Abstract
Because the ANN model has the ability to fit the data of the complex nonlinear, a large number of field data about gas flooding are collected to train the ANN. Using the formulas, we translate the weighting matrix of ANN which possesses good fitting precision to the weight of single factors for fuzzy mathematic. Then, the fuzzy mathematic method is used to implement comprehensive evaluation to quantized indicators.
The new screening methodology could determine exactly the potential of gas flooding in candidate reservoirs. To overcome the slow convergence of ANN and easy trapping in the local minimum, a large number of field data about gas flooding are collected to train the ANN. Results indicate that the ANN is reliable, and it's mean-square error(MSE) is smaller than 0.01. The weight coefficients of those indicators for fuzzy mathematic are calculated with ANN. The results are as follows: the oil viscosity is the highest score (its weight coefficient is 0.3), followed by the reservoir thickness, reservoir pressure, and reservoir permeability. Moreover, others indicators play a relatively small role. By using the above screening methodology, seven field cases from Ordos Basin are used to evaluate the potential of gas flooding. Results from the field application, indicate that the recovery factors of gas flooding in the field cases are sequential consistency with those from our methodology.
