Use of Pest for Improving a Computer Model of Wairakei-Tauhara

Abstract

This paper describes the use of PEST, Parameter ESTimation, for improving a model of Wairakei-Tauhara. The model used a simple regular rectangular grid and a water table top boundary. These two simplifications were made to try to reduce the model run-time so that each forward simulation required by PEST could be run in a reasonable time. For comparison, this model will be referred to as the ‘rectangular model’ in this paper. The naming convention for the other Wairakei reservoir models is by the total number of blocks, with the most recent being the 41458-block model.
In this project we have established an effective work-flow that allows automatic parameter estimation against observations, including both natural state temperatures and pressure and enthalpy data from the production history. Permeabilities, porosities and upflows at the bottom of the model are included in the parameter set to be estimated. Work has been done in formulating the objective function, measuring the fit of model results to observations, so the inversion target better matches how a modeller visually assesses the results.
The work-flow was able to produce a well-calibrated model of Wairakei for uncertainty analysis, data worth analysis and future scenario forecasting.