In this paper, a novel reduced-order active disturbance rejection control (RADRC) approach is presented for multivariable chemical process systems. By requiring little information on plant model, RADRC readily decouples multivariable chemical process systems through actively estimating and rejecting the combined effect of the internal plant dynamics and external disturbances. Compared to active disturbance rejection control, RADRC decreases phase lag in the observer by employing a reduced-order extended state observer. The comparative studies on two chemical process problems show excellent performance of RADRC in the presence of significant unknown disturbances and unmodeled dynamics.