EARTH

[31013-24-12] Well Log Correlation in Presence of Local Variations

○Chanchai Apiwatsakulchai¹, Suwat Athichanagorn¹ (1. Chulalongkorn University)

Chairman: ：Dawan Wiwattanadate （Chulalongkorn University）

Keywords:Well Log Correlation, Similarity Measurement, Data Representation, Symbolic Aggregate approXimation (SAX)

Subsurface exploration is known to be an ever-challenging activity requiring an inter-disciplinary body of knowledge. Varying techniques are commonly used in the data acquisition process in order to pinpoint a section of interest, which may be within a large area. Well logging, for example, is a data acquisition technique shared in both mineral and hydrocarbon exploration. The technique is commonly performed using multiple wells so as to achieve a large area coverage. Therefore, a systematic approach of well log correlation is needed. This study presents an integrated process for well log correlation in presence of local variations typically found in naturally recorded signals. To exploit any benefits given by different treatments of the signals, correlations are performed in two spaces: discrete and continuous spaces. The discrete space data representation is performed using Symbolic aggregate ApproXimation (SAX) algorithm while the continuous space is the data’s original space (real-valued data space). To assess suitability of all prospective correlations, quantitative similarity measurement is then performed on every sliding-window-based signal subsequences and the resulting similarity scores are ranked. Finally, the best correlation for a given section of interest is determined from a combined treatment of continuous and discrete-based similarity measurement. To demonstrate the effectiveness of this proposed process, real field signals were used in example correlations. This integrated process helps formalize well log correlation through the use of quantitative measurement, thereby reducing ambiguity in the correlated result frequently arising when a conventional approach is used. This process is also resilient to signal with local variations as the variations have minimal effects to the final result when appropriate parameters are used.