The nature of a geothermal reservoir depends on the geological structure and the deep heat and mass inputs. Development of a numerical model for predicting the performance of a geothermal field requires parameter values to be deduced by calibrating the model against field data. Our study investigates this model calibration or inverse geothermal modelling problem as a problem in statistical inference within a Bayesian framework.

Bayes' rule is used to determine the probability density for the distribution of parameters in a model of a geothermal reservoir. A Markov chain Monte Carlo (MCMC) sampling technique is used to explore and interpret the posterior distribution. The outputs of MCMC sampling are summarized by calculating sample expectation values of statistics about the parameters. The reservoir simulator TOUGH2 is adopted to simulate the forward problem. TOUGH2 is designed for non-isothermal, multiphase flows in three dimensional porous and fractured media, and has been applied to many problems including geothermal modelling and nuclear waste disposal.

The methodology is applied to determining the reservoir parameters near a single well by matching enthalpy and pressure data measured over three months. It is also applied to a 3D reservoir model, matching measurements of natural state temperature versus depth in several wells. In both cases, the model parameters are recovered successfully. We conclude that our model calibration is better than that achieved in previous work using manual methods.

Press the Back button in your browser, or search again.

Copyright 2006, University of Auckland: Readers who download papers from this site should honor the copyright of the original authors and may not copy or distribute the work further without the permission of the original publisher.