images