Real-Time GR logs Estimation While Drilling Using Surface Drilling Data; AI Application

Gamma-ray logging (GR) is one of the most crucial measurements to evaluate oil and gas reservoirs and identify the formation lithology. Logging while drilling (LWD) offers direct downhole measurements. LWD tools being placed a considerable distance above the drill bit which might result in a measurement of already penetrated formations. In this study, two artificial intelligence (AI) techniques, including support vector machine (SVM), and random forests (RF) were applied to predict a synthetic GR log using surface drilling parameters. A total of 4609 data entries from three wells in the Middle East were used to train, test, and validate the models. The data from wells 1 and 2 were used to build the AI models. Unseen data points from well 3 were then used to validate the model. The performance of the models was assessed in terms of average absolute percentage error (AAPE) and correlation coefficient (R). Results showed that both SVM and RF-produced models were able to predict the GR log with high accuracies. SVM slightly outperforms RF in prediction GR logs with R of 0.99 and AAPE of 0.34% in the training set, and with R of 0.98 and AAPE of 1.49% in the testing set. For the validation, SVM predicted GR log with R and AAPE of 0.98, and 1.42%. The presented models assist drilling engineers to real-time predict GR log and identify the formation lithology while the bit drilling the same formation.