Intelligent temperature control method for sintering ignition furnace

Abstract

The invention discloses an intelligent temperature control method for a sintering ignition furnace, and the method comprises the steps: collecting sintering production historical data, carrying out the filtering processing of the collected data, constructing a sample database, and dividing the sample database into a training data set, a verification data set, and a to-be-measured data set; selecting samples of which the sintering yield and the air permeability index of the sintering material layer are both ranked from the first 50% from the training data set for modeling training; performing model training on the sample by using an improved IGWO-XGBoost algorithm to obtain a prediction model; utilizing the prediction model to predict the to-be-measured data set to obtain a latest ignition furnace temperature target value, and storing the latest ignition furnace temperature target value in a database; and comparing the newest ignition furnace temperature target value with the last statistical period target value, if the absolute value of the variable quantity exceeds 3 DEG C and lasts for more than 30 seconds, transmitting the temperature target value to a controller for temperature adjustment, otherwise, not performing temperature adjustment. According to the invention, the labor cost is reduced, and the real-time accurate setting of the temperature target value of the ignition furnace is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of sintering processes, and in particular to an intelligent temperature control method for a sintering ignition furnace. Background Art

[0002] During the production process of the sintering process, the temperature control of the ignition furnace is an important part. If the temperature is too high, the surface of the mixture layer will melt, resulting in poor air permeability of the sintering layer; if the temperature is too low, the ignition effect on the surface of the sintering layer will be poor, and the combustion degree of the mixture will not be enough, affecting the quality of the sinter. Both of the above situations will affect the output and quality of the sinter, so the ignition furnace temperature must be controlled within the optimal temperature range.

[0003] Currently, the target values of the sintering ignition temperature in each ironmaking plant of the domestic metallurgical industry are mostly set manually. The on-site workers determine the temperature target values based on experience. This method has uncertainty and non-quantifiability, and is prone to cause deviation in ignition temperature control, resulting in a decrease in the quality and output of the sinter, and high labor costs; during the actual production process, parameters such as the temperature, humidity, and composition of the raw materials, and the trolley speed will change, and the operator cannot adjust the ignition temperature target value in real time, and cannot ensure the production stability throughout the period and a long-term high product qualification rate. Summary of the Invention

[0004] In order to overcome the above problems existing in the prior art, the present invention proposes an intelligent temperature control method for a sintering ignition furnace.

[0005] The technical solution adopted by the present invention to solve its technical problems is: an intelligent temperature control method for a sintering ignition furnace, including the following steps: Step 1, collect historical sintering production data, filter the collected data, construct a sample database, and divide the sample database into a training data set, a verification data set, and a data set to be tested; Step 2, select samples with the top 50% in both the sintering finished product rate and the sintering layer air permeability index from the training data set in Step 1 for modeling training; Step 3, use the improved IGWO-XGBoost algorithm to train the model for the samples selected in Step 2, use the influencing variables of the mixture temperature, humidity, composition, layer thickness, and trolley speed as input variables, and the appropriate value of the sintering ignition temperature as the output variable, and verify the model through the verification data set to obtain a prediction model; Step 4, use the prediction model obtained in Step 3 to predict the data set to be tested in Step 1, obtain the latest ignition furnace temperature target value, and store it in the database; Step 5: Compare the latest ignition furnace temperature target value obtained in Step 4 with the target value in the previous statistical period. If the absolute value of the change amount exceeds 3°C and lasts for more than 30 seconds, transmit the temperature target value to the controller for temperature adjustment; otherwise, no temperature adjustment is performed.

[0006] In the above intelligent temperature control method for a sintering ignition furnace, the calculation process of the sintered ore finished product rate in Step 2 is: Sintered ore finished product rate = Screened upper part after screening of the fired part / Fired part * 100%; The calculation formula for the permeability index of the sintering material layer is: where Pe is the permeability index; Q is the main extraction flow of the sintering machine; A is the effective area of the sintering machine; h is the thickness of the sintering machine material layer; P is the negative pressure in the main flue; n and m are gas characteristic constants.

[0007] In the above intelligent temperature control method for a sintering ignition furnace, the improved IGWO-XGBoost algorithm in Step 3 is specifically: Use the improved Grey Wolf Optimization algorithm IGWO to optimize the parameters of the XGBoost model, take the Root Mean Square Error RMSE as the scoring index, determine the best parameter combination on the validation set, and train the optimal model with the best parameter combination.

[0008] In the above intelligent temperature control method for a sintering ignition furnace, the improved Grey Wolf Optimization algorithm includes: Using non-linear decrease to replace linear decrease for the convergence factor, and the specific expression is: where is the maximum value of the convergence factor, is the maximum number of iterations, and t is the number of iterations.

[0009] According to the intelligent temperature control method for a sintering ignition furnace described in Claim 1, the improved Grey Wolf Optimization algorithm includes: Introducing an inertia weight w for the update of the grey wolf position, and the expression is: where t is the number of iterations, is the maximum inertia weight, is the minimum inertia weight, A is the coefficient vector, and D is the distance between the grey wolf and the prey.

[0010] The beneficial effects of the present invention are as follows: 1) It fully considers factors directly affecting the appropriate value of the ignition furnace temperature, such as the temperature, humidity, composition, layer thickness of the mixture, and the trolley speed. The improved IGWO-XGBoost algorithm is used to train high-quality samples, deeply mining the internal relationship between the influencing factors and the appropriate value of the ignition temperature and forming knowledge. Thus, when the influencing factors change, the appropriate value of the ignition temperature can be adjusted dynamically in real time to achieve production stability throughout the period and a long-term high product qualification rate.

[0011] 2) There is no need to manually set the target value of the ignition temperature, saving labor costs, achieving feedforward regulation, and improving the quality and output of sintered products.

[0012] 3) It not only combines the operation experience of on-site personnel but also integrates the idea of optimized setting into the control, reducing labor costs and achieving real-time and accurate setting of the target value of the ignition furnace temperature. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a schematic flow chart of the present invention; Figure 2 is a temperature trend chart of the sintering ignition furnace when a certain steel plant uses the control method of the present invention; Figure 3 is a temperature trend chart of the sintering ignition furnace when a certain steel plant does not use the control method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be described in detail below in conjunction with the drawings and specific embodiments.

[0015] This embodiment provides an intelligent control scheme for the temperature of a sintering ignition furnace based on data driving. This scheme collects historical production data to form a sample library, then screens out high-quality samples from it, and uses machine learning algorithms to fit the mapping relationship between influencing factors such as the temperature, humidity, composition, layer thickness of the mixture, and the trolley speed and the required ignition temperature, so as to accurately predict the appropriate temperature target value of each influencing factor at different values in real time and achieve feedforward control to eliminate the lag effect.

[0016] The control process includes six steps: data collection, construction of a sample database, screening of high-quality samples, establishment of a prediction model, prediction of the temperature target value, and control of the ignition temperature. The specific process is as Figure 1 shown, including: S1. Data collection Collect historical data of sintering production. The main parameters collected are the temperature, humidity, composition, layer thickness of the mixture, trolley speed, and sintering ignition temperature. The server collects data from the programmable logic controller (PLC) once per second and stores the data in the database.

[0017] S2. Construct a sample database Based on the collected original sample data, in order to make the data real, effective and smooth, average value filtering is performed on the data every 60 seconds. Then, a sample database is constructed for the processed production data. The sample data that has completed ignition temperature control is divided into a training data set and a validation data set according to a ratio of 8:2, and the sample data that has not completed ignition temperature control is used as the data set to be tested.

[0018] S3. Select high-quality samples The selection of high-quality cases is determined by the sintered ore yield and the sintering bed permeability index. Both of these indicators are positive indicators. After sorting from large to small, the top 50% are taken as high-quality samples, that is, the sample data with good ignition temperature control is used for modeling and training.

[0019] Sintered ore yield = (the part above the sieve after screening the fired part (deducting the powder part)) / the fired part * 100%.

[0020] The sintering bed permeability index adopts the currently commonly used Voice formula: Where Pe is the permeability index, Q is the main extraction flow rate of the sintering machine (m³ / min), A is the effective area of the sintering machine (㎡), h is the thickness of the sintering bed of the sintering machine (mm), P is the negative pressure of the main flue (Pa), n and m are gas characteristic constants, and generally both are taken as 0.6.

[0021] S4. Establish an optimal prediction model Use the improved IGWO-XGBoost algorithm to train the model for the high-quality sample data set selected in step three. The influencing variables such as the temperature, humidity, composition, bed thickness, and trolley speed of the mixture are used as input variables, and the appropriate value of the sintering ignition temperature is used as the output variable. The extreme gradient boosting model XGBoost (eXtreme Gradient Boosting) is an ensemble machine learning algorithm based on decision trees, and it performs very well in terms of parallel computing efficiency, controlling overfitting, and prediction generalization ability. The Grey Wolf Optimizer (GWO) is a new type of swarm intelligence optimization algorithm. It is a heuristic algorithm based on the social hierarchy and hunting strategy of grey wolves in nature and is used to solve various optimization problems. The specific steps are as follows: 1) Social hierarchy stratification. Initialize the individual positions of a group of grey wolves, where each position represents a potential solution to a problem. The dimension of the solution and the population size depend on the specific problem. Calculate the fitness value corresponding to each individual's position according to the fitness function, and select the top three grey wolves, denoted as the alpha wolf, beta wolf, and delta wolf, and the rest are denoted as omega wolves. The alpha wolf is the leader of the group and is responsible for making key decisions such as the hunting direction and location. The beta wolf assists the alpha wolf and can replace its role in the absence of the alpha wolf. The delta wolf obeys the commands of the alpha wolf and beta wolf, and the omega wolf has the lowest status.

[0022] 2) Surrounding the prey. During the hunting process, the grey wolf group will first surround the prey. Mathematically, this surrounding behavior can be represented by the following formula: where, , , linearly decreases from 2 to 0, and are random numbers within the interval [0, 1], D is the distance between the grey wolf and the prey, t is the iteration number, is the position vector of the prey, is the position vector of the current grey wolf, and A and C are coefficient vectors.

[0023] 3) Hunting attack. After surrounding the prey, the grey wolves participate in the hunting attack according to their social hierarchy. The alpha wolf plays a leading role in the hunting process, and other wolves gradually approach the prey under the guidance of the alpha wolf. This process is simulated by updating the positions of the grey wolf individuals. The position update formula comprehensively considers the position information of the alpha wolf, beta wolf, and delta wolf, enabling the entire group to continuously approach the prey (optimal solution).

[0024] This system has improved the grey wolf optimization algorithm GWO in terms of both the convergence factor and position update, as follows: 1. Adaptive convergence factor optimization The optimal solution of GWO is closely related to the coefficient vector A, and the convergence factor directly affects the value of A. Here, we use non-linear decrease instead of linear decrease for the convergence factor, and more potential optimal solutions can be found. The expression is as follows: where, is the maximum value of the convergence factor, is the maximum number of iterations.

[0025] 2. Grey wolf position update strategy To improve the ability of the Grey Wolf Optimization algorithm to jump out of local optima, an inertia weight w is introduced for updating the positions of grey wolves. The value range of the inertia weight w is [0.3, 0.8]. As the number of iterations increases, the inertia weight will also change: where t is the number of iterations, is the maximum inertia weight, is the minimum inertia weight.

[0026] Use the improved Grey Wolf Optimization algorithm IGWO to optimize the important parameters of the XGBoost model, such as the number of decision trees, the depth of the tree, and the learning rate. Use the root mean square error RMSE as the scoring metric to determine the best parameter combination on the validation set, train the optimal model with the best parameter combination, and save it for later prediction use; Root mean square error RMSE calculation formula: where, represents the predicted value, represents the actual value, and n represents the number of samples.

[0027] Facts show that the improved IGWO-XGBoost model can accelerate model convergence, quickly determine the global optimal solution, and improve the accuracy of the model.

[0028] S5, Prediction of temperature target value Use the optimal model generated in step four to predict the dataset to be measured in step two, obtain the latest ignition furnace temperature target value, and store it in the database; S6, Ignition temperature control Compare the latest sintering ignition furnace temperature target value with the target value in the previous statistical period. If the absolute value of the change amount exceeds 3°C and lasts for 30 seconds, transmit the temperature target value to the PLC for temperature control operations; otherwise, do not adjust the temperature.

[0029] After the system is put into use in a certain steel plant ( Figure 2 ) compared with before it was put into use ( Figure 3 ), through the comparative analysis of the ignition furnace temperature control effects, it is known that after using the control method of the present invention, the ignition furnace temperature control effect has been greatly improved. The control error has been optimized from about ±130°C to within ±15°C, and a large amount of labor costs have been saved.

[0030] The above embodiments are only exemplary embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art can make various modifications or equivalent replacements to the present invention within the essence and protection scope of the present invention, and such modifications or equivalent replacements should also be regarded as falling within the protection scope of the present invention.

Claims

1. An intelligent temperature control method for a sintering ignition furnace, characterized in that, It includes the following steps: Step 1: Collect the historical data of sintering production, filter the collected data, construct a sample database, and divide the sample database into a training data set, a validation data set, and a data set to be measured; Step 2: Select the samples with the top 50% in both the sintering finished product rate and the sintering bed permeability index from the training data set in Step 1 for model training; Step 3: Use the improved IGWO-XGBoost algorithm to train the model with the samples selected in Step 2. Take the influencing variables such as the temperature, humidity, composition, bed thickness, and trolley speed of the mixed material as input variables, and the appropriate value of the sintering ignition temperature as the output variable. Then verify the model through the validation data set to obtain a prediction model; Step 4: Use the prediction model obtained in Step 3 to predict the data set to be measured in Step 1, obtain the latest target value of the ignition furnace temperature, and store it in the database; Step 5: Compare the latest target value of the ignition furnace temperature obtained in Step 4 with the target value in the previous statistical cycle. If the absolute value of the change amount exceeds 3°C and lasts for more than 30 seconds, transmit the temperature target value to the controller for temperature adjustment; otherwise, do not perform temperature adjustment.

2. The temperature intelligent control method for a sintering ignition furnace according to claim 1, wherein In Step 2, the calculation process of the sintered ore finished product rate is: Sintered ore finished product rate = Screened part above the screen / Burned part * 100%; The calculation formula for the sintering bed permeability index is: Where, Pe is the permeability index; Q is the main extraction flow of the sintering machine; A is the effective area of the sintering machine; h is the thickness of the sintering bed of the sintering machine; P is the negative pressure of the main flue; n, m are gas characteristic constants.

3. A temperature intelligent control method for a sintering ignition furnace according to claim 1, characterized in that, In Step 3, the improved IGWO-XGBoost algorithm is specifically: Use the improved grey wolf optimization algorithm IGWO to optimize the parameters of the XGBoost model. Take the root mean square error RMSE as the scoring index, determine the best parameter combination on the validation set, and perform optimal model training with the best parameter combination.

4. The temperature intelligent control method for a sintering ignition furnace according to claim 3, characterized in that, The improved grey wolf optimization algorithm includes: Using non-linear decrease to replace linear decrease for the convergence factor, and the specific expression is: wherein, is the maximum value of the convergence factor, is the maximum number of iterations, and t is the number of iterations.

5. A temperature intelligent control method for a sintering ignition furnace according to claim 1, characterized in that, The improved grey wolf optimization algorithm includes: Introducing an inertia weight w for the position update of grey wolves, and the expression is: where t is the number of iterations, is the maximum inertia weight, is the minimum inertia weight, A is the coefficient vector, and D is the distance between the grey wolf and the prey.

Images