Data driven modeling and nonlinear model predictive control design for a rotary cement kiln

Abstract

Cement production is an energy intensive process, on the other hand the product quality directly influences the economic benefit. In this work a nonlinear model predictive control is designed to achieve an optimal compromise between energy consumption, production volume, and product quality. Based on measurements from a 100 t/h rotary cement kiln a non-linear autoregressive NARMAX-model is identified, and cross validation of this model shows good accuracy for control design. A prediction of the most influential disturbance (quality of the feed material) is utilized, product quality can be defined as a set-point, and the optimization criterion is defined using time-varying performance weights. This design achieves good transient response while still guaranteeing that the desired production volume is met. Validation results of the model with measured data and simulation results for the closed-loop operation demonstrate the functionality of the proposed methodology.