ABSTRACT

The characterisation of formation heterogeneities requires a multidisciplinary study of data acquired using a large number of numerical geophysical and geological measurements and a rigorous evaluation of the precision and accuracy of the data. Another essential aspect of the appraisal of any measurement is the quality assessment and quality control of the data. In this work multivariate statistical techniques and an artificial neural network (ANN) are used to identify heterogeneities in complex formations and to evaluate the boundaries they generate. The precision and accuracy of the data from different sources are very important and are considered here by using sample support in the integration of measurements at different scales. We use examples from two wells to show the differences in characterisation obtained with each technique. Multivariate statistical analyses are initially used to group the petrophysical, geophysical and geological parameters extracted from the downhole measurements into distinct geologically definable zones. This technique has the advantage of being quasi-independent of any predetermined ideas we have about the whole data set, and has shown to be very reliable in formation characterisation. Thus the result obtained here is a useful basis for comparison with that obtained from the neural network. An artificial neural network is used to characterise the different lithology sequences present in each well. Neural networks are relatively new tools and have proved very useful in applications where conventional computing methods are inadequate. Another application is the possibility of determining quantitative petrophysical parameters from wireline logs and core data in uncored intervals. We also show the ability of the neural network to construct petrophysical logs. The results are presented as a comparison between the two techniques. We show that both methods are very encouraging. When comparing the ANN derived petrophysical parameter logs with actual core measurements we see a good match. Low quality petrophysical measurements can be determined by a mismatch between the responses.

This content is only available via PDF.
You can access this article if you purchase or spend a download.