Automated calibration of geothermal models

Abstract

In order to achieve a reliable and realistic model prediction, a geothermal reservoir model needs to undergo an extensive calibration process. Since most models, if not all, are complex, manual fitting of the numerous parameters included in this calibration procedure is a laborious, repetitive and frustrating process. This thesis discusses methods for the automatic calibration of geothermal reservoir models that use the TOUGH2 multi-phase flow simulation software. The inverse modelling software suite PEST is utilised to identify the values of the calibration parameters that achieve the best fit between the model results and measured data. In order to gain further improvement in the model fit to the data, a few special techniques are employed, including (i) systematic subdivision of the model domain into smaller units by introducing additional rock types, (ii) freezing of very low sensitivity parameters, (iii) employing the user intervention and SVD-assist functionalities, and (iv) employing the pilot point method in PEST. The methods are demonstrated using models of two geothermal fields, namely the Northern Negros Geothermal Project in the Philippines and the Mokai Geothermal Field in New Zealand. The calibration process involves the matching of the natural state temperatures and pressures taken from well test measurements. Also the production response of the field during exploitation period, quantified by measurements of production enthalpies and pressure trends, are used to calibrate the model.