Continuous composition representation and multiscale solution method for refinery simulation and optimization

Because of the high economic importance of oil-related products, there is a continuing need to improve the efficiency and profitability of refinery processes. In the recent past model-based methods have gained large popularity in the refining industry and have already contributed largely to ecologically and economically more beneficial processes. When modeling refinery processes, one faces the complex composition of the processed materials (e.g. crude oil). The number of components present in crude oils quickly reaches several hundreds. Standard modeling and simulation approaches are not optimized for this type of mixtures.

Currently, the complex composition is reduced to a set of fictitious so-called pseudocomponents, which are treated as real components. The loss of information and model accuracy associated with this model reduction step is hard to quantify. Thus, the results obtained from these reduced models may not be reliable. An alternative approach to characterize oil mixtures is to represent its species by a continuous density function which reflects its full complexity. This leads to sets of continuously formulated equations. By formulating the models in this continuous way, it is possible to exploit modern discretization techniques for its solution. Using error estimation and adaptation concepts, the reliability of the simulation results can therefore be improved. A tailored multilevel approach using an adaptive wavelet- Galerkin discretization scheme is presented. The presentation will address the continuous modeling as well as the solution approach. The applicability of the techniques presented will be illustrated with several examples.