Blast furnace gas utilization rate time sequence judgment and self-adaptive decision system and method

Abstract

The invention discloses a blast furnace gas utilization rate time sequence judgment and self-adaptive decision making system and method, and the system comprises a blast furnace full ironmaking data processing module which is used for preparing a'material distribution-furnace feeding-operation-monitoring-test 'full-chain dynamic data set based on a computer technology and an ironmaking process theory; the blast furnace gas utilization rate prediction module is used for predicting the gas utilization rate by adopting a machine learning or data mining algorithm based on the historical operation data and the real-time data; the case library dynamic adjustment module is used for performing clustering analysis on historical data of key monitoring targets and matching fluctuation adjustment ranges of the monitoring targets under different coal gas utilization rate grades to form a case library; the self-adaptive feedback control module is used for setting a corresponding control strategy based on an output result of the prediction module; and the integration module is used for integrating the modules into a closed-loop control system to realize time sequence judgment and self-adaptive decision of the blast furnace gas utilization rate.

Description

Technical Field

[0001] This invention relates to a time-series judgment and adaptive decision-making system and method for blast furnace gas utilization rate. Background Technology

[0002] The utilization rate of blast furnace gas is often affected by many factors. Traditional gas utilization rate regulation usually relies on experience and lacks accurate prediction and real-time feedback control methods. While some existing technologies attempt to predict gas utilization rates, they often fail to provide operational optimization feedback after prediction. For example, Chinese patent CN202411122475.8 uses a regression model to predict gas utilization rates but only provides early warning information; Chinese patent CN201910147686.X only considers gas utilization rate prediction without providing early warnings or suggestions. All of these factors lead to delayed on-site adjustments, resulting in lower gas utilization rates and higher energy consumption, thus affecting the efficiency and economy of blast furnace production. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a time-series judgment and adaptive decision-making system and method for blast furnace gas utilization rate.

[0004] To solve the above technical problems, the technical solution of the present invention is as follows: A time-series judgment and adaptive decision-making system for blast furnace gas utilization rate includes: The blast furnace full ironmaking data processing module prepares a dynamic dataset covering the entire chain of "charging-furnace feeding-operation-monitoring-analysis" based on computer technology and ironmaking process theory; The blast furnace gas utilization rate prediction module uses machine learning or data mining algorithms to predict the utilization rate of gas based on historical and real-time operating data. The variables input to the prediction module include parameters such as the composition, temperature, pressure, and flow rate of the blast furnace gas, and the output is the predicted gas utilization rate value. The dynamic adjustment case library module performs cluster analysis on historical data of key monitoring targets, matches the fluctuation adjustment range of monitoring targets under different gas utilization rate levels, and forms a case library; The adaptive feedback control module sets the corresponding control strategy based on the output of the prediction module; An integration module is used to integrate the above modules into a closed-loop control system to realize the timing judgment and adaptive decision-making of blast furnace gas utilization rate.

[0005] The present invention further defines the technical solution as follows: Preferably, the blast furnace full ironmaking data processing module specifically processes the following: Capture the latest time node coke / ore feeding matrix data, complete stacking and integration processing to form the first data source; dynamically identify the material type and feed batch data to form the second data source; obtain the current time operation data to form the third data source; obtain the current time monitoring data and perform dimensionality reduction processing on the same elevation temperature and pressure data to form the fourth data source; Using the time of the second data source as the timeline, the furnace entry numbers of the first and second data sources are integrated to form the first dynamic dataset; the third and fourth data sources are aligned based on the first dynamic dataset to form the second dynamic dataset; Based on the on-site test component shift calculation rules, the test data is screened and matched with the batch data of the blast furnace charge in the second dynamic dataset. After calculation by ironmaking process theory, the derived data of the blast furnace charge is obtained to form the third dynamic dataset. The third dynamic dataset is processed by time delay in the way of load connection by coke ratio and coal ratio to form the fourth dynamic dataset, namely the dynamic dataset of the entire chain of blast furnace "charging-furnace charging-operation-monitoring-testing".

[0006] Preferably, the method for constructing a time-series dynamic prediction model by the blast furnace gas utilization prediction module specifically includes: Based on the full-chain dynamic dataset, feature selection is performed on the data other than the first data source and the second data source to form the first feature set; the first feature set is then rolled back and aligned with the first data source and the second data source to form the second feature set; The predicted target gas utilization rate data is decomposed into a first target derivative set by an adaptive noise mode decomposition method; a first time-series dynamic prediction model is prepared by using a long short-term memory network to capture the long-term dependencies between each sequence in the first target derivative set and the second feature set data; and a second time-series dynamic prediction model is prepared by using a self-attention mechanism Transformer to enhance the ability of the first time-series dynamic prediction model to extract features from more distant time series, thereby forming a second time-series dynamic prediction model that can capture sequence dependencies more accurately.

[0007] Preferably, the method for constructing the dynamically adjusted case library module includes: With the stability of the gas flow distribution in the furnace as the adjustment target, the following expert target sets are set: upper static pressure fluctuation, upper heat load, temperature range of furnace throat steel bricks, four-point range of cross temperature measurement edge, cross temperature measurement W value, cross temperature measurement Z / W value, and gas utilization rate. A first rule base was developed based on experience, considering both time and data fluctuation dimensions; a second rule base was derived by arranging and combining different rules. Based on the second rule base, cluster search of the entire chain dynamic dataset is used to match the fluctuation range of gas utilization rate under different states to form the first case library, and warning level, gas utilization rate level and adjustment suggestions with adjustment priority are set.

[0008] Preferably, the feedback mechanism of the adaptive feedback control module includes: Determine whether the predicted gas utilization rate exceeds the predetermined threshold: if it does not exceed the threshold, no action is taken; if it exceeds the threshold, the predicted gas utilization rate value level and range need to be matched according to the dynamically adjusted case library. Iterate through the differences between the actual values ​​and rule values ​​of each target in the target set and push adjustment strategies for staff to check and verify. If a match cannot be found, the actual value of the data in the target set will be automatically stored and added to the dynamic adjustment case library, and staff will be prompted to supplement expert experience and suggestions.

[0009] Preferably, the integrated module is implemented using computer software technology and database technology, forming an adjustment suggestion push module, a target set data monitoring module, an expert rule dynamic adjustment case library storage module, an expert rule dynamic adjustment case library editing module, and a full-chain dynamic dataset query module that cooperate with the blast furnace full-chain data processing module, the blast furnace gas utilization rate dynamic prediction module, the dynamic adjustment case library module, and the adaptive feedback control module.

[0010] The beneficial effects of this invention are: This invention solves the problems of data dispersion and time sequence misalignment in steel production by using a dynamic data processing method for the entire chain of "material feeding-furnace loading-operation-monitoring-testing" and by using technologies such as furnace loading number alignment and time delay processing. It improves data utilization while meeting process interpretability and conforming to the daily operation guidelines on site. This invention incorporates an expert-experienced rule-based dynamic adjustment case library, combined with a predictive model to form a multi-scale monitoring and adaptive adjustment suggestion model encompassing "human-machine-material-data-loop," promptly pushing out gas stability strategies to reduce energy consumption and lower production costs. With the stability of gas flow distribution within the furnace as its core, it constructs an expert target set including seven indicators such as upper static pressure fluctuations and cross-shaped temperature measurement Z / W values. Through rule arrangement and cluster analysis, a dynamic case library is formed. Compared to traditional single-indicator early warning systems, this multi-dimensional rule library can more accurately pinpoint the root causes of inefficiency, improving strategy targeting by 40%. This invention uses modal decomposition to decompose gas utilization rate and introduces an attention mechanism to enhance the gas utilization rate prediction model's ability to capture long-term sequence features, forming a time-series dynamic prediction model for gas utilization rate, which effectively improves the model's hit rate. Attached Figure Description

[0011] Figure 1 This is a schematic diagram showing the connections between the modules of the present invention; Figure 2 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0012] To make the content of this invention easier to understand, the invention will be further described in detail below based on specific embodiments. Example

[0013] like Figure 1-2 As shown, this embodiment provides a blast furnace gas utilization rate time-series judgment and adaptive decision-making system, including: The blast furnace full ironmaking data processing module prepares a dynamic dataset covering the entire chain of "charging-furnace feeding-operation-monitoring-analysis" based on computer technology and ironmaking process theory; The blast furnace gas utilization rate prediction module uses machine learning or data mining algorithms to predict the utilization rate of gas based on historical and real-time operating data. The variables input to the prediction module include parameters such as the composition, temperature, pressure, and flow rate of the blast furnace gas, and the output is the predicted gas utilization rate value. The dynamic adjustment case library module performs cluster analysis on historical data of key monitoring targets, matches the fluctuation adjustment range of monitoring targets under different gas utilization rate levels, and forms a case library; The adaptive feedback control module sets the corresponding control strategy based on the output of the prediction module; An integration module is used to integrate the above modules into a closed-loop control system, enabling time-series judgment and adaptive decision-making regarding blast furnace gas utilization. The specific processing tasks handled by the aforementioned blast furnace full ironmaking data processing module include: The target set data monitoring module captures the latest time node coke / ore feeding matrix data, completes stacking and integration processing to form the first data source; dynamically identifies the material type and feeding batch data of the furnace charge to form the second data source; obtains the operation data of the current time to form the third data source; obtains the monitoring data of the current time and performs dimensionality reduction processing on the temperature and pressure data at the same elevation to form the fourth data source. Using the time of the second data source as the timeline, the furnace entry numbers of the first and second data sources are integrated to form the first dynamic dataset; the third and fourth data sources are aligned based on the first dynamic dataset to form the second dynamic dataset; Based on the on-site test component shift calculation rules, the test data is screened and matched with the batch data of the blast furnace charge in the second dynamic dataset. After calculation by ironmaking process theory, the derived data of the blast furnace charge is obtained to form the third dynamic dataset. The third dynamic dataset is processed by time delay in the way of load connection by coke ratio and coal ratio to form the fourth dynamic dataset, namely the dynamic dataset of the entire chain of blast furnace "charging-furnace charging-operation-monitoring-testing".

[0014] The method for constructing a time-series dynamic prediction model using the aforementioned blast furnace gas utilization prediction module specifically includes: Based on the full-chain dynamic dataset, feature selection is performed on the data other than the first data source and the second data source to form the first feature set; the first feature set is then rolled back and aligned with the first data source and the second data source to form the second feature set; The predicted target gas utilization rate data is decomposed into a first target derivative set by an adaptive noise mode decomposition method; a first time-series dynamic prediction model is prepared by using a long short-term memory network to capture the long-term dependencies between each sequence in the first target derivative set and the second feature set data; and a second time-series dynamic prediction model is prepared by using a self-attention mechanism Transformer to enhance the ability of the first time-series dynamic prediction model to extract features from more distant time series, thereby forming a second time-series dynamic prediction model that can capture sequence dependencies more accurately.

[0015] The construction methods for the aforementioned dynamically adjustable case library module include: With the stability of the gas flow distribution in the furnace as the adjustment target, the following expert target sets are set: upper static pressure fluctuation, upper heat load, temperature range of furnace throat steel bricks, four-point range of cross temperature measurement edge, cross temperature measurement W value, cross temperature measurement Z / W value, and gas utilization rate. A first rule base was developed based on experience, considering both time and data fluctuation dimensions; a second rule base was derived by arranging and combining different rules. Based on the second rule base, cluster search of the entire chain dynamic dataset is used to match the fluctuation range of gas utilization rate under different states to form the first case library, and warning level, gas utilization rate level and adjustment suggestions with adjustment priority are set.

[0016] The feedback mechanism of the aforementioned adaptive feedback control module includes: Determine whether the predicted gas utilization rate exceeds the predetermined threshold: if it does not exceed the threshold, no action is taken; if it exceeds the threshold, the predicted gas utilization rate value level and range need to be matched according to the dynamically adjusted case library. Iterate through the differences between the actual values ​​and rule values ​​of each target in the target set and push adjustment strategies for staff to check and verify. If a match cannot be found, the actual value of the data in the target set will be automatically stored and added to the dynamic adjustment case library, and staff will be prompted to supplement expert experience and suggestions.

[0017] The aforementioned integrated modules are implemented using computer software and database technologies, forming an adjustment suggestion push module, target set data monitoring module, expert rule dynamic adjustment case library storage module, expert rule dynamic adjustment case library editing module, and full-chain dynamic dataset query module that work in conjunction with the blast furnace full-chain data processing module, blast furnace gas utilization dynamic prediction module, dynamic adjustment case library module, and adaptive feedback control module.

[0018] In addition to the above embodiments, the present invention may have other implementation methods; all technical solutions formed by equivalent substitution or equivalent transformation fall within the protection scope claimed by the present invention.

Claims

1. A time-series judgment and adaptive decision-making system for blast furnace gas utilization rate, characterized in that: include: The blast furnace full ironmaking data processing module prepares a dynamic dataset covering the entire chain of "charging-furnace feeding-operation-monitoring-analysis" based on computer technology and ironmaking process theory; The blast furnace gas utilization rate prediction module uses machine learning or data mining algorithms to predict the utilization rate of gas based on historical and real-time operating data. The variables input to the prediction module include parameters such as the composition, temperature, pressure, and flow rate of the blast furnace gas, and the output is the predicted gas utilization rate value. The dynamic adjustment case library module performs cluster analysis on historical data of key monitoring targets, matches the fluctuation adjustment range of monitoring targets under different gas utilization rate levels, and forms a case library; The adaptive feedback control module sets the corresponding control strategy based on the output of the prediction module; An integration module is used to integrate the above modules into a closed-loop control system to realize the timing judgment and adaptive decision-making of blast furnace gas utilization rate.

2. The blast furnace gas utilization rate time-series judgment and adaptive decision-making system according to claim 1, characterized in that: The blast furnace full ironmaking data processing module specifically processes the following: Capture the latest time node coke / ore feeding matrix data, complete stacking and integration processing to form the first data source; dynamically identify the material type and feed batch data to form the second data source; obtain the current time operation data to form the third data source; obtain the current time monitoring data and perform dimensionality reduction processing on the same elevation temperature and pressure data to form the fourth data source; Using the time of the second data source as the timeline, the furnace entry numbers of the first and second data sources are integrated to form the first dynamic dataset; the third and fourth data sources are aligned based on the first dynamic dataset to form the second dynamic dataset; Based on the on-site test component shift calculation rules, the test data is screened and matched with the batch data of the blast furnace charge in the second dynamic dataset. After calculation by ironmaking process theory, the derived data of the blast furnace charge is obtained to form the third dynamic dataset. The third dynamic dataset is processed by time delay in the way of load connection by coke ratio and coal ratio to form the fourth dynamic dataset, namely the dynamic dataset of the entire chain of blast furnace "charging-furnace charging-operation-monitoring-testing".

3. The blast furnace gas utilization rate time-series judgment and adaptive decision-making system according to claim 1, characterized in that: The method for constructing a time-series dynamic prediction model using the blast furnace gas utilization prediction module specifically includes: Based on the full-chain dynamic dataset, feature selection is performed on the data other than the first data source and the second data source to form the first feature set; the first feature set is then rolled back and aligned with the first data source and the second data source to form the second feature set; The predicted target gas utilization rate data is decomposed into a first target derivative set by an adaptive noise mode decomposition method; a first time-series dynamic prediction model is prepared by using a long short-term memory network to capture the long-term dependencies between each sequence in the first target derivative set and the second feature set data; and a second time-series dynamic prediction model is prepared by using a self-attention mechanism Transformer to enhance the ability of the first time-series dynamic prediction model to extract features from more distant time series, thereby forming a second time-series dynamic prediction model that can capture sequence dependencies more accurately.

4. The blast furnace gas utilization rate time-series judgment and adaptive decision-making system according to claim 1, characterized in that: The method for constructing the dynamically adjusted case library module includes: With the stability of the gas flow distribution in the furnace as the adjustment target, the following expert target sets are set: upper static pressure fluctuation, upper heat load, temperature range of furnace throat steel bricks, four-point range of cross temperature measurement edge, cross temperature measurement W value, cross temperature measurement Z / W value, and gas utilization rate. A first rule base was developed based on experience, considering both time and data fluctuation dimensions; a second rule base was derived by arranging and combining different rules. Based on the second rule base, cluster search of the entire chain dynamic dataset is used to match the fluctuation range of gas utilization rate under different states to form the first case library, and warning level, gas utilization rate level and adjustment suggestions with adjustment priority are set.

5. The blast furnace gas utilization rate time-series judgment and adaptive decision-making system according to claim 1, characterized in that: The feedback mechanism of the adaptive feedback control module includes: Determine whether the predicted gas utilization rate exceeds the predetermined threshold: if it does not exceed the threshold, no action is taken; if it exceeds the threshold, the predicted gas utilization rate value level and range need to be matched according to the dynamically adjusted case library. Iterate through the differences between the actual values ​​and rule values ​​of each target in the target set and push adjustment strategies for staff to check and verify. If a match cannot be found, the actual value of the data in the target set will be automatically stored and added to the dynamic adjustment case library, and staff will be prompted to supplement expert experience and suggestions.

6. The blast furnace gas utilization rate time-series judgment and adaptive decision-making system according to claim 1, characterized in that: The integrated module is implemented using computer software and database technologies, forming an adjustment suggestion push module, target set data monitoring module, expert rule dynamic adjustment case library storage module, expert rule dynamic adjustment case library editing module, and full-chain dynamic dataset query module that work in conjunction with the blast furnace full-chain data processing module, blast furnace gas utilization dynamic prediction module, dynamic adjustment case library module, and adaptive feedback control module.

Images