Time-lapse inversion of Controlled Source Electromagnetics using vertical sources and receivers

Abstract

Knowledge of spatial and temporal distribution of fluids in the subsurface is crucial in a wide range of applications. During the production of crude oil typically high saline produced formation water is injected into the reservoir layer, aiming to push the oil towards production wells. While oil is commonly seen as an electrical insulator, the injected saline brines are characterised by low electrical resistivity. Thus, electromagnetic (EM) methods and especially Controlled Source Electromagnetics (CSEM) attracted an increasing interest to monitor these resistivity changes inside the reservoir over time. This thesis mainly reports on numerical aspects of modelling and inversion of land based CSEM with particular focus towards hydrocarbon monitoring applications. Most of the presented developments were inspired by a superordinate research project including CSEM field surveys across an actively producing onshore oil field in Northern Germany. In producing oil fields there exists a large number of steel-cased wells. Such existing oil field infrastructure and especially the presence of metal casings significantly alters the propagation of EM fields in the subsurface. Their spatially unfavourable dimensions effectively prohibits a straightforward implementation into the modelling grid. Thus I developed a new modelling approach allowing consideration of such thin but vertically extended highly conductive structures including their mutual interaction. The developed methodology had been implemented into existing modelling and inversion codes. Using the new approach to investigate the influence of metal casings on CSEM data shows that they act as additional inductively coupled vertical electric dipole sources at depth and thereby increase resolution capabilities at depth. The presence of metal casings can thus be exploited by optimising the source receiver layout in such a way that the strength of these additional vertical dipole sources is maximised. An additional working package of the superordinate project was the measurement of vertical electric fields in a shallow observation well. However, measurements of vertical electric fields requires long measurement dipoles to achieve satisfactory signalto-noise ratios. Such extended dipoles span several modelling cells and are therefore in conflict with assumptions usually made for modelling, that receivers can be represented as point dipoles. I therefore expanded the modelling and inversion codes to consider the physical receiver dimensions. The new algorithm implicitly considers imperfect alignment of the receiver with the corresponding field component. Without the consideration of this effect inversion of vertical electric field measurements is likely to cause erroneous results. Finally I discuss different aspects of time-lapse inversion required to track changes in fluid saturation over time. The cascaded inversion scheme is applied to synthetic timelapse data for a simplified oilfield undergoing brine flushing. The influence of various inversion parameters in particular different regularisation techniques are examined. Surface based sources and receivers typically provide low sensitivity towards deep targets in highly conductive backgrounds. Despite that using additional constraints, in particular a model weighting scheme together with energised steel casings allowed to track resistivity changes inside the reservoir based on synthetic time-lapse data.