|
Отправлено: 15.04.10 01:06. Заголовок: Prediction of Poisson’s Ratio and Young’s Modulus Parameters for Hydrocarbon Reservoirs using NN
Prediction of Poisson’s Ratio and Young’s Modulus Parameters for Hydrocarbon Reservoirs using Artificial Neural Networks By Bandar Duraya Al-Anazi, King Abdulaziz City for Science and Technology, Oil and Gas Center, Mohsen Saemi, Research Institute of Petroleum Industry (RIPI), and Ammal Al-Anazi, Saudi Aramco ABSTRACT Determination of rock elastic properties plays an important role in various geomechanical applications such as hydraulic fracture design, sand production control and wellbore stability analysis. Th ese elastic properties are often reliably determined from laboratory tests on cores extracted from wells under simulated reservoir conditions. Unfortunately, these tests are expensive, time consuming and most of the wells have limited core data. On the other hand, logs are often available and provide a continuous record compared to cores where only discrete values are obtained. Empirical equations are used to estimate elastic properties from logs however empirical corrections must be performed to calibrate the dynamic calculation to core-measured values. Due to the complexity of the reservoirs, these models may not be of practical use and consequently extensive data preprocessing and understanding of the geology of the region and the tool limitations are required. Alternatively, artificial neural network has the potential to model complex nonlinear underlying dependency between high dimensional input logs and elastic properties. Th e potential of the neural network has been demonstrated by developing Poisson’s ratio and Young’s modulus interpretation models in a hydrocarbon reservoir using log-based density and acoustic measurements and core-measured porosity, minimum horizontal stress, pore pressure and overburden stress. Learning and prediction performance was performed using correlation coeffi cient, root mean squared error, absolute average error and maximum absolute error. The result shows that artificial neural network can successfully be used to construct elastic interpretation models that can be employed to interpret new input data with a minimum error rate. Keywords: Poisson’s ratio, Young’s modulus, Logs, Core, Artifi cial intelligence ================== Получил от автора письмом Saudi Aravia Oil &Gas Iss 13 Думаю есть у них на сайте
|