Minimum variance tuning of PI controllers using hybrid genetic algorithms

One of the main confronts in control engineering is the assessment of close loop performance. Harris ascertains a performance index where the best performance is assumed to be attained by a minimum variance controller.

This research spotlights on the tuning of the illustrious and most frequently used PI controller to achieve minimum variance conditions. The optimization problem is embrarked upon two different approached. The first approach uses enumerative search optimization for its simplicity. The second approach applies an exploited hybrid genetic algorithm that is developed to generate vigorous and premium results. The algorithm amalgamates the genetic operations of selection, crossover, and mutation with Newton's search inside successively expanding and contracting parameter domains using alternating logarithmic and linear mappings. Finally, the obtained PI parameters and tested and simulated with data from three control loops at Falconbridge Smelter in Sudbury and compared with the existing tuning parameters. The new parameters yield optimal results.