Premium
Tracer model identification using artificial neural networks
Author(s) -
Akin Serhat
Publication year - 2005
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/2004wr003838
Subject(s) - tracer , artificial neural network , matrix (chemical analysis) , geothermal gradient , porosity , fracture (geology) , network model , computer science , geology , data mining , geotechnical engineering , artificial intelligence , materials science , geophysics , physics , nuclear physics , composite material
The derivation of transport parameters from tracer tests conducted in geothermal systems will depend strongly on the conceptual and mathematical model that is fitted to the data. Depending on the model employed the estimation of transport parameters (porosity and dispersivity of the fracture network, porosity of the matrix) may result in a significant variation in dispersivity. If the results from such tracer tests are to be used in parameter selection for larger‐scale models, it is crucial that the tracer test is itself interpreted with an appropriate model. In order to tackle this problem, artificial neural network (ANN) technology is proposed. A dual‐layer neural network model was trained using synthetic tracer test data generated using analytical one‐dimensional homogeneous, double‐porosity pseudosteady state, multifracture, and fracture matrix models. The developed model was then used to identify several actual tracer tests conducted in various geothermal reservoirs reported in the literature. In most cases it was observed that the model successfully identified a wide variety of reservoir models. In some cases the model decreased the number possible models to two. It was also observed that ANN results were in accord with least squares analysis.