Subspace Modeling And Predictive Control Of Multivariate Molten Iron Quality Indices In Blast Furnace Ironmaking Processes

The steel is the lifeblood of national economy,where blast furnace ironmaking is one of the key units.Blast furnace(BF)ironmaking is a complex physical and chemical process in which molten iron is reduced from solid iron ores and other iron-compounds.The internal state of BF and the molten iron quality(MIQ)that is the final product of ironmaking process should be monitored in real time and controlled effectively to realize high production and low energy consumption of ironmaking process.However,the internal environment of the BF is extremely harsh.Due to high temperature,high pressure,multi-phase multi-field coupling and coexistence of solid,liquid and gas,it is hard to monitor the internal state and MIQ of the BF in real time.In addition,it is hard to realize operation control and optimization of BF ironmaking process.At present,four molten iron quality indices,namely the Si content([Si]),the phosphorus content([P]),the sulfur content([S])and the molten iron temperature(MIT)are widely used to indirectly reflect the operation performance and internal state of the BF.But the MIQ indices are hard to measure online and off-line laboratory testing is time consuming,usually once an hour.Furthermore,mechanism model(white-box model)of MIQ is hard to establish due to the complexity of BF.Therefore,it is necessary to establish an effective data-driven predictive model of MIQ indices to provide real-time predictive value of MIQ indices.In addition,blast furnace operation implements are still based on the experience and intuition of skilled operators today,which is limited to the detection and control technology.So,in order to achieve high production and low energy consumption of BF ironmaking,it is necessary to establish a data-driven model of the MIQ indices and control them based on the model.Considering these problem,this thesis carries out research on subspace modeling and predictive control of multivariate MIQ indices in blast furnace ironmaking and conduct experiment on 2#blast furnace of Liuzhou Iron and Steel Company in Guangxi province with the support of National Natural Science Foundations,"High-performance operation control and implement technology of large blast furnace"(61290323)and"Experimental verification platform construction and application verification of large blast furnace high-performance operation control"(61290321).The specific works are as follows:(1)Firstly,considering that too many variables and multivariate coupling in ironmaking process make it hard to select the modeling input variables,a data-driven hybrid method combining canonical correlation analysis(CCA)and correlation analysis(CA)is proposed to identify the most influential controllable and measurable variables as the modeling inputs from multitudinous factors that affect the MIQ indices,namely flow rate of cold air,pressure drop,flow rate of rich oxygen and volume of coal injection.(2)Secondly,considering the time-varying characteristics of operation condition,especially when operation condition is volatile,such as burden change and blowing-down in BF ironmaking process,an online subspace input-output model for the prediction of multivariate MIQ indices is presented based on recursive subspace identification with forgetting factory.The parameters of the identified model can be updated adaptively according to the operation condition based on the recursive algorithm,which can guarantee accurate and stable prediction result of MIQ indices.The established model is further applied to the predictive control of multivariate MIQ indices as a predictor.Because the model parameters can be updated according to the operation condition using the latest input and output data,the adaptive predictive control of multivariate MIQ indices can be realized.At last the experiment is conducted based on the actual industrial data.The control performance is compared with that of predictive control based on the subspace identification model and the results verified that the proposed method can ensure accurate prediction and satisfactory control performance of MIQ indices when operation condition changes.(3)Finally,considering the nonlinear dynamic characteristics of ironmaking and the drawbacks of linear modeling and control methods when operation condition is stable,a data-driven nonlinear subspace modeling for the prediction and control of multivariate MIQ indices is proposed.First,a Hammerstein model for the prediction of MIQ indices is established using the LS-SVM based nonlinear subspace identification method.Then the piecewise cubic Hermite interpolating polynomial method is used to further simplify the model which contains high dimensional kernel function.Compared to the original Hammerstein model,it has been shown that this simplified model can not only significantly reduce the computational complexity,but can also exhibit a good reliability and accuracy for a stable prediction of MIQ indices.After that,the established nonlinear subspace model is applied to nonlinear predictive control of the multivariate MIQ indices as a predictor.The nonlinear optimization problem can be solved by genetic algorithm which has the global search capability.The root mean square errors,set-points tracking and anti-jamming test results show that the proposed modeling method has better estimation accuracy and the proposed control method can be used to control complex nonlinear process effectively.