Автоматизація процесу керування термічною обробкою залізорудних обкотишів з використанням прогнозуючих Anfis-моделей

The thesis is devoted to solving of the urgent scientific task of increasing the efficiency of automated process control of thermal treatment of iron-ore pellets on conveyor induration furnace by the development of algorithms and control methods based on predictive neuro-fuzzy models. Realization of this goal allows decreasing energy costs, increasing productivity of conveyer induration furnace, and improving quality characteristics of the final product.

Analysis of the automation of control processes of thermal treatment of iron-ore pellets showed that along with the urgency of the problem and quite a lot of research and development in this direction, the existing systems and methods of control do not fully meet modern requirements and does not always provide an effective solution of complex problems in pellets production (compliance desired temperature of pelletizing, minimize energy costs), which leads to lower technical and economic parameters of the process. Promising direction of solving the problem of improving the efficiency of automated process control of thermal treatment of pellets is an approach based on a combination of methods of intellectual identification and model predictive control.

On the basis of experimental data the research of the efficiency of models and methods for rapid identification of the processes of thermal treatment of pellets in technological areas of induration furnace are conducted. Improved dynamic models of these processes in a class of autoregressive neuro-fuzzy predictors with taking into consideration the speed of moving grate, height of the pellets layer, energy consumption, parameters of gas flow distribution and temperature regimes in the previous zones are suggested. As a result of research of received neuro-fuzzy models their basic properties have been determined: the standard deviation between experimental values and values calculated on the basis of model - up to 0,65 °С; the normalized root mean square error of approximation NRMSE up to 1,5 %; the optimum period of prediction - 3 steps; regression depth - 2-3 signals.

Improved approach to automation of control process of thermal treatment of pellets based on the combination of the intellectual identification using autoregressive ANFIS-models and methods of model predictive control is proposed. With using this approach new structures of the predictive control system of thermal treatment of pellets in the technological areas of conveyer induration furnace as well as algorithms for their operation are developed. Computer simulation of the proposed algorithms of predicting control in nominal operating modes, at the presence of disturbances (changing pellets layer height, speed of moving grate) have proved the possibility of their application in conditions of technological process of thermal treatment of iron-ore pellets. Analysis of simulation results showed that the use of appropriate algorithms and predictive control systems allow reduce the standard deviation of manipulated variable (temperature in the technological areas) from the set value to 8 %, energy up to 0.8 % compared with existing systems, implemented on the basis of PI-regulators.

Developed in the thesis mathematical models, methods, structures, algorithms of predictive control of thermal treatment of pellets on conveyor induration furnace were used in the projects of the state enterprise “Kryvyi Rih Institute of Automatics”. Calculation of cost-effectiveness implementation of the results of the thesis for the conveyor induration furnace with annual capacity of 2.25 million tons upon reducing the consumption of natural gas to 0.8 % in zone preheating and 0.64 % in the cooking zone shows that the annual economic effect amounts to 186 thousand UAH per year.

Keywords: automation, pellets thermal treatment, conveyor induration furnace, neuro-fuzzy model, predictive control.

Размещено на Allbest.ru