Speaker
Description
Traditionally, in faulted/sedimentary systems, geothermal injection wells are drilled away from known faults to reduce the risk of induced seismicity, e.g., in the VITO geothermal project in Mol, Belgium. Unless there is high layer/matrix permeability in the area, this can limit the well capacity and result in prohibitively high injection pressures that can induce high stress changes on smaller potentially unknown faults in the area. The use of Artificial Intelligence (AI) and Machine Learning (ML) techniques is growing rapidly in O&G and geothermal applications. AI techniques have been used for well placement studies in geothermal; for well flowrates in field management; for fractured reservoir structural characterisation; and for seismic risk analysis in EGS systems. Yet, in sedimentary geothermal systems, where financial margins are extremely slim, AI techniques have not been leveraged to maximize heat production while simultaneously reducing the seismic risk in prone areas.
As part of the HEU-URGENT project, we will develop a risk management toolbox to allow for more effective placement of geothermal wells in doublet systems to maximize heat production while minimizing the risk of induced seismicity. We seek to apply and extend the ML frameworks tested in previous works targeting O&G applications to a geothermal setting. The novel part of this analysis is in the inclusion of mechanical effects, with a focus not only on maximizing well output but also on minimizing the seismic impact. ML-based control may be needed to ensure that thermal stresses on faults remain below critical thresholds, as temperature change is mostly a non-reversible process. The balance between thermal and pressure stress on faults will have to be managed in the ML coupling to maximize heat output and minimize seismic risk. In this presentation, we present an update of the toolbox from the first 3 months of the HEU-URGENT project.
