Data-Predictive Control of Multi-Timescale Processes

Many chemical processes have multi-timescale dynamics. This paper proposes a data-based control approach for these processes based on the behavioral systems theory. A data re-sampling method and an innovative multi-optimization horizon data predictive control design are developed to deal with different timescales. The optimization horizons with small to large time intervals are used to predict and optimize future steps from near to distant future and deal with the dynamics of different timescales. The time intervals in different optimization horizons are aligned and the control action based on the dynamics of different timescales are optimized simultaneously. An illustrative example is included to demonstrate the proposed approach.