Swab and surge pressure fluctuations are decisive during drilling for oil. The axial movement of the pipe in the wellbore causes pressure fluctuations in wellbore fluid; these pressure fluctuations can be either positive or negative, corresponding to the direction of the movement of the pipe. For example, if the drill string is lowering down in the borehole, the drop is positive (surge pressure), and if the drill string is pulling out of the hole, the drop is negative (swab pressure). The intensity of these pressure fluctuations depends on the speed of the lowering down (tripping in) or withdrawing the pipe out (tripping out). High tripping speed corresponds to higher pressure fluctuations and can lead to fracturing the well formation. Low tripping speed leads to a slow operation, causing non-productive time, thus increasing the overall well budget. Researchers used mathematical equations and physics to understand the phenomena and have provided many empirical, mathematical, and physics-based models. This paper starts with a literature study on the swab and surge pressures. After that, this paper concludes with a proposal for a new approach. The new approach proposes developing new models that are more robust, using field data, as we have access to field data from drilling operations. Research using field data would provide data-driven methodologies as new solutions for the rate of penetration, reservoir management, and drilling optimization. The expected outcome will improve the performance of the tripping in and tripping out process within drilling and well construction, and will further reduce the risk related to swab and surge pressures.

During drilling operations for oil and gas, swab and surge pressure occur while tripping in and out of a wellbore. High tripping speed can lead to fracturing the well's formation, whereas low tripping speed can increase non-productive time and cost. Hence, there is a need to predict surge/swab pressure accurately. Several analytical and machine learning models have already been developed to predict surge/swab pressure. However, these existing models use numerical calculations to generate the data. This paper explored four supervised machine learning models, i.e., Linear Regression, XGBoost, Feedforward Neural Network (FFNN), and Long-Short-Term Memory (LSTM). In this study, actual field data from the Norwegian Continental Shelf provided by an Exploration & Production company is utilized to develop the four machine learning models. The results indicated that XGBoost was the best-performing model with an R2-score of 0.99073. Therefore, this trained model can be applied during a tripping operation to regulate tripping speed where repetitive surge/swab pressure calculation is needed.

The axial movement of pipe in and out of the well generates positive (surge) and negative (swab) pressures that will impact the well pressure. When the swab and surge effects cause well pressures outside the allowable operational limits, wellbore instability (well collapse/well fracture), kick, and induced drill string sticking issues will occur. The problems increase the operational and nonproductive time-related costs. Consequently, the drilling budget rises significantly. It is therefore, imperative to predict the differential pressures in order to mitigate the problems. Even though several models have been developed in the past, models work for the considered experimental setup and conditions. In this paper, a simple analytical model was derived for eccentric/concentric annuli. The fluid rheological behaviors were assumed to be described by power law and yielded power-law. The model is derived based on a steady state condition, and the effects of tripping speed, the power-law fluids, the yield-power-law fluids rheological parameters, and well geometries are considered. The model is compared with experimental data from the literature and with the existing model. Parametric sensitivity studies have been conducted. Results show that the model prediction exhibited quite good performance, with an average percentile error deviation of 9.9% and 6.2% for the power-law and yield-power-law fluids, respectively. However, more testing is required to determine the model’s limitations and application.

Drilling operation is the major cost factor for the industry. Appropriately designed operations are essential for successful drilling. Optimized drilling operations also allow enhanced drilling performance and reduce the drilling cost. This is achieved by increasing the bit life (minimizing premature bit wear), drilling faster that allows reducing drilling time, and also reducing tripping operations. The paper presents two parts. The first part compares the parametric physic-based swab and surge simulation results obtained from Bingham Plastic, Power Law, and Robertson and Stiff models. The aim is to show how the model predictions deviate from each other. Two 80-20 Oil Water ratios (OWR) and two 90-10 OWR oil-based drilling fluids with 1.96 sg and 2.0 sg were considered in vertical- and deviated wells. Simulation results analysis revealed that the deviations depend on the drilling fluid’s physical and rheological parameters as well as the well trajectory. Moreover, the model’s predictions were inconsistent. Data-driven machine learning (ML) modeling makes up the second section. Data-driven modeling was done using both software-generated datasets and field datasets. Results show that the Random Forest Regressor (RF), Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), LightGBM, XGboost, and Multivariate regression models predict the training and test datasets with higher R-squared and minimum root means square values. Deploying the ML model in real-time applications and the planning phase would have the potential application of artificial intelligence for well planning and optimization processes.

Drilling operations are the major cost factor for the oil industry. Appropriately designed operations are essential for successful drilling. Optimized drilling operations also enhance drilling performance and reduce drilling costs. This is achieved by increasing the bit life (minimizing premature bit wear), drilling more quickly, which reduces drilling time, and also reducing tripping operations. This paper is presented in two parts. The first part compares the parametric physics-based swab and surge simulation results obtained from the Bingham plastic, power law, and Robertson–Stiff models. The aim is to show how the model’s predictions deviate from each other. Two 80:20 oil/water ratio (OWR) oil-based drilling fluids and two 90:10 OWR oil-based drilling fluids, 1.96 sg and 2.0 sg, were considered in vertical and deviated wells. Analysis of the simulation results revealed that the deviations depend on the drilling fluid’s physical and rheological parameters as well as the well trajectory. Moreover, the model’s predictions were inconsistent. Data-driven machine learning (ML) modeling is the focus of the second section. Data-driven modeling was performed using both software-generated datasets and field datasets. The results show that the random forest regressor (RF), artificial neural network (ANN), long short-term memory (LSTM), LightGBM, XGBoost, and multivariate regression models predicted the training and test datasets with higher R-squared and minimum mean square error values. Deploying the ML model in real-time applications and the planning phase would lead to potential applications of artificial intelligence for well planning and optimization processes.

This paper aims to extract a mathematical model from real-time data obtained during drilling operations in the oil and gas industry. The algorithm employed for model extraction is the Group Method for Data Handling (GMDH). The study utilizes data from tripping and drilling operations conducted on the Norwegian Continental Shelf, which offers vast drilling data. The data is generated by multiple sensors positioned along the wellbore, transmitting information to the surface through a wired drill pipe telemetry system. Despite the significant costs associated with generating and transmitting real-time drilling data through wired systems, much of it remains unused. This paper begins by demonstrating the data acquisition, collection, visualization, cleaning, statistical summary, and partitioning process. Subsequently, it introduces the GMDH algorithm for mathematical model extraction. Additionally, a software tool named GMDH-Oil is developed and presented. Finally, the paper includes experimentation involving the application of the tool to real-time data, along with the corresponding results obtained.