Method for predicting optimal humidification amount of blast furnace based on multi-objective genetic algorithm

Abstract

The invention relates to a method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm, which comprises the following steps of: training an RBF (Radial Basis Function) neural network by taking the hot air pressure, the pressure difference, the humidification amount, the coke amount and the sulfur content of the previous furnace as input variables to obtain a molten iron physical temperature prediction model; the material speed, the air volume, the top pressure, the humidification amount, the coke amount and the silicon content of the molten iron of the previous furnace serve as input variables, and an RBF neural network is trained to obtain a fuel ratio prediction model and a coal ratio prediction model; the humidification amount is used as a decision variable, and the molten iron physical temperature, the fuel ratio and the coal ratio are used as constraint conditions to construct a multi-objective optimization model; and solving the multi-objective optimization model by utilizing a genetic algorithm to obtain the optimal humidification amount. Based on the multi-target genetic algorithm, prediction of the optimal humidification amount of the target blast furnace can be achieved under different production conditions and environment change conditions, and the carbon dioxide emission of the blast furnace is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of blast furnaces, and particularly to a method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm. Background Art

[0002] In modern steel production, as an important metallurgical device, the production efficiency and product quality of a blast furnace directly affect the entire production process and economic benefits. The humidification amount of a blast furnace is one of the important parameters affecting its operating performance and smelting effect. In blast furnace production, humidified blowing can be used as a regulating means, which has two aspects of influence on the fuel ratio of the blast furnace. On the one hand, it absorbs the heat of the hearth, which will increase the fuel ratio. Only when used in combination with high blast temperature and high oxygen enrichment can it play a role in improving the smelting intensity, promoting the stable and smooth operation of the furnace condition, maintaining a reasonable thermal regime, and reducing the fuel ratio. On the other hand, the hydrogen generated by the decomposition of humidified water participates in the reduction, which is beneficial to improving the reduction ability of the gas and is beneficial to reducing the coke ratio and fuel ratio. At present, the methods for determining the optimal humidification amount of a blast furnace mostly rely on experience and tests. This method does not consider the influence caused by environmental changes and is difficult to provide accurate prediction of the humidification amount. Summary of the Invention

[0003] To solve the above problems, an object of the embodiments of the present invention is to provide a method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm.

[0004] A method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm includes:

[0005] Taking the hot blast pressure, differential pressure, humidification amount, coke amount, and sulfur content of the previous furnace as input variables, and training an RBF neural network to obtain a molten iron physical temperature prediction model;

[0006] Taking the feeding rate, air volume, top pressure, humidification amount, coke amount, and silicon content of the molten iron of the previous furnace as input variables, and training an RBF neural network to obtain a fuel ratio prediction model;

[0007] Taking the feeding rate, air volume, top pressure, humidification amount, coke amount, and silicon content of the molten iron of the previous furnace as input variables, and training an RBF neural network to obtain a coal ratio prediction model;

[0008] Using the humidification amount as a decision variable, and taking the molten iron physical temperature, fuel ratio, and coal ratio as constraint conditions to construct a multi-objective optimization model;

[0009] Using a genetic algorithm to solve the multi-objective optimization model to obtain the optimal humidification amount.

[0010] Preferably, the molten iron physical temperature prediction model is:

[0011]

[0012] Among them, m 3 represents the hot air pressure, m 5 represents the differential pressure, m 9 represents the humidification amount, n 1 represents the sulfur content of the previous heat of hot metal, m 11 represents the amount of coke, b 1 represents the bias of the physical temperature prediction model of hot metal, w i1 represents the weight of the i-th hidden neuron of the physical temperature prediction model of hot metal, φ i (X p1 ) represents the output of the j-th neuron in the hidden layer of the physical temperature prediction model of hot metal.

[0013] Preferably, the fuel ratio prediction model is:

[0014]

[0015] Among them, m 1 represents the feeding rate, m 2 represents the blast volume, m 4 represents the top pressure, m 9 represents the humidification amount, m 11 represents the amount of coke, n 2 represents the silicon content of the previous heat of hot metal, b 2 represents the bias of the fuel ratio prediction model, w i2 represents the weight of the i-th hidden neuron of the fuel ratio prediction model, φ i (X p2 ) represents the output of the j-th neuron in the hidden layer of the fuel ratio prediction model.

[0016] Preferably, the coal ratio prediction model is:

[0017]

[0018] Among them, m 1 represents the feeding rate, m 2 represents the blast volume, m 4 represents the top pressure, m 9 represents the humidification amount, m 11 represents the amount of coke, n 2 represents the silicon content of the previous heat of hot metal, b 3 represents the bias of the coal ratio prediction model, w i3 represents the weight of the i-th hidden neuron of the coal ratio prediction model, φ i (X p3 ) represents the output of the j-th neuron in the hidden layer of the coal ratio prediction model.

[0019] Preferably, the multi-objective optimization model is:

[0020]

[0021] where m 9 represents the humidification amount.

[0022] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:

[0023] The present invention relates to a method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm. Compared with the prior art, the present invention is based on a multi-objective genetic algorithm and can realize the prediction of the optimal humidification amount of the target blast furnace under different production conditions and environmental changes, reducing the carbon dioxide emissions of the blast furnace.

[0024] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following preferred embodiments are specifically exemplified below and described in detail in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0026] Figure 1 Schematic diagram of the prediction effect of the temperature prediction model provided by the present invention;

[0027] Figure 2 Comparison chart of the model prediction and the actual fuel ratio provided by the present invention;

[0028] Figure 3 Comparison chart of the model prediction and the actual coal ratio provided by the present invention;

[0029] Figure 4 Schematic diagram of the Pareto front of the multi-objective optimization model provided by the present invention;

[0030] Figure 5 Industrial test result diagram provided by the present invention;

[0031] Figure 6 Influence diagram of different humidifications on the K value provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0033] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, the meaning of "a plurality" is two or more unless otherwise specifically defined.

[0034] In the present invention, unless otherwise clearly specified and defined, the terms "mounted", "connected", "coupled", "fixed", etc. shall be construed in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0035] Please refer to Figure 1-6 , a method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm, comprising:

[0036] Step 1: Taking the hot blast pressure, differential pressure, humidification amount, coke amount, and sulfur content of the previous furnace as input variables, training an RBF neural network to obtain a prediction model for the physical temperature of hot metal;

[0037] Among them, the prediction model for the physical temperature of hot metal is:

[0038]

[0039] Among them, m 3 represents the hot blast pressure, m 5 represents the differential pressure, m 9 represents the humidification amount, n 1 represents the sulfur content of the hot metal in the previous furnace, m 11 represents the coke amount, b 1 represents the bias of the prediction model for the physical temperature of hot metal, w i1 represents the weight of the i-th hidden neuron of the prediction model for the physical temperature of hot metal, φi (X p1 ) represents the output of the j-th neuron in the hidden layer of the physical temperature prediction model of hot metal.

[0040] Step 2: Use the material feeding speed, air volume, top pressure, humidification amount, coke amount, and silicon content of the previous furnace of hot metal as input variables, and train an RBF neural network to obtain a fuel ratio prediction model;

[0041] Among them, the fuel ratio prediction model is:

[0042]

[0043] Among them, m 1 represents the material feeding speed, m 2 represents the air volume, m 4 represents the top pressure, m 9 represents the humidification amount, m 11 represents the coke amount, n 2 represents the silicon content of the previous furnace of hot metal, b 2 represents the bias of the fuel ratio prediction model, w i2 represents the weight of the i-th hidden neuron of the fuel ratio prediction model, φ i (X p2 ) represents the output of the j-th neuron in the hidden layer of the fuel ratio prediction model.

[0044] Step 3: Use the material feeding speed, air volume, top pressure, humidification amount, coke amount, and silicon content of the previous furnace of hot metal as input variables, and train an RBF neural network to obtain a coal ratio prediction model;

[0045] In Step 3, the coal ratio prediction model is:

[0046]

[0047] Among them, m 1 represents the material feeding speed, m 2 represents the air volume, m 4 represents the top pressure, m 9 represents the humidification amount, m 11 represents the coke amount, n 2 represents the silicon content of the previous furnace of hot metal, b 3 represents the bias of the coal ratio prediction model, w i3 represents the weight of the i-th hidden neuron of the coal ratio prediction model, φ i (X p3 ) represents the output of the j-th neuron in the hidden layer of the coal ratio prediction model.

[0048] Step 4: Use the humidification amount as the decision variable, and construct a multi-objective optimization model with the physical temperature of hot metal, fuel ratio, and coal ratio as constraints;

[0049] Step 5: Solve the multi-objective optimization model using the genetic algorithm to obtain the optimal humidification amount.

[0050] In Step 5, the multi-objective optimization model is as follows:

[0051]

[0052] where m 9 represents the humidification amount.

[0053] Next, the present invention further describes a method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm in combination with specific embodiments:

[0054] ① Constraint condition modeling

[0055] In blast furnace production, the hearth temperature is often used to measure whether the control of fuel ratio and coal ratio is appropriate. When the fuel ratio and coal ratio increase, the heat generated by fuel combustion increases, and the hearth temperature rises. When the fuel ratio and coal ratio decrease, the heat generated by fuel combustion decreases, and the hearth temperature drops. There are two methods to measure the hearth temperature, namely the physical temperature of hot metal and the silicon content of hot metal.

[0056] An RBF neural network is used to establish the correlation model between the physical temperature of hot metal and the silicon content of hot metal. Through model testing, the hot blast pressure (m 3 ), differential pressure (m 5 ), humidification (m 9 ), coke amount (m 11 ), and sulfur content of the previous furnace (n 1 ) are used as input variables, a total of 5 variables, to establish a prediction model for the physical temperature of hot metal. The correlation between the silicon content of hot metal and humidification is not obvious, which is consistent with the actual situation encountered in production: in actual production, when the humidification is adjusted, the physical temperature of hot metal changes significantly, but the silicon content basically does not change. Therefore, only the physical temperature of hot metal is used as a constraint condition in the humidification prediction model, and the prediction curve is as Figure 1 shown. Introducing production experience, an RBF neural network-based correlation model for the physical temperature of hot metal is established as shown in the following formula.

[0057]

[0058] The prediction result of the physical temperature model of hot metal is as Figure 1 shown. The predicted output of the obtained model is basically consistent with the actual output. The relative error between the predicted output and the actual output of the correlation model of the physical temperature of hot metal is 0.14%, and the relative error is less than 5.0%, indicating that the prediction model test has a good effect.

[0059] ② Objective condition modeling

[0060] An RBF neural network is used to establish the correlation model between fuel ratio and coal ratio. After debugging the model structure, according to the consistency between the predicted value and the actual value, the material feeding speed (m 1 ), air volume (m 2 ), top pressure (m 4 ), humidification (m 9 ), coke amount (m 11 ), and silicon content in the molten iron of the previous furnace (n 2 ), a total of 6 variables are used as input variables to establish a fuel ratio prediction model. The material feeding speed (m 1 ), air volume (m 2 ), top pressure (m 4 ), humidification (m 9 ), coke amount (m 11 ), and silicon content in the molten iron of the previous furnace (n 2 ), a total of 6 variables are used as input variables to establish a coal ratio prediction model, as shown in the following formula.

[0061]

[0062] The predicted output and actual output of the fuel ratio prediction model are as Figure 2 shown, and the relative error is 0.52%. The predicted output and actual output of the coal ratio prediction model are as Figure 3 shown, and the relative error is 1.67%. The relative errors of both models are less than 5.0%, indicating that the test performance of the prediction model is good.

[0063] The performance of the prediction model is verified by the absolute error, relative error, and hit rate of the prediction model, as shown in Table 1.

[0064] As can be seen from Table 1, the relative error and hit rate of the model prediction are within a reasonable control range. The hit rates of the coal ratio and molten iron physical temperature models are 93.41% and 100% respectively, and the relative errors are 1.67% and 0.14% respectively, showing excellent results. The hit rate of the fuel ratio model is 84.61%, and the relative error is 0.52%. The prediction effect of the model is good. According to production experience, the absolute error of the molten iron physical temperature model prediction is 2.74 °C, the absolute error of the coal ratio model prediction is 2.91 kg·tHM -1 , and the absolute error of the fuel ratio model prediction is 3.64 kg·tHM -1 . The absolute error is small, and the prediction performance of the constraint condition and target condition modeling is good.

[0065] Table 1 Model performance indicators

[0066]

[0067] ③ Multi-objective optimization results and analysis of the NSGA-II algorithm

[0068] NSGA-II multi-objective optimization model: During the optimization process, humidification is used as a decision variable, and the physical temperature of hot metal is used as a constraint condition. The reasonable constraint range of the physical temperature of hot metal is controlled within 1500 ± 10 °C, which can ensure the smooth operation of the furnace condition while maintaining a low fuel consumption level. In addition, according to actual production experience, the variable order of the model is adjusted. The reasonable constraint range of humidification in the decision variables is 10 g·m - 3 - 20 g·m - 3, and the multi-objective optimization model is shown as follows.

[0069]

[0070] Given that there are many decision variables and objective functions, the present invention has conducted different tests on the population size from 100 - 1000. The population size corresponding to the model is set to 90, the maximum number of iterations is 200, the crossover probability is set to 0.85, and the mutation probability is 0.15. The Pareto solution set of the decision variables is solved, and the obtained Pareto front is as Figure 4 shown. To facilitate the representation of the maximum coal ratio and the minimum fuel ratio, the vertical coordinate uses the coal ratio as a negative number, and the horizontal coordinate is the fuel ratio.

[0071] Figure 4 The Pareto front of the multi-objective optimization model shown indicates that at point A, the fuel ratio is relatively low, but the coal ratio replacing the coke ratio is relatively small; at point B, the fuel ratio is relatively high, but the coal ratio replacing the coke ratio is relatively large. According to production experience, the key to improving the carbon ratio is to reduce the fuel ratio. Therefore, the fuel ratio values of 3 points on the Pareto front that are closer to point A are selected as the optimal solution set, as shown in Table 2. Among the 3 optimal solution sets, the fuel ratio index has been significantly improved. Among them, in the first solution set, when the fuel ratios are similar, the coal ratio replacing the coke ratio is relatively large. Therefore, the multi-objective optimization value 1 is taken as the optimal solution, that is, when the humidification is 12.98 g·m - 3, the best fuel ratio is 535.81 kg·tHM -1 , and the hearth temperature index is at an appropriate level. The model predicts that the fuel ratio is reduced by 4.19 kg·tHM -1 .

[0072] Table 2 Multi-objective optimization results

[0073]

[0074]

[0075] The present invention is expected to reduce CO 2 emissions by 39,992.33 tons per year, and the calculation process is shown as follows:

[0076]

[0077] In the formula, the effective volume of the blast furnace is 4150 m 3 , and the average utilization coefficient is 2.113 t·m -3 ·d -1 (considering the influence of the blowing-down effect), the average carbon content of the fuel is 81.33%, the molar mass of carbon atom is 12 g·mol -1 , and the molar mass of oxygen atom is 16 g·mol -1 .

[0078] In order to verify the prediction effect of the model, 146 groups of industrial tests were carried out on 8 # blast furnaces, and the changes of fuel ratio and coal ratio under different humidification amounts were tested, as Figure 5 shown. It is difficult to obtain the correlation between the fuel ratio, coal ratio and the change of humidification amount in Figure 5 , and it is necessary to convert the fuel ratio and coal ratio into the index K first before comparison. In addition, Figure 5 the hearth temperatures and gas utilization rates corresponding to the fuel ratio and coal ratio in Figure 5 are different, and it is necessary to convert the fuel ratio and coal ratio data in

[0079] to the conditions of the same hearth temperature and gas utilization rate.

[0080] Let the ratio of the fuel ratio to the coal ratio be K, then the calculation method of K is as follows:

[0081] K = Fr / PCR

[0082] In the above formula, Fr—the fuel ratio; PCR—the coal ratio. 2 It can be seen from the above formula that the smaller the K value, the lower the fuel ratio of the blast furnace, the higher the coal ratio, the more optimized the fuel structure, and the better the effect of reducing CO

[0083] Fitting the Figure 5 trend line by the least square method gives Figure 6 . From the Figure 6 146 groups of data, 37 groups of data with the smallest K value (about 25% of the total data volume) are selected, and the average humidification amount and the arithmetic mean of K of these 37 groups of data are used as the optimal solution in the industrial data. After calculation, the arithmetic mean of the humidification amount is 13.28 g·m - 3, the arithmetic mean of the K value is 3.89, the average fuel ratio at this time is 536.49 kg·t -1 , and the average coal ratio is 138.01 kg·t -1 . The absolute error between the optimal humidification amount setting value predicted by the model and the industrial test result is 0.30 g·m -3. The set value of the humidification amount obtained from the industrial test is relatively consistent with the set value of the humidification amount predicted by the model, and the model has good prediction performance. The fuel ratio obtained from the industrial test is 536.49 kg·t -1 compared with 540.00 kg·t -1 which is reduced by 3.51 kg·t -1 . The annual CO 2 emission can be reduced by 33501.93 t / year, and the calculation process is shown in the following formula.

[0084]

[0085] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technical solution that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm, characterized in that: include: The hot air pressure, pressure difference, humidification amount, coke amount, and sulfur content of the previous furnace are used as input variables to train the RBF neural network to obtain the prediction model of molten iron physical temperature; The material speed, air volume, top pressure, humidification amount, coke amount, and silicon content of the previous furnace of molten iron are used as input variables to train the RBF neural network to obtain the fuel ratio prediction model; The material speed, air volume, top pressure, humidification amount, coke amount, and silicon content of the previous furnace of molten iron are used as input variables to train the RBF neural network to obtain the coal ratio prediction model; A multi-objective optimization model was constructed using the humidification amount as the decision variable and the molten iron physical temperature, fuel ratio and coal ratio as constraints. The multi-objective optimization model is solved by using a genetic algorithm to obtain the optimal humidification amount.

2. The method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm according to claim 1, characterized in that: The molten iron physical temperature prediction model is: Among them, m3 represents the hot air pressure, m5 represents the pressure difference, m9 represents the humidification amount, n1 represents the sulfur content of the previous furnace molten iron, and m 11 represents the amount of coke, b1 represents the bias of the molten iron physical temperature prediction model, and w i1 represents the weight of the ith hidden neuron in the molten iron physical temperature prediction model, φ i (X p1 ) represents the output of the jth neuron in the hidden layer of the molten iron physical temperature prediction model.

3. The method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm according to claim 1, characterized in that: The fuel ratio prediction model is: Among them, m1 represents the material speed, m2 represents the air volume, m4 represents the top pressure, m9 represents the humidification amount, m 11 represents the amount of coke, n2 represents the silicon content of the previous hot metal, b2 represents the bias of the fuel ratio prediction model, wi2 represents the weight of the i-th hidden neuron in the fuel ratio prediction model, φ i (X p2 ) represents the output of the jth neuron in the hidden layer of the fuel ratio prediction model.

4. The method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm according to claim 1, characterized in that: The coal ratio prediction model is: Among them, m1 represents the material speed, m2 represents the air volume, m4 represents the top pressure, m9 represents the humidification amount, m 11 represents the amount of coke, n2 represents the silicon content of the previous furnace of molten iron, b3 represents the bias of the coal ratio prediction model, wi3 represents the weight of the i-th hidden neuron in the coal ratio prediction model, φ i (X p3 ) represents the output of the jth neuron in the hidden layer of the coal ratio prediction model.

5. A method for predicting the optimal humidification amount of a blast furnace based on a multi-objective genetic algorithm according to any one of claims 2 to 4, characterized in that: The multi-objective optimization model is: Among them, m9 represents the humidification amount.

Images