| Title | PREDICTIVE UNCERTAINTY ESTIMATES IN GEOTHERMAL RESERVOIR MODELS USING LINEAR ANALYSIS |
|---|---|
| Authors | J. Omagbon, M. O`Sullivan, J. O`Sullivan and C. Walker |
| Year | 2015 |
| Conference | New Zealand Geothermal Workshop |
| Keywords | Sensitivity, parameter estimation, uncertainty analysis, reservoir modelling, PEST, Monte Carlo, linear analysis. |
| Abstract | Deterministic simulation models of geothermal systems are now used extensively, not only to understand the different processes occurring within the geothermal reservoir, but also to predict future responses. Accurate model results are critical for the short- and long-term management of a geothermal field, particularly in determining the production strategy and ongoing management and development of the field. However, model predictions are often affected by uncertainties in input data, model parameters, and by incomplete knowledge of the underlying physics. A deterministic simulation assumes one set of input conditions, and generates one result without considering uncertainties. When making decisions based on these models, managers of geothermal fields must include some estimate of uncertainty to reflect the imperfect information at their disposal. In this work, we present a method for estimating the uncertainty in the prediction of a reservoir model, thus taking an important step towards the development of a more robust decision support tool. A synthetic 2D model that included six faults was developed and then used to generate production data against which the working model could be calibrated. Calibration was carried out using the gradient based Levenberg Marquardt optimization algorithm with Tikhonov regularization. The calibration process determined the possible parameter values for the faults and background permeability of the working model using temperature, pressure and tracer recovery curves. Assuming a parameter probability distribution based on the result of the local sensitivity information at the optimum parameter values, a Monte Carlo analysis was then used to investigate the uncertainty interval of the model prediction of four different forecast scenarios. |