CN-121978964-A - Ferrous metallurgy sintering flue gas regulation and control method and system based on space-time cooperative framework

CN121978964ACN 121978964 ACN121978964 ACN 121978964ACN-121978964-A

Abstract

The invention provides a method and a system for regulating and controlling ferrous metallurgy sintering flue gas based on space-time collaborative architecture, which relate to the technical field of ferrous metallurgy sintering computer aided design, and the method comprises the following steps of S1, arranging an array sensor, collecting multidimensional physical parameters in the sintering flue gas circulation process, and carrying out data processing; S2, establishing a space-time collaborative architecture model to realize the perception and dynamic prediction of a flue gas flow field, S3, constructing a multi-target collaborative optimization loss function to optimize the emission parameters of gas component concentration, the process energy consumption parameters of section flow velocity and static pressure, the sinter quality parameters of flue gas temperature and physical consistency constraint indexes, and S4, optimizing a dynamic control vector of sintering flue gas to control an actuating mechanism to realize the regulation and control of the sintering flue gas. According to the invention, the flue gas flow field of the sintering furnace is predicted by the space-time collaborative framework model, the uniformity of the flue gas flow field of the sintering furnace is optimized, the mobility of flue gas in the smoke hood and the sintering material layer is improved, and the running power consumption of the main exhaust fan is reduced.

Inventors

LI JIE
XUE TAO
LIU YUXIN
YANG AIMIN
WANG LIYA
LIU WEIXING

Assignees

华北理工大学

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (9)

1. A ferrous metallurgy sintering flue gas regulation and control method based on a space-time cooperative framework is characterized by comprising the following steps of: S1, establishing a distributed sensing network, arranging an array sensor at a flow field key node in an air inlet circulating smoke hood of a sintering machine, collecting a section flow velocity, a static pressure, a smoke temperature and a gas component concentration reflecting sintering working conditions as multidimensional physical parameters, and carrying out physical consistency reconstruction on the multidimensional physical parameters by utilizing edge computing equipment according to mass conservation and energy conservation law to generate flow field data meeting a hydrodynamic continuity equation; S2, acquiring flow field data output in the step S1, establishing a space-time collaborative architecture model, mapping the flow field data into a dynamic graph structure by using a graph injection force network, extracting a space flow field feature vector representing the distribution features of the flow speed and the static pressure space of a section, recursively deducing the time sequence evolution trend of the temperature of the flue gas and the concentration of the gas components by using a selective state space model, and outputting a time sequence dynamic feature vector; S3, inputting the space-time fusion feature vector of the flue gas flow field output in the step S2 into a prediction mapping network, outputting the prediction results of the section flow velocity, the static pressure, the flue gas temperature and the gas component concentration, further constructing a multi-objective collaborative optimization loss function, optimizing the emission parameters of the gas component concentration, the process energy consumption parameters of the section flow velocity and the static pressure, the sinter quality parameters of the flue gas temperature and the physical consistency constraint indexes; And S4, optimizing a dynamic control vector comprising the deflection angle of the guide vane and the opening of the regulating valve according to the prediction result output in the step S3, generating a regulating instruction, driving the actuating mechanism to change the section flow velocity and the static pressure distribution so as to optimize the uniformity of the flue gas temperature field, and realizing closed-loop regulation and control on the sintering flue gas circulation process.
2. The method for regulating and controlling ferrous metallurgy sintering flue gas based on space-time cooperative framework as claimed in claim 1, wherein the step S1 specifically comprises the following substeps: S11, establishing a distributed sensing network, adopting a distributed networking architecture, arranging an array sensor at a key node of a flow field in an air inlet circulating smoke hood on a sintering machine, and collecting multidimensional physical parameters reflecting sintering conditions, including section flow velocity, static pressure, smoke temperature and gas component concentration; S12, establishing a multi-stage progressive data pretreatment system by using edge computing equipment, sequentially processing the acquired multi-dimensional physical parameters, namely identifying and removing outliers of the multi-dimensional physical parameters acquired by the sensor, dynamically adjusting a filtering window according to signal fluctuation characteristics by adopting an average filtering method to filter high-frequency electromagnetic interference noise, mapping measurement data to a physical feasible domain, and outputting sintering parameter characteristics with physical consistency.
3. The method for regulating and controlling ferrous metallurgy sintering flue gas based on space-time cooperative framework according to claim 2, wherein the step S12 is characterized in that the sintering parameter with physical consistency is outputted specifically as follows: ; Wherein, the X is a sintering parameter data matrix; a characteristic projection matrix which is a sintering parameter; The potential feature matrix is subjected to dimension reduction; The weight coefficient is a constraint term; a fluid network node association matrix constructed for kirchhoff's law based on the fluid network; The target vector is the mass conservation residual error under the steady state of the system; And F is a matrix norm, so as to meet the physical consistency constraint penalty term of the law of hydrodynamic conservation.
4. The ferrous metallurgy sintering flue gas regulating and controlling method based on space-time cooperative framework according to claim 1, wherein the step S2 specifically comprises the following substeps: S21, establishing a graph attention network dynamic space model, receiving flow field data output in the step S1, mapping sensor nodes into graph attention network nodes, and constructing a dynamic graph structure representing a flue gas flow field topological relation; S22, receiving the space flow field feature vector output in the step S21, inputting a selective state space model, capturing thermal inertia and flow field hysteresis effect in the sintering process by using the recursion characteristic of a state space equation, and outputting dynamic time sequence evolution features capable of reflecting the flow field evolution trend; S23, constructing a self-adaptive gating mechanism capable of learning, automatically acquiring gating weights of space features and time sequence features according to the steady or transient state working condition of the current flue gas flow field, carrying out weighted fusion on the space velocity distribution features and the dynamic time sequence evolution features according to the gating weights, and outputting a flue gas flow field space-time state fusion feature vector 。
5. The ferrous metallurgy sintering flue gas regulating and controlling method based on space-time cooperative framework according to claim 1, wherein the step S3 specifically comprises the following substeps: S31, constructing a virtual sintering flow field simulation environment, generating a sintering flow field simulation data set covering various extreme working conditions by using fluid mechanics simulation CFD, and performing supervision pre-training on the space-time collaborative architecture model established in the step S2 to obtain a fluid dynamics physical evolution rule of the sintering flow field; s32, performing migration learning adjustment on the pre-trained space-time collaborative architecture model by utilizing historical operation data of a sintering site, constructing a multi-objective collaborative optimization loss function, and fitting nonlinear characteristics and noise distribution in an actual sintering process; S33, when the space-time collaborative architecture model is deployed and operated after being adjusted in the step S32, an online increment learning mechanism is started; And S34, adopting an optimization algorithm combining the dynamic term and the self-adaptive learning rate, and iteratively updating the weight parameters of the space-time collaborative architecture model according to a back propagation algorithm to realize self-adaptive tracking of the drift of the sintering working condition.
6. The method for controlling the sintering flue gas of ferrous metallurgy based on space-time cooperative framework according to claim 5, wherein the step S33 is specifically: Caching and triggering sintering parameter operation data, and storing a sintering condition flow data sample set According to the sintering condition flow data sample set Obtaining an incremental learning loss measurement value representing online prediction bias of a space-time collaborative architecture model The method for updating sintering parameters by adopting a small batch random gradient descent method comprises the following steps: ; ; Wherein, the Incremental learning loss metric values for characterizing online prediction bias of the space-time collaborative architecture model; Optimizing a total loss function for multi-objective synergy; predicting an error weight for carbon monoxide; mean square error for carbon monoxide emission peak capture; Predicting error weight for energy consumption; The average absolute error of the operation energy consumption prediction of the main exhaust fan is obtained; Predicting error weight for quality; the average absolute error of the sintering quality index prediction is obtained; Losing weight for physical constraint; θ is a training parameter for representing the mapping relation between sintering parameters and a flue gas flow field in a space-time collaborative architecture model; the weight parameters of the updated space-time collaborative architecture model are adapted to the current sintering working condition; the weight parameters of the space-time collaborative architecture model at the historical moment before updating are obtained; learning step length for online self-adaptive fine adjustment; is a gradient operator with respect to parameter θ; and the data sample set is a sintering working condition stream data sample set.
7. The method for regulating and controlling ferrous metallurgy sintering flue gas based on space-time cooperative framework as claimed in claim 1, wherein the step S4 is specifically: S41, setting a dynamic control vector of a sintering flue gas distributor, wherein the dynamic control vector comprises a guide vane real-time deflection angle and a flow regulating valve opening, and configuring preset curing physical structure parameters as basic geometric constraints; S42, constructing a closed-loop control model, inputting sintering parameter characteristics with real-time physical consistency into a space-time collaborative architecture model, and outputting a fine adjustment instruction aiming at a dynamic control vector of a sintering flue gas distributor to realize real-time control of working condition fluctuation in the sintering process.
8. The method for controlling the ferrous metallurgy sintering flue gas based on the space-time collaborative architecture according to claim 7, wherein the closed-loop control model in the step S42 is specifically: ; Wherein, the A weighted total cost function for multi-dimensional dynamic regulation and control of the sintering flue gas circulation system; the carbon monoxide emission reduction weight is as follows; is an emission reduction objective function of carbon monoxide; Is the energy consumption weight; Is an energy consumption objective function; is the weight of sintering quality; Is a sintering quality objective function; control vectors for the joint actuators; A physical consistency constraint lower limit for a physical action range of the executing mechanism; constraint upper limit for physical consistency of physical action range of the executing mechanism; Is constrained by the conditional symbol.
9. A sintering flue gas control system for implementing the ferrous metallurgy sintering flue gas control method based on space-time collaborative architecture according to any one of claims 1 to 8, characterized in that it comprises an array sensor module, an edge data processing module, a collaborative module and an execution module; The array sensor module adopts a distributed space gridding architecture and comprises a plurality of flow velocity sensors, pressure sensors and temperature sensors which are arranged in an air inlet circulating smoke hood of the sintering machine, wherein the sensors are arranged at grid nodes of an inlet of a main sintering flue, the periphery of the inner wall of the smoke hood and the space above a sintering material layer according to the topological structure of smoke flow, and a physical sensing network for capturing three-dimensional flow field original signals is constructed; The system comprises an array sensor module, an edge data processing module, a principal component analysis operation, a flow field data stream analysis module, a data analysis module and a data analysis module, wherein the array sensor module is used for receiving original multidimensional physical parameters acquired by the array sensor module, performing outlier rejection, self-adaptive filtering and principal component analysis operation of mass conservation physical consistency constraint, and reconstructing an original signal comprising noise and measurement deviation into the flow field data stream meeting a physical continuity equation; The system comprises a coordination module, a time sequence evolution prediction unit, a multi-objective optimization unit, a control instruction and a control module, wherein the coordination module is used for running a space-time coordination architecture model and comprises a space feature extraction unit, a time sequence evolution prediction unit and a multi-objective optimization unit; The execution module is a physical execution terminal of the system and comprises an electric deflector array and an electric flow regulating valve group which are arranged at the position of a key flow passage in the sintering smoke hood, and the execution module is in communication connection with the coordination module and is used for responding to a control instruction and realizing reconstruction of the flow direction of the sintering smoke and distribution of local flow by changing the physical section and the guide angle of the smoke flow in real time.

Description

Ferrous metallurgy sintering flue gas regulation and control method and system based on space-time cooperative framework Technical Field The invention relates to the technical field of ferrous metallurgy sintering computer aided design, in particular to a ferrous metallurgy sintering flue gas regulation and control method and system based on a space-time collaborative framework. Background In the ferrous metallurgy process, the sintering process is a link with high energy consumption and high pollution, and the generated flue gas contains a large amount of pollutants such as dust, sulfur dioxide, nitrogen oxides and the like. The traditional sintering flue gas treatment mostly adopts a terminal treatment mode, and the energy consumption is relatively high. In recent years, sintering flue gas circulation technology is widely used, and heat energy recovery and pollutant emission reduction are realized by reintroducing part of flue gas into a sintering material layer. However, the existing sintering machine generally adopts a fixed structure on the design of a smoke circulating hood, and the dynamic distribution characteristic of smoke on the surface of a material layer is not fully considered when a smoke inlet is arranged, so that a flow field is uneven and local high-speed airflow impacts the material layer, and the uniformity and the efficiency of a sintering reaction are affected. In order to optimize the flow field, the prior art mostly adopts computational fluid dynamics CFD simulation to design a flue gas distribution device. The method simulates a flow field by constructing an idealized model and determines structural parameters based on simulation results. However, the fluid dynamic CFD method has static limitation, and cannot capture the dynamic changes of working condition parameters such as air quantity, temperature and material layer resistance in the sintering process in real time, so that the deviation between a design result and actual operation is large. In addition, the fluid dynamics CFD simulation takes a long time, real-time data accumulated in the production process are not effectively utilized, and the application effect of the fluid dynamics CFD simulation on an actual sintering line is limited. The prior art, namely the hydrodynamic CFD, has the following defects in the design of a sintering flue gas distribution device that the static design cannot adapt to the dynamic working condition. The fluid dynamics CFD is based on an idealized model, and cannot reflect dynamic changes of parameters such as air quantity, temperature, material layer characteristics and the like in the sintering process in real time, so that the flow field of the flue gas distribution device is unevenly distributed in the actual operation, and the sintering quality is affected. The invention collects the flue gas flow field data in real time through the array sensor, and dynamically predicts the flow field state by utilizing the cooperative model, thereby realizing the real-time adjustment of the flue gas distribution device. The fluid dynamic CFD simulation needs multiple iterations and a large amount of calculation, the design efficiency is low, and the optimization period of sintering equipment is prolonged. According to the invention, the trained intelligent model is used as a proxy model, so that the flow field performance is rapidly predicted, and the optimal structural parameters are directly output by combining an optimization algorithm such as a genetic algorithm, so that the design time is shortened. And thirdly, neglecting real-time data utilization. The fluid dynamics CFD is independent of actual production data, and cannot be analyzed and optimized by utilizing operation data accumulated by a sintering line, so that information waste is caused. According to the invention, the sensor data is collected and preprocessed in real time through the edge data processing module and is input into the judging module, so that dynamic regulation and control based on real data are realized. In addition, conventional PID feedback control strategies also have limitations in sintering complex flow fields in terms of actual operational control. The sintering process has hysteresis and multivariable coupling characteristics, PID is based on current error feedback, memory and look-ahead prediction on long-sequence historical information are lacked, control response hysteresis is caused, and thermal inertia fluctuation is difficult to eliminate. Meanwhile, the traditional single-point feedback mechanism ignores the characteristic of uneven spatial distribution of the flow field, and aims at local flow field distortion such as edge air leakage, so that the uniformity regulation and control of the whole flow field are difficult to realize under complex working conditions. In view of the above drawbacks, with the development of artificial intelligence technology, methods such as graphical annotation force networks and sele