Research On Predictive Control Method Of Calcination Temperature In Rotary Kiln

In recent years, Chinese metallurgical industry maintains a high growth rate, and every year consuming large quantities of lime. Therefore, the research on lime production process for better optimization control, so as to improving the product quality and saving energy, is of great economic significance and social meaning.Based on the analysis of the lime production process, the calcination temperature, as one of the most important control parameters of the rotary kiln, was selected for the controlled object, controlled by adjusting the gas flow. And by using lime kiln historical data of Tiantie group, the article established the state space model, between the calcination temperature and the gas flow, through the MOESP subspace methods. We changed it into CARIMA model form, for the subsequent simulations of control algorithm.Lime kiln production process consists of complex physical and chemical reaction, with large inertia, multivariable, strong coupling, characteristics of time-varying. And using conventional control methods is difficult to obtain satisfactory control effect. The subspace predictive control strategy is one of the hot spot of predictive control study. This subspace predictive control algorithm is deduced. Considering the existence of time-varying character-ristics in the lime calcination process, a variable forgetting factor recursive strategy was proposed. This strategy used the matching error of model to calculate the time-varying forgetting factor, which was got into the construction of the data matrix. Thus the data will follow the system information better, and the optimal control will be obtained. Required lower triangular matrix was updated by a series of Givens rotation, avoiding a computation of a new RQ decomposition and saving some computational time.The simulation results showed that the control strategy achieved a better control effect. In a certain extent, this solved the decline problem of the conventional control algorithm performance due to the model mismatch, when the controlled object parameter changed. And liked the generalized predictive control, the strategy could compensate for the interference and reduce the deviation. Compared the two methods, response of the strategy is faster. Also obtained from the simulation, the adaptive ability of variable forgetting factor is stronger than the fixed forgetting factor.