Abstract
In order to more accurately identify multistage fracturing horizontal well (MFHW) parameters and address the heterogeneity of reservoirs and the randomness of well-production data, a new method based on the PSO-RBF neural network model is proposed. First, the GPU parallel program is used to calculate the bottomhole pressure of a multistage fracturing horizontal well. Second, most of the above pressure data are imported into the RBF neural network model for training. In the training process, the optimization function of the global optimal solution of the PSO algorithm is employed to optimize the parameters of the RBF neural network, and eventually, the required PSO-RBF neural network model is established. Third, the resulting neural network is tested using the remaining data. Finally, a field case of a multistage fracturing horizontal well is studied by using the presented PSO-RBF neural network model. The results show that in most cases, the proposed model performs better than other models, with the highest correlation coefficient, the lowest mean, and absolute error. This proves that the PSO-RBF neural network model can be applied effectively to horizontal well parameter identification. The proposed model has great potential to improve the prediction accuracy of reservoir physical parameters.
Highlights
Reservoir description is an important means of estimating the physical properties of a reservoir in the petroleum industry. ese reservoir physical parameters require better methods to improve the prediction accuracy to further enhance the subsequent success rate of exploration and development
Permeability is defined as a measure of the ability of a porous medium to allow fluid to pass through it. e concept of permeability is important for determining accurate reservoir description, simulation, and management. erefore, prior to any modeling or calculation, the permeability of porous media must be determined. e earliest method of permeability prediction is the empirical correlation between permeability and other petrophysical properties such as porosity and water saturation. ese correlations have made some successes in sandstone reservoirs
In order to illustrate the effect of parameter settings of inertial weight and learning factor on the prediction results of the particle swarm optimization algorithm (PSO)-RBF model, a classic particle swarm algorithm with fixed parameter values and an improved particle swarm algorithm with parameter value changing were used to predict the permeability and the half-length of the fracture, respectively. e basic parameters of the two algorithms are shown in Table 3. e permeability prediction data and fracture length prediction data of the two algorithms are shown in Tables 4 and 5, respectively. e prediction results of the two algorithms are shown in Figures 8–11, respectively
Summary
Reservoir description is an important means of estimating the physical properties of a reservoir in the petroleum industry. ese reservoir physical parameters require better methods to improve the prediction accuracy to further enhance the subsequent success rate of exploration and development. Ese reservoir physical parameters require better methods to improve the prediction accuracy to further enhance the subsequent success rate of exploration and development. The artificial neural network (ANN) has been applied to the reservoir field, solving many highly complex nonlinear problems [5]. It is used to solve many different types of problems in petroleum engineering, such as reservoir characterization, permeability prediction, vertical multiphase flow downhole pressure prediction, and dissolved gas driven flow prediction [6,7,8]. In order to overcome the shortcoming, in this study, an MFHW parameter identification model based on the PSO-RBF (particle swarm optimization, PSO; radial basis function, RBF) neural network is proposed. E parameter identification model based on the PSO-RBF neural network has high learning ability and flexibility when encountering complex problems, which makes the proposed model applicable to most engineering practices E parameters of the radial basis neural network are designed by using the particle swarm algorithm to obtain higher prediction accuracy. e parameter identification model based on the PSO-RBF neural network has high learning ability and flexibility when encountering complex problems, which makes the proposed model applicable to most engineering practices
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
