Automatic Well Test Interpretation Method for Circular Reservoirs With Changing Wellbore Storage Using One-Dimensional Convolutional Neural Network

In order to develop reservoirs rationally, accurate reservoir parameters are usually obtained through well test analysis. However, a good deal of well test data with changing wellbore storage characteristics bring difficulties to the current well test interpretation, so it is important to find a valid interpretation method for changing well storage reserves data. This paper proposed an automatic well test interpretation method based on one-dimensional convolutional neural network (1D CNN) for circular reservoir with changing wellbore storage. Compared with two-dimensional convolutional neural network (2D CNN), 1D CNN significantly reduces the computational complexity and time cost. The CNN takes pressure change and pressure derivative data of the log–log plot as input and reservoir parameters as output of network. This method applies two 1D CNNs respectively to fit two types of reservoir parameters, one type includes C_De^2s, C_αD, and C_ϕD and the other type is boundary distance R. In addition, the training samples of the two networks are different according to different parameters. The two-network approach reduces the difficulty of extracting curve characteristics and improves interpretation ability. The effectiveness of this method is proved by the field data in Daqing oilfield. The method greatly improves the working efficiency of well test interpreters and can be widely used.