Novel model reduction techniques for refinery-wide energy optimisation

The oil refining industry mainly uses linear programming (LP) modelling tools for refinery optimisation and planning purposes, on a daily basis. LPs are attractive from the computational time point of view; however these models have limitations such as the nonlinearity of the refinery processes is not taken into account. In addition, building the LP model can be an arduous task that requires collecting large amounts of data. The main aim of this work is to develop approximate models to replace the rigorous ones providing a good accuracy without compromising the computational time, for refinery optimisation. The data for deriving approximate models has been generated from rigorous process models from a commercial software, which is extensively used in the refining industry. In this work we present novel model reduction techniques based upon optimal configuration of artificial neural networks to derive approximate models and demonstrate how these models can be used for refinery-wide energy optimisation.