Use of the Genetic Algorithm for Load Balancing of Sugar Beet Presses

obtaining, from the models, ratios to run the presses at and running the presses at those ratios. Results will be obtained during the next sugar processing campaign. On-line tests of direct optimal control are currently running at the Bardney Sugar Factory and their results will be evaluated at the end of current sugar campaign by calculating the content of dry substances in pressed pulp per unit throughput of sugar beet pulp processed by the press station. The preliminary evaluation based on data obtained so far indicates that the direct control improves performance of the press station by increasing the percentage moisture removed from the sugar beet residue during pressing by 0.623%-0.749%. This result corresponds with British Sugar plc initial estimates of 0.635% and 0.835% improvement based on figures obtained from two different sugar plants. Data analysis, making use of graphic tool XVGR available on X-terminals, will enable us to asses the optimisation systems efficiency. It will also facilitate definition of a control strategy for the proposed on-line optimisation systems and critical comparison of the approaches used. Another area of our current investigation related to this project is use of an on-line control system with a genetic algorithm which will enable tracking optima of the The aim of this research is to continually evolve an optimal solution to the problem which will limit the need for repeated initialization of the genetic algorithm from a random starting point in the search space when changes in the system cause variation of the parameters of the press station model. Fogarty T C (1988) " Rule-based optimisation of combustion in multiple burner furnaces and boiler plants " in Engineering Applications of Artificial Intelligence, vol:1, iss:3, p.203-209. Fogarty T C (1989a) " An Incremental Genetic Algorithm for Real-time Learning " in 'Proceedings of the Sixth International Workshop on Machine Learning', p.416-419. Fogarty T C (1989b) " Learning new rules and adapting old ones with the genetic algorithm " in 'Artificial faster without a relatively long period of prior data collection with random variation of speeds of the presses. On the other hand there is a need to monitor sliced weight and Ph factor of press water and use these values for modification of the cost function. Another drawback of this method for some applications can be the direct evaluation of weak offspring may have dangerous consequences on the system controlled. The system identification approach has not …