engleski

Adaptive on-line optimisation of a chemical reactor system

Adaptive on-line optimisation of a system defined as a series of two chemical nonisothermal continuous stirred tank reactors (CSTR) with a recycle loop is presented. The method of estimation of local gradient under dynamic conditions and steepest ascent is applied. Considered are various forms of linear dynamic and static input output models with time dependent parameters. Model parameters are estimated by the recursive least squares method. A local gradient is determined by: a) under assumption of pseudo steady state, b) with dynamic equations for gradient tracking. The second method proved to be more stable as it avoids high gain of model errors due to ill conditioned steady state equations. The proposed algorithm enables process optimisation within time corresponding to a process time constant.

adaptive optimisation ; parameter estimation ; gradient estimation ; discrete modelling ; optimal control

nije evidentirano