| Popis: |
Many reservoir problems require multiple regression solutions. As the training set is often small in practice, it is important to quantify the relative relevance of all input attributes (or drivers) and select only the significant ones. This paper compares the performance of two techniques for ranking drivers for a fractured reservoir characterization problem. The techniques are rank order correlation and fuzzy curves. The case study contains 49 geological and geophysical drivers linking to a fractured intensity indicator. The results of the case study show that the fuzzy curve method has many practical limitations in the presence of noisy data and poorly populated data regions, and its ranking results are less conclusive. On the other hand, the rank correlation technique gives more satisfactory results and is more robust. An improved version of rank correlation technique, namely “optimized piecewise rank correlation” (or OPRAC), is also presented. |
Full Text from ScienceDirect