Fredy Alex Villaorduña Artola

Neural net ensembles for lithology recognition

Lithology recognition is a common task found in the petroleum exploration field. Roughly speaking, it is a problem of classifying rock types, based on core samples obtained from well drilling programs. In this paper we evaluate the performance of different ensemble systems, specially developed for the task of lithology recognition, based on well data from a major petroleum company. Among the procedures for creating committee members we applied Driven Pattern Replication (DPR), Bootstrap and ARC-X4 techniques. With respect to the available combining methods, Averaging, Plurality Voting, Borda Count and Fuzzy Integrals were selected. The paper presents results obtained with ensembles derived from these different methods, evaluating their performance against the single neural network classifier. The results confirm the effectiveness of applying ensembles in real world classification problems.

Time-lapse critical reflection: does it really work in seismic monitoring of low porosity and high effective stress conditions?

We present an evaluation of the time lapse critical reflection method from the point of view of sensitivity and uncertainty analysis. The purpose of this analysis is to establish the link between changes in the p-wave velocity in the reservoir, due to fluid substitution, and the critical distance (XC), a dynamic parameter. Two static parameters of the overburden are considered in the analysis: its thickness and its effective or equivalent P-wave velocity. Stochastic uncertainty analysis by means of Monte Carlo simulations was carried out to determine the sensitivity of XC to velocity contrast between that of the reservoir and the corresponding one to the overburden (incident medium). Results show that the effect of the velocity values in each medium, overburden and reservoir (reflecting medium), on estimates of XC depends on the velocity contrast between the two media. It turns out that the greater the velocity difference between the two media, the greater the effects associated with the reservoir P-wave velocity. On the other hand, the dependence of XC on overburden velocity (Vrms of the incident medium) is just the opposite. From the inversion viewpoint, predicting changes of the reservoir P-wave velocity from changes in critical distance is strongly dependent of the uncertainty in the initial estimate of the reservoir P-wave velocity, followed by degree of importance by the uncertainty in the overburden velocity. Other sources of uncertainty in this analysis turned out to be negligible. Geomechanical effects have not been accounted for in this analysis.

Time-lapse seismic modeling assisted by numerical reservoir simulation of water and gas flooding scenarios in oil reservoirs

In this project, we link fluid flow simulation results to time-lapse seismic through rock physics and modeling. Our goal is to examine the main effects of permeability barriers on seismic response using fluid flow simulations to generate pressure and saturation fields. To explore this problem, we have carried out water and gas injection numerical experiments in a simple reservoir model which has vertical and horizontal variations of porosity as well as permeability barriers. In each experiment, we change the barrier permeability values. By using fluid substitution theory, Gassmann and patchy models, and Batzle and Wangs empirical relationship we model the main seismic parameters, such as acoustic impedance and compressional velocity. After that, we generate synthetics seismograms and some contrast sections to compare the seismic images prior and after fluid injection events in subsequent time periods to analyze possible differences in the seismic parameters due to changes in barriers properties. The results show that barriers can increase fluid pore pressure changing the bulk modulus in the regions with barriers. The results also show that water flow does not have significant impact on seismic response when the barrier is present. On the other hand, gas flow and the degree of impermeability of the barrier can help us to understand how the barriers act on the seismic response and thus to reduce uncertainty. Finally, this paper presents a methodology to examine barriers effects on seismic response.

Time Lapse Critical Reflection: Sensitivity And Uncertainty Analysis

We present an evaluation of the time lapse critical reflection method from the point of view of sensitivity and uncertainty analysis. The purpose of this analysis is to establish the link between changes in the p-wave velocity in the reservoir, due to fluid substitution, and the critical offset xc (a dynamic parameter). Two static parameters of the overburden are considered in the analysis: its thickness and its velocity (vRMS). Stochastic uncertainty analysis by means of Monte Carlo simulations was carried out to determine the sensitivity of xc to velocity contrast. Results show that the effect of the velocity values in each medium, overburden and reservoir, on estimates of xc depends on the velocity contrast between the two media. It turns out that the greater the velocity difference between the two media, the greater the effects associated with the reservoir P-wave velocity. On the other hand, the dependence on overburden velocity is just the opposite. From the inversion viewpoint, that is, predicting changes of the reservoir P-wave velocity from changes in critical offset is a strongly dependent of the uncertainty in the initial estimate of the reservoir Pwaver velocity, followed by its degree importance by the uncertainty in the overburden velocity. Other sources of uncertainty in this analysis turned out to be negligible. Geomechanical effects have not been accounted for in this analysis.