| Title | DATA-WORTH ANALYSIS: DESIGNING A MONITORING PLAN FOR ROTORUA THAT REDUCES UNCERTAINTY |
|---|---|
| Authors | K. Dekkers, M. Gravatt, O.J. Maclaren, R. Nicholson, J. O’Sullivan, M. O’Sullivan |
| Year | 2021 |
| Conference | New Zealand Geothermal Workshop |
| Keywords | Reservoir modelling, Rotorua, Uncertainty quantification, Data-worth analysis |
| Abstract | Developing an accurate geothermal model requires model calibration to match available data. The data is expensive to gather and, therefore, usually sparse. It is common to use geothermal models to make predictions that aid in managing the geothermal field sustainably. However, these model predictions are uncertain due to model complexity and sparse data. This paper discusses how we can quantify that uncertainty and how we can design monitoring plans to reduce the uncertainty in model predictions in the context of the Rotorua geothermal field. The design of monitoring plans for uncertainty reduction falls under the area of data-worth analysis. From 1950 to 1986, the state of the reservoir of Rotorua was deteriorating. Therefore, the government and the Bay of Plenty Regional Council collected valuable monitoring data that helps to sustainably manage the geothermal field. There are many individual users of the geothermal reservoir, and it is not known exactly how much production has occurred in the past from various users. Uncertainty in production rates is an unusual situation compared to other geothermal reservoirs, as usually the production and reinjection are known. We have hence developed a new geothermal model for Rotorua that considers the uncertainty in production and reinjection. Using the new model, we can make model predictions that have uncertainty bands that provide better information for reservoir management. In this paper, we consider uncertainty and data-worth analyses for Rotorua using our new model. We present three scenarios of monitoring plans for taking additional measurements in Rotorua and compare how measuring new data in different locations reduces the uncertainty in model predictions. The results of these monitoring scenarios show the benefit of data-worth analysis using simulation models: we can assess the possible effect of new monitoring plans before spending any money on these further measurements. |