A systematic methodology is presented for refinery modelling and optimisation on the basis of much more detail molecular information than conventional lumped methods. The novelty of this work is to incorporate models based on microscopic understanding into optimisation to achieve macroscopic im- provement. A molecular matrix based on homologous series of hydrocarbon compounds (B. Peng, Ph.D. Thesis, UMIST, 1999, pp. 22-41) is used to characterise refining streams. A transformation method is developed for obtaining molecular compositions from stream bulk properties. Molecular models are then built for several processes. Molecular modelling and optimisation are finally integrated into an overall refinery optimisation framework. Results show that molecular information can provide much better un-derstanding and new insights into refining operation. Consequently optimisation based on molecular modelling can greatly improve total profit while making quality products satisfying environmental regu-lations.