Adaptive geno-fuzzy control of process plants via perpetual evolution

摘要

This paper presents an adaptive mechanism based on genetic algorithms for basic fuzzy controllers. The concept of "perpetual evolution" is proposed as a novel strategy for using genetic search to tune fuzzy controllers in real-time. In this way, a population of potential fuzzy controllers evolves continuously over the time responding to changes in the environment. At a given time, the best fitting controller in the latest population is accepted as the solution. Adaptability for any real-time process changes is achieved by using an online neural network model for evaluating trial solutions. The proposed technique is demonstrated through simulations by applying to the control a chemical reactor.