To our understanding a good Scheduling-System should solve 3 basic problems: (1) To close the ' scheduling-gap' to the monthly LP, (2) so that at no time a tank has an overflow or an underflow and (3) so that the units are running as smooth as possible. The most Scheduling-Systems we know are more or less event driven simulation systems. A simulation algorithm starts at time zero and works forward on the time axis, each event produces decisions by given rules to find a feasible schedule. Our approach is completely different. The heart of our system is a 3-period full refinery LP-model with some extra features and extensions. The period lengths are 3 days, 4 days and a variable restperiod, e.g. to the end of the month. The week of the first two periods is the essential scheduling time span. This LP is embedded in a C-programme which (1) generates the necessary dynamic LP-structure, (2) communicates with the LP in a recurrent way and (3) produces state of the art graphical output on a surface where the schedule can be manipulated by the scheduler without affecting the overall LP-solution. Open inventories and qualities are transferred automatically from the refinery date base into the scheduling system and the schedules can be sent over the refinery network, stored or printed.
展开▼
