Abstract: PGQ (process goose queue) is a novel approach to deal with the decomposition and coordination optimization of complex industrial processes. For the PGQ formation adjustment problem in the absence of process disturbance, we propose a predictive control algorithm based on a hierarchically distributed model. This is based on Hankel matrices of input and output data, and uses subspace identification to directly obtain impulse response sequences of the system for model predictive controller design. The alumina carbonation decomposition process is used as a case study to demonstrate the effectiveness of the proposed approach.