A hot blast stove automatic control method, device and equipment based on machine learning

Abstract

The present application provides a hot blast stove automatic control method, device and equipment based on machine learning, comprising: in response to the hot blast stove starting instruction, starting the equipment according to the average air-fuel ratio and the maximum gas flow of the last combustion, and dividing the running process into multiple stages according to the dome temperature; in the current combustion stage, real-time collection of running data and input of corresponding target prediction model, the model is trained by historical stage temperature data, and the predicted dome temperature and flue temperature are output; combining the predicted temperature, the historical air-fuel ratio and the gas flow parameter, the current optimal gas flow value is dynamically calculated, so as to accurately control the hot blast stove running. The method realizes the staged temperature prediction and flow dynamic optimization of the combustion process through the machine learning model, effectively improves the combustion efficiency, reduces the energy consumption, enhances the system stability, and realizes the automatic and efficient control of the hot blast stove.

Description

Technical Field

[0001] This invention relates to the field of automatic control of hot blast stoves, and in particular to an automatic control method, apparatus and equipment for hot blast stoves based on machine learning. Background Technology

[0002] A hot blast stove is a device that provides hot air to aid combustion in a blast furnace. In early industrial production, it was mostly operated manually, and the combustion control of the hot blast stove varied due to differences in worker operation. The hot blast stove itself is a nonlinear system with a large time lag, and its control process is complex. Various influencing factors are interdependent, making it difficult to describe with a precise mathematical model. Achieving automatic control of the hot blast stove can reduce manual intervention, improve combustion efficiency, reduce energy consumption, and simultaneously reduce operational difficulty and labor intensity. Summary of the Invention

[0003] The purpose of this invention is to provide an automatic control method, device, and equipment for hot blast stoves based on machine learning to alleviate the technical problems of low combustion efficiency, high energy consumption, high operation difficulty, and high labor intensity caused by manual control of hot blast stoves in the prior art. This reduces manual intervention, improves combustion efficiency, reduces energy consumption, and at the same time reduces operation difficulty and labor intensity.

[0004] In a first aspect, embodiments of the present invention provide an automatic control method for a hot blast stove based on machine learning, comprising: responding to a start command of the hot blast stove, calculating the average air-fuel ratio and maximum gas flow rate of the hot blast stove during the previous combustion; starting the hot blast stove based on the average air-fuel ratio and the maximum gas flow rate; monitoring the dome temperature of the hot blast stove in real time, and determining the current combustion stage based on the dome temperature; wherein the combustion stage is a plurality of continuous operating stages divided based on a dome temperature threshold; and collecting the current operating data of the hot blast stove in the current combustion stage; and inputting the current operating data into a target corresponding to the current combustion stage. The target prediction model outputs the predicted dome temperature and predicted flue temperature of the current operating data. The target prediction model is trained based on the following steps: obtaining the target historical operating temperature of the hot blast stove and the flue connected to the hot blast stove corresponding to the current combustion stage; training the preset initial target prediction model with the target historical operating temperature until the preset termination condition is reached to obtain the target prediction model; dynamically adjusting the current gas flow rate according to the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate to control the hot blast stove to operate with the adjusted gas flow rate.

[0005] In a preferred embodiment of the present invention, the combustion stage includes: the initial combustion stage, the steady combustion stage, and the temperature drop stage of the hot air furnace; the target prediction model includes: a first target prediction model corresponding to the initial combustion stage; a second target prediction model corresponding to the steady combustion stage; and a third target prediction model corresponding to the temperature drop stage.

[0006] Among them, the first objective prediction model, the second objective prediction model and the third objective prediction model are all constructed based on gradient boosting decision trees.

[0007] In a preferred embodiment of the present invention, the first target prediction model, the second target prediction model and the third target prediction model are all constructed based on gradient boosting decision trees.

[0008] In a preferred embodiment of the present invention, the construction steps of the first target prediction model include: dividing the historical operating temperature of the first target corresponding to the initial combustion stage into a training set and a test set based on a preset division ratio; initializing a preset weak learner and setting the number of iterations; calculating the negative gradient of each sample in the training set; fitting a CART decision tree based on the number of iterations, the samples, and the negative gradient; calculating the minimum number of leaf nodes in each CART decision tree; updating the weak learner based on the minimum value to obtain a strong learner; and testing the strong learner using the test set until a preset test standard is reached to obtain the first target prediction model.

[0009] In a preferred embodiment of the present invention, when the current combustion stage is the initial combustion stage, the step of dynamically adjusting the current gas flow rate based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate includes: determining whether the predicted dome temperature exceeds a first threshold or the predicted flue temperature exceeds a second threshold; if yes, adjusting the current gas flow rate of the hot blast stove to a first low gas flow rate; the first low gas flow rate is less than the maximum gas flow rate, and the difference between the first low gas flow rate and the maximum gas flow rate is a first preset step size; if no, continuing to drive the hot blast stove with the maximum gas flow rate as the adjusted gas flow rate.

[0010] In a preferred embodiment of the present invention, when the current combustion stage is the steady combustion stage; the current operating data includes: the steady dome temperature and the steady flue temperature corresponding to the steady combustion stage; after the step of collecting the current operating data of the hot blast stove in the current combustion stage, the method includes: setting the time when the steady dome temperature reaches a third threshold as a first time, and setting the time when the steady flue temperature reaches the second threshold as a second time; calculating the steady temperature rise slope of the flue based on the third threshold, the second threshold, the first time, and the second time; after the step of inputting the current operating data into the target prediction model corresponding to the current combustion stage and outputting the predicted dome temperature and the predicted flue temperature of the current combustion stage, the method includes: determining whether the predicted dome temperature exceeds the first threshold or If the predicted flue gas temperature exceeds the second threshold, adjust the current gas flow rate to a second low gas flow rate; the second low gas flow rate is less than the current gas flow rate, and the difference between the two is the first preset step size; if not, calculate the predicted temperature rise slope based on the second threshold, the predicted flue gas temperature, the current combustion time, and the first time; determine if the predicted temperature rise slope is less than the steady temperature rise slope; if so, adjust the current gas flow rate of the hot blast stove to a high gas flow rate; the high gas flow rate is greater than the current gas flow rate, and the difference between the two is the second preset step size; if not, adjust the current gas flow rate of the hot blast stove to a third low gas flow rate; the third low gas flow rate is less than the current gas flow rate, and the difference between the two is the second preset step size.

[0011] In a preferred embodiment of the present invention, when the current combustion stage is the temperature decrease stage; after inputting the current operating data into the target prediction model corresponding to the current combustion stage and outputting the predicted dome temperature and predicted flue temperature of the current combustion stage, the method includes: determining whether the predicted dome temperature exceeds the first threshold or the predicted flue temperature exceeds the second threshold; if yes, adjusting the current gas flow rate to a fourth low gas flow rate; the fourth low gas flow rate is less than the current gas flow rate, and the difference between the fourth low gas flow rate and the current gas flow rate is the second preset step size; if no, determining whether the predicted flue temperature is lower than the fourth threshold; the fourth threshold is less than the second threshold; if yes, adjusting the current gas flow rate of the hot blast stove to the high gas flow rate.

[0012] In a preferred embodiment of the present invention, after the step of dynamically adjusting the current gas flow rate based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate to control the hot blast stove to operate with the adjusted gas flow rate, the method includes: determining whether the combustion time of the hot blast stove has reached the second time; if so, controlling the hot blast stove to stop combustion.

[0013] Secondly, embodiments of the present invention also provide an automatic control device for a hot blast stove based on machine learning, comprising: a start-up module, configured to respond to a start-up command of the hot blast stove and calculate the average air-fuel ratio and maximum gas flow rate of the hot blast stove during the previous combustion; an acquisition module, configured to start the hot blast stove based on the average air-fuel ratio and the maximum gas flow rate; monitor the dome temperature of the hot blast stove in real time and determine the current combustion stage based on the dome temperature; wherein the combustion stage is a plurality of continuous operation stages divided based on a dome temperature threshold; and collect the current operation data of the hot blast stove during the current combustion stage; and a prediction module, configured to input the current operation data into the current... The target prediction model corresponding to the combustion stage outputs the predicted dome temperature and predicted flue temperature of the current operating data. The target prediction model is trained based on the following steps: obtaining the target historical operating temperature of the hot blast stove and the flue connected to the hot blast stove corresponding to the current combustion stage; training the preset initial target prediction model with the target historical operating temperature until the preset termination condition is reached to obtain the target prediction model; the control module is used to dynamically adjust the current gas flow rate according to the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate, so as to control the hot blast stove to operate with the adjusted gas flow rate.

[0014] Thirdly, embodiments of the present invention also provide an electronic device, which includes a processor and a memory. The memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the above-mentioned automatic control method for a hot blast stove based on machine learning.

[0015] The embodiments of the present invention have the following beneficial technical effects:

[0016] This invention provides an automatic control method, apparatus, and device for a hot blast stove based on machine learning, comprising: responding to a start command of the hot blast stove, calculating the average air-fuel ratio and maximum gas flow rate of the hot blast stove during the previous combustion; starting the hot blast stove based on the average air-fuel ratio and the maximum gas flow rate; monitoring the dome temperature of the hot blast stove in real time, and determining the current combustion stage based on the dome temperature; wherein the combustion stage is a plurality of continuous operation stages divided based on a dome temperature threshold; and collecting the current operation data of the hot blast stove in the current combustion stage; and inputting the current operation data into a corresponding input method. The target prediction model outputs the predicted dome temperature and predicted flue temperature based on the current operating data. This model is trained using the following steps: obtaining the target historical operating temperatures of the hot blast stove and the flue connected to it for the current combustion stage; training the initial target prediction model using these historical operating temperatures until a preset termination condition is met, thus obtaining the target prediction model; and dynamically adjusting the current gas flow rate based on the predicted dome temperature, predicted flue temperature, average air-fuel ratio, and maximum gas flow rate to control the hot blast stove to operate with the adjusted gas flow rate. This method, by combining a machine learning prediction model trained with historical operating data, achieves accurate temperature prediction and dynamic gas flow rate optimization control during the hot blast stove combustion stage, significantly improving combustion efficiency, reducing energy consumption, and enhancing system operational stability. Attached Figure Description

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

[0018] Figure 1 A flowchart illustrating an automatic control method for a hot blast stove based on machine learning, provided in an embodiment of the present invention;

[0019] Figure 2 A schematic diagram illustrating the construction process of a first target prediction model provided in an embodiment of the present invention;

[0020] Figure 3 A schematic flowchart of an automatic control device for a hot blast stove based on machine learning is provided in an embodiment of the present invention.

[0021] Figure 4 This is a flowchart illustrating an electronic device provided in an embodiment of the present invention.

[0022] Icons: 31-Startup module; 32-Acquisition module; 33-Prediction module; 34-Control module; 41-Processor; 42-Memory; 43-Bus; 44-Communication interface. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0024] The hot blast stove itself is a nonlinear system with a large time lag. Its control process is complex, and various influencing factors are mutually restrictive, making it difficult to describe with a precise mathematical model. Achieving automatic control of the hot blast stove can reduce manual intervention, improve combustion efficiency, reduce energy consumption, and at the same time reduce the difficulty of operation and labor intensity.

[0025] Based on this, embodiments of the present invention provide an automatic control method, apparatus, and device for a hot blast stove based on machine learning. This method, by combining a machine learning prediction model trained with historical operating data, achieves accurate temperature prediction and dynamic gas flow optimization control during the combustion stage of the hot blast stove, significantly improving combustion efficiency, reducing energy consumption, and enhancing system operational stability. For ease of understanding, an automatic control method for a hot blast stove based on machine learning is first described in detail.

[0026] Example 1

[0027] In an embodiment of the present invention, Figure 1 This is a flowchart illustrating an automatic control method for a hot blast stove based on machine learning, provided in an embodiment of the present invention.

[0028] Depend on Figure 1 As seen, the method includes:

[0029] Step S101: In response to the start command of the hot blast stove, calculate the average air-fuel ratio and maximum gas flow rate of the hot blast stove during the last combustion.

[0030] Step S102: Based on the above average air-fuel ratio and the above maximum gas flow rate, start the above hot blast stove; monitor the dome temperature of the above hot blast stove in real time, and determine the current combustion stage according to the dome temperature; wherein, the above combustion stage is a number of continuous operation stages divided based on the dome temperature threshold; and collect the current operation data of the above hot blast stove in the above current combustion stage.

[0031] The current operating data includes outdoor temperature at time t during the current combustion stage, gas calorific value, gas flow rate, air flow rate, air-fuel ratio, dome temperature, flue temperature, CO content in flue gas, oxygen content in flue gas, and the duration of the previous air supply.

[0032] In some examples, the combustion stages mentioned above include: the initial combustion stage, the steady combustion stage, and the temperature drop stage of the hot blast stove; the target prediction model mentioned above includes: the first target prediction model corresponding to the initial combustion stage; the second target prediction model corresponding to the steady combustion stage; and the third target prediction model corresponding to the temperature drop stage.

[0033] Step S103: Input the current operating data into the target prediction model corresponding to the current operating data, and output the predicted dome temperature and predicted flue temperature of the current operating data; the target prediction model is trained based on the following steps: obtain the target historical operating temperature of the hot blast stove and the flue connected to the hot blast stove corresponding to the current combustion stage; train the preset initial target prediction model with the target historical operating temperature until the preset termination condition is reached, and obtain the target prediction model.

[0034] In this embodiment, when the current combustion stage is the initial combustion stage, the step of dynamically adjusting the current gas flow rate based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate includes: determining whether the predicted dome temperature exceeds a first threshold or the predicted flue temperature exceeds a second threshold; if so, adjusting the current gas flow rate of the hot blast stove to a first low gas flow rate; the first low gas flow rate is less than the maximum gas flow rate, and the difference between the first low gas flow rate and the maximum gas flow rate is a first preset step size; if not, continuing to drive the hot blast stove with the maximum gas flow rate as the adjusted gas flow rate.

[0035] Furthermore, after the aforementioned initial combustion stage, the combustion process of the hot blast stove enters a steady combustion stage. When the current combustion stage is the steady combustion stage; the aforementioned current operating data includes: the steady dome temperature and the steady flue temperature corresponding to the steady combustion stage; after the step of collecting the current operating data of the hot blast stove in the current combustion stage, the method includes: setting the time when the steady dome temperature reaches a third threshold as a first time, and setting the time when the steady flue temperature reaches the second threshold as a second time; calculating the steady temperature rise slope of the flue based on the third threshold, the second threshold, the first time, and the second time; inputting the aforementioned current operating data into the target prediction model corresponding to the current combustion stage, and outputting the predicted dome temperature and predicted flue temperature of the current combustion stage, the method includes: determining whether the predicted dome temperature is... If the predicted flue gas temperature exceeds the first threshold or the predicted flue gas temperature exceeds the second threshold, adjust the current gas flow rate to the second low gas flow rate. The second low gas flow rate is less than the current gas flow rate, and the difference between the two is the first preset step size. If not, calculate the predicted temperature rise slope based on the second threshold, the predicted flue gas temperature, the current combustion time, and the first time. Determine if the predicted temperature rise slope is less than the steady temperature rise slope. If so, adjust the current gas flow rate of the hot blast stove to the high gas flow rate. The high gas flow rate is greater than the current gas flow rate, and the difference between the two is the second preset step size. If not, adjust the current gas flow rate of the hot blast stove to the third low gas flow rate. The third low gas flow rate is less than the current gas flow rate, and the difference between the two is the second preset step size.

[0036] Furthermore, after the aforementioned steady combustion stage, the combustion process of the hot blast stove enters a temperature decrease stage. When the current combustion stage is the aforementioned temperature decrease stage; after inputting the aforementioned current operating data into the target prediction model corresponding to the aforementioned current combustion stage, and outputting the predicted dome temperature and predicted flue temperature of the aforementioned current combustion stage, the aforementioned method includes: determining whether the aforementioned predicted dome temperature exceeds the aforementioned first threshold or the aforementioned predicted flue temperature exceeds the aforementioned second threshold; if yes, adjusting the aforementioned current gas flow rate to a fourth low gas flow rate; the aforementioned fourth low gas flow rate is less than the aforementioned current gas flow rate, and the difference between the aforementioned fourth low gas flow rate and the aforementioned current gas flow rate is the aforementioned second preset step size; if no, determining whether the aforementioned predicted flue temperature is lower than the fourth threshold; the aforementioned fourth threshold is less than the second threshold; if yes, adjusting the aforementioned current gas flow rate of the hot blast stove to the aforementioned high gas flow rate.

[0037] Step S104: Based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate, dynamically adjust the current gas flow rate to control the hot blast stove to operate with the adjusted gas flow rate.

[0038] In actual operation, after dynamically adjusting the current gas flow rate based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate to control the hot blast stove to operate with the adjusted gas flow rate, the method includes: determining whether the combustion time of the hot blast stove has reached the second time; if so, controlling the hot blast stove to stop combustion.

[0039] Here, the steps described above are used to indicate the end of this combustion cycle and the start of the oven shut-off process.

[0040] This invention provides an automatic control method for a hot blast stove based on machine learning, comprising: responding to a start command of the hot blast stove, calculating the average air-fuel ratio and maximum gas flow rate of the hot blast stove during the previous combustion; starting the hot blast stove based on the average air-fuel ratio and the maximum gas flow rate; monitoring the dome temperature of the hot blast stove in real time, and determining the current combustion stage based on the dome temperature; wherein the combustion stage is a plurality of continuous operation stages divided based on a dome temperature threshold; and collecting the current operation data of the hot blast stove in the current combustion stage; and inputting the current operation data into a target preset corresponding to the current operation data. The prediction model outputs the predicted dome temperature and predicted flue temperature of the current operating data. The target prediction model is trained based on the following steps: obtaining the target historical operating temperatures of the hot blast stove and the flue connected to it corresponding to the current combustion stage; training the preset initial target prediction model using the target historical operating temperatures until a preset termination condition is met, thus obtaining the target prediction model; and dynamically adjusting the current gas flow rate based on the predicted dome temperature, predicted flue temperature, average air-fuel ratio, and maximum gas flow rate to control the hot blast stove to operate with the adjusted gas flow rate. This method, by combining a machine learning prediction model trained with historical operating data, achieves accurate temperature prediction and dynamic gas flow rate optimization control during the hot blast stove combustion stage, significantly improving combustion efficiency, reducing energy consumption, and enhancing system operational stability.

[0041] Example 2

[0042] In this embodiment, the construction processes of the first target prediction model, the second target prediction model, and the third target prediction model are based on the same principle. For ease of understanding, the construction process of the first target prediction model will be introduced first.

[0043] in, Figure 2This is a schematic diagram illustrating the construction process of a first target prediction model provided in an embodiment of the present invention.

[0044] Depend on Figure 2 As seen, the method includes:

[0045] Step S201: Based on the preset division ratio, divide the first target historical operating temperature corresponding to the above-mentioned initial combustion stage into a training set and a test set; and initialize the preset weak learner and set the number of iterations.

[0046] Among them, the first target historical operating temperature includes the outdoor temperature at time t during the initial combustion stage, the calorific value of the gas, the gas flow rate, the air flow rate, the air-fuel ratio, the dome temperature, the flue temperature, the CO content in the flue gas, the oxygen content in the flue gas, and the duration of the previous air supply.

[0047] Furthermore, the data labels corresponding to the historical operating temperature of the first target mentioned above are the crown temperature and flue temperature at time t+3. The data is labeled, and data without labels is removed.

[0048] Furthermore, based on an 8:2 ratio, the historical operating temperatures of the first target are divided into a training set and a test set.

[0049] Step S202: Calculate the negative gradient of each sample in the training set above.

[0050] The training set mentioned above is represented as: The losses are as follows: The number of iterations is W, and i is the sample label. N is the number of samples. The formula for initializing the weak learner is: .

[0051] Where xN is the input of the Nth sample, yN is the output of the Nth sample, and c is the normalization constant. It is a weak learner.

[0052] Furthermore, the negative gradient is calculated based on the following formula:

[0053]

[0054] in, It is a negative gradient.

[0055] Step S203: Fit a CART decision tree based on the above number of iterations, the above samples, and the above negative gradient.

[0056] Here, using , A CART decision tree can be fitted to obtain the first... A decision tree, and its corresponding leaf node region ,in The number of leaf nodes.

[0057] Step S204: Calculate the minimum number of leaf nodes for each of the above CART decision trees.

[0058] Among them, for ,calculate

[0059]

[0060] in, The minimum number of leaf nodes for each of the above CART decision trees.

[0061] Step S205: Update the weak learner based on the minimum value to obtain the strong learner.

[0062] Step S205 above is implemented based on the following formula:

[0063]

[0064] in, For a strong learner, t represents time.

[0065] Step S206: Test the strong learner using the test set until the preset test standard is met, and obtain the first target prediction model.

[0066] This invention provides a process for constructing a first target prediction model, comprising: dividing the historical operating temperature of the first target corresponding to the initial combustion stage into a training set and a test set based on a preset division ratio; initializing a preset weak learner and setting the number of iterations; calculating the negative gradient of each sample in the training set; fitting a CART decision tree based on the number of iterations, the samples, and the negative gradient; calculating the minimum number of leaf nodes in each CART decision tree; updating the weak learner based on the minimum value to obtain a strong learner; and testing the strong learner using the test set until a preset testing standard is met to obtain the first target prediction model. This method, through the iterative optimization mechanism of gradient boosting decision trees combined with dynamic verification of the training and test sets, constructs a high-precision first target prediction model, significantly improving the prediction accuracy and model generalization ability of key parameters in the combustion process, thereby achieving more precise hot blast stove operation control.

[0067] Example 3

[0068] In addition to the above embodiments, for ease of understanding, this application also provides the following practical examples.

[0069] First, in the initial combustion stage, calculate the average air-fuel ratio of the previous combustion cycle as the initial air-fuel ratio, and the maximum gas flow rate as the initial gas flow rate. Starting from the 1st minute of combustion, perform the following operations: The current time is the t1-th minute of combustion. Using the above first target prediction model, predict the vault temperature and flue gas temperature at the (t1 + 3)-th minute in the future. If the predicted vault temperature exceeds °C, the first threshold (upper limit of the vault temperature), or the predicted flue gas temperature exceeds °C, the second threshold (target value of the flue gas temperature when the hot blast stove is well burned), then the gas flow rate is reduced by ×10,000 Nm3 / h. Otherwise, the gas flow rate remains unchanged. If the CO content in the current flue gas is higher than PPm, the air-fuel ratio is increased by . Otherwise, if the oxygen content in the current flue gas is higher than %, the air-fuel ratio is reduced by . If neither condition is met, the air-fuel ratio remains unchanged.

[0070] Then, when the vault temperature reaches °C at the t-th minute of combustion, and the temperature that the flue gas temperature needs to reach at the end of combustion is minutes, and the entire combustion cycle is °C and the entire combustion cycle is minutes, then the rising slope k of the flue gas temperature is:

[0071] . <000018�>Furthermore, in the stable combustion stage of the hot blast stove, that is, the stable combustion stage, the current time is the t1-th minute of combustion. Using the above second target prediction model, predict the vault temperature and flue gas temperature T' at the (t1 + 3)-th minute in the future. If the predicted vault temperature exceeds °C or the predicted flue gas temperature exceeds °C, then the gas flow rate is reduced by ×10,000 Nm3 / h. Otherwise, calculate the predicted rising slope k1 of the flue gas temperature through the following formula:

[0073] ,

[0074] If k1 < k, the gas flow rate is increased by ×1,000 Nm3 / h. If k1 > k, the gas flow rate is reduced by ×1,000 Nm3 / h. If k1 = k, the gas flow rate remains unchanged. If the CO content in the current flue gas is higher than PPm, the air-fuel ratio is increased by . Otherwise, if the oxygen content in the current flue gas is higher than %, the air-fuel ratio is reduced by . If neither condition is met, the air-fuel ratio remains unchanged.

[0075] Furthermore, during the temperature drop phase of combustion in the hot blast stove, the current time is minute t1 of combustion, i.e., the flue gas temperature reaches... After reaching a certain temperature, the third objective prediction model described above is used to predict the crown temperature and flue temperature at the next minute t1+3. If the predicted crown temperature exceeds a certain value... ℃ or predicted flue temperature exceeding At ℃, the gas flow rate decreases. 10,000 Nm3 / h; otherwise, if the flue temperature is not lower than If the temperature is below ℃ (the lowest temperature in the flue when the hot blast stove is fully burned), the gas flow rate remains constant; if it is below this temperature... At ℃, the gas flow rate increases. If neither of these conditions is met, the gas flow rate remains unchanged. If the current CO content in the flue gas is higher than [a certain value] Nm³ / h... PPm, then the air-fuel ratio increases Otherwise, if the oxygen content in the current flue gas is higher than If the air-fuel ratio decreases by %, the air-fuel ratio will also decrease. If neither condition is met, the air-fuel ratio remains unchanged.

[0076] S4: When the combustion time reaches After a few minutes, this combustion cycle ends, and the oven begins to shut off.

[0077] Furthermore, due to the significant temperature difference between outdoor temperatures in northern winters and summers, the data collection period for hot air furnace operation should be no less than one year.

[0078] Furthermore, since frequent adjustments to the hot blast stove can affect its performance, the time interval for calling the automatic control method of the hot blast stove is 3 minutes.

[0079] Furthermore, the first target prediction model, the second target prediction model, and the third target prediction model are trained once a month.

[0080] Example 4

[0081] Based on the above embodiments, Figure 3 This is a schematic diagram of a machine learning-based automatic control device for a hot blast stove, provided as an embodiment of the present invention.

[0082] Depend on Figure 3 As seen, the device includes:

[0083] The start-up module 31 is used to respond to the start-up command of the hot blast stove and calculate the average air-fuel ratio and maximum gas flow rate of the hot blast stove during the last combustion.

[0084] The acquisition module 32 is used to start the hot blast stove based on the average air-fuel ratio and the maximum gas flow rate; monitor the dome temperature of the hot blast stove in real time, and determine the current combustion stage based on the dome temperature; wherein the combustion stage is a series of continuous operation stages divided based on the dome temperature threshold; and collect the current operation data of the hot blast stove in the current combustion stage.

[0085] The prediction module 33 is used to input the current operating data into the target prediction model corresponding to the current combustion stage, and output the predicted dome temperature and predicted flue temperature of the current operating data. The target prediction model is trained based on the following steps: obtaining the target historical operating temperature of the hot blast stove and the flue connected to the hot blast stove corresponding to the current combustion stage; training the preset initial target prediction model with the target historical operating temperature until the preset termination condition is reached, and obtaining the target prediction model.

[0086] The control module 34 is used to dynamically adjust the current gas flow rate based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate, so as to control the hot blast stove to operate with the adjusted gas flow rate.

[0087] The module includes a startup module 31, an acquisition module 32, a prediction module 33, and a control module 34.

[0088] In one embodiment, the prediction module 33 is further configured to divide the first target historical operating temperature corresponding to the initial combustion stage into a training set and a test set based on a preset division ratio; initialize a preset weak learner and set the number of iterations; calculate the negative gradient of each sample in the training set; fit a CART decision tree based on the number of iterations, the samples, and the negative gradient; calculate the minimum number of leaf nodes in each CART decision tree; update the weak learner based on the minimum value to obtain a strong learner; and test the strong learner using the test set until a preset test standard is reached to obtain the first target prediction model.

[0089] In one embodiment, when the current combustion stage is the initial combustion stage, the step of dynamically adjusting the current gas flow rate based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate includes: determining whether the predicted dome temperature exceeds a first threshold or the predicted flue temperature exceeds a second threshold; if so, adjusting the current gas flow rate of the hot blast stove to a first low gas flow rate; the first low gas flow rate is less than the maximum gas flow rate, and the difference between the first low gas flow rate and the maximum gas flow rate is a first preset step size; if not, continuing to drive the hot blast stove with the maximum gas flow rate as the adjusted gas flow rate.

[0090] In one embodiment, when the current combustion stage is the steady combustion stage; the current operating data includes: the steady dome temperature and the steady flue temperature corresponding to the steady combustion stage; the control module 34 is further configured to set the time when the steady dome temperature reaches a third threshold as a first time, and the time when the steady flue temperature reaches a second threshold as a second time; calculate the steady temperature rise slope of the flue based on the third threshold, the second threshold, the first time, and the second time; after inputting the current operating data into the target prediction model corresponding to the current combustion stage and outputting the predicted dome temperature and the predicted flue temperature of the current combustion stage, the method includes: determining whether the predicted dome temperature exceeds the first threshold or whether the predicted flue temperature exceeds the second threshold. A second threshold is set; if so, the current gas flow rate is adjusted to a second low gas flow rate; the second low gas flow rate is less than the current gas flow rate, and the difference between the two is the first preset step size; if not, the predicted temperature rise slope is calculated based on the second threshold, the predicted flue temperature, the current combustion time, and the first time; it is determined whether the predicted temperature rise slope is less than the steady temperature rise slope; if so, the current gas flow rate of the hot blast stove is adjusted to a high gas flow rate; the high gas flow rate is greater than the current gas flow rate, and the difference between the two is the second preset step size; if not, the current gas flow rate of the hot blast stove is adjusted to a third low gas flow rate; the third low gas flow rate is less than the current gas flow rate, and the difference between the two is the second preset step size.

[0091] In one embodiment, the control module 34 is further configured to determine whether the predicted dome temperature exceeds the first threshold or the predicted flue temperature exceeds the second threshold; if so, adjust the current gas flow rate to a fourth low gas flow rate; the fourth low gas flow rate is less than the current gas flow rate, and the difference between the fourth low gas flow rate and the current gas flow rate is the second preset step size; if not, determine whether the predicted flue temperature is lower than the fourth threshold; the fourth threshold is less than the second threshold; if so, adjust the current gas flow rate of the hot blast stove to the high gas flow rate.

[0092] In one embodiment, the control module 34 is further configured to determine whether the combustion time of the hot air furnace has reached the second time; if so, control the hot air furnace to stop combustion.

[0093] The automatic control device for hot blast stove based on machine learning provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned automatic control method for hot blast stove based on machine learning. For the sake of brevity, any parts not mentioned in the embodiments of the automatic control device for hot blast stove based on machine learning can be referred to the corresponding content in the aforementioned method embodiments.

[0094] This invention also provides an electronic device, such as... Figure 4 The diagram shows the structure of the electronic device, which includes a processor 41 and a memory 42. The memory 42 stores machine-executable instructions that can be executed by the processor 41. The processor 41 executes the machine-executable instructions to implement the above-mentioned automatic control method for hot blast stove based on machine learning.

[0095] exist Figure 4 In the illustrated embodiment, the electronic device further includes a bus 43 and a communication interface 44, wherein the processor 41, the communication interface 44, and the memory 42 are connected via the bus.

[0096] The memory 42 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 44 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus can be an ISA bus, PCI bus, or EISA bus, etc. The aforementioned bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0097] Processor 41 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 41 or by software instructions. Processor 41 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory. The processor 41 reads the information in the memory 42 and, in conjunction with its hardware, completes the steps of the machine learning-based automatic control method for hot blast stoves described in the foregoing embodiment.

[0098] This invention also provides a machine-readable storage medium storing machine-executable instructions. When these machine-executable instructions are invoked and executed by a processor, they cause the processor to implement the aforementioned automatic control method for a hot blast stove based on machine learning. For specific implementation details, please refer to the foregoing method embodiments, which will not be repeated here.

[0099] The computer program products of the machine learning-based hot blast stove automatic control method, machine learning-based hot blast stove automatic control device, and electronic device provided in the embodiments of the present invention include a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the machine learning-based hot blast stove automatic control method described in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0100] Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention.

[0101] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0102] Finally, it should be noted that the above embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the scope of the technology disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention.

Claims

1. An automatic control method for a hot blast stove based on machine learning, characterized in that, include: In response to the start-up command of the hot blast stove, calculate the average air-fuel ratio and maximum gas flow rate of the hot blast stove during the last combustion. Based on the average air-fuel ratio and the maximum gas flow rate, the hot blast stove is started. The dome temperature of the hot blast stove is monitored in real time, and the current combustion stage is determined based on the dome temperature; wherein the combustion stage is a series of continuous operation stages divided based on the dome temperature threshold; and the current operation data of the hot blast stove in the current combustion stage is collected. The current operating data is input into the target prediction model corresponding to the current combustion stage, and the predicted dome temperature and predicted flue temperature of the current operating data are output. The target prediction model is trained based on the following steps: obtaining the target historical operating temperature of the hot air furnace and the flue connected to the hot air furnace corresponding to the current combustion stage; training the preset initial target prediction model with the target historical operating temperature until the preset termination condition is reached, and obtaining the target prediction model. Based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate, the current gas flow rate is dynamically adjusted to control the hot blast stove to operate with the adjusted gas flow rate. The combustion stage includes: the steady combustion stage of the hot blast stove; When the current combustion stage is the steady combustion stage; the current operating data includes: the steady dome temperature and the steady flue temperature corresponding to the steady combustion stage; After the step of collecting the current operating data of the hot blast stove during the current combustion stage, the method includes: The time when the steady dome temperature reaches the third threshold is set as the first time, and the time when the steady flue temperature reaches the second threshold is set as the second time. The steady temperature rise slope of the flue is calculated based on the third threshold, the second threshold, the first time, and the second time. After inputting the current operating data into the target prediction model corresponding to the current combustion stage, and outputting the predicted dome temperature and predicted flue temperature for the current combustion stage, the method includes: Determine whether the predicted dome temperature exceeds a first threshold or whether the predicted flue temperature exceeds a second threshold. If so, adjust the current gas flow rate to the second low gas flow rate; the second low gas flow rate is less than the current gas flow rate, and the difference between the two is a first preset step size; If not, calculate the predicted temperature rise slope based on the third threshold, the predicted flue temperature, the current combustion time, and the first time. Determine whether the predicted temperature rise slope is less than the steady temperature rise slope; If so, adjust the current gas flow rate of the hot blast stove to a high gas flow rate; the high gas flow rate is greater than the current gas flow rate, and the difference between the high gas flow rate and the current gas flow rate is a second preset step size; If not, adjust the current gas flow rate of the hot blast stove to the third low gas flow rate; the third low gas flow rate is less than the current gas flow rate, and the difference between the two is the second preset step size.

2. The automatic control method for a hot blast stove based on machine learning according to claim 1, characterized in that, The combustion stages include: the initial combustion stage of the hot air furnace, the steady combustion stage, and the temperature drop stage; the target prediction model includes: a first target prediction model corresponding to the initial combustion stage; a second target prediction model corresponding to the steady combustion stage; and a third target prediction model corresponding to the temperature drop stage.

3. The automatic control method for a hot blast stove based on machine learning according to claim 2, characterized in that, The first target prediction model, the second target prediction model, and the third target prediction model are all constructed based on gradient boosting decision trees.

4. The automatic control method for a hot blast stove based on machine learning according to claim 3, characterized in that, The steps for constructing the first target prediction model include: Based on a preset division ratio, the first target historical operating temperature corresponding to the initial combustion stage is divided into a training set and a test set; and a preset weak learner is initialized and the number of iterations is set. Calculate the negative gradient for each sample in the training set; A CART decision tree is fitted based on the number of iterations, the samples, and the negative gradient. Calculate the minimum number of leaf nodes for each of the CART decision trees; The weak learner is updated based on the minimum value to obtain the strong learner; The strong learner is tested using the test set until it reaches a preset test standard, thus obtaining the first target prediction model.

5. The automatic control method for a hot blast stove based on machine learning according to claim 2, characterized in that, When the current combustion stage is the initial combustion stage; The step of dynamically adjusting the current gas flow rate based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate includes: Determine whether the predicted dome temperature exceeds the first threshold or the predicted flue temperature exceeds the second threshold; If so, adjust the current gas flow rate of the hot blast stove to the first low gas flow rate; the first low gas flow rate is less than the maximum gas flow rate, and the difference between the first low gas flow rate and the maximum gas flow rate is the first preset step size; If not, continue to drive the hot blast stove using the maximum gas flow rate as the adjusted gas flow rate.

6. The automatic control method for a hot blast stove based on machine learning according to claim 5, characterized in that, When the current combustion stage is the temperature decrease stage; After inputting the current operating data into the target prediction model corresponding to the current combustion stage, and outputting the predicted dome temperature and predicted flue temperature for the current combustion stage, the method includes: Determine whether the predicted dome temperature exceeds the first threshold or the predicted flue temperature exceeds the second threshold; If so, adjust the current gas flow rate to the fourth low gas flow rate; the fourth low gas flow rate is less than the current gas flow rate, and the difference between the two is the second preset step size; If not, determine whether the predicted flue temperature is lower than the fourth threshold; the fourth threshold is less than the second threshold. If so, adjust the current gas flow rate of the hot blast stove to the high gas flow rate.

7. The automatic control method for a hot blast stove based on machine learning according to claim 6, characterized in that, After the step of dynamically adjusting the current gas flow rate based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate to control the hot blast stove to operate with the adjusted gas flow rate, the method includes: Determine whether the combustion time of the hot air furnace has reached the second time; If so, control the hot air furnace to stop combustion.

8. An automatic control device for a hot blast stove based on machine learning, characterized in that, include: The start-up module is used to respond to the start-up command of the hot blast stove and calculate the average air-fuel ratio and maximum gas flow rate of the hot blast stove during the last combustion. The acquisition module is used to start the hot blast stove based on the average air-fuel ratio and the maximum gas flow rate; monitor the dome temperature of the hot blast stove in real time, and determine the current combustion stage based on the dome temperature; wherein the combustion stage is a number of continuous operation stages divided based on the dome temperature threshold; and collect the current operation data of the hot blast stove in the current combustion stage. The prediction module is used to input the current operating data into the target prediction model corresponding to the current combustion stage, and output the predicted dome temperature and predicted flue temperature of the current operating data; the target prediction model is trained based on the following steps: obtaining the target historical operating temperature of the hot air furnace and the flue connected to the hot air furnace corresponding to the current combustion stage; training the preset initial target prediction model with the target historical operating temperature until the preset termination condition is reached, and obtaining the target prediction model; The control module is used to dynamically adjust the current gas flow rate based on the predicted dome temperature, the predicted flue temperature, the average air-fuel ratio, and the maximum gas flow rate, so as to control the hot blast stove to operate with the adjusted gas flow rate. The combustion stage includes: the steady combustion stage of the hot blast stove; When the current combustion stage is the steady combustion stage; the current operating data includes: the steady dome temperature and the steady flue temperature corresponding to the steady combustion stage; The control module is also used for: The time when the steady dome temperature reaches the third threshold is set as the first time, and the time when the steady flue temperature reaches the second threshold is set as the second time. The steady temperature rise slope of the flue is calculated based on the third threshold, the second threshold, the first time, and the second time. The control module is also used for: Determine whether the predicted dome temperature exceeds a first threshold or whether the predicted flue temperature exceeds a second threshold. If so, adjust the current gas flow rate to the second low gas flow rate; the second low gas flow rate is less than the current gas flow rate, and the difference between the two is a first preset step size; If not, calculate the predicted temperature rise slope based on the third threshold, the predicted flue temperature, the current combustion time, and the first time. Determine whether the predicted temperature rise slope is less than the steady temperature rise slope; If so, adjust the current gas flow rate of the hot blast stove to a high gas flow rate; the high gas flow rate is greater than the current gas flow rate, and the difference between the high gas flow rate and the current gas flow rate is a second preset step size; If not, adjust the current gas flow rate of the hot blast stove to the third low gas flow rate; the third low gas flow rate is less than the current gas flow rate, and the difference between the two is the second preset step size.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory, the memory storing computer-executable instructions that can be executed by the processor, the processor executing the computer-executable instructions to implement the machine learning-based automatic control method for hot blast stoves as described in any one of claims 1 to 7.

Images