Abstract The use of isenthalpic flash has become of interest for the simulation of some heavy oil recovery processes where large temperature changes are experienced. For these thermal simulations energy can be used as a primary variable. This leads to thousands or millions of individual multiphase isenthalpic flash calculations. Robust and efficient algorithms for multiple-phase isenthalpic flash are required to improve the efficiency of compositional simulations for thermal recovery.
The general framework on state function based flash specifications proposed by Michelsen (1999) is applied to general multiphase isenthalpic flash. An extension to the general multiphase case is presented for the two-phase Jacobian. This can be formulated as either a partial Newton or full Newton procedure. An extended solution strategy is presented to deal with many of the problems found in isenthalpic flash. This incorporates Q function maximisation for instances where other methods are not convergent.
Narrow boiling mixtures can be dealt with in the majority of cases without any significant difficulty. This is true of the direct substitution algorithm and the proposed solution procedure. The vast majority of examples can be solved without using Q function maximisation. The challenges associated with multiphase calculations in the Newton steps are investigated. In particular, inadequate initial estimate of the equilibrium type may lead to non-convergent iteration. This can usually be solved by introduction of a new phase and/or elimination of an existing phase. The speed of the method is analysed for a large number of specifications and is found to be only slightly more expensive than isothermal flash in the majority of cases.