Blast furnace burden structure optimization method based on intelligent prediction mixed regulation and control

Abstract

The invention discloses a blast furnace burden structure optimization method based on intelligent prediction mixed regulation and control, and belongs to the technical field of blast furnace smelting. The blast furnace burden with alkalinity of 1.05-1.25 and pellet proportion of 100% is prepared by taking high-grade ultralow-titanium vanadium titano-magnetite fine powder as a raw material through material mixing, green pellet preparation, roasting, cooling, intelligent burden prediction and dynamic regulation and control. According to the method, an intelligent prediction model and closed-loop feedback control are combined for the first time, molten drop performance fluctuation is reduced through accurate regulation, the coke ratio is reduced by 3% or above, carbon emission is further reduced, the advantages that the pellet proportion is 100% and the molten drop interval is shortened by 30% or above in the prior art are reserved, and meanwhile the operation stability and adaptability of the blast furnace are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of blast furnace smelting, and relates to an optimization method for the burden structure of a blast furnace based on intelligent prediction and hybrid regulation. Specifically, it relates to an optimization method for the burden structure of a blast furnace using high-grade and ultra-low-titanium vanadium-titanium magnetite, and realizes energy conservation and emission reduction in the total process on the premise of obtaining the best burden melting and dripping performance. Background Art

[0002] At present, the blast furnace method is the most commonly used and technically mature method for processing vanadium-titanium magnetite resources in China. In this method, vanadium-titanium magnetite is first agglomerated, and then the valuable elements in the ore are selectively separated through blast furnace smelting. In the traditional blast furnace smelting process of vanadium-titanium magnetite, a burden structure with a high proportion of sinter and a small amount of pellet and lump ore is adopted. However, a large amount of greenhouse gases and air pollutants are generated in the production process of sinter. According to the relevant data analysis of actual production and "Best Available Techniques for Pollution Prevention and Control in the Iron and Steel Industry" at present, the energy consumption of the grate-kiln pelletizing process is only 50% of that of the sintering process, and the advanced value is only 1 / 3 of that of the sintering process.

[0003] To solve the above problems, the prior art has proposed new technologies for optimizing and improving blast furnace control to increase the proportion of pellets charged into the furnace. For example, Chinese Patent CN114959258A discloses a smelting method for a blast furnace with a high proportion of pellet ore, which realizes the smooth operation of the blast furnace with a high proportion of pellets by controlling burden distribution and blowing; Chinese Patent CN117701797A discloses a discharge self-feeding control method based on a burden structure with a high proportion of pellet ore, which realizes the precise distribution of pellet ore on the cross-section of the blast furnace. Although the existing patent technologies (such as CN114959258A, CN117701797A) have increased the proportion of pellets through burden distribution control or discharge optimization, they do not involve intelligent prediction and dynamic regulation based on real-time data of melting and dripping performance, resulting in insufficient adaptability of the burden structure.

[0004] In summary, on the one hand, the existing technologies still have the problems of high energy consumption and large pollution emissions due to the reliance on sinter; on the other hand, the adjustment of the burden structure to increase the proportion of pellet ore mostly depends on static proportioning or manual experience, lacking the ability to respond in real time to the dynamic changes during blast furnace smelting. Therefore, it is imperative to further increase the proportion of pellet ore charged into the furnace and develop an optimization method for the burden structure of a blast furnace based on intelligent prediction and hybrid regulation. Summary of the Invention

[0005] In view of the above problems, the present invention provides a method for realizing the smelting of a burden structure with a high proportion of pellets of high-grade and ultra-low-titanium vanadium-titanium magnetite. Based on intelligent prediction and hybrid regulation, the component ratio of pellet ore is improved, which not only greatly increases the use proportion of high-grade and ultra-low-titanium vanadium-titanium magnetite pellets in the burden structure, but also improves the melting and dripping performance of the burden, and realizes that the proportion of pellets in the burden reaches 100%.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] An optimization method for the burden structure of a blast furnace based on intelligent prediction and hybrid regulation, comprising the following steps:

[0008] Step 1. Mix high-grade ultra-low titanium type vanadium-titanium magnetite iron concentrate, basic flux, and bentonite to prepare an acidic iron concentrate mixture and a basic iron concentrate mixture respectively;

[0009] Step 2. Pelletize the acidic iron concentrate mixture and the basic iron concentrate mixture to obtain acidic green pellets and basic green pellets respectively;

[0010] Step 3. After drying the acidic green pellets and the basic green pellets, perform roasting to obtain acidic oxidized pellets and basic oxidized pellets;

[0011] Step 4. Cool the acidic oxidized pellets and the basic oxidized pellets to obtain the cooled acidic oxidized pellets and basic oxidized pellets;

[0012] Step 5. Mix the cooled acidic oxidized pellets and basic oxidized pellets through intelligent prediction and dynamic adjustment of the burden to obtain a dynamically optimized mixed burden, i.e., the blast furnace burden;

[0013] The pellet ratio of the blast furnace burden is 100%.

[0014] The intelligent prediction and dynamic regulation process of the burden includes:

[0015] a. Data collection and preprocessing: Real-time collect the data of the droplet temperature, pressure, and slag fluidity in the blast furnace through a sensor network, and transmit it to the central processor for noise filtering and feature extraction;

[0016] b. Droplet performance prediction: Input the processed data into a pre-trained LSTM model to predict the change of the droplet interval in the next 10 minutes under the current burden ratio;

[0017] c. Dynamic ratio adjustment: If the predicted droplet interval exceeds the preset threshold of ±5°C, the model generates a ratio adjustment signal to control the mixing ratio of the basic oxidized pellet ore and the acidic oxidized pellet ore, and adjusts the conveying amount of the charging materials in real time through the distributed control DSC system, and adjusts to the final burden basicity as the target value as needed;

[0018] d. Closed-loop verification and iterative optimization: Re-collect the droplet data after adjustment, compare it with the prediction result, update the model parameters to improve the prediction accuracy, form a closed-loop process of "monitoring - prediction - regulation - verification", and finally regulate to obtain the dynamically optimized mixed pellets.

[0019] In Step 1, the flux is one or several compositions of quicklime, limestone, and dolomite;

[0020] The composition of the acidic iron ore concentrate mixture, by mass fraction, includes: TFe: 58% - 64%, FeO: 18% - 25%, SiO2: 4% - 8%, CaO: 1% - 9%, MgO: 2% - 3%, Al2O3: 1% - 1.5%, V2O5: 0.5% - 1.5%, TiO2: 1% - 4%, Cr2O3: 0 - 0.5%;

[0021] The composition of the basic iron ore concentrate mixture, by mass fraction, includes: TFe: 55% - 58%, FeO: 21% - 24%, SiO2: 5% - 8%, CaO: 6% - 10%, MgO: 2% - 3%, Al2O3: 0.8% - 1.8%, V2O5: 0.5% - 1.0%, TiO2: 2% - 4%, Cr2O3: 0.18% - 0.48%.

[0022] In Step 2, the pellet size of the green pellets is 6 mm - 40 mm.

[0023] In Step 3, the roasting temperature is regulated to achieve a compressive strength of more than 2000 N / piece and a reduction expansion rate of less than 20% for acidic oxidized pellets and basic oxidized pellets;

[0024] The basicity of the acidic oxidized pellets is less than 0.5, and the basicity of the basic oxidized pellets is 1.15 - 1.4.

[0025] In Step 4, the cooling method is natural cooling.

[0026] In Step 5, the basicity of the blast furnace burden is 1.05 - 1.25.

[0027] The present invention provides an optimization method for the blast furnace burden structure based on intelligent prediction and hybrid regulation, achieving technological breakthroughs through the following innovative means:

[0028] Real-time monitoring and data acquisition of the melting drop performance: Deploy multi-modal sensors (including temperature sensors, pressure sensors, and optical imaging devices) at key positions in the blast furnace to collect data on the melting drop temperature range, melting drop speed, and slag-iron separation state in real time.

[0029] Construction of an intelligent prediction model: Use machine learning algorithms (such as LSTM neural networks) to train historical melting drop data and burden ratios, establish a melting drop performance prediction model, and dynamically predict the changing trend of the melting drop range under different ratios.

[0030] Closed-loop feedback control mechanism: Input real-time monitoring data into the prediction model to generate adjustment instructions for the mixing ratio of basic pellet ore and acidic pellet ore, and realize dynamic ratio optimization through the automated batching system to ensure that the melting drop interval is stable within the target range (reduced by more than 30% compared with the traditional structure).

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0032] For the first time, the intelligent prediction model is combined with closed-loop feedback control to realize the dynamic optimization of the burden structure, which is significantly different from the existing static ratio technology.

[0033] Reduce the fluctuation of melting drop performance through precise control, reduce the coke ratio by more than 3%, and further reduce carbon emissions.

[0034] Retain the advantages of 100% pellet ratio and more than 30% reduction in melting drop interval in the original technology, and at the same time improve the stability and adaptability of blast furnace operation.

[0035] Increase the pellet ratio of the blast furnace burden. The proportion of pellet ore in this burden structure is 100%. Compared with the traditional technology, it reduces the energy consumption and pollution emissions of the agglomeration process. Because the grade of pellet ore is much higher than that of sintered ore, it further increases the blast furnace utilization coefficient and reduces the slag ratio.

[0036] The solution provided by the present invention pays attention to the melting drop performance of the burden structure. The optimized burden structure has a more sufficient indirect reduction effect, higher gas utilization rate, better permeability and other indicators during the smelting process. In the best case, the melting drop interval is reduced by more than 30% compared with the traditional burden structure. Detailed implementation manners

[0037] The present invention provides an optimization method for the blast furnace burden structure based on intelligent prediction and hybrid control, including the following steps:

[0038] 1. Mix high-grade ultra-low titanium vanadium-titanium magnetite iron concentrate with basic fluxes and bentonite to prepare acidic iron concentrate mixture and basic iron concentrate mixture respectively. The composition of the acidic iron concentrate mixture by mass fraction includes: TFe: 58% - 64%, FeO: 18% - 25%, SiO2: 4% - 8%, CaO: 1% - 9%, MgO: 2% - 3%, Al2O3: 1% - 1.5%, V2O5: 0.5% - 1.5%, TiO2: 1% - 4%, Cr2O3: 0 - 0.5%; The composition of the basic iron concentrate mixture by mass fraction includes: TFe: 55% - 58%, FeO: 21% - 24%, SiO2: 5% - 8%, CaO: 6% - 10%, MgO: 2% - 3%, Al2O3: 0.8% - 1.8%, V2O5: 0.5% - 1.0%, TiO2: 2% - 4%, Cr2O3: 0.18% - 0.48%. The fluxes are one or several combinations of quicklime, limestone, and dolomite;

[0039] 2. Use the acidic iron concentrate mixture and the basic iron concentrate mixture to make pellets to obtain acidic green pellets and basic green pellets; The pellet sizes of the acidic green pellets and the basic green pellets are both 6mm - 40mm;

[0040] 3. Dry the green pellets and roast the green pellets in a high-temperature muffle furnace in an air atmosphere to obtain finished acidic oxidized pellets and basic oxidized pellets. The roasting temperature is adjusted according to the average compressive strength of the finished acidic oxidized pellets and basic oxidized pellets being above 2000N / piece, and the expansion rate is less than 20%; The basicity of the obtained acidic oxidized pellets is less than 0.5, and the basicity of the basic oxidized pellets is 1.15 - 1.4;

[0041] 4. Take out the acidic oxidized pellets and basic oxidized pellets from the furnace and cool them naturally;

[0042] 5. Mix the cooled acidic oxidized pellets and basic oxidized pellets through intelligent prediction and dynamic control of furnace charge:

[0043] a. Data collection and preprocessing: Real-time collect the data of the droplet temperature, pressure, and slag fluidity in the blast furnace through a sensor network, and transmit it to the central processor for noise filtering and feature extraction;

[0044] b. Prediction of droplet performance: Input the processed data into a pre-trained LSTM model to predict the change of the droplet interval in the next 10 minutes under the current furnace charge ratio;

[0045] c. Dynamic ratio adjustment: If the predicted droplet interval exceeds the preset threshold (±5°C), the model generates a ratio adjustment signal, controls the mixing ratio of basic pellet ore and acidic pellet ore, and adjusts the feeding amount of the furnace charge in real time through the distributed control DSC system, and adjusts to the target value of the final furnace charge basicity as needed;

[0046] d. Closed-loop verification and iterative optimization: After adjustment, remelt the droplet data, compare it with the prediction results, update the model parameters to improve the prediction accuracy, and form a closed-loop process of "monitoring - prediction - regulation - verification". Finally, obtain the dynamically optimized mixed burden, that is, the blast furnace burden with an alkalinity of 1.05 - 1.25 and a pellet ratio of 100%.

[0047] When the present invention monitors and collects the droplet performance in real time, multi-modal sensors (including temperature sensors, pressure sensors, and optical imaging devices) are deployed at key positions in the blast furnace to collect data on the droplet temperature range, droplet velocity, and slag-iron separation state in real time.

[0048] When the present invention constructs an intelligent prediction model, machine learning algorithms (such as LSTM neural networks) are used to train historical droplet data and burden ratios, establish a droplet performance prediction model, and dynamically predict the change trend of the droplet interval under different ratios.

[0049] When the present invention is in the closed-loop feedback control mechanism, the real-time monitoring data is input into the prediction model to generate an adjustment instruction for the mixing ratio of basic pellets and acidic pellets, and the dynamic ratio optimization is realized through an automatic batching system to ensure that the droplet interval is stably within the target range (reduced by more than 30% compared with the traditional structure).

[0050] The present invention will be further described below in conjunction with embodiments.

[0051] Embodiment 1

[0052] 1. Mix vanadium-titanium magnetite iron concentrate, basic flux, and bentonite, and prepare the following two kinds of mixed concentrates with different mass ratios by changing the addition amount of calcium oxide. The first is an acidic iron concentrate mixture: TFe: 58.83%, FeO: 24.59%, SiO2: 6.55%, CaO: 2.17%, MgO: 2.35%, Al2O3: 1.23%, V2O5: 0.78%, TiO2: 2.74%, Cr2O3: 0.36%. The second is a basic iron concentrate mixture: TFe: 55.67%, FeO: 23.27%, SiO2: 6.20%, CaO: 7.43%, MgO: 2.23%, Al2O3: 1.17%, V2O5: 0.74%, TiO2: 2.60%, Cr2O3: 0.34%.

[0053] 2. Use the acidic iron concentrate mixture and the basic iron concentrate mixture to make pellets. The particle sizes of the obtained acidic green pellets and basic pellets are both 6 mm - 40 mm;

[0054] 3. Dry the green pellets, and roast the green pellets in a high-temperature muffle furnace under an air atmosphere to obtain acid oxidized pellets with an alkalinity of 0.33 and basic oxidized pellets with an alkalinity of 1.2. The suitable temperature means that the average compressive strength of the finished oxidized pellets is above 2000 N / piece, and the reduction expansion rate is less than 20%;

[0055] 4. Take out the acid oxidized pellets and basic oxidized pellets from the furnace and let them cool naturally;

[0056] 5. Mix the cooled acid oxidized pellets and basic oxidized pellets through intelligent prediction and dynamic control of the furnace charge:

[0057] a. Data collection and preprocessing: Real-time collect the data of the droplet temperature, pressure and slag fluidity in the blast furnace through a sensor network, and transmit it to the central processor for noise filtering and feature extraction;

[0058] b. Prediction of droplet performance: Input the processed data into a pre-trained LSTM model to predict the change of the droplet interval in the next 10 minutes under the current furnace charge ratio;

[0059] c. Dynamic ratio adjustment: If the predicted droplet interval exceeds the preset threshold (±5°C), the model generates a ratio adjustment signal to control the mixing ratio of basic pellet ore and acid pellet ore, and adjusts the conveying amount of the furnace charge in real time through a distributed control DSC system;

[0060] d. Closed-loop verification and iterative optimization: Re-collect the droplet data after adjustment, compare it with the prediction results, update the model parameters to improve the prediction accuracy, form a closed-loop process of "monitoring - prediction - control - verification", and finally control to obtain the dynamically optimized mixed furnace charge, which is the blast furnace charge with a pellet ratio of 100% and an alkalinity of 1.1.

[0061] Determine the droplet performance of the furnace charge structure according to the national standard of the People's Republic of China GB / T34211-2017 "Determination Method for High Temperature Load Reduction Softening Melting and Droplet Performance of Iron Ore". The softening start temperature is 1090°C, the softening end temperature is 1174°C, the melting start temperature is 1221°C, the dripping temperature is 1348°C, the maximum pressure difference is 25.35 kPa, the softening interval is 84°C, the droplet interval is 127°C, and the total characteristic value is 1689.35 kPa·°C.

[0062] Example 2

[0063] 1. Mix the vanadium-titanium magnetite iron concentrate, basic flux, and bentonite to make the following two kinds of mixed concentrates with the following mass ratios. The first is the acidic iron concentrate mixture: TFe: 57.82%, FeO: 24.17%, SiO2: 6.43%, CaO: 3.86%, MgO: 2.28%, Al2O3: 1.21%, V2O5: 0.77%, TiO2: 2.70%, Cr2O3: 0.35%. The second is the basic iron concentrate mixture: TFe: 55.67%, FeO: 23.27%, SiO2: 6.20, CaO: 7.43%, MgO: 2.23%, Al2O3: 1.17%, V2O5: 0.74%, TiO2: 2.60%, Cr2O3: 0.34%.

[0064] 2. Use the acidic iron concentrate mixture and the basic iron concentrate mixture to make pellets. The sizes of the obtained acidic green pellets and basic pellets are both 6 mm to 40 mm;

[0065] 3. Dry the green pellets and roast the green pellets in a high-temperature muffle furnace under an air atmosphere to obtain acidic oxidized pellets with an alkalinity of 0.6 and basic oxidized pellets with an alkalinity of 1.2. The suitable temperature means that the average compressive strength of the finished oxidized pellets is above 2000 N / piece, and the reduction expansion rate is less than 20%;

[0066] 4. Take out the acidic oxidized pellets and basic oxidized pellets from the furnace and perform natural cooling;

[0067] 5. Mix the cooled acidic oxidized pellets and basic oxidized pellets through intelligent prediction and dynamic regulation of the furnace charge:

[0068] a. Data collection and preprocessing: Real-time collect the data of the droplet temperature, pressure, and slag fluidity in the blast furnace through a sensor network, and transmit it to the central processor for noise filtering and feature extraction;

[0069] b. Prediction of droplet performance: Input the processed data into a pre-trained LSTM model to predict the change of the droplet interval in the next 10 minutes under the current furnace charge ratio;

[0070] c. Dynamic ratio adjustment: If the predicted droplet interval exceeds the preset threshold (±5 °C), the model generates a ratio adjustment signal, controls the mixing ratio of basic pellet ore and acidic pellet ore, and adjusts the feeding amount of the furnace charge in real time through a distributed control DSC system;

[0071] d. Closed-loop verification and iterative optimization: Re-collect the droplet data after adjustment, compare it with the prediction results, update the model parameters to improve the prediction accuracy, form a closed-loop process of "monitoring - prediction - regulation - verification", and finally regulate to obtain a dynamically optimized mixed furnace charge, which is the blast furnace charge with a pellet ratio of 100% and an alkalinity of 1.1.

[0072] The softening and dripping properties of the burden structure were determined according to the national standard of the People's Republic of China GB / T 34211-2017 "Determination method for high-temperature burden reduction softening, melting and dripping properties of iron ore". The softening start temperature was 1080 °C, the softening end temperature was 1161 °C, the melting start temperature was 1203 °C, the dripping temperature was 1364 °C, the maximum pressure difference was 29.6 kPa, the softening interval was 81 °C, the dripping interval was 161 °C, and the total characteristic value was 2105.9 kPa·°C.

[0073] Example 3

[0074] 1. Mix the vanadium-titanium magnetite iron concentrate, basic flux, and bentonite to make the following two kinds of mixed concentrates with mass ratios. The first is the acidic iron concentrate mixture: TFe: 57.08%, FeO: 23.86%, SiO2: 6.35%, CaO: 5.08%, MgO: 2.28%, Al2O3: 1.20%, V2O5: 0.76%, TiO2: 2.66%, Cr2O3: 0.35%. The second is the basic iron concentrate mixture: TFe: 55.67%, FeO: 23.27%, SiO2: 6.20%, CaO: 7.43%, MgO: 2.23%, Al2O3: 1.17%, V2O5: 0.74%, TiO2: 2.60%, Cr2O3: 0.34%.

[0075] 2. Pelletize using the acidic iron concentrate mixture and the basic iron concentrate mixture. The particle sizes of the obtained acidic green pellets and basic pellets are both 6 mm to 40 mm;

[0076] 3. Dry the green pellets and roast them in a high-temperature muffle furnace in an air atmosphere to obtain acidic oxidized pellets with a basicity of 0.8 and basic oxidized pellets with a basicity of 1.2. The appropriate temperature means that the average compressive strength of the finished oxidized pellets is above 2000 N / piece and the reduction expansion rate is less than 20%;

[0077] 4. Take out the acidic oxidized pellets and basic oxidized pellets from the furnace and perform natural cooling;

[0078] 5. Mix the cooled acidic oxidized pellets and basic oxidized pellets through intelligent prediction and dynamic regulation of the burden:

[0079] a. Data collection and preprocessing: Real-time collect the dripping temperature, pressure, and slag fluidity data in the blast furnace through a sensor network and transmit them to the central processor for noise filtering and feature extraction;

[0080] b. Dripping property prediction: Input the processed data into the pre-trained LSTM model to predict the change of the dripping interval in the next 10 minutes under the current burden ratio;

[0081] c. Dynamic ratio adjustment: If the predicted melting-droplet interval exceeds the preset threshold (±5°C), the model generates a ratio adjustment signal to control the mixing ratio of basic pellets and acidic pellets, and the feeding amount of the materials charged into the furnace is adjusted in real time through the distributed control DSC system;

[0082] d. Closed-loop verification and iterative optimization: After adjustment, the melting-droplet data is collected again, compared with the predicted results, and the model parameters are updated to improve the prediction accuracy, forming a closed-loop process of "monitoring - prediction - regulation - verification". Finally, the dynamically optimized mixed burden is obtained, which is the blast furnace burden with a pellet ratio of 100% and an alkalinity of 1.1.

[0083] The melting-droplet performance of the burden structure was measured according to the national standard of the People's Republic of China GB / T34211-2017 "Determination method for high-temperature load reduction softening melting-droplet performance of iron ore". The softening start temperature was 1076°C, the softening end temperature was 1159°C, the melting start temperature was 1196°C, the dripping temperature was 1345°C, the maximum pressure difference was 28.4 kPa, the softening interval was 83°C, the melting-droplet interval was 149°C, and the total characteristic value was 2394.45 kPa·°C.

[0084] Comparative example

[0085] A traditional burden structure of acidic pellets plus basic sinter was adopted.

[0086] 1. Mix the vanadium-titanium magnetite iron concentrate, basic flux, and bentonite to make a mixed concentrate with the following mass ratio: TFe: 58.83%, FeO: 24.59%, SiO2: 6.55%, CaO: 2.17%, MgO: 2.35%, Al2O3: 1.23%, V2O5: 0.78%, TiO2: 2.74%, Cr2O3: 0.36%.

[0087] 2. Pelletize the above materials to obtain green pellets with a pellet size of 6 mm to 40 mm;

[0088] 3. Dry the green pellets and roast them in a high-temperature muffle furnace in an air atmosphere to obtain finished oxidized pellets with an alkalinity of 0.33. The appropriate temperature means that the average compressive strength of the finished oxidized pellets is above 2000 N / piece and the reduction expansion rate is less than 20%;

[0089] 4. Take out the roasted materials from the furnace and cool them naturally;

[0090] 5. Mix the above pelletized ore with sintered ore to obtain blast furnace burden with an alkalinity of 1.1. The chemical composition and weight percentage of the above sintered ore are as follows: TFe: 52.95, FeO: 19.82, SiO2: 5.11, CaO: 9.15, MgO: 2.7, Al2O3: 2.4, V2O5: 0.34, TiO2: 6.78, Cr2O3: 0.28.

[0091] Determine the melting and dripping properties of the burden structure according to the national standard of the People's Republic of China GB / T34211-2017 "Determination Method for High Temperature Load Reduction Softening Melting and Dripping Properties of Iron Ores". The softening start temperature is 1040 °C, the softening end temperature is 1137 °C, the melting start temperature is 1219 °C, the dripping temperature is 1408 °C, the maximum pressure difference is 25.8 kPa, the softening interval is 97 °C, the melting and dripping interval is 189 °C, and the total characteristic value is 1643.69 kPa·°C.

[0092] After adopting the intelligent control system of the present invention, the fluctuation range of the melting and dripping interval of the blast furnace in a steel plant is reduced from ±15 °C of the traditional process to ±5 °C, the molten iron output is increased by 1.5%, and the energy consumption per ton of iron is reduced by 4.2%, verifying the feasibility and advancement of the present solution.

Claims

1. A method for optimizing blast furnace charge structure based on intelligent prediction and mixed regulation, characterized in that: The following steps are involved: Step 1. Mixing high-grade ultra-low titanium vanadium-titanium magnetite iron ore concentrate with alkaline flux and bentonite to prepare an acidic iron ore concentrate mixture and an alkaline iron ore concentrate mixture respectively; Step 2. pelletizing the acidic iron concentrate mixture and the alkaline iron concentrate mixture to obtain acidic green balls and alkaline green balls respectively; Step 3. Drying the acidic green balls and the alkaline green balls and then calcining them to obtain acidic oxidized balls and alkaline oxidized balls; Step 4. Cooling the acidic oxidation pellets and the alkaline oxidation pellets to obtain cooled acidic oxidation pellets and alkaline oxidation pellets; Step 5. The cooled acidic oxidation pellets and alkaline oxidation pellets are mixed by intelligent prediction and dynamic adjustment of furnace charge to obtain a dynamically optimized mixed furnace charge, i.e., a blast furnace charge; The pellet ratio of the blast furnace charge is 100%.

2. The method for optimizing blast furnace charge structure based on intelligent prediction and mixed control according to claim 1, characterized in that: The process of intelligent prediction and dynamic control of furnace charge includes: a. Data collection and preprocessing: The sensor network collects the data of droplet temperature, pressure and slag fluidity in the blast furnace in real time and transmits it to the central processor for noise filtering and feature extraction; b. Prediction of droplet performance: Input the processed data into the pre-trained LSTM model to predict the droplet range change in the next 10 minutes under the current charge ratio; c. Dynamic ratio adjustment: If the predicted droplet range exceeds the preset threshold of ±5°C, the model generates a ratio adjustment signal to control the mixing ratio of alkaline oxidized pellets and acidic oxidized pellets, and adjusts the amount of material delivered to the furnace in real time through the distributed control DSC system, and adjusts the final charge basicity to the target value as needed; d. Closed-loop verification and iterative optimization: After adjustment, the droplet data is collected again, compared with the prediction results, and the model parameters are updated to improve the prediction accuracy, forming a closed-loop process of "monitoring-prediction-control-verification", and finally the mixed pellets are dynamically optimized.

3. The method for optimizing blast furnace charge structure based on intelligent prediction and mixed control according to claim 2, characterized in that: In step 1, the flux is one or a combination of quicklime, limestone, and dolomite; The components of the acidic iron ore concentrate mixture include, by mass fraction, TFe: 58% to 64%, FeO: 18% to 25%, SiO2: 4% to 8%, CaO: 1% to 9%, MgO: 2% to 3%, Al2O3: 1% to 1.5%, V2O5: 0.5% to 1.5%, TiO2: 1% to 4%, and Cr2O3: 0 to 0.5%; The components of the basic iron ore concentrate mixture include, by mass fraction, TFe: 55% to 58%, FeO: 21% to 24%, SiO2: 5% to 8%, CaO: 6% to 10%, MgO: 2% to 3%, Al2O3: 0.8% to 1.8%, V2O5: 0.5% to 1.0%, TiO2: 2% to 4%, and Cr2O3: 0.18% to 0.48%.

4. The method for optimizing blast furnace charge structure based on intelligent prediction and mixed control according to claim 2, characterized in that: In step 2, the pellet size of the raw balls is 6 mm to 40 mm.

5. The method for optimizing blast furnace charge structure based on intelligent prediction and mixed control according to claim 2, characterized in that: In step 3, the roasting temperature is controlled to achieve a compressive strength of the acidic oxidation pellets and the alkaline oxidation pellets of more than 2000N / piece and a reduction expansion rate of less than 20%; The alkalinity of acidic oxidation pellets is less than 0.5, and the alkalinity of alkaline oxidation pellets is 1.15-1.

4.

6. The method for optimizing blast furnace charge structure based on intelligent prediction and mixed control according to claim 2, characterized in that: In step 4, the cooling method is natural cooling.

7. The method for optimizing blast furnace charge structure based on intelligent prediction and mixed control according to claim 2, characterized in that: In step 5, the basicity of the blast furnace charge is 1.05 to 1.25.