In the development of Maximus, considerable effort has been directed
towards making the algorithms numerically efficient with fast execution.
This effort was expended for two principal reasons:
1.To make feasible the solution of more complex and computationally difficult problems (e.g.
production optimisation or stochastic analysis) without the need to 'dumb down'
the physical models, for example by simplifying the equations or by introducing
approximate methods such as PVT lookup tables.
2. To make feasible the solution of large network systems also without the need
to reduce the fidelity of the physical models.
As a result, Maximus is able to solve quite large systems very quickly. The following model shows a large gathering network with 131 wells and 200 flowlines/pipeline branches. The production from these wells is all fed to a Central Processing Facility. The network problem is completely pressure specified and takes about 40 seconds to achieve the first solution (i.e. from internal initial guesses) and about 10 seconds for subsequent simulations on a standard Laptop with a 2 GHz Pentium 4 processor.
