|
Simulator Reservoir Neural Network
Overview:
Sirenn, plug-in for Petrel*, is a very powerful and flexible tool to build proxy
models of reservoir simulators. The proxy models are developed using artificial
neural networks, a technique that has been expanding rapidly in petroleum engineering
applications, as in other fields requiring huge computer power like aerospace (flight
simulation). Sirenn can be used for many applications: the Well Placement Optimization,
Field Development Scheduling, History Matching with Multiple Models, Global Optimization
of Oil Production Systems, Screening and Designing Improved Oil Recovery Methods.
Useful links:
Features:
Sirenn allows bypassing time consuming reservoir simulations, while maintaining
a good approximation of results. This can be very useful, particularly for inverse
problems, sensitivity analysis and uncertainty analysis which, all of them, requires
huge amounts of reservoir simulations. With Sirenn, you will be able to handle cases
that could not be handled using directly reservoir simulators. Neuro-Simulation
techniques are the best suited approach as they are very well adapted to represent
nonlinear phenomena.
|
Sirenn is a very powerful and flexible tool to build proxy models of reservoir simulators;
very useful for History Matching studies. The proxy models are developed using artificial
neural networks. This tool identifies the best neural network for your problem with
the option "Optimal neural network". This tool allows you to save computation time
while keeping high accuracy of your results.
|
|
|
|
|
Sirenn aims to improve History Matching workflows using a proxy model based on neural
network to avoid time consuming reservoir simulation. For example, Sirenn allows
to history match all production data (Pressure, water-cut …) or maximize the production
(wells placement, production rates …)
|
|
|
|
|
Sirenn shows its superiority over other proxies. With three new cases, which were
not used to learning, the water cut predictions provides by Sirenn are extremely
close to those provides by Eclipse*.
|
|
|
|
|
