Artificial neural network assisted prediction of dissolution spatial distribution in the volcanic weathered crust: A case study from Chepaizi Bulge of Junggar Basin, northwestern China

摘要

The reservoir rocks in the volcanic weathered crust are characterized by plentiful dissolution pores, caverns and fractures. However, as an essential control factor of the reservoir distribution, volcanic dissolution is often ignored in quantitative studies. In this paper, we propose a complete artificial intelligence workflow to model and predict the volcanic dissolution distribution by integrating the image logging semi-quantitative analysis with seismic prediction. A synthetic reservoir property named DCV (Dissolution Comprehensive Values) is generated as logging curves from XRMI resistivity pseudo-pictures to indicate the intensity of dissolution. These curves are then used as the target for the seismic-based prediction in an artificial intelligence framework. A multi-layer artificial neural network (ANN) model is constructed in order to map the seismic attributes into the DCV. The methodology is demonstrated on a real case study from the Junggar basin in northwestern China. The maps and profiles from the prediction represent the dissolution values in the volcanic weathered crust. The predictions are consistent with the geological volcanic dissolution model provided by geological knowledge. It is concluded that the spatial dissolution distribution in volcanic weathered crust can be reliably predicted by this integrated method.