| Title | Sensitivity Analysis, Parameter Estimation and Uncertainty Propagation in a Numerical Model of the Ngatamariki Geothermal Field, New Zealand |
|---|---|
| Authors | Hyungsul MOON, Jonathon CLEARWATER, Peter FRANZ, Irene WALLIS and Lutfhie AZWAR |
| Year | 2014 |
| Conference | Stanford Geothermal Workshop |
| Keywords | Ngatamariki, inverse modelling, reservoir management, statistical analysis |
| Abstract | In March 2013, Mighty River Power commissioned an 82MWe binary power plant in the Ngatamariki geothermal field, 17 km north east of Taupo, New Zealand. Previously, a full field numerical model and several simpler process models of the Ngatamariki system were built and tested prior to the decision to proceed with resource development. The sensitivity of these models was tested over a limited domain by varying individual parameters of interest. Conservative forward scenario runs were simulated by a reservoir model with reduced areal extent of permeable reservoir. In this study, the sensitivity of model results to parameter values is analysed in more detail with the inverse modelling package iTOUGH2 and an optimal model is then used to propagate the uncertainties in scenario forecasting by Monte Carlo simulation. This results in more rigorous statistics on what parameters drive model behaviour and quantitative bounds on the confidence of model predictions. This study found that the fundamental parameters in determining the temperature and pressure in the natural state were fracture permeabilities, while for production scenarios the important parameters driving the system response were fracture spacing and matrix porosity. These influential parameters were varied over reasonable ranges in a series of Monte Carlo model simulations runs in order to include some rigorous statistical uncertainty in model forecasts. The uncertainty (two standard deviations) in the enthalpy forecast after a 50 year production run due to uncertainty in reservoir porosity and fracture spacing was +/- 40kJ/kg from the mean enthalpy prediction. The impact of uncertainty in reservoir size on modelled enthalpy evolution was also addressed with iTOUGH, with results indicating reservoir size could contribute a +/- 8kJ/kg range over a 50 year production scenario. |