Prediction of accumulated outflow from a fractured hydrocarbon reservoir using metamodeling

Three metamodels have been established for predicting accumulated outflow from a fractured hydrocarbon reservoir over a planning horizon of 18 years. The models have been validated and compared based on the amount of accumulated outflow prediction using dynamic reservoir simulation. The reservoir model was set up first and its basic parameters (porosity, permeability and saturation of water) were estimated. The accumulated outflow of a new well has been estimated as a function of the reservoir parameters using a Hydrocarbon in Place (HIP) equation. Low potential points were excluded using HIP equation and 25 high potential points were chosen for calibrating the metamodels. The three metamodels used are a quadratic model, a multiplicative model and a radial basis function model. These metamodels have been tested at 7 new test points and the predictions have been compared with those obtained by dynamic simulation. Prediction results show that all the three models can fairly predict the accumulated outflow of the reservoir studied in this paper while their computation time is much less than that of dynamic simulation. The results also show that in our case study the RBF model outperforms the other two models.