PERMEABILITY PREDICTION WITH ARTIFICIAL NEURAL-NETWORK MODELING IN THE VENTURE GAS-FIELD, OFFSHORE EASTERN CANADA

Estimating permeability from well log information in uncored borehole intervals is an important yet difficult task encountered in many earth science disciplines. Most commonly, permeability is estimated fr om v arious well log curves using either empirical relationships or some fo rm of multiple linear regression (MLR). More sophisticated, multiple n onlinear regression (MNLR) techniques are not as common because of dif ficulties associated with choosing an appropriate mathematical model a nd with analyzing the sensitivity of the chosen model to the various i nput variables. However, the recent development of a class of nonlinea r optimization techniques known as artificial neural networks (ANNs) d oes much to overcome these difficulties. We use a back-propagation ANN (BP-ANN) to model the interrelationships between spatial position, si x different well logs, and permeability. Data from four wells in the V enture gas field (offshore eastern Canada) are organized into training and supervising data sets for BP-ANN modeling. Data from a fifth well in the same field are retained as an independent data set for testing . When applied to this test data, the trained BP-ANN produces permeabi lity values that compare well with measured values in the cored interv als. Permeability profiles calculated with the trained BP-ANN exhibit numerous low permeability horizons that are correlatable between the w ells at Venture. These horizons likely represent important, intra-rese rvoir barriers to fluid migration that are significant for future rese rvoir production plans at Venture. For discussion, we also derive pred ictive equations using conventional statistical methods (i.e., MLR, an d MNLR) with the same data set used for BP-ANN modeling. These example s highlight the efficacy of BP-ANNs as a means of obtaining multivaria te, nonlinear models fur difficult problems such as permeability estim ation.