In the oil production it important the monitoring of the reservoir parameters (permeability, porosity, saturation, pressure, etc) for its management. Changes in the reservoir dynamic parameters induce variations in reservoir flow, as for example, losses in the pressure, making it difficult the process of extraction of the oil. The fluid injection increases the internal energy and pressure of the reservoir, stimulating the movement of the oil in the direction of the extracting wells. The crosswell electromagnetic method can become in efficient technique in monitoring the injection processes, considering the fact that the percolation of conductive fluid through the sediment is a very sensitive. This thesis presents the results of a very efficient algorithm of electromagnetic tomography applied to synthetic data. The imaging scheme assumes a cylindrical symmetry around a source consisting of a magnetic dipole. During the process of imaging we used 21 transmitters and 21 receivers distributed within two wells 100 meters apart. For the forward problem solution it was used the finite element method applied to the Helmholtz equation for the secondary electric field. It will be demonstrated that the algorithm obtained is not under restrictions imposed by Born and Rytov approximations, therefore, the algorithm can be efficiently applied for any situation as a electric conductivity contrasts as large as 2 to 100, frequencies as 0.1 to 1000.0 kHz and scatterers of any dimensions. The inverse problem was solved using the stabilized Marquardt scheme. This scheme employs a technique that seeks the solution iteratively. The inverted synthetic data, with added Gaussian noise, are the magnetic vertical component, separated in its respective real (in-phase) and imaginary (quadrature) parts. Without constrains, the inverse tomography problem is totally unstable. To stabilize the inverse solution, absolute and relative constraints have been used. The use of these constraints allows producing high definition images. The results show that the resolution is better in the vertical direction than the horizontal and it is also a function of source operating frequency. The position and attitude of the target are recovered well. These results show that constraints can attenuate or eliminate the poor resolution.
|
