A method and system for predicting the real-time temperature of an arc furnace after melting and cleaning

Abstract

The present invention belongs to the field of intelligent metallurgical industry, and discloses a method and system for predicting the real-time temperature of an electric arc furnace after purging. An infrared thermal imager and an argon blowing gun are installed on the top of the electric arc furnace to obtain the temperature of the slag surface and the exposed steel liquid surface in real time; power consumption, oxygen blowing amount, carbon powder amount, lime amount, furnace weight, flue gas composition and temperature, and natural gas amount are collected in real time through a variety of sensors; a molten pool temperature mechanism prediction model is established; based on the collected historical data and recurrent neural network structure, a molten pool temperature dynamic prediction model is established; the current process parameters and mechanism prediction values ​​are input in real time, and the steel liquid surface temperature is output through the dynamic prediction model. The system of the present invention includes an infrared thermal imager, an argon blowing gun, a data acquisition unit, a database, a mechanism model unit, a neural network prediction unit and a control system. The system can realize real-time and high-precision prediction of the steel liquid surface temperature after purging the electric arc furnace, and improve the control accuracy and energy efficiency of the production process.

Description

Technical Field

[0001] The present invention relates to the field of intelligent metallurgical industry, and in particular to a method and system for predicting the real-time temperature of an arc furnace after melting. Background Art

[0002] Electric arc furnace is a key equipment widely used in the steel smelting industry, mainly used for melting scrap steel and producing high-quality steel. In the electric arc furnace smelting process, accurate control and prediction of the molten pool temperature are of great significance for optimizing the smelting process, improving product quality, reducing energy consumption and extending equipment life. However, due to the complexity of the internal environment of the electric arc furnace, continuous and accurate measurement and prediction of the molten pool temperature has always been a major challenge faced by this field. The existing method is mainly the thermocouple method, which measures the molten pool temperature by intermittently inserting a high-temperature thermocouple. Although this method is relatively accurate, it will interrupt the production process. Based on known process parameters, a mathematical model or machine learning is established to estimate the end temperature of the electric arc furnace. This is difficult to adapt to complex smelting conditions, and the accuracy of the model is difficult to guarantee. In summary, it is difficult for the above methods to obtain accurate and continuous molten pool temperature information.

[0003] With the development of artificial intelligence and big data technology, it has become a trend to apply machine learning methods to parameter prediction in industrial process control. However, there is a lack of continuous molten pool temperature data in the actual production process of electric arc furnaces, so it is difficult to model them using data-driven methods. Therefore, developing a method and system that can obtain continuous molten pool temperature data and model the post-melting temperature of an electric arc furnace based on continuous data, and realizing real-time, accurate and safe prediction, has great practical application value for improving the automation level and economic benefits of steel smelting. Summary of the invention

[0004] The object of the present invention is to provide a method and system for predicting the real-time temperature of an arc furnace after purging, aiming to achieve the acquisition of continuous data on the temperature of the arc furnace molten pool and the high-precision prediction of the molten pool temperature.

[0005] To achieve the above object, the present invention provides the following technical solutions.

[0006] A method for predicting the real-time temperature of an arc furnace after melting and cleaning, comprising the following steps:

[0007] S1. Install monitoring equipment: Install an infrared thermal imager on the top of the electric arc furnace so that its field of view covers the entire slag surface; install an argon blowing gun on the top of the electric arc furnace, with the tip of the argon blowing gun 30cm~50cm away from the slag surface;

[0008] S2. Obtaining the slag surface temperature: Using the infrared thermal imager installed in step S1, the average temperature of the slag surface is obtained at intervals of time t;

[0009] S3. Exposing the molten steel surface and measuring the temperature: blowing argon gas through the argon blowing gun installed in step S1 at intervals of time t to disperse the slag and expose the molten steel surface; using the infrared thermal imager installed in step S1, the average temperature of the exposed molten steel surface is obtained at intervals of time t and transmitted to the database in real time;

[0010] S4. Collect real-time data, including: power consumption, oxygen blowing amount, carbon powder blowing amount, lime amount, furnace weight, flue gas composition and temperature, natural gas amount, average temperature of slag surface, and transmit them to the database in real time;

[0011] S5. Establish a molten pool temperature prediction mechanism model, input the real-time data collected in step S4, obtain real-time temperature mechanism prediction value, and transmit it to the database in real time;

[0012] S6. Based on the recurrent neural network structure, a dynamic prediction model for the molten pool temperature is established; the database is read, and the real-time data collected in step S4 and the real-time temperature mechanism prediction value obtained in step S5 are used as input, and the average temperature of the exposed steel surface obtained in step S3 is used as output; data from 5000 to 10000 heats are collected;

[0013] S7. Real-time prediction of current molten steel surface temperature: input the real-time data collected in step S4 at the current moment and the real-time temperature mechanism prediction value obtained in step S5; output the average temperature of the molten steel surface at the current moment through the model operation established in step S6.

[0014] According to the comparison between the predicted average temperature of the molten steel surface at the current moment and the temperature required by the process, the process parameters are adjusted, specifically:

[0015] When the average temperature of the molten steel surface is lower than the temperature required by the process, increase the power supply, increase the amount of oxygen blowing and the amount of carbon powder sprayed to meet the temperature required by the process; when the average temperature of the molten steel surface is higher than the temperature required by the process, reduce the power supply, reduce the amount of oxygen blowing and the amount of carbon powder sprayed to meet the temperature required by the process.

[0016] The method of obtaining the average temperature of the slag surface at intervals of time t is characterized by comprising the following steps:

[0017] A1. Every t seconds, a thermal image covering the entire slag surface is collected by the infrared thermal imager;

[0018] A2. Perform image processing on the collected thermal images, including: identifying and excluding non-slag areas in the image; performing temperature conversion on each pixel in the processed effective area;

[0019] A3. Calculate the arithmetic mean of the pixel temperatures in all valid areas in step A2 to obtain the average slag surface temperature at the current moment;

[0020] A4. Record the obtained average temperature value together with the corresponding timestamp and transmit it to the database;

[0021] A5. Repeat steps A1 to A4 to achieve continuous temperature measurement every t seconds.

[0022] The method of obtaining the average temperature of the exposed steel liquid surface at intervals of time t comprises the following steps:

[0023] B1. Every t seconds, argon is blown onto the slag surface through an argon blowing gun to expose the molten steel surface;

[0024] B2. Immediately after completing step B1, a thermal image covering the exposed steel surface area is collected by the infrared thermal imager;

[0025] B3. Perform image processing on the collected thermal images, including: identifying and locating the exposed steel surface area in the image; excluding the non-steel surface area in the image; and performing temperature conversion on each pixel point in the identified effective steel surface area;

[0026] B4. Calculate the arithmetic mean of the temperatures of all valid pixels in step B3 to obtain the average temperature of the exposed steel surface at the current moment;

[0027] B5. Record the obtained average temperature value of the molten steel surface together with the corresponding timestamp and transmit it to the database;

[0028] B6. Repeat steps B1 to B5 to achieve continuous steel surface temperature measurement every t seconds.

[0029] The method of establishing a molten pool temperature prediction mechanism model comprises the following steps:

[0030] C1. Divide the molten pool into three regions: molten steel, slag and gas phase, establish the heat balance equation of the arc furnace molten pool, including arc heating, chemical reaction heat release and heat loss; and perform modeling;

[0031] C2. Calculate the effective heating contribution of arc power to the molten pool based on the arc characteristics between the electrode and the molten pool, and construct an arc heating model;

[0032] C3. Establish the heat model of chemical reactions of decarbonization, dephosphorization and desulfurization, and calculate the heat of reaction of each reaction;

[0033] C4. Establish a heat dissipation loss model, including heat conduction loss of the furnace wall, heat conduction loss of the furnace cover, radiation heat loss and heat carried away by flue gas;

[0034] C5. Establish a heat change model for materials entering and leaving the molten pool, including continuous feeding, slag discharge, and exhaust;

[0035] C6. Integrate the above sub-models to construct a complete molten pool temperature prediction mechanism model, including establishing the differential equation of the molten pool temperature changing with time and determining the boundary conditions and initial conditions of the model;

[0036] C7. Optimize the model parameters using measured data and use the least squares method to fit the parameters;

[0037] C8. Input the real-time collected process parameters into the molten pool temperature prediction mechanism model, solve the molten pool temperature prediction mechanism model equation, and output the temperature mechanism prediction value at the current moment.

[0038] The real-time data collection includes:

[0039] Collecting power consumption: The power consumption of the electric arc furnace is measured and recorded in real time through the power metering device installed on the power supply system of the electric arc furnace;

[0040] Collect oxygen blowing amount: Use the gas flow meter installed on the oxygen supply pipeline to measure and record the amount of oxygen blown into the arc furnace in real time;

[0041] Collect the amount of carbon powder injected: The mass flow meter installed on the carbon powder injection system measures and records the amount of carbon powder injected into the electric arc furnace in real time;

[0042] Collecting lime quantity: Using the weighing sensor installed on the lime feeding system, the amount of lime added to the electric arc furnace is measured and recorded in real time;

[0043] Collecting furnace weight: The weight change of the entire furnace body is measured and recorded in real time through the weighing sensor system installed under the supporting structure of the electric arc furnace;

[0044] Collecting flue gas composition: Using the flue gas online analyzer installed on the flue of the electric arc furnace, the content of each component in the flue gas is measured and recorded in real time;

[0045] Collecting flue gas temperature: Using a high-temperature thermocouple or optical fiber temperature measurement system installed on the flue of the arc furnace, the flue gas temperature is measured and recorded in real time;

[0046] Collecting natural gas volume: Using the gas flow meter installed on the natural gas supply pipeline, the volume of natural gas injected into the electric arc furnace is measured and recorded in real time;

[0047] Collecting the average temperature of the slag surface: The average temperature of the slag surface is acquired and recorded every t seconds by the infrared thermal imaging system.

[0048] The method for establishing a dynamic prediction model for molten pool temperature based on a recurrent neural network structure includes: collecting historical data of 5,000 to 10,000 heats; arranging the collected data in chronological order to form a time series data set; selecting Transformer as the recurrent neural network structure, and outputting a predicted value of the average temperature of the molten steel surface at the current moment every t seconds.

[0049] A system for predicting the real-time temperature of an arc furnace after melting and cleaning, comprising the following components:

[0050] Infrared thermal imager: installed on the top of the electric arc furnace to collect thermal images of the slag surface and exposed steel liquid surface;

[0051] Argon blowing gun: installed on the top of the electric arc furnace, used to blow argon to expose the steel liquid surface;

[0052] Data acquisition unit: including lime weighing sensor 1, electric energy meter 2, infrared thermal imager 3, flue gas composition monitor 5, flue gas temperature monitor 6, oxygen mass flow meter 7, carbon powder mass flow meter 8, natural gas mass flow meter 9, furnace weight sensor 10, used for real-time collection of power consumption, oxygen blowing amount, carbon powder blowing amount, lime amount, furnace weight, flue gas composition and temperature, and natural gas amount data;

[0053] Database: used to store real-time data obtained from infrared thermal imager, argon blowing gun and data acquisition unit;

[0054] Mechanism model unit: Based on the established molten pool temperature prediction mechanism model, it inputs real-time data and outputs the temperature mechanism prediction value;

[0055] Neural network prediction unit: Based on the established recurrent neural network model, it inputs real-time data and temperature mechanism prediction value, and outputs the average temperature of the steel liquid surface at the current moment;

[0056] Control system: Integrates the above units, coordinates system operation, and displays the predicted results of the molten steel surface temperature to the operator. At the same time, it controls the power supply, oxygen blowing amount and carbon powder spraying amount according to the difference between the current average temperature of the molten steel surface and the temperature required by the process.

[0057] The installation position and angle of the infrared thermal imager and the argon blowing gun are adjustable to ensure that the maximum area of ​​the slag surface and the molten steel surface is covered.

[0058] The data acquisition unit is connected to the database via a wireless or wired network to achieve real-time data transmission; the control system includes a human-machine interface for displaying predicted temperature, historical data and alarm information, and allowing operators to manually adjust model parameters and system settings.

[0059] The present invention has the following beneficial effects:

[0060] The method and system of the present invention can realize real-time and continuous prediction of the temperature of the arc furnace molten pool. The system can continuously output the latest temperature prediction results without interrupting the production process. This high-frequency temperature monitoring capability greatly improves the control accuracy of the smelting process, enabling operators to make timely adjustments and optimize the production process. The present invention adopts a method combining a mechanism model and a data-driven model to significantly improve the accuracy of temperature prediction. In practical applications, the error between the predicted temperature and the actual measured temperature can be controlled within ±10°C, meeting the requirements for high-quality steel production. The present invention adopts a non-contact measurement method and realizes indirect measurement of the temperature of the molten steel through infrared thermal imaging technology and argon blowing technology. This method avoids the potential safety hazards of traditional inserted thermocouples and greatly reduces operational risks. At the same time, since the measurement process does not introduce foreign matter, it also reduces interference with the molten pool and ensures the stability of the smelting process. The system of the present invention has strong adaptability and can cope with various complex working conditions in the smelting process of the electric arc furnace. After adopting this system, the energy consumption of the electric arc furnace can be reduced by 5-8%, bringing considerable economic benefits to the enterprise.

[0061] The present invention provides accurate temperature prediction, so that operators can better control the temperature curve during the smelting process, thereby more accurately controlling the timing and amount of alloy element addition, and optimizing the metallurgical reactions such as deoxidation and desulfurization. The system of the present invention can be seamlessly integrated with other control systems of the electric arc furnace, laying the foundation for the full automation of the electric arc furnace smelting process. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 The present invention is a flow chart of the method for predicting the real-time temperature of an arc furnace after melting and cleaning.

[0063] Figure 2 The schematic diagram of the structure of the system for predicting the real-time temperature of an arc furnace after melting is shown in the present invention.

[0064] Figure 3 This is a schematic diagram of the installation of the data acquisition device of the present invention.

[0065] Numbers in the figure: 1. Lime weighing sensor; 2. Electric energy meter; 3. Infrared thermal imager; 4. Argon blowing gun; 5. Flue gas composition monitor; 6. Flue gas temperature monitor; 7. Oxygen mass flow meter; 8. Carbon powder mass flow meter; 9. Natural gas mass flow meter; 10. Furnace weight sensor. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0068] The present invention relies on the added infrared thermal imager 3 and argon blowing gun 4 to measure the slag surface temperature and the average temperature of the exposed steel liquid surface, and the training of the molten pool temperature prediction mechanism model and the molten pool temperature dynamic prediction model is realized through these two surface temperatures.

[0069] Example 1

[0070] This embodiment provides a method and system for predicting the real-time temperature of an arc furnace after melting. Figure 1 and Figure 2 , the specific implementation steps are as follows:

[0071] A method for predicting the real-time temperature of an arc furnace after melting is as follows:

[0072] Monitoring equipment installation: Install a high-resolution infrared thermal imager on the top of the electric arc furnace to ensure that its field of view completely covers the entire slag surface. The infrared thermal imager adopts a water-cooled shell and dust-proof design to adapt to high temperature and dust environments. An adjustable angle argon blowing gun is installed on the top of the electric arc furnace, and the gun head is kept at 30cm~50cm from the slag surface. The argon blowing gun is made of high-temperature resistant alloy material and has an automatic telescopic function.

[0073] The installed infrared thermal imager is used to collect a thermal image covering the entire slag surface every 5 seconds. The thermal image is processed in real time, including: identifying and excluding non-slag areas (such as furnace walls, electrodes, etc.); converting the temperature of each pixel in the effective area; and calculating the arithmetic mean of the temperature of all pixels in the effective area. The calculated average temperature value and the corresponding timestamp are transmitted to the database.

[0074] Every 5 seconds, argon is blown onto the slag surface through an argon blowing gun to form an exposed steel surface area with a diameter of about 50 cm. Immediately after the argon blowing is completed, a thermal image of the exposed steel surface is collected using an infrared thermal imager. The thermal image is processed, including: identifying and locating the exposed steel surface area; excluding non-steel surface areas such as residual slag and furnace walls; and converting the temperature of each pixel in the identified effective steel surface area. Calculate the arithmetic mean of the temperature of all pixels in the effective area. The calculated average temperature of the steel surface and the corresponding timestamp are transmitted to the database.

[0075] The parameters collected and transmitted to the database in real time include: electricity consumption, measured by an electric energy metering device; oxygen blowing amount, measured by a gas flow meter on an oxygen pipeline; carbon powder injection amount, measured by a mass flow meter on a carbon powder injection system; lime amount, measured by a weighing sensor on a lime feeding system; furnace body weight, measured by a weighing sensor system under the arc furnace support structure; flue gas composition, measured by an online flue gas analyzer on the flue to measure the contents of CO, CO2, O2, etc.; flue gas temperature, measured by a high-temperature thermocouple or optical fiber temperature measurement system on the flue; natural gas amount, measured by a gas flow meter on a natural gas pipeline; and the average slag surface temperature, obtained by the aforementioned method.

[0076] The molten pool is divided into three regions: molten steel, slag and gas phase, and a heat balance equation is established. An arc heating model is constructed to calculate the effective heating contribution of arc power to the molten pool. A chemical reaction heat model is established, including decarburization, dephosphorization, desulfurization and other reactions. A heat dissipation loss model is established, including heat conduction loss of the furnace wall and furnace cover, radiation heat loss, and heat carried away by flue gas. A heat change model for materials entering and leaving the molten pool is established. The above sub-models are integrated to construct a complete molten pool temperature prediction mechanism model. The model parameters are optimized and fitted using measured data.

[0077] Collect historical data of 5,000 to 10,000 heats, including all the real-time data mentioned above. Arrange the data in chronological order to form a time series data set. Select Transformer as the recurrent neural network structure. Take the collected real-time data and the temperature mechanism prediction value as input. Take the average temperature of the exposed steel liquid surface as output. Train the neural network model so that it can output the predicted value of the average temperature of the steel liquid surface at the current moment every 5 seconds.

[0078] Input the current process parameter data and mechanism model prediction value in real time. Calculate through the trained neural network model. Output and display the current average temperature prediction value of the molten steel surface. Compare the current average temperature of the molten steel surface with the process required temperature. When the average temperature of the molten steel surface is lower than the process required temperature, increase the power supply, increase the oxygen blowing amount and the amount of carbon powder sprayed to reach the process required temperature; when the average temperature of the molten steel surface is higher than the process required temperature, reduce the power supply, reduce the oxygen blowing amount and the amount of carbon powder sprayed to reach the process required temperature.

[0079] Furthermore, the present invention provides a specific embodiment. Monitoring equipment is installed on a 100-ton continuous charging AC arc furnace in a steel plant. An infrared thermal imager is selected with a resolution of 1024×1024 pixels, a temperature measurement range of 0°C to 2000°C, and a measurement accuracy of ±2%. The infrared thermal imager is installed on the top of the arc furnace, 3 meters away from the slag surface, with a field of view of 45° to ensure that the entire slag surface is covered. A customized high-temperature argon blowing gun is installed, and the barrel is made of high-temperature resistant molybdenum alloy material, with a length of 2 meters and an inner diameter of 20mm. The argon blowing gun is installed 40cm from the slag surface, and the angle can be adjusted within the range of ±15°.

[0080] The infrared thermal imager collects a thermal image every 5 seconds. Using the self-developed image processing algorithm, the thermal image is first subjected to noise reduction, using a Gaussian filter with a kernel size of 5×5. Then the non-slag area is identified and excluded by the threshold segmentation method. The temperature of each pixel in the effective area is converted, and the arithmetic mean of the temperature of all pixels in the effective area is calculated to obtain the average temperature of the slag surface.

[0081] Every 5 seconds, argon is blown onto the slag surface through an argon blowing gun for 3 seconds, forming an exposed steel surface area with a diameter of about 60 cm. Within 0.5 seconds after the completion of argon blowing, the infrared thermal imager collects the thermal image of the exposed steel surface. Using an image processing method similar to that for obtaining the slag surface temperature, a threshold is set to identify the steel surface area. The arithmetic mean of the temperature of all pixels in the effective area is calculated to obtain the average temperature of the steel surface.

[0082] The parameters collected and transmitted to the database in real time include: Power consumption: measured by an energy meter with a sampling frequency of 1Hz. Oxygen blowing volume: measured by a turbine flowmeter with a range of 0-5000Nm³ / h and an accuracy of ±1.5%. Carbon powder volume: measured by a mass flowmeter with a range of 0-500kg / h and an accuracy of ±0.5%. Lime volume: measured by a belt scale with a range of 0-2000kg / h and an accuracy of ±1%. Furnace weight: measured by 4 weighing sensors installed under the support structure of the electric arc furnace, each with a range of 0-50 tons and an accuracy of ±0.1%. Flue gas composition: The flue gas analyzer measures the content of CO, CO2, and O2 with a sampling frequency of 0.2Hz. Flue gas temperature: measured by a K-type thermocouple with a temperature range of 0-1300°C and an accuracy of ±0.75%. Natural gas volume: measured by turbine flowmeter, with a range of 0-5000Nm³ / h and an accuracy of ±1%. All data is stored in the SQL Server database.

[0083] Based on the principles of thermodynamics and heat transfer, a mechanism model including the following sub-models is established:

[0084] a) Arc heating model:

[0085] Among them, Q_arc is the effective arc heating power (kW), η is the arc heating efficiency (taken as 0.7), and P_electric is the input electric power (kW).

[0086] b) Chemical reaction thermal model:

[0087] Where Q_reaction is the reaction heat (kJ), ΔH_i is the reaction enthalpy of the ith reaction (kJ / kg), and m_i is the mass of the reactant (kg).

[0088] c) Heat dissipation loss model:

[0089] Q_loss = Q_wall + Q_radiation + Q_gas

[0090] Among them, Q_loss is the total heat dissipation loss (kW), Q_wall is the furnace wall heat loss (kW), Q_radiation is the radiation heat loss (kW), and Q_gas is the heat carried away by the flue gas (kW).

[0091] Integrate the above sub-models to construct the molten pool temperature prediction differential equation:

[0092]

[0093] Where T is the melt pool temperature (°C), t is the time (s), m is the melt pool mass (kg), and Cp is the melt pool specific heat capacity (kJ / (kg·°C)).

[0094] The fourth-order Runge-Kutta method is used to solve the differential equation, and the time step is set to 1 second.

[0095] Collect historical data of 8000 heats, including all the above real-time data. Select Transformer as the recurrent neural network structure, with the following parameters: input feature dimension: 20 (including all real-time data and mechanism model output), output dimension: 1 (predicted temperature), number of attention heads: 8, number of encoder layers: 6, feedforward neural network hidden layer dimension: 512, dropout rate: 0.1, activation function: ReLU, use Adam optimizer for model training, learning rate set to 0.0001, batch size of 64, and number of training rounds of 100.

[0096] The system executes a prediction cycle every 5 seconds. First, the process parameter data at the current moment is input into the mechanism model to obtain the temperature mechanism prediction value. Then, the process parameter data and the mechanism prediction value are input into the trained Transformer model together, and the average temperature prediction value of the steel liquid surface at the current moment is output. The prediction results are transmitted to the on-site HMI system in real time through the OPC UA protocol and displayed on a 15-inch touch screen. The display content includes real-time predicted temperature, temperature trend curve, key process parameters, etc.

[0097] The present invention also provides a system for predicting the real-time temperature of an arc furnace after melting and cleaning, comprising:

[0098] Infrared thermal imager: installed on the top of the electric arc furnace to collect thermal images of the slag surface and exposed steel liquid surface.

[0099] Argon blowing gun: installed on the top of the electric arc furnace, used to spray argon to expose the molten steel surface.

[0100] Data acquisition unit: including electric energy metering device, gas flow meter, mass flow meter, weighing sensor, flue gas online analyzer, high temperature thermocouple or optical fiber temperature measurement system, etc., used to collect various real-time process parameters.

[0101] Database: used to store real-time data obtained from various sensors and devices.

[0102] Mechanism model unit: Based on the established molten pool temperature prediction mechanism model, real-time data is input and temperature mechanism prediction value is output.

[0103] Neural network prediction unit: Based on the established recurrent neural network model, real-time data and temperature mechanism prediction values ​​are input, and the average temperature of the steel liquid surface at the current moment is output.

[0104] Control system: Integrates the above units, coordinates system operation, and displays the predicted results of the molten steel surface temperature to the operator. The system includes a human-machine interface to display the predicted temperature, historical data and alarm information, and allows operators to manually adjust model parameters and system settings; according to the difference between the average temperature of the molten steel surface at the current moment and the temperature required by the process, it controls the power supply, oxygen blowing amount and the amount of carbon powder sprayed.

[0105] The system features include: The infrared thermal imager and argon blowing gun can be installed at adjustable positions and angles to ensure the best measurement results. The data acquisition unit is connected to the database via a wireless or wired network to achieve real-time data transmission. The system has self-diagnosis and self-calibration functions and can adapt to different arc furnace operating conditions.

[0106] This embodiment achieves high-precision real-time prediction of the arc furnace molten pool temperature by combining advanced sensing technology, mechanism model and deep learning algorithm, laying the foundation for intelligent control of the arc furnace.

[0107] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0108] The method and system of the present invention can realize real-time and continuous prediction of arc furnace molten pool temperature. Through a data collection and processing cycle every 5 seconds, the system can continuously output the latest temperature prediction results without interrupting the production process. This high-frequency temperature monitoring capability greatly improves the control accuracy of the smelting process, allowing operators to make timely adjustments and optimize the production process.

[0109] The present invention adopts a method that combines the mechanism model and the data-driven model to significantly improve the accuracy of temperature prediction. The mechanism model takes into account the complex physical and chemical processes in the electric arc furnace, providing a solid theoretical basis for prediction. The neural network model based on the Transformer structure can capture hidden patterns and long-term dependencies in historical data, further improving the prediction accuracy. In practical applications, the error between the predicted temperature and the actual measured temperature can be controlled within ±10°C, meeting the requirements for the production of high-quality steel grades.

[0110] The present invention adopts a non-contact measurement method, and realizes indirect measurement of molten steel temperature through infrared thermal imaging technology and argon blowing technology. This method avoids the potential safety hazards of traditional inserted thermocouples and greatly reduces the operational risks. At the same time, since no foreign matter is introduced during the measurement process, the interference with the molten pool is also reduced, ensuring the stability of the smelting process.

[0111] The system of the present invention has strong adaptability and can cope with various complex working conditions in the electric arc furnace smelting process. By comprehensively analyzing a variety of process parameters, such as power consumption, oxygen blowing amount, carbon powder spraying amount, etc., the system can accurately capture the changing pattern of the molten pool temperature under different operating conditions. This comprehensive data analysis capability enables the system to maintain high-precision prediction effects in the production of different types of electric arc furnaces and different types of steel. Through real-time and accurate temperature prediction, the present invention can help operators control the energy input of the electric arc furnace more accurately. This can not only avoid the problem of overheating or insufficient temperature, but also optimize the use of power consumption, oxygen and natural gas, and significantly improve energy utilization efficiency. In practical applications, after adopting this system, the energy consumption of the electric arc furnace can be reduced by 5-8%, bringing considerable economic benefits to the enterprise.

[0112] The present invention provides accurate temperature predictions, allowing operators to better control the temperature curve during the smelting process, thereby more accurately controlling the timing and amount of alloy element additions and optimizing metallurgical reactions such as deoxidation and desulfurization. This directly leads to improved product quality, such as more uniform steel composition and more stable mechanical properties. Excessive temperature fluctuations will accelerate the erosion of arc furnace refractory materials and the consumption of electrodes. The present invention helps maintain a more stable operating temperature and reduce drastic temperature fluctuations by providing accurate temperature predictions. This not only extends the service life of refractory materials and electrodes, but also reduces the frequency of maintenance shutdowns and improves the overall efficiency of the equipment.

[0113] The system of the present invention can be seamlessly integrated with other control systems of the electric arc furnace, laying the foundation for the full automation of the electric arc furnace smelting process. By providing accurate real-time temperature data, the system can support more complex automatic control algorithms, such as automatic adjustment of electrode position, automatic control of feeding, etc., thereby further improving production efficiency and consistency of product quality. The system will accumulate a large amount of historical data during operation. These data are not only used to improve the prediction model, but also can be used to deeply analyze the smelting process and find potential optimization space. Through data mining, the key factors affecting the smelting quality can be found, providing data support for process improvement and innovation.

Claims

1. A method for predicting the real-time temperature of an arc furnace after melting. It is characterized in that The following steps are involved: S1. Install monitoring equipment: Install an infrared thermal imager on the top of the electric arc furnace so that its field of view covers the entire slag surface; install an argon blowing gun on the top of the electric arc furnace, with the argon blowing gun tip 30cm to 50cm away from the slag surface; S2. Obtaining the slag surface temperature: Using the infrared thermal imager installed in step S1, the average temperature of the slag surface is obtained at intervals of time t; S3. Expose the molten steel surface and measure the temperature: at intervals of t, argon is blown by the argon blowing gun installed in step S1 to disperse the slag and expose the molten steel surface; using the infrared thermal imager installed in step S1, the average temperature of the exposed molten steel surface is obtained at intervals of t and transmitted to the database in real time; S4. Collect real-time data, including: power consumption, oxygen blowing amount, carbon powder blowing amount, lime amount, furnace weight, flue gas composition and temperature, natural gas amount, average temperature of slag surface, and transmit them to the database in real time; S5. Establish a molten pool temperature prediction mechanism model, input the real-time data collected in step S4, obtain a real-time temperature mechanism prediction value, and transmit it to the database in real time; The method of establishing a molten pool temperature prediction mechanism model comprises the following steps: C1. Divide the molten pool into three regions: molten steel, slag and gas phase, establish the heat balance equation of the arc furnace molten pool, including arc heating, chemical reaction heat release and heat dissipation loss; and perform modeling; C2. According to the arc characteristics between the electrode and the molten pool, the effective heating contribution of the arc power to the molten pool is calculated, and an arc heating model is constructed; C3. Establish the heat model of chemical reactions of decarbonization, dephosphorization and desulfurization, and calculate the heat of reaction of each reaction; C4. Establish a heat dissipation loss model, including heat conduction loss of the furnace wall, heat conduction loss of the furnace cover, radiation heat loss and heat carried away by flue gas; C5. Establish a heat change model for materials entering and leaving the molten pool, including continuous feeding, slag discharge, and exhaust; C6. Integrate the above models to construct a complete prediction mechanism model of molten pool temperature, including establishing the differential equation of molten pool temperature changing with time and determining the boundary conditions and initial conditions of the model; C7. Use measured data to optimize model parameters and use the least squares method to fit the parameters; C8. Input the real-time collected process parameters into the molten pool temperature prediction mechanism model, solve the molten pool temperature prediction mechanism model equation, and output the temperature mechanism prediction value at the current moment; S6. Based on the recurrent neural network structure, a dynamic prediction model for the molten pool temperature is established; the database is read, and the real-time data collected in step S4 and the real-time temperature mechanism prediction value obtained in step S5 are used as input, and the average temperature of the exposed steel liquid surface obtained in step S3 is used as output; data from 5000 to 10000 heats are collected; S7. Real-time prediction of the current molten steel surface temperature: input the real-time data collected in step S4 at the current moment and the real-time temperature mechanism prediction value obtained in step S5; through the model calculation established in step S6, output the average temperature of the molten steel surface at the current moment.

2. The method for predicting the real-time temperature of an arc furnace after melting according to claim 1, It is characterized in that According to the comparison between the predicted average temperature of the molten steel surface at the current moment and the temperature required by the process, the process parameters are adjusted, specifically: When the average temperature of the molten steel surface is lower than the temperature required by the process, increase the power supply, increase the amount of oxygen blowing and the amount of carbon powder sprayed to meet the temperature required by the process; when the average temperature of the molten steel surface is higher than the temperature required by the process, reduce the power supply, reduce the amount of oxygen blowing and the amount of carbon powder sprayed to meet the temperature required by the process.

3. The method for predicting the real-time temperature of an arc furnace after melting according to claim 1 or 2, It is characterized in that The method of obtaining the average temperature of the slag surface at intervals of time t is characterized by comprising the following steps: A1. Every t seconds, a thermal image covering the entire slag surface is collected by the infrared thermal imager; A2. Perform image processing on the collected thermal images, including: identifying and excluding non-slag areas in the image; performing temperature conversion on each pixel point in the processed effective area; A3. Calculate the arithmetic mean of the temperature of all pixels in the effective area in step A2 to obtain the average temperature of the slag surface at the current moment; A4. Record the obtained average temperature value together with the corresponding timestamp and transmit it to the database; A5. Repeat steps A1 to A4 to achieve continuous temperature measurement every t seconds.

4. The method for predicting the real-time temperature of an arc furnace after melting and cleaning according to claim 3, It is characterized in that The method of obtaining the average temperature of the exposed steel liquid surface at intervals of time t comprises the following steps: B1. Every t seconds, argon is blown onto the slag surface through an argon blowing gun to expose the molten steel surface; B2. Immediately after completing step B1, a thermal image covering the exposed steel surface area is collected by the infrared thermal imager; B3. Perform image processing on the collected thermal images, including: identifying and locating the exposed steel surface area in the image; excluding the non-steel surface area in the image; and performing temperature conversion on each pixel point in the identified effective steel surface area; B4. Calculate the arithmetic mean of the temperatures of all valid pixels in step B3 to obtain the average temperature of the exposed steel surface at the current moment; B5. Record and transmit the obtained average temperature value of the molten steel surface together with the corresponding timestamp to the database; B6. Repeat steps B1 to B5 to achieve continuous steel liquid surface temperature measurement every t seconds.

5. The method for predicting the real-time temperature of an arc furnace after melting and cleaning according to claim 1, It is characterized in that The real-time data collection includes: Collecting power consumption: The power consumption of the electric arc furnace is measured and recorded in real time through the power metering device installed on the power supply system of the electric arc furnace; Collect oxygen blowing amount: Use the gas flow meter installed on the oxygen supply pipeline to measure and record the amount of oxygen blown into the arc furnace in real time; Collect the amount of carbon powder injected: The mass flow meter installed on the carbon powder injection system measures and records the amount of carbon powder injected into the electric arc furnace in real time; Collecting lime quantity: Using the weighing sensor installed on the lime feeding system, the amount of lime added to the electric arc furnace is measured and recorded in real time; Collecting furnace weight: The weight change of the entire furnace body is measured and recorded in real time through the weighing sensor system installed under the supporting structure of the electric arc furnace; Collect flue gas composition: Use the flue gas online analyzer installed on the flue of the arc furnace to measure and record the content of each component in the flue gas in real time; Collecting flue gas temperature: The flue gas temperature is measured and recorded in real time through a high-temperature thermocouple or optical fiber temperature measurement system installed on the flue of the arc furnace; Collecting natural gas volume: Using the gas flow meter installed on the natural gas supply pipeline, the volume of natural gas injected into the electric arc furnace is measured and recorded in real time; Collecting the average temperature of the slag surface: The average temperature of the slag surface is acquired and recorded every t seconds by the infrared thermal imaging system.

6. The method for predicting the real-time temperature of an arc furnace after melting and cleaning according to claim 1, It is characterized in that The method for establishing a dynamic prediction model for molten pool temperature based on a recurrent neural network structure includes: collecting historical data of 5,000 to 10,000 heats; arranging the collected data in chronological order to form a time series data set; selecting Transformer as the recurrent neural network structure, and outputting a predicted value of the average temperature of the molten steel surface at the current moment every t seconds.

7. A system for predicting the real-time temperature of an arc furnace after melting, using the method according to any one of claims 1 to 6, It is characterized in that Includes the following components: Infrared thermal imager: installed on the top of the electric arc furnace to collect thermal images of the slag surface and exposed steel liquid surface; Argon blowing gun: installed on the top of the electric arc furnace, used to blow argon to expose the steel liquid surface; The data acquisition unit includes a lime weighing sensor (1), an electric energy meter (2), an infrared thermal imager (3), a flue gas composition monitor (5), a flue gas temperature monitor (6), an oxygen mass flow meter (7), a carbon powder mass flow meter (8), a natural gas mass flow meter (9), and a furnace weight sensor (10), and is used to collect data on power consumption, oxygen blowing amount, carbon powder blowing amount, lime amount, furnace weight, flue gas composition and temperature, and natural gas amount in real time; Database: used to store real-time data obtained from infrared thermal imager, argon blowing gun and data acquisition unit; Mechanism model unit: Based on the established molten pool temperature prediction mechanism model, it inputs real-time data and outputs the temperature mechanism prediction value; Neural network prediction unit: Based on the established recurrent neural network model, it inputs real-time data and temperature mechanism prediction value, and outputs the average temperature of the steel liquid surface at the current moment; Control system: Integrates the above units, coordinates system operation, and displays the predicted results of the molten steel surface temperature to the operator. At the same time, it controls the power supply, oxygen blowing amount and carbon powder spraying amount according to the difference between the current average temperature of the molten steel surface and the temperature required by the process.

8. The system for predicting the real-time temperature of an arc furnace after melting and cleaning according to claim 7, It is characterized in that The installation position and angle of the infrared thermal imager and the argon blowing gun are adjustable to ensure that the maximum area of ​​the slag surface and the molten steel surface is covered.

9. The system for predicting the real-time temperature of an arc furnace after melting and cleaning according to claim 7, It is characterized in that The data acquisition unit is connected to the database via a wireless or wired network to achieve real-time data transmission; the control system includes a human-machine interface for displaying predicted temperature, historical data and alarm information, and allowing operators to manually adjust model parameters and system settings.

Images