How Data-Driven Modeling Methods like Neural Networks can help to integrate different Types of Data into Reservoir Management

摘要

Additionally to proven deterministic modeling techniques,rnlike numerical reservoir simulation, so called "data-driven"rnmodeling techniques can support petroleum engineers inrnreservoir management tasks.rnData-driven modeling means that the underlyingrnrelationship among measured data is calculated by the modelrnitself and no a priori knowledge of the physical systemrngoverning the data behavior is needed.rnNeural Networks are such data-driven models and "learn"rnthe underlying model from data. Neural Networks arerncomputational models broadly inspired by the organization ofrnthe human brain. The most important features of a NeuralrnNetwork are its abilities to learn, to associate, and to be errortolerant.rnMany papers have been published during the last yearsrndemonstrating the wide range of possible applications ofrnNeural Networks in Exploration & Production. An example inrnreservoir management issues will be presented, with regard tornoptimizing the injection-production ratio in a Middle Eastrnreservoir.