Intelligent optimization method of fsFBG temperature in quasi-distributed steelmaking electric furnace based on expert model

Abstract

The quasi-distributed steel furnace fsFBG temperature intelligent optimization method based on the expert model first installs femtosecond laser point-by-point direct writing fiber Bragg grating sensors at key positions of the furnace to collect data, and demodulates through different methods to ensure accurate acquisition of temperature information; then establishes an expert model and integrates multiple independent gradient boosting tree models; then conducts in-depth data analysis to mine temperature data to identify key factors. Generate an optimization strategy based on the expert model and data analysis, and determine the best temperature control strategy under various working conditions through high-performance computing; then feed back the optimization strategy to the furnace control system through the real-time data system and continuously monitor the actual temperature, and use the PID controller to reduce the deviation; then convert the adjusted control strategy into precise instructions to regulate the power of the heating element and the flow of the cooling system to achieve precise temperature control; finally, regularly evaluate the operation effect of the furnace, compare the temperature control and energy consumption, and continuously optimize the expert model and control algorithm.

Description

Technical Field

[0001] The invention relates to the field of fiber grating sensors, and in particular to a quasi-distributed steel furnace fsFBG temperature intelligent optimization method based on an expert model. Background Art

[0002] In modern industrial production, steel electric furnaces, as an important heating equipment, are widely used in metallurgy, chemical industry, machinery manufacturing and other fields. However, traditional temperature control methods for steel electric furnaces often have problems such as low precision, high energy consumption, and poor stability, making it difficult to meet increasingly stringent production requirements. With the continuous advancement of science and technology, fiber grating sensor technology has gradually been widely used in the field of temperature monitoring due to its advantages such as high precision, anti-electromagnetic interference, and corrosion resistance. Femtosecond laser point-by-point direct writing fiber grating sensor (fsFBG) provides a new solution for temperature monitoring of steel electric furnaces with its high resolution and fast response.

[0003] At the same time, traditional temperature control methods usually use simple control algorithms. Although they can achieve temperature regulation to a certain extent, their control accuracy is often difficult to meet the requirements of high-precision production for complex steel furnace systems. For example, in some material processing processes that have extremely strict temperature requirements, small temperature fluctuations may lead to significant changes in product performance. Summary of the invention

[0004] To solve the above problems, the present invention proposes a quasi-distributed steel furnace fsFBG temperature intelligent optimization method based on an expert model. This scheme collects data by installing a femtosecond laser fiber grating sensor in the steel furnace, establishes an integrated expert model, conducts in-depth data analysis, generates optimization strategies, implements real-time monitoring and adjustment, realizes precise temperature control, and regularly evaluates and improves to ensure efficient and stable operation of the furnace and optimize energy consumption management.

[0005] The quasi-distributed steel furnace fsFBG temperature intelligent optimization method based on the expert model has the following specific steps:

[0006] Step 1: Collect the relevant data of the steel electric furnace by collecting sensors, install multiple femtosecond laser point-by-point direct writing fiber Bragg grating sensors at key positions of the steel electric furnace, demodulate different signals by different methods, and ensure accurate acquisition of temperature information in a complex electric furnace operation environment;

[0007] Step 2: Establish an expert model, combine the collected data, fuse multiple independent gradient boosting tree models to obtain a more powerful integrated model, and calculate accurate temperature data;

[0008] Step 3: In-depth data analysis: using big data technology to mine temperature data, identify key factors affecting temperature distribution, and provide data support for optimized control;

[0009] Step 4: Generate optimization strategies, based on expert models and data analysis, and determine the best temperature control strategies under various working conditions through high-performance computing;

[0010] Step 5: Real-time monitoring and adjustment: optimize the strategy to the electric furnace control system through real-time data system feedback, and continuously monitor the actual temperature to reduce the deviation between feedback and setting through PID controller;

[0011] Step 6, precise temperature control, converting the control strategy adjusted by the PID controller into precise control instructions to regulate the heating element power of the electric furnace and the flow of the cooling system to ensure that the temperature reaches the optimal state;

[0012] Step 7: Effect evaluation and improvement: regularly evaluate the operation effect of the electric furnace, compare temperature control and energy consumption, and continuously optimize the expert model and control algorithm.

[0013] Further, the process of collecting the relevant data of the steel electric furnace collected by the sensor in step 1 is expressed as:

[0014] An array of femtosecond laser point-by-point direct-writing fiber Bragg grating sensors is installed at key positions of the steel electric furnace. Different filtering devices are used to output the spectral characteristics of the sensor, and spectral signal decoding is achieved by monitoring the peak value of the filtered signal. Different filtering methods include matched grating method, edge filtering method, and tunable FP filter method.

[0015] Furthermore, the process of establishing the expert model in step 2 is expressed as follows:

[0016] Step 2.1, three independent gradient boosting tree models are set up to predict the temperature of the data collected by the matched grating method, the edge filtering method, and the tunable FP filter method, respectively. Each model is obtained by training the boosting tree algorithm;

[0017] Step 2.2, train each expert model. The following is the implementation method of a single gradient boosting tree expert model;

[0018] Step 2.2.1, initialize the model. The prediction value of the initialization model is a constant, and the prediction value is described as the mean f0(x) of all target values:

[0019]

[0020] In the formula, n is the number of samples, x is the sample, and y is i is the true value of the temperature of the i-th sample, L is the loss function, and γ is a constant;

[0021] Step 2.2.2, iteratively build the tree and calculate the negative gradient of the mth iteration:

[0022]

[0023] In the formula, r im is the negative gradient calculated in each iteration, indicating the direction and degree of the gap between the current model prediction value and the true value, which is used to guide the construction of the next tree. m-1 (x i ) indicates that before the mth iteration, the model has i The predicted value is trained using the training data and negative gradient to train a regression tree h m (x), whose leaf node area is denoted by J m ;

[0024] Step 2.2.3, calculate the value of the leaf node:

[0025]

[0026] In the formula, Υ jm represents the optimal prediction value of the jth leaf node region in the mth iteration, so that the loss in the leaf node region is minimized, Υ represents the value of the leaf node, R jm Represents the regression tree h obtained in the mth iteration training m The jth leaf node region of (x), J m Represents the leaf node area of ​​the regression tree obtained by the mth iteration training, and each leaf node area corresponds to a predicted value;

[0027] Step 2.2.4, update the model:

[0028]

[0029] Where I(·) is the indicator function, which is 1 if x is in the leaf node region and 0 otherwise;

[0030] Step 2.2.5, after M iterations, output a single gradient boosting tree expert model f(x):

[0031]

[0032] In step 2.3, the prediction results of each expert model are integrated and the final prediction value is calculated using the averaging method.

[0033] Further, the process of deep data analysis in step 3 is expressed as follows:

[0034] The collected data are deeply mined, and the key factors with strong correlation with temperature are analyzed through the Pearson correlation coefficient formula, among which the key factors are electric furnace power and ambient temperature.

[0035] Furthermore, the process of generating the optimization strategy in step 4 can be expressed as follows:

[0036] Based on the key factors found in step 3, an optimization model is constructed to minimize the deviation between the temperature and the set value. The optimization formula can be expressed as:

[0037]

[0038] In the formula, u is the control variable representing the furnace power and ambient temperature, T i (u) represents the temperature of the ith observation point under the control variable u, T set Indicates the set target temperature.

[0039] The quasi-distributed steel furnace fsFBG temperature intelligent optimization method based on the expert model of the present invention has beneficial effects. The technical effects of the present invention are:

[0040] 1. The present invention adopts advanced femtosecond laser point-by-point direct writing fiber Bragg grating sensor, which can accurately obtain temperature information in complex environments, ensuring high accuracy and reliability of monitoring;

[0041] 2. The present invention establishes a powerful integrated expert model by fusing multiple gradient boosting tree models, improves the accuracy of prediction and decision-making, and enhances the intelligence of the system;

[0042] 3. The solution of the present invention includes real-time monitoring, adjustment and effect evaluation links, which can continuously optimize control strategies and algorithms, ensuring long-term optimization of electric furnace operation efficiency and energy management. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of the present invention;

[0044] Figure 2 It is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0046] The present invention proposes a quasi-distributed steel furnace fsFBG temperature intelligent optimization method based on an expert model. The invention flow chart is as follows: Figure 1 As shown, the steps of the present invention are described in detail below.

[0047] Step 1: Collect sensors to collect relevant data of the steel furnace. Install multiple femtosecond laser point-by-point direct writing fiber Bragg grating sensors at key positions of the steel furnace. Demodulate different signals by different methods to ensure accurate acquisition of temperature information in a complex furnace operation environment. The system structure diagram is shown in the figure. Figure 2 As shown;

[0048] An array of femtosecond laser point-by-point direct-writing fiber Bragg grating sensors is installed at key positions of the steel electric furnace. Different filtering devices are used to output the spectral characteristics of the sensor, and spectral signal decoding is achieved by monitoring the peak value of the filtered signal. Different filtering methods include matched grating method, edge filtering method, and tunable FP filter method.

[0049] Step 2: Establish an expert model, combine the collected data, fuse multiple independent gradient boosting tree models to obtain a more powerful integrated model, and calculate accurate temperature data;

[0050] Step 2.1, three independent gradient boosting tree models are set up to predict the temperature of the data collected by the matched grating method, the edge filtering method, and the tunable FP filter method, respectively. Each model is obtained by training the boosting tree algorithm;

[0051] Step 2.2, train each expert model. The following is the implementation method of a single gradient boosting tree expert model;

[0052] Step 2.2.1, initialize the model. The prediction value of the initialization model is a constant, and the prediction value is described as the mean f0(x) of all target values:

[0053]

[0054] In the formula, n is the number of samples, x is the sample, and y is i is the true value of the temperature of the i-th sample, L is the loss function, and γ is a constant;

[0055] Step 2.2.2, iteratively build the tree and calculate the negative gradient of the mth iteration:

[0056]

[0057] In the formula, r im is the negative gradient calculated in each iteration, indicating the direction and degree of the gap between the current model prediction value and the true value, which is used to guide the construction of the next tree. m-1 (x i ) indicates that before the mth iteration, the model has i The predicted value is trained using the training data and negative gradient to train a regression tree h m (x), whose leaf node area is denoted by J m ;

[0058] Step 2.2.3, calculate the value of the leaf node:

[0059]

[0060] In the formula, Υ jmrepresents the optimal prediction value of the jth leaf node region in the mth iteration, so that the loss in the leaf node region is minimized, Υ represents the value of the leaf node, R jm Represents the regression tree h obtained in the mth iteration training m The jth leaf node region of (x), J m Represents the leaf node area of ​​the regression tree obtained by the mth iteration training, and each leaf node area corresponds to a predicted value;

[0061] Step 2.2.4, update the model:

[0062]

[0063] Where I(·) is the indicator function, which is 1 if x is in the leaf node region and 0 otherwise;

[0064] Step 2.2.5, after M iterations, output a single gradient boosting tree expert model f(x):

[0065]

[0066] In step 2.3, the prediction results of each expert model are integrated and the final prediction value is calculated using the averaging method.

[0067] Step 3: In-depth data analysis: using big data technology to mine temperature data, identify key factors affecting temperature distribution, and provide data support for optimized control;

[0068] The collected data are deeply mined, and the key factors with strong correlation with temperature are analyzed through the Pearson correlation coefficient formula, among which the key factors are electric furnace power and ambient temperature.

[0069] Step 4: Generate optimization strategies, based on expert models and data analysis, and determine the best temperature control strategies under various working conditions through high-performance computing;

[0070] Based on the key factors found in step 3, an optimization model is constructed to minimize the deviation between the temperature and the set value. The optimization formula can be expressed as:

[0071]

[0072] In the formula, u is the control variable representing the furnace power and ambient temperature, T i (u) represents the temperature of the ith observation point under the control variable u, T set Indicates the set target temperature.

[0073] Step 5: Real-time monitoring and adjustment: optimize the strategy to the electric furnace control system through real-time data system feedback, and continuously monitor the actual temperature to reduce the deviation between feedback and setting through PID controller;

[0074] Step 6, precise temperature control, converting the control strategy adjusted by the PID controller into precise control instructions to regulate the heating element power of the electric furnace and the flow of the cooling system to ensure that the temperature reaches the optimal state;

[0075] Step 7: Effect evaluation and improvement: regularly evaluate the operation effect of the electric furnace, compare temperature control and energy consumption, and continuously optimize the expert model and control algorithm.

[0076] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A quasi-distributed steel furnace fsFBG temperature intelligent optimization method based on an expert model, the specific steps are as follows, and the characteristics are as follows: Step 1: Collect the relevant data of the steel electric furnace by collecting sensors, install multiple femtosecond laser point-by-point direct writing fiber Bragg grating sensors at key positions of the steel electric furnace, demodulate different signals by different methods, and ensure accurate acquisition of temperature information in a complex electric furnace operation environment; Step 2: Establish an expert model, combine the collected data, fuse multiple independent gradient boosting tree models to obtain a more powerful integrated model, and calculate accurate temperature data; The process of establishing the expert model in step 2 is as follows: Step 2.1, three independent gradient boosting tree models are set up to predict the temperature of the data collected by the matched grating method, the edge filtering method, and the tunable FP filter method, respectively. Each model is obtained by training the gradient boosting tree algorithm; Step 2.2, train each expert model. The following is the implementation method of a single gradient boosting tree expert model; Step 2.2.1, initialize the model. The prediction value of the initialization model is a constant, and the prediction value is described as the mean f0(x) of all target values: In the formula, n is the number of samples, x is the sample, and y is i is the true value of the temperature of the i-th sample, L is the loss function, and γ is a constant; Step 2.2.2, iteratively build the tree and calculate the negative gradient of the mth iteration: In the formula, r im is the negative gradient calculated in each iteration, indicating the direction and degree of the gap between the current model prediction value and the true value, which is used to guide the construction of the next tree. m-1 (x i ) indicates that before the mth iteration, the model has i The predicted value is trained using the training data and negative gradient to train a regression tree h m (x), whose leaf node area is denoted by J m ; Step 2.2.3, calculate the value of the leaf node: In the formula, Υ jm represents the optimal prediction value of the jth leaf node region in the mth iteration, so that the loss in the leaf node region is minimized, Υ represents the value of the leaf node, R jm Represents the regression tree h obtained in the mth iteration training m The jth leaf node region of (x), J m Represents the leaf node area of ​​the regression tree obtained by the mth iteration training, and each leaf node area corresponds to a predicted value; Step 2.2.4, update the model: Where I(·) is the indicator function, which is 1 if x is in the leaf node region and 0 otherwise; Step 2.2.5, after M iterations, output a single gradient boosting tree expert model f(x): Step 2.3, integrate the prediction results of each expert model and calculate the final prediction value using the average method; Step 3: In-depth data analysis: using big data technology to mine temperature data, identify key factors affecting temperature distribution, and provide data support for optimized control; Step 4: Generate optimization strategies, based on expert models and data analysis, and determine the best temperature control strategies under various working conditions through high-performance computing; Step 5: Real-time monitoring and adjustment: optimize the strategy to the electric furnace control system through real-time data system feedback, and continuously monitor the actual temperature to reduce the deviation between feedback and setting through PID controller; Step 6, precise temperature control, converting the control strategy adjusted by the PID controller into precise control instructions to regulate the heating element power of the electric furnace and the flow of the cooling system to ensure that the temperature reaches the optimal state; Step 7: Effect evaluation and improvement: regularly evaluate the operation effect of the electric furnace, compare temperature control and energy consumption, and continuously optimize the expert model and control algorithm.

2. The quasi-distributed steel furnace fsFBG temperature intelligent optimization method based on expert model according to claim 1 is characterized by: The process of collecting the relevant data of the steel electric furnace collected by the sensor in step 1 is expressed as follows: An array of femtosecond laser point-by-point direct-writing fiber Bragg grating sensors is installed at key positions of the steel electric furnace. Different filtering devices are used to output the spectral characteristics of the sensor, and spectral signal decoding is achieved by monitoring the peak value of the filtered signal. Different filtering methods include matched grating method, edge filtering method, and tunable FP filter method.

3. The quasi-distributed steelmaking electric furnace fsFBG temperature intelligent optimization method based on expert model according to claim 1 is characterized by: The process of deep data analysis in step 3 is as follows: The collected data are deeply mined, and the key factors with strong correlation with temperature are analyzed through the Pearson correlation coefficient formula, among which the key factors are electric furnace power and ambient temperature.

4. The quasi-distributed steelmaking electric furnace fsFBG temperature intelligent optimization method based on expert model according to claim 1 is characterized by: The process of generating the optimization strategy in step 4 is as follows: Based on the key factors found in step 3, an optimization model is constructed to minimize the deviation between the temperature and the set value. The optimization formula is expressed as: In the formula, u is the control variable representing the furnace power and ambient temperature, T i (u) represents the temperature of the ith observation point under the control variable u, T set Indicates the set target temperature.

Images