Static Optimisation of a Fermenter Using a Learning Automaton

摘要

This paper deals with the static optimisation of a stirred tank fermenter using a random search technique. This random search technique is based on a learning automaton operating in a random environment constituted by the process under consideration. The control actions of the automaton are associated with a discretisation of the manipulated variable: the inlet substrate flow rate. At each sampling instant, a reinforcement scheme which uses the realisations of the function to be minimised is used to adjust the probability distribution. The control action is selected according to the probability distribution of the vector of discrete control actions. An averaging procedure is introduced to deal with noisy measurements and to add robustness to the control strategy. Simulation results illustrate different aspects as well as the performance of this optimisation technique.