engleski

ADAPTIVE ON-LINE OPTIMISATION OF CHEMICAL PROCESS BY ENHANCED RLS ESTIMATION

On-line adaptive optimisation of chemical reactor system is analysed. Optimisation is based on local adaptive linear discrete time dependent models and steepest ascent algorithm. Local gradients are evaluated by the dynamic equation model. Model parameters are estimated and adapted to process variations by use of the enhanced recursive least square (RLS ) algorithm. Estimation is enhanced by introduction of uncorrelated perturbations of control variables simultaneously with optimisation steps. Tested are pseudo random binary sequences ( PRBS ) and harmonic functions. Significant improvement in accuracy of estimation and stability of optimisation is proven.

process optimisation; adaptive model; recursive estimation; steepest ascent

nije evidentirano