Formation-Grain-Size Prediction Whilst Drilling: A Key Factor in Intelligent Sand Control Completions

Abstract

More than 80% of oil and gas reservoirs especially in HP/HT wells and deep water environments of the Atlantic Margin are known to be highly unconsolidated requiring some form of advanced, complex well geometry sand control completions. A key factor in an optimum sand control completion is a thorough knowledge of formation sand profile, which is traditionally determined by sieve analysis of samples obtained from cores, production lines, etc. In complex wellbore geometries, the samples obtained through these traditional processes may be totally unreliable. In this paper, a neural network technique for predicting grain-size distribution from wireline logs is presented. The paper briefly introduces neural networks and discusses the implementation of the network and its application to Smart Well Sand Control Completions. The network is recommended for application for either real-time prediction (that is, while drilling) where input data may be fed directly to the neural network prediction model, or for prediction at office base depending on the requirements, urgency and purpose of the data. Benefits of the new model include real time prediction of formation grain-size and more reliable prediction of grain size in complex well geometries where the use of traditional sand sampling would be difficult. There is potential for application in the prediction of plugging tendency of screen systems and gravelpacks.

UCS Neural Network Model for Real Time Sand Prediction

“…Though normal distribution of data is not particularly important to neural network [5], it is however important that the distribution be approximately symmetrical. In fact, evidence from previous work [6] has shown that neural networks learn better and faster on transformed normally distributed data. The transformation of the data used in this study was achieved by using several functions such as inverse, logarithm, square and square root in a "trial and observe" method, in which each of these functions was used to transform all the data sets in turn and the skewness and kurtosis observed each time.…”

Predicting sand size distribution based on well logs of east Malaysia basins

“…Due to those issues, some efforts were directed towards finding alternative techniques to estimate the sand size by exploiting data from wireline logging [12]. M.B Oyeneyin [13] analyzed an equation to forecast the formation sand size in Niger Delta. This empirical equation was taken to correlate the sand formation dimension to some petrophysical properties such as permeability, porosity, and water saturation.…”

Sand Production Forecasts for Unconsolidated Gas Reservoirs

“…These are known to be largely due to changes in operational or reservoir conditions. It is difficult and almost impossible to measure sand production levels especially with reservoir pressure depletion and corresponding increase in Water-Gas ratio(WGR) [3][4][5].…”