Real-time Dynamic Optimisation by Integrated Two-Time-Scale Scheme

This paper deals with the problem of uncertainties in optimal control of real process. The measurement-based optimisation is used to treat variations in terminal constraints, model mismatch and process disturbances. It is assumed that this process will be carried out several times in a row and so that run-to-run optimisation can be performed. The paper presents an integrated two-time-scale control where constraints in optimisation problem are adopted between runs and the pre-computed optimal inputs are corrected according to the on-line output measurements during the run. Moreover, the proposed control approach has been implemented to control level transition in two connected tanks with liquid interaction. The results uncover better convergence properties with the resulting control scheme than individual schemes dealing either with run-to-run adaptation or with neighbouring extremal corrections inside the run.