Application of neural network technique for logging fluid identification in low resistance reservoir

In recent years, artificial-neural-network (ANN) technology has been applied successfully to many petroleum engineering problems, including reservoir logging fluid identification. In this paper, we present the application of ANN technology to judge the type of fluid of reservoir sandstones. We demonstrate this with an ANN model that uses the well logs associated with known fluid type from well test conclusion as input and produces predictions of water/(oil + water) ratio, a key reservoir fluid property used in oilfield to evaluate the type of reservoir fluid. We set the output vector as x and y, so that the train sample with fluid type can be reflected to a two-dimensional crossplot and create four point of intersections represent oil, oil & water, water and dry layer respectively. With this trained crossplot, inputting well logs of the layer to be identified, using Euclidean distance to calculate the distance between the result and the four fluid type crossing points and find the shortest one, we can obtain the fluid type of this layer. The result of this research indicates that this method is quite effective and gets satisfying prediction precision for the low resistance reservoir logging fluid identification.