Determination of the formation tops is an important and critical parameter while drilling a hydrocarbon well since it is one of the main factors affecting selection of the casing setting depths and drilling fluid design. During the field exploration and delineation phase and based on the geological data, the formation tops are estimated with low accuracy because of data limitations.

In this study, a potential alternative technique for predicting formation tops is introduced. This technique involves application of artificial neural networks (ANN) and the use of a combination of the drilling mechanical parameters and the rate of penetration (ROP) to provide an accurate prediction of the formation tops. Incorporating the drilling mechanical parameters in this technique is suggested to help in predicting the true increase or decrease in the ROP regardless of the fluctuation on the other drilling parameters.

Field data from two vertical oil wells (Well-A and Well-B) from the Middle East were used in this study. Seventy percent of the data from Well-A (4,436 data points) was used to train the ANN model, which was then tested on the remaining 30% of the data for Well-A (1,900 data points) and validated using the data from Well-B (6,569 data points).

The sensitivity analysis confirmed that using a ANN model that consists of 25 neurons, one hidden layer, and with the Levenberg-Marquardt backpropagation function as the training function, is the optimum for predicting the formation tops with correlation coefficients (R) of 0.94 and 0.98 for the testing and validation data of Well-A and Well-B, respectively. The developed ANN model showed high accuracy in estimating the formation tops for both the testing and validation datasets of Well-A and Well-B, respectively.

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