Converter gas feasible recovery amount prediction method based on multi-source heterogeneous data

Abstract

The invention discloses a converter gas feasible recovery amount prediction method based on multi-source heterogeneous data, and belongs to the technical field of complementary energy recovery in the iron and steel industry. The method comprises the following steps: analyzing a converter gas generation mechanism to obtain influence factors; the method comprises the following steps: collecting corresponding primary machine and secondary machine data, and preprocessing the primary machine data into a data set taking a furnace as a unit; fusing the first-level data and the second-level data to form a converter gas feasible recovery quantity prediction data set; a neural network prediction method is utilized, the influence factors serve as an input layer, the gas feasible recovery amount serves as an output layer, and a gas feasible recovery amount prediction model is established; signals of converter smelting steel types, blowing starting time, blowing ending time and the like in three-level machine steelmaking production plan data are collected, the converter production process type is determined, and according to operation factors corresponding to converter production of the type, accurate prediction of the feasible recovery amount of converter gas is achieved through the prediction model.

Description

Technical Field

[0001] The present invention relates to the technical field of waste energy recovery in the steel industry, and in particular to a method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data. Background Art

[0002] Recovering energy carriers and various energies lost during the production process is an indispensable part of energy conservation in the steel industry. As one of the three major by-product gases of steel enterprises, converter gas has the characteristics of intermittent, complex, and uncertain recovery due to the periodicity and complexity of converter production. It is very easy to cause the converter gas cabinet to be full and rejected, and the gas to be released, making the dynamic balance control of the by-product gas in the whole plant more difficult. Accurate prediction of converter gas recovery volume is an important basis for optimizing the balance control of by-product gas, which can effectively improve the converter gas recovery rate and reduce gas release.

[0003] The amount of converter gas recovered is mainly related to the amount of converter gas generated, the conditions for converter gas recovery, and whether the converter gas tank is full during the recovery period. Among them, the amount of converter gas generated is mainly affected by the raw material factors, product factors and operation factors of steelmaking, and the converter gas recovery conditions are mainly affected by the CO concentration and O2 concentration in the blowing stage. For steel companies, the recovery conditions remain basically unchanged, and the amount of converter gas that meets the recovery conditions is the feasible recovery amount of converter gas.

[0004] At present, most of the predictions about converter gas generation are based only on the historical data of converter gas generation. Among them, Chinese patent application number CN202310626080.0 discloses a method for predicting converter gas generation, which extracts the intermittent duration characteristics from the original converter gas generation data and classifies the data according to the intermittent duration characteristics; then, according to the generation data characteristics obtained by intermittent classification, the CPSO-Elman generation prediction model and the Elman intermittent duration prediction model are established; Chinese patent application number CN201410066954.2 is a long-term prediction method for converter gas generation in metallurgical enterprises based on steelmaking rhythm estimation, which uses the model matching method to extract the time domain characteristics and frequency domain characteristics of the steelmaking rhythm from the on-site converter gas generation data, and then uses the improved fuzzy C-means clustering method to fuse the data, and finally inversely constructs the long-term prediction value of converter gas generation. The above prediction method only makes predictions from a data-driven method, without considering the influence of converter process parameters (raw materials, products, operations) on the amount of converter gas generated. In addition, the CO concentration change patterns of converter gas in the early, middle and late stages of blowing are different, and the gas recoverable sections are different, which in turn affects the recovery amount of converter gas.

[0005] Some gas generation prediction methods take into account the impact of CO concentration on gas recovery during the blowing stage. Among them, Chinese patent application number CN202210589223.0 discloses a converter gas generation and supply adjustment method and system based on gas cabinet position. Based on the on-site historical data, the characteristic curve of converter flue gas flow rate and CO and O2 concentrations changing with blowing time is fitted, and then the converter gas recovery is predicted. Chinese patent application number CN201110070892.9 discloses a by-product gas generation prediction method based on the grey multi-factor MGM (l, n) model of principal component analysis. According to the on-site converter gas generation and the CO flow rate, CO2 flow rate, N2 flow rate and O2 consumption data of flue gas components, the principal component analysis of the converter gas before, during and after is carried out. Then, according to the obtained principal components and converter gas generation, the converter gas generation MGM (l, 3) prediction model is established to predict its generation. The above prediction method takes into account the differences in the composition of converter gas at different blowing stages, but does not consider the impact of the raw materials, products and operating conditions of converter smelting on the amount of converter gas generated and the gas composition.

[0006] And Chinese patent application number CN202310024846.8 discloses a method for predicting the amount of converter gas generated based on material carbon balance. Through the carbon balance theory in converter production, a method for predicting the amount of converter gas generated based on the converter production plan is established. This method takes into account the raw materials and product factors of converter production, but does not consider the influence of operating conditions such as smoke hood height, fan speed, annular seam opening, etc. on the amount of converter gas generated. The converter recovery time is simplified according to a fixed value. Summary of the invention

[0007] 1. Problem to be solved

[0008] In order to achieve accurate prediction of the feasible recovery amount of converter gas, the present invention proposes a method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data. The method obtains the influencing factors (raw materials, products, operations) of the converter gas recovery amount through the analysis of the converter gas generation mechanism, and preprocesses and integrates the on-site primary and secondary data into a converter gas feasible recovery amount prediction data set. A converter gas feasible recovery amount prediction model is established using a neural network method, and the feasible recovery amount of converter gas is predicted in combination with the converter production plan, providing a basis for the dynamic balance scheduling optimization of converter gas.

[0009] 2. Technical solution

[0010] In order to solve the above problems, the technical solution adopted by the present invention is as follows:

[0011] The method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data proposed in the present invention has the following specific implementation steps:

[0012] Step 1: Establish a converter gas energy-quality balance model through converter gas generation mechanism analysis to obtain the raw materials, products, operations and recovery influencing factors of the feasible converter gas recovery amount.

[0013] Step 2: Based on the above factors, the corresponding primary and secondary machine data are collected respectively, and the primary machine data is preprocessed into a data set based on furnaces;

[0014] Step 3: According to the blowing time in the primary and secondary data, the primary and secondary data are integrated to form a prediction data set of the feasible recovery amount of converter gas;

[0015] Step 4: Using the neural network prediction method, taking the above-mentioned influencing factors as the input layer and the feasible recovery amount of coal gas as the output layer, and using the above-mentioned prediction data set, establish a prediction model for the feasible recovery amount of coal gas;

[0016] Step 5: Collect signals such as the type of steel being smelted in the converter, the start time of blowing, and the end time of blowing from the third-level machine data of the steelmaking production plan, determine the type of converter production process, and use the prediction model to accurately predict the feasible recovery amount of converter gas based on the operating factors corresponding to this type of converter production.

[0017] The method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data provided by the present invention comprises the following steps:

[0018] The first step is to analyze the influencing factors

[0019] By analyzing the mechanism of converter gas generation, the converter gas quality conservation model and converter heat balance model are established:

[0020] Converter gas recovery volume V gas,rec It is mainly affected by the amount of converter flue gas generated, converter gas recovery conditions and gas cabinet position.

[0021] Among them, the converter flue gas generation includes the theoretical furnace gas generation V Fgas And the cold air volume V air N2 in.

[0022]

[0023] Furnace gas generation V Fgas The main process is the oxidation reaction between C in the molten iron and the blowing O2 to produce CO and CO2. Based on the converter heat balance, we know that:

[0024]

[0025] Among them, Q iron =m iron ×[c ironS ×(t ironfreeze -tenvir )+Q ironMelt +c ironL ×(t iron -t ironfreeze )]

[0026] Q oxi =29177m Si +6593m Mn +11637m C(co) +34824m C(co2)

[0027] +35874m P +1620m SiO2 +4249m Fe ( FeO) +6459m Fe ( Fe2O3)

[0028] Q steel =m steel ×[c steelS ×(t steelfreeze -t envir )+Q steelMelt +c steelL ×(t steel -t steelfreeze )]

[0029] Q slag =m slag ×[c slagS ×(t slag -t envir )+Q slagMelt ].

[0030] Cold air volume V at furnace mouth air It is mainly affected by the furnace mouth pressure and dust removal system control. Based on the OG system pressure balance, it can be known that:

[0031]

[0032] Where P0 = P2-(ΔP Vfp +ΔP twr +ΔP vntr +ΔP rec -P fan )

[0033] ΔP vntr =S vntr (0.78V air +V Fgas ) 2

[0034]

[0035] Comprehensive analysis shows that the converter gas generation V is mainly affected by raw material factors (molten iron charge, scrap steel charge, feed amount, molten iron temperature, molten iron C content, molten iron Si content, molten iron Mn content, molten iron P content), product factors (end steel temperature, molten steel C content, molten steel P content, molten steel Si content, molten steel Mn content) and operating factors (active smoke hood height, annular seam opening, fan speed) and other factors.

[0036] The feasible recovery amount of converter gas is also affected by the recovery conditions, namely the CO concentration and O2 concentration of the converter gas during the blowing stage. The CO concentration of the converter gas during the blowing stage is mainly related to the molten iron composition, molten iron temperature, oxygen blowing intensity and oxygen lance height.

[0037] Step 2: Multi-source data collection and preprocessing

[0038] The multi-source data includes the time series data of the first-level machine, the data of the second-level machine in units of furnaces, and the production plan data of the third-level machine. Among them, the above-mentioned raw material factors and product factors are collected from the second-level machine, the operating factors and recycling condition factors are time series data that fluctuate during the smelting process, and the field data are collected in the first-level machine. The data of the third-level machine is based on the converter blowing plan to provide support for subsequent model predictions.

[0039] Due to the different data structures, the first-level machine data needs to be preprocessed. The blowing period is mainly intercepted according to the first-level data O2 flow rate, and then the relevant operating factors and gas flow data of the feasible recovery period are intercepted according to the CO and O2 concentration recovery limits to form the first-level machine training data set.

[0040] The third step is to fuse the data to form a prediction data set for the feasible recovery of coal gas.

[0041] The prediction data set comes from the first-stage machine training set and the second-stage machine data set after the above preprocessing. According to the blowing start time and blowing end time of the second-stage data, the first-stage machine data set of the corresponding period is intercepted, and then the gas flow rate of the period is summed, and other operating factors such as hood height, fan speed, annular seam opening are averaged, and the processing results are fused with the corresponding second-stage machine data set to form a prediction data set for the feasible recovery of gas.

[0042] Step 4: Establish a prediction model for feasible gas recovery

[0043] The factors affecting the feasible recovery of gas analyzed above (such as molten iron weight, scrap steel weight, molten iron C content, molten steel C content, total weight of feed, smoke hood height, annular seam opening, fan speed, etc.) are used as the input layer of the neural network prediction model, and the feasible recovery of gas is used as the output layer. The neural network structures such as LM-BP, RBF, and CNN are used to train the model using the above-established feasible recovery of gas prediction data set to establish a feasible recovery of gas prediction model.

[0044] Step 5: Combine production plan to achieve accurate prediction of feasible gas recovery volume

[0045] The information on the start and end time of blowing, type of steel to be smelted, etc. in the three-stage converter steelmaking production plan is read, and combined with the operational experience data of smelting this type of steel, a prediction model input data set is constructed. The feasible gas recovery amount prediction model established above is used to predict the feasible gas recovery amount of this furnace, and then the gas flow rate is calculated according to the blowing period.

[0046] The present invention relates to the theoretical analysis of factors affecting converter gas generation and recovery, multi-source heterogeneous data fusion, and neural network prediction of the feasible recovery amount of converter gas, and is a method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data. The present invention obtains the factors affecting the feasible recovery amount of converter gas through the converter gas generation mechanism and heat and mass balance analysis, and then uses the primary and secondary field historical data (multi-source heterogeneous data) to fuse the multi-source structure as a historical data set for deep learning, and adopts LM-BP, RBF and other neural network methods to predict the feasible recovery amount, and finally combines the production plan of the tertiary machine converter to predict the feasible recovery amount of the converter. The prediction result can effectively guide the balanced scheduling of converter gas, and can also be used to analyze the impact of the production plan on the feasible recovery amount of converter gas.

[0047] 3. Beneficial effects

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The present invention analyzes the mechanism of converter steelmaking gas generation and heat and mass balance theory, combines multi-source heterogeneous data such as the time series data of the first-level machine, the production data of the second-level machine and the production plan of the third-level machine, and uses a neural network prediction method to establish a method for predicting the feasible recovery amount of converter gas. The method takes into account the influence of multiple factors such as raw materials, products, operations and recovery in the converter production process, greatly improves the prediction accuracy of converter gas, and provides technical support for gas balance scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments, but it should be understood that these drawings are designed only for explanation purposes and are not intended to limit the scope of the present invention. In addition, unless otherwise specified, these drawings are intended only to conceptually illustrate the structural configurations described herein and are not necessarily drawn to scale.

[0051] Figure 1 It is a flow chart of the method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data of the present invention;

[0052] Figure 2The prediction method of the present invention is a primary machine interface of a converter gas system in a steel plant;

[0053] Figure 3 It is a schematic diagram of data processing of a primary machine in the prediction method of the present invention;

[0054] Figure 4 The prediction result (training set) of the BP neural network method in Example 1 of the present invention;

[0055] Figure 5 The prediction results (test set) of the BP neural network method used in Example 1 of the present invention;

[0056] Figure 6 The gas flow prediction results of three furnaces combined with the production plan in Example 1 of the present invention;

[0057] Figure 7 The total gas recovery flow rate prediction and actual value comparison of the production plan in Example 1 of the present invention;

[0058] Figure 8 This is the data interface of the third-level machine of a steel plant in Example 2 of the present invention;

[0059] Fig. 9 The prediction result of the feasible gas recovery flow rate combined with the production plan in Example 2 of the present invention;

[0060] Fig.10 The comparison of the feasible recovery amount of converter gas under different recovery start and end concentrations of the prediction method in Example 3 of the present invention;

[0061] Fig.11 This is the prediction result of the feasible recovery flow rate of converter gas with different starting and ending recovery concentrations in Example 3 of the present invention. DETAILED DESCRIPTION

[0062] Exemplary embodiments of the present invention are described in detail below. Although these exemplary embodiments are described in sufficient detail to enable those skilled in the art to implement the present invention, it should be understood that other embodiments may be implemented and various changes may be made to the present invention without departing from the spirit and scope of the present invention. The more detailed description of the embodiments of the present invention below is not intended to limit the scope of the claimed invention, but is merely for illustration and does not limit the description of the features and characteristics of the present invention, in order to propose the best mode for performing the present invention, and is sufficient to enable those skilled in the art to implement the present invention. Therefore, the scope of the present invention is limited only by the appended claims.

[0063] The method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data proposed in the present invention has the following specific implementation steps: Figure 1 As shown:

[0064] Step 1: Establish a converter gas energy-quality balance model through converter gas generation mechanism analysis to obtain the raw materials, products, operations and recovery influencing factors of the feasible converter gas recovery amount.

[0065] Step 2: Based on the above factors, the corresponding primary and secondary machine data are collected respectively, and the primary machine data is preprocessed into a data set based on furnaces;

[0066] Step 3: According to the blowing time in the primary and secondary data, the primary and secondary data are integrated to form a prediction data set for the feasible recovery amount of converter gas;

[0067] Step 4: Using the neural network prediction method, taking the above-mentioned influencing factors as the input layer and the feasible recovery amount of coal gas as the output layer, and using the above-mentioned prediction data set, establish a prediction model for the feasible recovery amount of coal gas;

[0068] Step 5: Collect signals such as the type of steel being smelted in the converter, the start time of blowing, and the end time of blowing from the third-level machine data of the steelmaking production plan, determine the type of converter production process, and use the prediction model to accurately predict the feasible recovery amount of converter gas based on the operating factors corresponding to this type of converter production.

[0069] Example 1

[0070] Taking the converter production situation of a steel plant A as an example, the method for predicting the feasible recovery amount of converter gas of the present invention is described in detail.

[0071] Step 1: By analyzing the mechanism of converter gas generation, the converter gas generation mainly includes the furnace gas generated by oxidation of C in molten iron during the converter blowing stage and the air sucked in from the gap between the furnace mouth and the smoke hood. From the perspective of converter heat balance, the analysis shows that the factors affecting the converter gas generation mainly include: molten iron charge, scrap steel charge, total feed weight, molten iron temperature, molten iron C content, molten iron Si content, molten iron Mn content, molten iron P content, final steel temperature, molten steel C content, molten steel Si content, molten steel Mn content, molten steel P content, etc.

[0072] The amount of air sucked in from the gap between the furnace mouth and the smoke hood is mainly affected by the dust removal system and the furnace pressure at the furnace mouth. Taking the OG dust removal system as an example, the present invention analyzes the pressure balance of the OG system and finds that the factors affecting the air volume mainly include: the height of the active smoke hood, the circumference of the smoke hood, the annular gap opening, the fan speed, the furnace gas production, etc.

[0073] The feasible recovery of coal gas is not only affected by the coal gas production, but also by the coal gas recovery conditions. Through the analysis of the CO concentration and O2 change characteristics of coal gas at different stages of blowing, it can be seen that the change of CO concentration in the blowing stage is mainly related to the C, Si, Mn, P content in the molten iron, the molten iron temperature, the oxygen blowing intensity, and the oxygen lance height.

[0074] Comprehensive analysis shows that the feasible recovery amount of converter gas is mainly affected by factors such as molten iron charge, scrap steel charge, total charge weight, molten iron temperature, molten iron C content, molten iron Si content, molten iron Mn content, molten iron P content, terminal steel temperature, molten steel C content, molten steel Si content, molten steel Mn content, molten steel P content, oxygen blowing intensity, oxygen lance height, movable smoke hood height, smoke hood circumference, annular seam opening, and fan speed.

[0075] Step 2: The molten iron charge, scrap steel charge, total feed weight, molten iron temperature, molten iron C content, molten iron Si content, molten iron Mn content, molten iron P content in the above analysis are raw material factors; the final steel temperature, molten steel C content, molten steel Si content, molten steel Mn content, molten steel P content are product factors. These influencing factors can be collected from the secondary machine database, as shown in Table 1. The data set is in units of furnaces. Oxygen blowing intensity, oxygen lance height, active smoke hood height, smoke hood circumference, annular seam opening, fan speed, etc. are operating factors (see Figure 2 ), where the hood perimeter is a fixed value for the site, and the rest of the data changes with time and needs to be collected from the primary database, as shown in Table 2. This data set is a time series data set.

[0076] Table 1 Data set of the secondary converter gas machine collected

[0077]

[0078] Table 2 Collected converter gas primary machine data set

[0079]

[0080] Comparing the data in Table 1 and Table 2, we can see that the data structures are different and the primary machine data needs to be preprocessed. Figure 3 As shown in the figure, the variation of O2 flow rate, gas flow rate, CO concentration and O2 concentration during converter blowing is shown, where the vertical axis represents the values ​​of the above four variables, and the units of the variables are km 3 / h, km 3 / h, %, and %. Figure 3 It can be seen that the above operating factor data collected from the first-stage machine are intercepted into a furnace sequence according to the O2 flow rate (see Figure 3 The vertical dotted line O2>0 part), and then cut the feasible recovery period according to the CO concentration and O2 concentration (see Figure 3 The intersection of the horizontal dotted line and the 30% CO concentration line), the flue gas flow rate and oxygen flow rate are weighted summed, and the hood height, annular seam opening, oxygen lance height, and fan speed take the average value of this interval.

[0081] Step 3: According to the blowing start time, the secondary machine data set and the processed primary machine data set are coupled to form a converter gas feasible recovery amount prediction data set, as shown in Table 3.

[0082] Table 3 Feasible recycling volume prediction data set after data fusion

[0083]

[0084] Step 4: Use the 4th to n-1th columns of the above-established data set (a total of n columns) as the model input, and the nth column as the model output, and use the LM-BP neural network method to establish a feasible recovery model for converter gas. Now take the one-month operation data of a converter in a steel plant as an example. After data preprocessing, there are a total of 531 heats. After randomly shuffling the data, it is divided into 2 groups, of which 425 groups of data are training sets for model training, and the remaining 106 groups of data are verification test samples for model verification testing. The prediction results of the model training set and the prediction set are shown in Figure 4 and Figure 5 As shown in the figure, the model error MAPE of the training set and the prediction set are 1.80% and 1.76% respectively, and the model accuracy is 98.2% and 98.24% respectively, which shows high model accuracy.

[0085] Step 5: Collect signals such as the steelmaking blowing plan number, steel type, blowing start time, blowing end time, and molten iron charge in the steel plant's tertiary machine production plan, determine the feed amount, scrap steel charge amount, hood height, annular seam opening and other data based on the steel type to form the model input, and bring them into the above-trained prediction model to predict the feasible recovery amount of the converter, and calculate the converter gas flow rate during the recoverable period in combination with the blowing period.

[0086] Table 4 is the blowing plan of the converter during the period of 2:35-3:10 on a certain day of the steel plant. It is put into the converter gas recovery prediction model to obtain the feasible gas recovery amount of the three furnaces. The comparison between the prediction results and the actual values ​​is shown in Table 5.

[0087] Table 4 Converter Steelmaking Production Schedule

[0088] First furnace Second furnace No. 3 furnace Opening time 2:35 2:41 2:57 Steel tapping time 2:54 3:05 3:12

[0089] Table 5 Comparison of predicted and actual values ​​of feasible recovered gas volume of 3 furnaces

[0090] First furnace Second furnace No. 3 furnace <![CDATA[Predicted value (m 3 )]]> 49449.5 40685.8 43445.5 <![CDATA[Actual value (m 3 )]]> 49981.5 40360.8 43514.9 Difference -532 325 -69.4 error% -1.06％ 0.81％ -0.16％

[0091] The predicted feasible recovery results of each furnace are averaged to the feasible recovery period, and the converter gas flow rate during the feasible recovery period can be obtained. Figure 6 It can be seen that the flow rate of a single furnace during the feasible recovery period is 220-270km 3 / h fluctuation. The actual value of the feasible recovery volume of the first furnace fluctuates around the predicted value, but the difference is not large; the feasible recovery periods of the second and third furnaces are slightly different, and the flow fluctuations within this period are not large. The actual flow rate is basically consistent with the predicted flow rate line, indicating that the method for the feasible recovery volume of converter gas proposed in the present invention has high accuracy.

[0092] Figure 7 The comparison results of the predicted value and actual value of gas recovery flow rate during the production planning period are given. It can be seen from the figure that during this production planning period, there is a period of simultaneous blowing of two furnaces, and the gas flow rate during this period is about 500km 3 / h. During the prediction period, the actual value of the gas flow rate is very close to the predicted value and they are basically consistent, which proves that the prediction method has high accuracy. The predicted converter gas recovery flow rate can effectively guide the converter gas and the whole process gas control.

[0093] Example 2

[0094] In the feasible recovery amount prediction method of this embodiment, different neural network prediction methods can be used. In the embodiment 1, the traditional BP neural network prediction method is not used. In this embodiment, the RBF radial basis network is used. Taking the converter production of a steel plant B as an example, the feasible recovery amount prediction method of the converter gas of the present invention is described in detail.

[0095] Step 1: Same as Example 1, through the analysis of the converter gas generation mechanism and the pressure balance analysis of the converter dust removal system, the influencing factors of the converter gas generation amount are obtained: 1) raw material factors - molten iron charge, scrap steel charge, total weight of feed, molten iron temperature, molten iron C content, molten iron Si content, molten iron Mn content, molten iron P content; 2) product factors - end point steel temperature, molten steel C content, molten steel Si content, molten steel Mn content, molten steel P content; 3) operation factors - movable smoke hood height, smoke hood circumference, annular seam opening, fan speed.

[0096] The feasible recovery amount of converter gas is also restricted by the gas recovery conditions: that is, the CO and O2 recovery concentration limits at the start and end of the blowing stage. The CO concentration change characteristics during the blowing process are mainly affected by operating factors - oxygen blowing intensity and oxygen lance height.

[0097] From the above analysis, it can be seen that the feasible recovery amount of converter gas is mainly affected by the above-mentioned raw material factors, product factors, operating factors and recovery conditions.

[0098] Step 2: As in Example 1, the above-mentioned operating factors and timing factor data such as CO concentration, O2 concentration and gas flow rate are collected from the on-site primary machine, and the data are converted according to the method of Example 1 to form a primary machine data set in units of furnaces.

[0099] Step 3: According to the start time of blowing, the pre-treated secondary machine data set and the converted primary machine data set are coupled to form a feasible converter gas recovery amount prediction data set for model training and model verification.

[0100] Step 4: Taking the data of the converter from August to September 2024 as an example, the data set sorted out by the above method is a 1062×19 matrix. After the above data are randomly shuffled, they are divided into training sets and test validation sets, which are 850 groups and 212 groups respectively. The model adopts a radial basis function (RBF) neural network, in which the first 18 columns of the data set are the above-mentioned influencing factor data, which are used as the input layer neurons of the RBF neural network, the hidden layer is 1 layer, and its neuron kernel function is a Gaussian function. The input information is spatially mapped and transformed. The 19th column is the feasible recovery of the converter gas as the output layer of the model, and the error function uses the mean absolute percentage error MAPE. After calculation, the MAPE of the training set and the prediction set of the feasible recovery of the converter are 1.53% and 1.90% respectively, and the model accuracy is 98.47% and 98.1% respectively, which is very high.

[0101] Examples 1 and 2 illustrate that the method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data proposed by the present invention can achieve higher model accuracy by using different neural network prediction models.

[0102] Step 5: Read the plan number, steel grade, converter start time, converter end time, planned steel output and other signals in the production plan of the converter three-stage machine, and establish a model mean library containing data such as feed amount, scrap steel loading amount, smoke hood height, annular seam opening, etc. according to the previous historical data. Read the parameters in the corresponding model mean library according to the planned steel grade, as well as the molten iron volume, molten iron temperature and molten iron composition treated by KR as model input, and bring them into the above prediction model to predict the feasible recovery amount of the converter of the corresponding plan number, and calculate the converter gas flow rate during the recoverable period in combination with the blowing period.

[0103] Figure 8 This is the steelmaking plan for a group of three-stage machines in the steel plant. The blowing start time of this furnace is 08:49, and the converter steelmaking end time is 09:18. According to the three-stage machine steelmaking steel type (see Figure 8 The average value library is called (in the box), and the molten iron volume, molten iron temperature, and molten iron C, Si, Mn, and P contents are read from the KR primary machine data. The prediction model is used to predict that the feasible recovery amount of converter gas of the furnace is 44324.7m 3 , averaging it to the blowing recovery period (08:50-09:01), the gas flow rate recovered in this section is 221.6km 3 / h, for comparison with the actual results, see Fig. 9 As shown. Fig. 9It can be seen that the prediction accuracy of gas flow in the feasible recovery period is very high, which can effectively guide on-site gas control and gas dispatch.

[0104] Example 3

[0105] The method for predicting the feasible recovery amount of converter gas in this embodiment can be used to predict the feasible recovery amount of gas under different recovery conditions under the same raw materials, products and operating conditions. Taking the converter production of a steel plant C as an example, the method for predicting the feasible recovery amount of converter gas of the present invention is described in detail.

[0106] Step 1: As in Examples 1 and 2, the raw material factors, product factors and operation factors of the converter gas generation amount are obtained through the analysis of the converter gas generation mechanism and the pressure balance analysis of the converter dust removal system, totaling 16.

[0107] In addition, the feasible recovery amount of converter gas is also restricted by the gas recovery conditions: that is, the starting and ending recovery concentration limits of CO and O2 in the blowing stage. Enterprises can adjust the recovery CO concentration at the beginning and end of the blowing stage according to their own needs to adjust the recovered converter gas volume and the recovered gas calorific value.

[0108] Step 2: As in Example 2, the above 13 raw material and product factors are screened out from the secondary machine database, and the above 3 operating factors as well as oxygen flow rate, oxygen lance height, CO concentration, and O2 concentration are screened out from the primary machine database. According to the same method, the primary machine database is converted to convert the time series data into data per furnace. Among them, the feasible recovery amount of converter gas can be obtained by intercepting according to different start and end recovery concentration limits.

[0109] Step 3: Taking the converter C of a steel plant as an example, the data of June 2024 are selected, and the data of the first-level machine is processed according to the above method and merged with the data of the second-level machine according to the start time of blowing to form a prediction data set of the feasible recovery amount of converter gas, totaling 500 data. The feasible recovery amount of converter gas is intercepted by different CO concentration limits, which are 25%, 30%, 35% and 40% respectively. Fig.10 The comparison results of the feasible recovery amount of converter gas under 20 groups of different CO concentration limits are given. It can be seen that under the same recovery conditions, the feasible recovery amount of each furnace is different. 3 To 55000m 3 Fluctuation. For the same furnace, when the CO concentration starts and ends at the recovery concentration from 25% to 40%, the reduction in the feasible recovery amount of converter gas is also different. Therefore, if you need to analyze the impact of different recovery conditions on the feasible recovery amount of converter gas, you need to predict them separately.

[0110] Step 4: After randomly shuffling the data set sorted out in step 3, divide it into a training set and a test validation set, which are 400 groups and 100 groups respectively. The model uses a radial basis function (RBF) neural network, and the error function uses the mean absolute percentage error MAPE. After calculation, when the starting and ending recovery concentrations are 25%, 30%, 35% and 40%, the MAPE of the model test set is 2.49%, 2.59%, 3.5% and 4.17% respectively, which are very accurate, proving the accuracy of the model.

[0111] Step 5: Read the blowing plan of the converter in the third-stage machine. This time, the furnace with the blowing start time of 14:44 and the blowing end time of 15:00 is selected. The same as Example 2, after calling the steel grade mean value library and the molten iron related parameters of the KR first-stage machine according to the steel grade, use the above-established prediction model with CO start and end concentrations of 25% and 35% to predict the feasible recovery flow rate of converter gas under different start and end recovery concentrations. The prediction results are as follows: Fig.11 As shown. Fig.11 It can be seen that the feasible recovery periods of different CO starting and ending recovery concentrations in the same furnace are different, resulting in different feasible recovery volumes. Using the prediction method of the present invention, the feasible recovery flow rates predicted by different CO starting and ending recovery concentrations are highly accurate, which can effectively guide the site to adjust the recovery gas volume and gas calorific value by adjusting the recovery conditions.

[0112] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications can be made without departing from the technical principle of the present invention. These improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data, characterized in that: The following steps are involved: Through the analysis of converter gas generation mechanism, the converter gas energy-quality balance model is established to obtain the raw materials, products, operation and recovery influencing factors of the feasible recovery amount of converter gas; Collect the corresponding primary and secondary machine data respectively according to the influencing factors, and preprocess the primary machine data into a data set with furnace as the unit; According to the blowing time in the primary and secondary data, the primary and secondary data are integrated to form a feasible converter gas recovery amount prediction data set; By using the neural network prediction method, taking the influencing factors as the input layer and the feasible recovery volume of gas as the output layer, and using the prediction data set, a prediction model for the feasible recovery volume of gas is established; Collect the converter steel type, blowing start time, and blowing end time signals from the third-level machine data of the steelmaking production plan to determine the converter production process type. Based on the operating factors corresponding to this type of converter production, use the prediction model to achieve accurate prediction of the feasible recovery amount of converter gas.

2. The method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data according to claim 1 is characterized in that: The specific steps are: The first step is to analyze the influencing factors By analyzing the mechanism of converter gas generation, the converter gas mass conservation model and converter heat balance model are established, and the main influencing factors of converter gas generation volume V are analyzed. Step 2: Multi-source data collection and preprocessing According to the analysis of the main influencing factors in the first step, the raw material factors and product factors that affect the feasible recovery of converter gas are based on the furnace, and the field data need to be collected from the secondary machine, and the operating factors and recovery condition factors are time series data. The field data is collected in the primary machine; The blowing period is intercepted according to the primary data O2 flow rate, and then the relevant operating factors and gas flow data of the feasible recovery period are intercepted according to the CO and O2 concentration recovery limits to form the primary machine training data set; The third step is to fuse the data to form a prediction data set for the feasible recovery of coal gas. According to the blowing start time and blowing end time of the secondary data, the primary machine data set of the corresponding period is intercepted, and then the gas flow rate of the period is summed up. The remaining operating factors such as hood height, fan speed, and annular seam opening are processed by average value, and the processing results are merged with the corresponding secondary machine data set to form a gas feasible recovery amount prediction data set; Step 4: Establish a prediction model for feasible gas recovery The factors influencing the feasible recovery of gas analyzed in the third step are used as the input layer of the neural network prediction model, and the feasible recovery of gas is used as the output layer. The neural network structure is used to train the model using the feasible recovery of gas prediction data set established above, and a feasible recovery of gas prediction model is established. Step 5: Combine production plan to achieve accurate prediction of feasible gas recovery volume The start and end time of blowing and the type of steel to be smelted in the converter steelmaking production plan are read, and combined with the operational experience data of smelting this type of steel, a prediction model input data set is constructed. The feasible gas recovery amount prediction model established in the fourth step is used to predict the feasible gas recovery amount of this furnace, and then the gas flow rate is calculated according to the blowing period.

3. The method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data according to claim 2 is characterized in that: Converter gas recovery volume V gas,rec It is mainly affected by the converter flue gas generation, converter gas recovery conditions and gas cabinet position. The converter flue gas generation includes the theoretical gas generation V Fgas And the cold air volume V air N2 in Furnace gas generation V Fgas It is mainly produced by the oxidation reaction between C in the molten iron and the blowing O2 to produce CO and CO2. Based on the heat balance of the converter, it can be known that: where Q iron = m iron × [c ironS × (t ironfreeze - t envir ) + Q ironMelt + c ironL × (t iron - t ironfreeze )] Q oxi =29177m Si +6593m Mn +11637m C(co) +34824m C(co2) +35874m P +1620m SiO2 +4249m Fe(FeO) +6459m Fe(Fe2O3) Q steel =m steel ×[c steelS ×(t steelfreeze -t envir )+Q steelMelt +c steelL ×(t steel -t steelfreeze )] Q slag =m slag ×[c slagS ×(t slag -t envir )+Q slagMelt ], Cold air volume V at furnace mouth air It is mainly affected by the furnace mouth pressure and dust removal system control. Based on the pressure balance of the OG system, it can be known that: Where P0 = P2-(ΔP Vfp +ΔP twr +ΔP vntr +ΔP rec -P fan ) ΔP vntr =S vntr (0.78V air +V Fgas ) 2 4. The method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data according to claim 2 is characterized in that: The factors affecting the feasible recovery amount of coal gas include but are not limited to the weight of molten iron, the weight of scrap steel, the C content of molten iron, the C content of molten steel, the total weight of feed materials, the height of the smoke hood, the annular gap opening and the fan speed.

5. The method for predicting the feasible recovery amount of converter gas based on multi-source heterogeneous data according to claim 2 is characterized in that: The neural network structures include but are not limited to LM-BP, RBF and CNN.

Images