| Title | Analysing cyclic well behaviour via time series analytics and reservoir simulation |
|---|---|
| Authors | P.M.B. Abrasaldo, S.J. Zarrouk, A.W. Kempa-Liehr |
| Year | 2022 |
| Conference | New Zealand Geothermal Workshop |
| Keywords | Geothermal energy, machine learning, feature engineering, time-series analytics, fractional dimension, reservoir modelling |
| Abstract | The geothermal energy industry has always been in constant flux with the advent of new technology and the rich experience of long-time geothermal operators. However, the industry’s growth has been tempered by the inherent risks associated with geothermal energy extraction, particularly by unfavourable drilling outcomes. In this study, we looked at the characteristics of a low permeability geothermal production well exhibiting weak discharge metrics relative to other nearby wells. Fluctuations in the maximum discharge pressure of the target well have caused interruptions in its utilisation resulting in additional costs and downtime for the geothermal operator. The performance of the target well was analysed using sensor data from the surface facility, and a workflow based on systematic time-series feature engineering for improvedutilisation of the well was proposed. A predictive model was trained using the extracted time-series features and evaluated to forecast the occurrence of low discharge pressure events in the target well. The same dataset was then used to develop a reservoir model that captured the behaviour of the target well observed during post-drilling completion tests and while producing into the surface facility. |