Comparison of some algorithms for hierarchical steady-state optimising control of interconnected industrial processes

A common method for controlling a dynamical industrial process is to split the control action into two layers: a follow-up control layer responsible for keeping the chosen process variables at their desired values (set-points) despite fast disturbances acting upon the process, and an optimising control layer responsible for determining optimal values of the set-points. Algorithms for the optimising control layer are considered in this paper. In practice, accurate models of industrial processes are rarely available and, therefore, these algorithms have to perform both optimisation and identification tasks. Since a natural two-step approach of iterating between successive solutions of steady-state optimisation and model parameter estimation rarely leads to the true process optimum, an integrated system optimisation and parameter estimation (ISOPE) method was proposed by Roberts (1979). Various ISOPE algorithms are now available, and the aim of the paper is to provide a comparative study of the most important algorithms