Hydrocarbon Show Database: Using Algorithms for Faster and Objective Analyses (2019)

Dit internship rapport is geschreven door oud-stagiair Victor Larède.

For the last 2 years EBN has been compiling the hydrocarbon (HC) show database. HC shows are defined as significant occurrences of HC gases or fluids in combination with favourable lithology (i.e. rocks that can be produced from). The database in assembled from Log-, (SW-)core and Test data. One of the methods of filling this database is the ‘Manual input’ method (QC1) where a person analyses and processes all the data. Filling the database using this method will take at least another 3 years to with the current number of relevant boreholes in the Netherlands. In this internship project, an attempt has been made into automating the gas log analysis part from QC1 by using computer algorithms. A code has been written which can perform the gas analysis workflow used in QC1. Furthermore, an investigation has been done to find a decent method for determining/calculating the background gas signal. A dataset has been analysed both manually and automatically. Assuming that the manual analysis is perfect, a misfit through an RMS error has been calculated for the automated methods to compare how successful the different methods are. The ‘drawn’ background gas method performed the best with an RMS-error of 0.31. the automated analysis using this method takes around 10-15 minutes. The project has resulted in tool (GAT) with a user-interface that can be used to perform an automated gas log analysis with major time savings. To find a correct manner to calculate the background gas signal more testing and research is needed.