| Title | Geothermal Data Mining -- Value in Antiques |
|---|---|
| Authors | Roland N. Horne |
| Year | 2008 |
| Conference | New Zealand Geothermal Workshop |
| Keywords | data analysis, nonparametric regression, wavelets |
| Abstract | Analysis of the production history in developed geothermal fields has been used extensively to forecast the future life of the reservoir. Mature reservoirs such as Wairakei offer a unique opportunity to take a retrospective look at how well reservoir engineering methods have been able to predict the future. In general, reservoir models have performed adequately at describing the bulk volume-pressure behaviour of reservoirs such as Wairakei, however there are other significant production behaviours such as thermal breakthrough from injection wells that have been almost impossible to predict. These fracture-dominated phenomena can still be modelled, however may need a different approach. The use of data mining procedures that allow the reservoir to reveal its own internal dependencies can provide very good indications of future behaviour, even for such events as thermal breakthrough. This paper will describe wellto- well correlation techniques based on nonparametric regression. The techniques have been applied successfully to real field histories. |