Calcination control method, device, equipment and medium of active lime calcination kiln

Abstract

The application discloses a calcination control method, device, equipment and medium of an active lime calcination kiln, and relates to the technical field of industry.The method comprises the following steps: constructing multi-source operation data by using temperature variables, process operation variables, flue gas variables and finished product quality detection data of the calcination kiln in a calcination process, pre-processing the multi-source operation data, and determining a state vector by feature construction; determining a current calcination maturity by using a calcination maturity soft measurement model, and determining a target risk; determining a current working condition mode by using an inductive working condition identification model, and determining a probability ranking result of each candidate cause by using a traceable root cause diagnosis model; determining a future action result vector of a candidate control action by using a cause-effect action deduction model, determining a target candidate control action by using a target boundary constraint; and determining an optimal control action based on the comprehensive cost of the target candidate control action, so as to perform a calcination control operation. The calcination maturity is accurately perceived, and the optimization control of the calcination kiln is realized.

Description

Technical Field

[0001] This invention relates to the field of industrial technology, and in particular to a method, apparatus, equipment and medium for controlling the calcination of active lime in a calcining kiln. Background Technology

[0002] Quicklime is a crucial auxiliary material in steelmaking processes, including converter slagging, desulfurization and dephosphorization, and ladle refining. Its quality stability directly impacts the production efficiency and metallurgical effects of the steelmaking process. The calcination process of quicklime is typically completed in rotary kilns, sleeve kilns, or double-chamber vertical kilns. The precision of thermal control during calcination directly determines core quality indicators of the finished lime, such as its activity and residual carbon dioxide content, and is also related to fuel consumption, pollutant emissions, and equipment lifespan. Steel companies have continuous, stable, and quality-constrained demands for quicklime. On the one hand, insufficient lime activity affects slagging speed, desulfurization and dephosphorization efficiency, and slag fluidity. On the other hand, excessively increasing the calcination temperature, while potentially reducing underburning, increases fuel consumption, reduces activity, exacerbates thermal damage to refractory materials, and may also lead to increased nitrogen oxide emissions.

[0003] Existing control schemes typically equate the temperature at a single point in the kiln with the calcination maturity of the material. However, the actual calcination effect is influenced by multiple factors, including material particle size, kiln speed, bed thickness, fuel calorific value, combustion air volume, kiln atmosphere, and residence time in the high-temperature zone. The same kiln temperature reading can result in significant differences in the degree of calcium carbonate decomposition under different feeding conditions and operating conditions. Adjusting fuel supply solely based on temperature can easily lead to problems such as the material reaching the required temperature but undergoing incomplete under-burning or excessively high temperatures causing over-burning and decreased activity, thus failing to consistently guarantee the quality of the finished product. Core quality indicators of the finished lime, such as activity and residual carbon dioxide, require multiple offline testing processes including sampling, cooling, sample preparation, and analysis, resulting in long testing cycles and making them unsuitable for real-time control feedback. Operators can only make judgments based on indirect characteristics such as kiln temperature, flue gas conditions, flame morphology, and finished product appearance, combined with experience. This leads to significant delays in the detection and control of quality anomalies, a high risk of producing substandard products, and large fluctuations in kiln conditions. Existing control schemes typically only output adjustment actions and cannot differentiate or diagnose the causes of quality anomalies. For example, the abnormal phenomenon of elevated residual carbon dioxide may be caused by a variety of reasons, such as insufficient heat in the calcination zone, excessive kiln speed, increased feed particle size, decreased fuel calorific value, insufficient combustion air supply, or a sudden increase in feed volume. The optimal control direction varies depending on the cause. If the root cause cannot be accurately located, incorrect control strategies are easily adopted, which may not only fail to improve quality but also exacerbate kiln condition disorder.

[0004] As can be seen from the above, how to accurately perceive the calcination maturity and achieve optimized control of the calcination kiln is an urgent problem to be solved. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method, apparatus, equipment, and medium for controlling the calcination of active lime in a calcining kiln, capable of accurately sensing the calcination maturity and achieving optimized control of the calcining kiln. The specific solution is as follows: In a first aspect, this application provides a method for controlling the calcination of an active lime calcining kiln, comprising: Multi-source operational data is constructed using temperature variables, process operation variables, flue gas variables, and finished product quality detection data during the calcination process of the active lime calcining kiln. The multi-source operational data is preprocessed and feature constructed to obtain trend features. The state vector is determined using the multi-source operational data and the trend features. Based on the state vector and using a soft measurement model of calcination maturity, the current calcination maturity is determined. Based on the current calcination maturity and a preset maturity threshold, the target risk is determined. The target risk includes under-calcination risk and over-calcination risk. Based on the state vector, the current calcination maturity, and the target risk, and using an inductive working condition identification model, the current working condition mode is determined. Based on the current working condition mode and the state vector, and using a root cause diagnosis model, the probability ranking results corresponding to each candidate cause are determined. The corresponding candidate control actions are then generated using the probability ranking results. Based on the state vector and the candidate control actions, and using a causal action inference model, the action result vector of each candidate control action within the future prediction window is obtained. Based on the action result vectors corresponding to the candidate control actions, the target candidate control actions are determined using target boundary constraints. The target boundary constraints include constraints based on the calcination zone temperature boundary, kiln pressure boundary, contaminant boundary, underburning / overburning risk boundary, kiln shell or refractory temperature boundary, and ring formation risk boundary. Based on the comprehensive cost corresponding to the target candidate control action, a weighted sum is obtained using a minimum intervention decision maker to obtain the optimal control action. Based on the optimal control action, the corresponding calcination control operation is performed on the active lime calcining kiln.

[0006] Optionally, the step of constructing multi-source operational data using temperature variables, process operation variables, flue gas variables, and finished product quality inspection data during the calcination process of the active lime calcining kiln, preprocessing and feature constructing the multi-source operational data to obtain trend features, and determining the state vector using the multi-source operational data and the trend features includes: Multi-source operational data was constructed using temperature variables, process operation variables, flue gas variables, and finished product quality testing data during the calcination process of the quicklime calcining kiln. Time alignment, missing value and outlier processing are performed on multi-source running data with different sampling periods to obtain processed running data, and trend features are constructed based on the processed running data; A state vector is constructed using the multi-source operational data, the trend features, the hysteresis quality label, and the sensor reliability. The temperature variables include kiln head temperature, kiln tail temperature, calcination zone temperature, preheating zone temperature, cooling zone temperature, kiln shell temperature, and refractory material temperature; the process operation variables include kiln speed, feed rate, feed particle size distribution, feed moisture, limestone calcium carbonate content, limestone impurity content, fuel flow rate, fuel calorific value, combustion air volume, primary air volume, secondary air volume, kiln pressure, and induced draft fan frequency; the flue gas variables include flue gas oxygen content, carbon monoxide concentration, carbon dioxide concentration, nitrogen oxide concentration, flue gas temperature, flue gas flow rate, and flue gas pressure; the finished product quality inspection data includes finished product activity, residual carbon dioxide content, calcium oxide content, finished product particle size, finished product temperature, and pulverization rate; the trend characteristics include the calcination zone temperature change rate, kiln head temperature change rate, residual carbon dioxide change trend, activity change trend, fuel adjustment range, kiln speed change rate, oxygen content change rate, and carbon monoxide rise rate; the hysteresis quality label is a label determined based on the finished product quality inspection data.

[0007] Optionally, the step of determining the current calcination maturity based on the state vector and using a soft-sensor model of calcination maturity, and determining the target risk based on the current calcination maturity and a preset maturity threshold, includes: The state vector corresponding to the sampling period within the target historical window is input into the calcination maturity soft measurement model to obtain the current calcination maturity; If the current calcination maturity exceeds the upper limit of the preset maturity threshold, then the target risk is characterized as over-calcination risk. If the current calcination maturity does not exceed the lower limit of the preset maturity threshold, then the target risk is characterized as under-calcination risk.

[0008] Optionally, the step of determining the current operating condition mode based on the state vector, the current calcination maturity, and the target risk using an inductive operating condition identification model, determining the probability ranking results corresponding to each candidate cause based on the current operating condition mode and the state vector using a root cause diagnosis model, and generating corresponding candidate control actions using the probability ranking results includes: Based on historical multi-source operating data, and using clustering models, time series classification models, or attention networks, an inductive operating condition identification model is determined. The state vector, the current calcination maturity, and the target risk are input into the inductive working condition identification model to obtain the current working condition mode; Based on the current operating mode and the state vector, the probability of each candidate cause in the cause set and the corresponding probability ranking result are determined using the root cause diagnosis model; the root cause diagnosis model is a model determined based on Bayesian network, causal graph model or neural network classifier. Based on the probability ranking results, target candidate causes whose probabilities exceed the target probability threshold are determined, and corresponding candidate control actions are generated using the target candidate causes.

[0009] Optionally, the step of inferring the action result vector of each candidate control action within a future prediction window based on the state vector and the candidate control actions using a causal action inference model, and determining the target candidate control action based on the action result vectors corresponding to the candidate control actions and using target boundary constraints, includes: The state vector and the candidate control actions are input into the causal action inference model to obtain the action result vector of each candidate control action within the future prediction window; the action result vector includes the change in maturity, the change in underburning risk, the change in overburning risk, the change in fuel consumption per unit area, the change in nitrogen oxides, the change in kiln pressure, the change in refractory temperature, and the change in ring formation risk. Target boundary constraints are constructed using preset process rules and the target safe operation boundary of the active lime calcining kiln. Target candidate control actions are determined based on the action result vectors corresponding to the candidate control actions and using the target boundary constraints.

[0010] Optionally, determining the target candidate control action based on the action result vector corresponding to the candidate control action and using the target boundary constraints includes: If the action result vector corresponding to the candidate control action violates any of the constraints in the target boundary constraints, then the candidate control action that violates any of the constraints in the target boundary constraints is eliminated to obtain the target candidate control action.

[0011] Optionally, the step of obtaining the optimal control action by weighted summation based on the comprehensive costs corresponding to the target candidate control actions using a minimum intervention decision maker includes: Determine the comprehensive cost corresponding to each of the target candidate control actions; the comprehensive cost includes the cost of underburning risk, the cost of overburning risk, the cost of fuel consumption per unit area, the cost of nitrogen oxide emissions, the cost of equipment risk, and the cost of weighted action amplitude; The optimal control action is determined based on the comprehensive cost and using the objective optimization function in the minimum intervention decision maker.

[0012] Secondly, this application provides a calcination control device for an active lime calcination kiln, comprising: The state vector determination module is used to construct multi-source operating data using temperature variables, process operation variables, flue gas variables and finished product quality detection data during the calcination process of the active lime calcining kiln, preprocess and feature construction of the multi-source operating data to obtain trend features, and determine the state vector using the multi-source operating data and the trend features. The target risk determination module is used to determine the current calcination maturity based on the state vector and using a soft measurement model of calcination maturity, and to determine the target risk based on the current calcination maturity and a preset maturity threshold; the target risk includes under-calcination risk and over-calcination risk; The control action generation module is used to determine the current working condition mode based on the state vector, the current calcination maturity and the target risk using an inductive working condition identification model, determine the probability ranking result corresponding to each candidate cause based on the current working condition mode and the state vector using a root cause diagnosis model, and generate corresponding candidate control actions using the probability ranking result. The target action determination module is used to perform inference based on the state vector and the candidate control actions using a causal action inference model to obtain the action result vector of each candidate control action within a future prediction window. Based on the action result vectors corresponding to the candidate control actions, the module determines the target candidate control actions using target boundary constraints. The target boundary constraints include constraints based on calcination zone temperature boundaries, kiln pressure boundaries, contaminant boundaries, under-burning / over-burning risk boundaries, kiln shell or refractory temperature boundaries, and ring formation risk boundaries. The optimal action determination module is used to obtain the optimal control action by weighted summation based on the comprehensive cost corresponding to the target candidate control action and using the minimum intervention decision-maker, and to perform corresponding calcination control operations on the active lime calcining kiln based on the optimal control action.

[0013] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute the computer program to implement the aforementioned calcination control method for an active lime calcining kiln.

[0014] Fourthly, this application provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned calcination control method for an active lime calcining kiln.

[0015] This application utilizes temperature variables, process operation variables, flue gas variables, and finished product quality inspection data from the calcination process of an active lime calcining kiln to construct multi-source operational data. This multi-source operational data undergoes preprocessing and feature construction to obtain trend features. A state vector is determined using the multi-source operational data and the trend features. Based on the state vector and a soft-sensor model of calcination maturity, the current calcination maturity is determined. A target risk is determined based on the current calcination maturity and a preset maturity threshold. The target risk includes under-calcination risk and over-calcination risk. Based on the state vector, the current calcination maturity, and the target risk, an inductive operating condition identification model is used to determine the current operating condition mode. Based on the current operating condition mode and the state vector, a root cause diagnosis model is used to determine the probability ranking of each candidate cause. The system first ranks the results and generates corresponding candidate control actions using the probability ranking results. Then, it uses a causal action deduction model to deduce the action result vectors of each candidate control action within a future prediction window. Based on the action result vectors corresponding to the candidate control actions, it determines the target candidate control action using target boundary constraints. These target boundary constraints include constraints based on calcination zone temperature boundaries, kiln pressure boundaries, contaminant boundaries, under-burning / over-burning risk boundaries, kiln shell or refractory temperature boundaries, and ring formation risk boundaries. Finally, it uses a minimum intervention decision maker to perform a weighted summation based on the comprehensive cost corresponding to the target candidate control action to obtain the optimal control action. Based on the optimal control action, it performs corresponding calcination control operations on the active lime calcining kiln.

[0016] As shown above, this application collects multi-source data on temperature, process operation, flue gas, and finished product quality. After preprocessing and trend feature construction, the data is integrated to form a standardized state vector. Using the state vector as input, the current calcination maturity is output through a soft measurement model of calcination maturity. Combined with a preset maturity threshold, the risks of under-calcination and over-calcination are determined. Based on the state vector, current calcination maturity, and risk identification, the current operating mode is determined. Then, a root cause diagnosis model is used to output the probability ranking of each candidate cause, thereby generating candidate control actions. A causal action inference model is used to predict the future action result vector of each candidate action. Finally, non-compliant actions are screened out through constraints of temperature, kiln pressure, pollutants, quality, equipment, and ring boundary, resulting in a set of compliant target candidate actions. In this way, the multi-dimensional comprehensive cost weighted calculation of the target candidate actions selects the optimal action with the lowest comprehensive cost for the calcination process, reducing fuel waste and avoiding secondary disturbances to the kiln condition caused by large adjustments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is a flowchart of a calcination control method for an active lime calcination kiln disclosed in this application; Figure 2 This is a schematic diagram of the calcination control device of an active lime calcination kiln disclosed in this application; Figure 3 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0019] 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.

[0020] Currently, existing control schemes typically only output adjustment actions and cannot differentiate or diagnose the causes of quality anomalies. For example, the abnormal phenomenon of elevated residual carbon dioxide may be caused by a variety of reasons, such as insufficient heat in the calcination zone, excessive kiln speed, increased feed particle size, decreased fuel calorific value, insufficient combustion air supply, or a sudden increase in feed volume. The optimal control direction varies depending on the cause. If the root cause cannot be accurately located, incorrect control strategies are easily adopted, which may not only fail to improve quality but also exacerbate kiln condition disturbances. To address this, this application provides a calcination control method for an active lime calcination kiln. It performs a multi-dimensional comprehensive cost weighted calculation of the target candidate actions and selects the optimal action with the lowest comprehensive cost for the calcination process, reducing fuel waste and avoiding secondary disturbances to the kiln condition caused by large adjustments.

[0021] See Figure 1 As shown in the figure, an embodiment of the present invention discloses a method for controlling the calcination of an active lime calcining kiln, comprising: Step S11: Construct multi-source operating data using temperature variables, process operation variables, flue gas variables, and finished product quality detection data during the calcination process of the active lime calcining kiln. Preprocess and feature construct the multi-source operating data to obtain trend features. Determine the state vector using the multi-source operating data and the trend features.

[0022] In this embodiment, temperature variables such as kiln head temperature, kiln tail temperature, calcination zone temperature, preheating zone temperature, cooling zone temperature, kiln shell temperature, and refractory material temperature are collected; process operation variables such as kiln speed, feed rate, feed particle size distribution, feed moisture, limestone calcium carbonate content, limestone impurity content, fuel flow rate, fuel calorific value, combustion air volume, primary air volume, secondary air volume, kiln pressure, and induced draft fan frequency are collected; flue gas variables such as flue gas oxygen content, carbon monoxide concentration, carbon dioxide concentration, nitrogen oxide concentration, flue gas temperature, flue gas flow rate, and flue gas pressure are collected; finished product quality inspection data such as finished product activity, residual carbon dioxide content, calcium oxide content, finished product particle size, finished product temperature, pulverization rate, and manual quality judgment results are collected; offline detection data are correlated with the corresponding operating window according to the sampling time period. Multi-source operating data is constituted based on the temperature variables, process operation variables, flue gas variables, and finished product quality inspection data.

[0023] Understandably, time alignment is performed on multi-source operational data from different sampling periods; sliding window statistics are used for continuous temperature and flue gas data, and a backtracking correlation method is used for offline quality data to map the quality results to their corresponding kiln operating states. Missing and outlier values ​​in the multi-source operational data are processed; short-term missing values ​​are compensated for using adjacent measuring points and historical trends; long-term missing values ​​are marked as low confidence; data that is significantly out of range or during maintenance is not included in model training. After the above data preprocessing, processed operational data is obtained, and trend features are constructed based on the processed operational data.

[0024] Furthermore, a state vector is constructed based on the kiln condition variables within the current timeframe and historical window. This state vector includes not only the current value but also the trend, hysteresis quality label, and sensor reliability. In other words, the state vector is constructed using the multi-source operational data, the trend characteristics, the hysteresis quality label, and the sensor reliability. The expression for the state vector is as follows: ; in, This is the state vector at the current time t; This refers to the kiln head temperature. This refers to the temperature at the kiln tail. The calcination zone represents the temperature; This refers to the temperature of the kiln shell. For kiln speed; Fuel flow rate; To increase the air volume for combustion; This refers to the feed rate; This refers to the oxygen content of the flue gas. This refers to the carbon monoxide concentration. This refers to the concentration of carbon dioxide. This refers to the concentration of nitrogen oxides. For kiln pressing; Characteristics of feed particle size; Moisture content of the feed material; The activity level of the finished product; Residual carbon dioxide; For sensor reliability; This refers to the trend characteristic.

[0025] Specifically, the process involves constructing multi-source operational data using temperature variables, process operation variables, flue gas variables, and finished product quality inspection data during the calcination process of an active lime calcining kiln. This multi-source operational data undergoes preprocessing and feature construction to obtain trend features. A state vector is then determined using the multi-source operational data and the trend features. This process includes: constructing multi-source operational data using temperature variables, process operation variables, flue gas variables, and finished product quality inspection data during the calcination process of an active lime calcining kiln; performing time alignment, missing value and outlier processing on multi-source operational data from different sampling periods to obtain processed operational data; constructing trend features based on the processed operational data; and constructing a state vector using the multi-source operational data, the trend features, hysteresis quality labels, and sensor reliability. The temperature variables include kiln head temperature, kiln tail temperature, calcination zone temperature, and preheating zone temperature. The temperature of the cooling zone, the kiln shell, and the refractory material; the process operating variables include kiln speed, feed rate, feed particle size distribution, feed moisture, limestone calcium carbonate content, limestone impurity content, fuel flow rate, fuel calorific value, combustion air volume, primary air volume, secondary air volume, kiln pressure, and induced draft fan frequency; the flue gas variables include flue gas oxygen content, carbon monoxide concentration, carbon dioxide concentration, nitrogen oxide concentration, flue gas temperature, flue gas flow rate, and flue gas pressure; the finished product quality inspection data includes finished product activity, residual carbon dioxide content, calcium oxide content, finished product particle size, finished product temperature, and pulverization rate; the trend characteristics include the calcination zone temperature change rate, kiln head temperature change rate, residual carbon dioxide change trend, activity change trend, fuel adjustment range, kiln speed change rate, oxygen content change rate, and carbon monoxide rise rate; the hysteresis quality label is a label determined based on the finished product quality inspection data.

[0026] In this embodiment, the action vector is defined as the upper-level control correction amount, rather than the lower-level valve opening. The expression for the action vector is as follows: ; in, This is the action vector at the current time t; Fuel quantity correction; Adjustment of combustion air volume; For kiln speed correction; Adjustment for feed rate; Corrected for target temperature; This is a correction to the kiln pressure control target. Using a correction vector can reduce the impact on the underlying control loop at the site. The action vector must satisfy the following constraints: ; ; in, This is the action vector at the current time t; , These are the lower and upper limits allowed for each component of the action vector, respectively; This represents the maximum allowable change in motion between adjacent control cycles. This is the action vector of the previous control cycle. This constraint means that each action must be within the allowable range of the equipment, and the change in action between adjacent control cycles must not exceed the allowable rate of change.

[0027] Step S12: Determine the current calcination maturity based on the state vector and using the calcination maturity soft measurement model; determine the target risk based on the current calcination maturity and the preset maturity threshold; the target risk includes under-calcination risk and over-calcination risk.

[0028] In this embodiment, the calcination maturity soft measurement model is used to estimate the material decomposition and sintering state online. Since activity and residual carbon dioxide are typically not available in real time, the maturity soft measurement model is trained using historical quality monitoring results and real-time kiln condition variables. Specifically, the state vector corresponding to the sampling period within the target historical window is input into the calcination maturity soft measurement model to obtain the current calcination maturity, as shown in the following formula: ; in, The current calcination maturity can be normalized to between 0 and 1; A soft measurement model for calcination maturity; This refers to the state vector within the historical window spanning the past L sampling periods, i.e., the state vector corresponding to the sampling period within the target historical window, where L is the length of the historical window. The current calcination maturity is described as follows: too low indicates a tendency towards under-calcination, moderate indicates normal maturity, and too high indicates a tendency towards over-calcination.

[0029] It is understood that the target risk is determined based on the current calcination maturity and a preset maturity threshold, whereby the preset maturity threshold includes a lower maturity limit and an upper maturity limit. If the current calcination maturity exceeds the upper maturity limit of the preset maturity threshold, the target risk is characterized as over-calcination risk; if the current calcination maturity does not exceed the lower maturity limit of the preset maturity threshold, the target risk is characterized as under-calcination risk. The formulas corresponding to under-calcination risk and over-calcination risk are as follows: ; ; in, To avoid the risk of burning; To avoid the risk of overheating; To correct the linear unit function; , These are the lower and upper limits of maturity in the preset maturity thresholds, respectively; Residual carbon dioxide; For residual carbon dioxide control limits; Target activity; The activity level of the finished product; The calcination zone represents the temperature; This is the upper limit of the safe temperature for the calcination zone; , , These are the weighting coefficients for the residual carbon dioxide exceeding the limit, the insufficient activity, and the calcination zone temperature exceeding the limit, respectively.

[0030] Specifically, determining the current calcination maturity based on the state vector and using a soft-sensing model of calcination maturity, and determining the target risk based on the current calcination maturity and a preset maturity threshold, includes: inputting the state vector corresponding to the sampling period within the target historical window into the soft-sensing model of calcination maturity to obtain the current calcination maturity; if the current calcination maturity exceeds the upper limit of maturity in the preset maturity threshold, the target risk is characterized as over-calcination risk; if the current calcination maturity does not exceed the lower limit of maturity in the preset maturity threshold, the target risk is characterized as under-calcination risk.

[0031] Step S13: Based on the state vector, the current calcination maturity, and the target risk, determine the current operating condition mode using an inductive operating condition identification model; based on the current operating condition mode and the state vector, determine the probability ranking results corresponding to each candidate cause using a root cause diagnosis model; and generate corresponding candidate control actions using the probability ranking results.

[0032] In this embodiment, the inductive operating condition identification model learns typical operating condition patterns from historical operating data. Training samples are constructed based on calcination maturity, under-calcination risk, over-calcination risk, temperature field, flue gas composition, and quality detection results. Clustering models, time series classification models, or attention networks are used to identify operating conditions. Then, the state vector, the current calcination maturity, and the target risk are input into the inductive operating condition identification model to obtain the current operating condition pattern, as shown in the following formula: ; in, This is the current operating mode; This is the inductive working condition identification model; This is the state vector within the historical window of the past L sampling periods; This represents the current calcination maturity. To avoid the risk of burning; To avoid the risk of overheating.

[0033] In one specific implementation, the current operating mode includes at least the following types: normal stable, low-temperature underfiring, insufficient residence time, high-temperature overfiring, uneven thermal field, insufficient combustion, air-fuel ratio mismatch, feed disturbance, abnormal kiln speed, and low sensor reliability. Typical performance and potential impacts of some of these operating modes are shown in Table 1 below.

[0034] Table 1 Typical Operating Conditions

[0035] Understandably, the root cause diagnostic model is used to deduce the most likely cause from abnormal results. A set of causes is established, which includes at least the following: insufficient heat in the calcination zone, excessive kiln speed, excessively slow kiln speed, decreased fuel calorific value, excessive fuel quantity, insufficient combustion air, excessive combustion air, increased feed particle size, increased feed moisture content, sudden increase in feed quantity, uneven heat field distribution, abnormal kiln pressure, and sensor malfunction. Based on the current operating mode and state vector, the probability corresponding to each candidate cause in the cause set is determined using the root cause diagnostic model, as shown in the following formula: ; in, Let be the probability that the i-th candidate cause is true in the current state; Let i be the i-th candidate cause in the cause set C; The current calcination maturity; To avoid the risk of burning; To avoid the risk of overheating; The root cause diagnosis model is implemented using Bayesian networks, causal graph scoring models, causal graph models, similar case retrieval models, or neural network classifiers. This is the state vector at the current time t; This refers to the current operating mode.

[0036] Furthermore, the candidate causes in the cause set are sorted based on the probabilities to obtain corresponding probability ranking results. Based on these ranking results, target candidate causes with probabilities exceeding the target probability threshold are identified, and corresponding candidate control actions are generated using these target candidate causes. These candidate control actions include, but are not limited to, fine-tuning of fuel quantity, combustion air, kiln speed, feed rate, and target temperature. For example, when residual carbon dioxide levels rise and maturity is low, instead of directly increasing fuel, the root cause is first determined. If both particle size increase and kiln speed increase exist simultaneously, insufficient residence time is identified. If fuel flow remains unchanged but calcination zone temperature decreases and carbon monoxide levels increase, incomplete combustion or decreased fuel calorific value is identified. If temperature measurement points contradict each other and quality detection is normal, the output sensor may be malfunctioning.

[0037] Specifically, the process of determining the current operating condition mode based on the state vector, the current calcination maturity, and the target risk using an inductive operating condition identification model, determining the probability ranking results corresponding to each candidate cause based on the current operating condition mode and the state vector using a root cause diagnosis model, and generating corresponding candidate control actions using the probability ranking results includes: determining an inductive operating condition identification model based on historical multi-source operating data using a clustering model, a time series classification model, or an attention network; inputting the state vector, the current calcination maturity, and the target risk into the inductive operating condition identification model to obtain the current operating condition mode; determining the probability of each candidate cause in the cause set and the corresponding probability ranking results based on the current operating condition mode and the state vector using a root cause diagnosis model; the root cause diagnosis model is a model determined based on a Bayesian network, a causal graph model, or a neural network classifier; determining target candidate causes whose probabilities exceed a target probability threshold based on the probability ranking results, and generating corresponding candidate control actions using the target candidate causes.

[0038] Step S14: Based on the state vector and the candidate control actions, and using the causal action inference model, perform inference to obtain the action result vector of each candidate control action within the future prediction window. Based on the action result vector corresponding to the candidate control action, determine the target candidate control action using target boundary constraints. The target boundary constraints include constraints based on the calcination zone temperature boundary, kiln pressure boundary, contaminant boundary, underburning / overburning risk boundary, kiln shell or refractory temperature boundary, and ring formation risk boundary.

[0039] In this embodiment, the causal action inference model is used to predict the consequences after the execution of candidate control actions. The causal action inference model is trained based on historical data, simulation data, and expert rules. Its inputs are the current state and candidate control actions, and its outputs are the changes in maturity, under-firing risk, over-firing risk, energy consumption, nitrogen oxides, kiln pressure, and equipment risk within a future window, as shown in the following formula: ; in, In the current state After executing candidate control action a, the action result vector within the next H sampling periods; This is the causal action inference model. The action result vector includes changes in maturity, underburning risk, overburning risk, fuel consumption per unit area, nitrogen oxides, kiln pressure, refractory temperature, and ring formation risk. The causal action inference model is used to answer the question of what would happen if certain adjustments were made. For example, reducing the kiln speed may reduce the underburning risk, but may also increase the overburning risk and yield loss; increasing fuel may improve maturity, but will increase energy consumption, nitrogen oxides, and refractory heat load; increasing combustion air may improve combustion, but excessive amounts may remove heat or increase local temperature.

[0040] It is understandable that target boundary constraints are constructed using preset process rules and the target safe operation boundary of the active lime calcining kiln. Whether an action is permitted to be executed is determined based on the process mechanism, equipment specifications, safety regulations, and environmental protection requirements. The target boundary constraints are shown in the following formula: ; ; ; ; ; in, The equivalent temperature or representative temperature of the calcined zone material, predicted by the soft measurement model of calcination maturity or the causal action inference model, at the predicted time t+H in the future; , These are the lower limit and upper limit of the permissible calcination temperature, respectively; This is used to prevent insufficient calcination temperature from causing incomplete decomposition of limestone and increasing the risk of under-calcination. Used to prevent overheating, decreased activity, damage to refractory materials, or increased energy consumption caused by excessively high calcination temperatures; The kiln pressure is the kiln head pressure, kiln tail pressure, or a representative kiln pressure obtained by merging multiple pressure points at the predicted time t+H in the future. , These are the lower and upper limits of the permissible kiln pressure, respectively; Predict the concentration of nitrogen oxides in flue gas at the future prediction time t+H; The upper limit or control target limit for nitrogen oxides; Predict the carbon monoxide concentration in flue gas at the predicted time t+H in the future; This refers to the upper limit of permissible or control target for carbon monoxide. The risk value of under-firing at the predicted time t+H can be calculated by comprehensively considering variables such as calcination zone temperature, residence time, limestone particle size, feed rate, kiln speed, residual carbon dioxide content, and soft measurement value of product activity. The maximum permissible risk level for undercooking; The overburning risk value at the predicted time t+H can be calculated by comprehensively considering variables such as the duration of high temperature in the calcination zone, fuel quantity, kiln speed, material residence time, the decreasing trend of product activity, and the risk of calcium oxide grain growth. The upper limit of the allowable risk of overheating; the risks of underheating and overheating can be represented by a normalized risk value between 0 and 1, or by a continuous substitution value or risk level score. The refractory heat load index is the kiln shell temperature, the equivalent temperature of the refractory material, or the refractory heat load index calculated by the kiln shell scanner and the refractory thermal model at the predicted time t+H in the future. This is the upper limit of the allowable temperature for the kiln shell or refractory material, used to prevent local overheating, increased refractory material corrosion, abnormal kiln lining peeling, or increased equipment safety risks caused by the candidate operation. The risk value of ring formation at the predicted time t+H can be calculated from variables such as kiln temperature distribution, duration of local high temperature, kiln speed, material adhesion tendency, flue gas composition, pressure difference change, kiln current fluctuation and similarity of historical ring formation samples. The upper limit of the allowable risk of ring formation is used to prevent control actions from causing local melting, material adhesion, reduced ventilation cross section in the kiln, or decreased long-term stable operation capability.

[0041] It is worth mentioning that the above-mentioned allowable upper and lower limits can be determined based on process design parameters, equipment manuals, safe operating procedures, environmental emission requirements, historical statistics of excellent operating conditions, and expert experience. For limestone of different grades, different particle size ranges, different fuel calorific values, and different production loads, the above boundary values ​​can also be set as dynamic boundaries, automatically calling the corresponding constraint parameters according to the current operating conditions. If the action result vector corresponding to the candidate control action violates any of the above target boundary constraints, the candidate control action that violates any of the target boundary constraints is eliminated to obtain the target candidate control action; if the candidate control action only causes a slight quasi-constraint risk, the action is retained and enters the subsequent minimum intervention decision, while adding a penalty to its comprehensive cost.

[0042] Specifically, the step of using the state vector and the candidate control actions to deduce the action result vector of each candidate control action within a future prediction window, and determining the target candidate control action based on the action result vectors corresponding to the candidate control actions and using target boundary constraints, includes: inputting the state vector and the candidate control actions into the causal action deduction model to obtain the action result vector of each candidate control action within a future prediction window; the action result vector includes changes in maturity, changes in underburning risk, changes in overburning risk, changes in fuel consumption per unit area, changes in nitrogen oxides, changes in kiln pressure, changes in refractory temperature, and changes in ring formation risk; constructing target boundary constraints using preset process rules and the target safe operation boundary of the active lime calcining kiln; and determining the target candidate control action based on the action result vectors corresponding to the candidate control actions and using the target boundary constraints.

[0043] Specifically, determining the target candidate control action based on the action result vector corresponding to the candidate control action and using the target boundary constraints includes: if the action result vector corresponding to the candidate control action violates any constraint in the target boundary constraints, then the candidate control action that violates any constraint in the target boundary constraints is eliminated to obtain the target candidate control action.

[0044] Step S15: Based on the comprehensive cost corresponding to the target candidate control action, a weighted sum is performed using a minimum intervention decision maker to obtain the optimal control action. Based on the optimal control action, the corresponding calcination control operation is performed on the active lime calcining kiln.

[0045] In this embodiment, the comprehensive cost corresponding to each target candidate control action is determined. The comprehensive cost includes the cost of underburning risk, the cost of overburning risk, the cost of fuel consumption per unit area, the cost of nitrogen oxide emissions, the cost of equipment risk, and the cost of weighted action amplitude. Then, based on the comprehensive cost and using the objective optimization function in the minimum intervention decision-maker, a weighted sum is performed to determine the optimal control action. The objective of the minimum intervention decision-maker is to select the action with the smallest action amplitude and the greatest reduction in comprehensive risk from the safe and feasible candidate actions. The objective optimization function is shown in the following equation: ; in, The optimal control action at the current time t; To execute candidate control actions Risks associated with burning raw food; To execute candidate control actions The risk of overheating afterwards; To execute candidate control actions Fuel consumption per unit afterward; To execute candidate control actions The risk of nitrogen oxides afterward; To execute candidate control actions Post-equipment risks; The range of weighted actions; , , , , , These are the weights corresponding to the costs of underburning risk, overburning risk, fuel consumption per unit area, nitrogen oxide emissions, equipment risk, and weighted action range. For example, when underburning risk is caused by excessive kiln speed, a slight reduction in kiln speed is preferred over a direct and significant increase in fuel; when underburning risk is caused by a decrease in fuel calorific value, fuel quantity compensation or target temperature correction is preferred; when overburning risk is caused by excessively slow kiln speed, a slight increase in kiln speed or a decrease in fuel is preferred over a direct reduction in all thermal parameters.

[0046] Specifically, the step of obtaining the optimal control action by weighted summation based on the comprehensive costs corresponding to the target candidate control actions using a minimum intervention decision-maker includes: determining the comprehensive cost corresponding to each of the target candidate control actions; the comprehensive cost includes underburning risk cost, overburning risk cost, fuel consumption cost, nitrogen oxide emission cost, equipment risk cost, and weighted action amplitude cost; and determining the optimal control action based on the comprehensive cost and using the target optimization function in the minimum intervention decision-maker.

[0047] Furthermore, for ease of understanding, the calcination control method provided in this application will be further explained below with reference to four typical specific embodiments. In the first specific embodiment, the tendency for under-calcination is caused by excessively high kiln speed. Currently, residual carbon dioxide is showing an upward trend, calcination maturity is low, and the temperature of the calcination zone is slightly low but has not significantly exceeded the limit. The inductive condition identification model judges it to be an insufficient residence condition, and the root cause diagnosis model outputs that excessively high kiln speed and excessively large feed particle size are the main causes. Causal action deduction shows that simply increasing fuel can reduce the risk of under-calcination, but will increase energy consumption and nitrogen oxides; slightly reducing the kiln speed can significantly reduce the risk of under-calcination with fewer side effects. Therefore, the system outputs a minimal intervention action of slightly reducing the kiln speed, keeping the fuel quantity unchanged, or slightly compensating.

[0048] In the second specific implementation, the overburning tendency is caused by excessive fuel. Currently, the calcination zone temperature remains consistently high, resulting in decreased activity but low residual carbon dioxide levels. The inductive condition identification model classifies it as a high-temperature overburning condition, while the root cause diagnosis model identifies the main causes as excessive fuel quantity and slow kiln speed. Causal action deduction shows that increasing the kiln speed will increase output but may lead to localized underburning, while slightly reducing fuel and maintaining the kiln speed can reduce the risk of overburning. Therefore, the output actions are to slightly reduce the fuel quantity, maintain the combustion air ratio, and observe the decline in maturity.

[0049] In the third specific implementation, uneven thermal field leads to both localized underburning and localized overburning. Currently, multiple temperature measurement points show abnormal differences, with localized higher kiln shell temperatures and fluctuating finished product quality. The inductive condition identification model identifies this as an uneven thermal field issue, while the root cause diagnosis model identifies flame deviation, fuel-air ratio mismatch, and material layer fluctuation as candidate causes. The system performs causal deduction on fuel quantity, combustion air, and kiln pressure targets, ultimately choosing to slightly adjust the combustion air distribution and kiln pressure targets instead of globally increasing fuel consumption.

[0050] In the fourth specific implementation, sensor malfunctions lead to a risk of misjudgment. The temperature of the calcination zone suddenly drops, but the trends in flue gas carbon dioxide, carbon monoxide, fuel quantity, and finished product quality do not show corresponding changes. The system's sensor reliability is judged to be low, and the root cause diagnosis model lists the temperature measurement point malfunction as a high-probability cause. Instead of immediately and significantly increasing fuel, the system prompts a re-verification of the measurement points and adopts conservative, small-scale control.

[0051] Furthermore, this solution can be applied to rotary lime kilns, sleeve lime kilns, double-chamber vertical kilns, and other high-temperature calcining furnaces. The state and motion variables for different kiln types can be replaced according to the on-site equipment. For example, for vertical kilns, the kiln speed can be replaced with the unloading speed or feeding rhythm; for multi-combustion beam vertical kilns, fuel distribution, air distribution, and reversing cycles can be incorporated into the motion space. This solution can also be extended to high-temperature processes such as pellet roasting, sintering, rotary kiln roasting, and heating furnace homogenization control, but maturity indicators and quality risk indicators should be redefined during extension. This application is primarily used for minimal intervention control of under-burning, over-burning, and maturity during the calcination process of active lime.

[0052] As shown above, this application collects multi-source data on temperature, process operation, flue gas, and finished product quality. After preprocessing and trend feature construction, the data is integrated to form a standardized state vector. Using the state vector as input, the current calcination maturity is output through a soft measurement model of calcination maturity. Combined with a preset maturity threshold, the risks of under-calcination and over-calcination are determined. Based on the state vector, current calcination maturity, and risk identification, the current operating mode is determined. Then, a root cause diagnosis model is used to output the probability ranking of each candidate cause, thereby generating candidate control actions. A causal action inference model is used to predict the future action result vector of each candidate action. Finally, non-compliant actions are screened out through constraints of temperature, kiln pressure, pollutants, quality, equipment, and ring boundary, resulting in a set of compliant target candidate actions. In this way, the multi-dimensional comprehensive cost weighted calculation of the target candidate actions selects the optimal action with the lowest comprehensive cost for the calcination process, reducing fuel waste and avoiding secondary disturbances to the kiln condition caused by large adjustments.

[0053] Accordingly, see Figure 2 As shown, this application also provides a calcination control device for an active lime calcination kiln, comprising: The state vector determination module 11 is used to construct multi-source operating data using temperature variables, process operation variables, flue gas variables and finished product quality detection data during the calcination process of the active lime calcining kiln, preprocess and feature construction of the multi-source operating data to obtain trend features, and determine the state vector using the multi-source operating data and the trend features. The target risk determination module 12 is used to determine the current calcination maturity based on the state vector and using a soft measurement model of calcination maturity, and to determine the target risk based on the current calcination maturity and a preset maturity threshold; the target risk includes under-calcination risk and over-calcination risk; The control action generation module 13 is used to determine the current working condition mode based on the state vector, the current calcination maturity and the target risk using an inductive working condition identification model, determine the probability ranking result corresponding to each candidate cause based on the current working condition mode and the state vector using a root cause diagnosis model, and generate corresponding candidate control actions using the probability ranking result. The target action determination module 14 is used to perform inference based on the state vector and the candidate control actions using a causal action inference model to obtain the action result vector of each candidate control action within the future prediction window, and to determine the target candidate control action based on the action result vector corresponding to the candidate control action and using target boundary constraints; the target boundary constraints include constraints based on the calcination zone temperature boundary, kiln pressure boundary, contaminant boundary, underburning / overburning risk boundary, kiln shell or refractory temperature boundary, and ring formation risk boundary; The optimal action determination module 15 is used to obtain the optimal control action by weighted summation based on the comprehensive cost corresponding to the target candidate control action and using the minimum intervention decision-maker, and to perform corresponding calcination control operations on the active lime calcining kiln based on the optimal control action.

[0054] In some specific embodiments, the state vector determination module 11 may specifically include: The operational data construction unit is used to construct multi-source operational data using temperature variables, process operation variables, flue gas variables, and finished product quality detection data during the calcination process of the active lime calcining kiln; The data processing unit is used to perform time alignment, missing value and outlier processing on multi-source running data with different sampling periods to obtain processed running data, and to construct trend features based on the processed running data. The state vector construction unit is used to construct a state vector using the multi-source operational data, the trend features, the hysteresis quality label, and the sensor credibility.

[0055] In some specific embodiments, the target risk determination module 12 may specifically include: The maturity determination unit is used to input the state vector corresponding to the sampling period within the target historical window into the calcination maturity soft measurement model to obtain the current calcination maturity. The first risk determination unit is used to characterize the target risk as over-burning risk if the current calcination maturity exceeds the upper limit of the preset maturity threshold. The second risk determination unit is used to characterize the target risk as under-calcination risk if the current calcination maturity does not exceed the lower limit of the preset maturity threshold.

[0056] In some specific embodiments, the control action generation module 13 may specifically include: The identification model determination unit is used to determine the inductive working condition identification model based on historical multi-source operating data and by using clustering models, time series classification models or attention networks. The mode determination unit is used to input the state vector, the current calcination maturity and the target risk into the inductive working condition identification model to obtain the current working condition mode; The probability determination unit is used to determine the probability of each candidate cause in the cause set and the corresponding probability ranking result based on the current working condition mode and the state vector and using the root cause diagnosis model; the root cause diagnosis model is a model determined based on Bayesian network, causal graph model or neural network classifier. The candidate action generation unit is used to determine the target candidate cause whose probability exceeds the target probability threshold based on the probability ranking result, and to generate the corresponding candidate control action using the target candidate cause.

[0057] In some specific embodiments, the target action determination module 14 may specifically include: The result vector determination unit is used to input the state vector and the candidate control action into the causal action inference model to obtain the action result vector of each candidate control action within the future prediction window. The target action determination unit is used to construct target boundary constraints using preset process rules and the target safe operation boundary of the active lime calcining kiln, and to determine the target candidate control action based on the action result vector corresponding to the candidate control action and using the target boundary constraints.

[0058] In some specific embodiments, the target action determination module 14 may specifically include: An action elimination unit is used to eliminate candidate control actions that violate any of the target boundary constraints if the action result vector corresponding to the candidate control action violates any of the constraints in the target boundary constraints, so as to obtain the target candidate control action.

[0059] In some specific embodiments, the optimal action determination module 15 may specifically include: The comprehensive cost determination unit is used to determine the comprehensive cost corresponding to each of the target candidate control actions; the comprehensive cost includes the cost of underburning risk, the cost of overburning risk, the cost of fuel consumption per unit area, the cost of nitrogen oxide emissions, the cost of equipment risk, and the cost of weighted action amplitude; The optimal action determination unit is used to determine the optimal control action based on the comprehensive cost and using the objective optimization function in the minimum intervention decision maker.

[0060] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the calcination control method for the active lime calcining kiln disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0061] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0062] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0063] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the calcination control method of the active lime calcining kiln executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0064] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned calcination control method for an active lime calcining kiln. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0065] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0066] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0067] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0068] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0069] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for controlling the calcination of active lime in a calcining kiln, characterized in that, include: Multi-source operational data is constructed using temperature variables, process operation variables, flue gas variables, and finished product quality detection data during the calcination process of the active lime calcining kiln. The multi-source operational data is preprocessed and feature constructed to obtain trend features. The state vector is determined using the multi-source operational data and the trend features. Based on the state vector and using a soft measurement model of calcination maturity, the current calcination maturity is determined. Based on the current calcination maturity and a preset maturity threshold, the target risk is determined. The target risk includes under-calcination risk and over-calcination risk. Based on the state vector, the current calcination maturity, and the target risk, and using an inductive working condition identification model, the current working condition mode is determined. Based on the current working condition mode and the state vector, and using a root cause diagnosis model, the probability ranking results corresponding to each candidate cause are determined. The corresponding candidate control actions are then generated using the probability ranking results. Based on the state vector and the candidate control actions, and using a causal action inference model, the action result vector of each candidate control action within the future prediction window is obtained. Based on the action result vectors corresponding to the candidate control actions, the target candidate control actions are determined using target boundary constraints. The target boundary constraints include constraints based on the calcination zone temperature boundary, kiln pressure boundary, contaminant boundary, underburning / overburning risk boundary, kiln shell or refractory temperature boundary, and ring formation risk boundary. Based on the comprehensive cost corresponding to the target candidate control action, a weighted sum is obtained using a minimum intervention decision maker to obtain the optimal control action. Based on the optimal control action, the corresponding calcination control operation is performed on the active lime calcining kiln.

2. The calcination control method for an active lime calcining kiln according to claim 1, characterized in that, The process involves constructing multi-source operational data using temperature variables, process operation variables, flue gas variables, and finished product quality inspection data during the calcination of active lime in a calcining kiln. This multi-source operational data undergoes preprocessing and feature construction to obtain trend features. A state vector is then determined using the multi-source operational data and the trend features, including: Multi-source operational data was constructed using temperature variables, process operation variables, flue gas variables, and finished product quality testing data during the calcination process of the quicklime calcining kiln. Time alignment, missing value and outlier processing are performed on multi-source running data with different sampling periods to obtain processed running data, and trend features are constructed based on the processed running data; A state vector is constructed using the multi-source operational data, the trend features, the hysteresis quality label, and the sensor reliability. The temperature variables include kiln head temperature, kiln tail temperature, calcination zone temperature, preheating zone temperature, cooling zone temperature, kiln shell temperature, and refractory material temperature; the process operation variables include kiln speed, feed rate, feed particle size distribution, feed moisture, limestone calcium carbonate content, limestone impurity content, fuel flow rate, fuel calorific value, combustion air volume, primary air volume, secondary air volume, kiln pressure, and induced draft fan frequency; the flue gas variables include flue gas oxygen content, carbon monoxide concentration, carbon dioxide concentration, nitrogen oxide concentration, flue gas temperature, flue gas flow rate, and flue gas pressure; the finished product quality inspection data includes finished product activity, residual carbon dioxide content, calcium oxide content, finished product particle size, finished product temperature, and pulverization rate; the trend characteristics include the calcination zone temperature change rate, kiln head temperature change rate, residual carbon dioxide change trend, activity change trend, fuel adjustment range, kiln speed change rate, oxygen content change rate, and carbon monoxide rise rate; the hysteresis quality label is a label determined based on the finished product quality inspection data.

3. The calcination control method for an active lime calcining kiln according to claim 1, characterized in that, The process of determining the current calcination maturity based on the state vector and using a soft-sensor model of calcination maturity, and determining the target risk based on the current calcination maturity and a preset maturity threshold, includes: The state vector corresponding to the sampling period within the target historical window is input into the calcination maturity soft measurement model to obtain the current calcination maturity; If the current calcination maturity exceeds the upper limit of the preset maturity threshold, then the target risk is characterized as over-calcination risk. If the current calcination maturity does not exceed the lower limit of the preset maturity threshold, then the target risk is characterized as under-calcination risk.

4. The calcination control method for an active lime calcining kiln according to claim 1, characterized in that, The process involves determining the current operating condition mode based on the state vector, the current calcination maturity, and the target risk using an inductive operating condition identification model; determining the probability ranking of each candidate cause based on the current operating condition mode and the state vector using a root cause diagnosis model; and generating corresponding candidate control actions using the probability ranking results. This includes: Based on historical multi-source operating data, and using clustering models, time series classification models, or attention networks, an inductive operating condition identification model is determined. The state vector, the current calcination maturity, and the target risk are input into the inductive working condition identification model to obtain the current working condition mode; Based on the current operating mode and the state vector, the probability of each candidate cause in the cause set and the corresponding probability ranking result are determined using the root cause diagnosis model; the root cause diagnosis model is a model determined based on Bayesian network, causal graph model or neural network classifier. Based on the probability ranking results, target candidate causes whose probabilities exceed the target probability threshold are determined, and corresponding candidate control actions are generated using the target candidate causes.

5. The calcination control method for an active lime calcining kiln according to claim 1, characterized in that, The process of inferring the action result vector of each candidate control action within a future prediction window based on the state vector and the candidate control actions using a causal action inference model, and determining the target candidate control action based on the action result vectors corresponding to the candidate control actions and using target boundary constraints, includes: The state vector and the candidate control actions are input into the causal action inference model to obtain the action result vector of each candidate control action within the future prediction window; the action result vector includes the change in maturity, the change in underburning risk, the change in overburning risk, the change in fuel consumption per unit area, the change in nitrogen oxides, the change in kiln pressure, the change in refractory temperature, and the change in ring formation risk. Target boundary constraints are constructed using preset process rules and the target safe operation boundary of the active lime calcining kiln. Target candidate control actions are determined based on the action result vectors corresponding to the candidate control actions and using the target boundary constraints.

6. The calcination control method for an active lime calcining kiln according to claim 1, characterized in that, The step of determining the target candidate control action based on the action result vector corresponding to the candidate control action and using the target boundary constraints includes: If the action result vector corresponding to the candidate control action violates any of the constraints in the target boundary constraints, then the candidate control action that violates any of the constraints in the target boundary constraints is eliminated to obtain the target candidate control action.

7. The calcination control method for an active lime calcining kiln according to any one of claims 1 to 6, characterized in that, The step of obtaining the optimal control action by weighted summation based on the comprehensive costs corresponding to the target candidate control actions using a minimum intervention decision maker includes: Determine the comprehensive cost corresponding to each of the target candidate control actions; the comprehensive cost includes the cost of underburning risk, the cost of overburning risk, the cost of fuel consumption per unit area, the cost of nitrogen oxide emissions, the cost of equipment risk, and the cost of weighted action amplitude; The optimal control action is determined based on the comprehensive cost and using the objective optimization function in the minimum intervention decision maker.

8. A calcination control device for an active lime calcination kiln, characterized in that, include: The state vector determination module is used to construct multi-source operating data using temperature variables, process operation variables, flue gas variables and finished product quality detection data during the calcination process of the active lime calcining kiln, preprocess and feature construction of the multi-source operating data to obtain trend features, and determine the state vector using the multi-source operating data and the trend features. The target risk determination module is used to determine the current calcination maturity based on the state vector and using a soft measurement model of calcination maturity, and to determine the target risk based on the current calcination maturity and a preset maturity threshold; the target risk includes under-calcination risk and over-calcination risk; The control action generation module is used to determine the current working condition mode based on the state vector, the current calcination maturity and the target risk using an inductive working condition identification model, determine the probability ranking result corresponding to each candidate cause based on the current working condition mode and the state vector using a root cause diagnosis model, and generate corresponding candidate control actions using the probability ranking result. The target action determination module is used to perform inference based on the state vector and the candidate control actions using a causal action inference model to obtain the action result vector of each candidate control action within a future prediction window. Based on the action result vectors corresponding to the candidate control actions, the module determines the target candidate control actions using target boundary constraints. The target boundary constraints include constraints based on calcination zone temperature boundaries, kiln pressure boundaries, contaminant boundaries, under-burning / over-burning risk boundaries, kiln shell or refractory temperature boundaries, and ring formation risk boundaries. The optimal action determination module is used to obtain the optimal control action by performing a weighted summation based on the comprehensive cost corresponding to the target candidate control action and using a minimum intervention decision-maker, and to perform corresponding calcination control operations on the active lime calcining kiln based on the optimal control action.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the calcination control method for an active lime calcining kiln as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the calcination control method for an active lime calcining kiln as described in any one of claims 1 to 7.

Images