The objective of this paper is to demonstrate how drilling parameter optimization in real-time provides a drilling team with an Edge-system that can continuously improve performance and avoid problems without the need for subject-matter experts. An Edge-system based on cloud technology with Model based reasoning in Artificial Intelligence (AI) is made to give real-time and forward advice for operational parameters, see (Lahlou et al, 2021) for description. The key enabler for such system is "automatic" auto-calibration of models to be used for multiple forward-looking and what-if to find optimal drilling parameters within the well envelope ahead. A simplified configuration has been made so that the rig-team can operate and maintain the system without the need for subject matter experts. "Automatic" Auto-calibration at stable conditions and/or during ramping conditions removes the need for such experts. Results from testing of the Edge-system on multiple wells from several operators will be presented both related to automatic auto-calibration of real-time prediction models and for optimization of drilling parameters. As expected, a major challenge has been to design a calibration algorithm that improves accuracy of calculations without being kicked out by any data quality issues, and without masking upcoming actual anomalies like kicks, losses and issues related to hole cleaning. This challenge has been approached by using a combination of time-delayed robust calibration methods and testing on a comprehensive set of data from diverse operations.

As part of the digital transformation in oil and gas industry, well construction move toward new efficient methods using digital twins of the wells. This paper will highlight how the drilling operations are monitored, how a digital twin of the well is utilized and how learnings are implemented for future wells. A Digital Twin is a digital copy of assets, systems and processes. A Digital Twin in drilling is an exact digital replica of the physical well during the whole drilling life cycle. Its functionality is based on advanced hydraulic and dynamic models processing in real time. By utilizing real-time data from the well, it enables automatic analysis of data and monitoring of the drilling operation and offer early diagnostic messages to detect early signs of problems or incidents. In the current study various actual operational cases will be presented related to different wells. This includes using digital twin during drilling under challenging circumstances such as conditions when using MPD techniques. Also, various diagnostic messages which gave early signs of problems during running in the hole, pulling out of the hole and drilling will be presented. High restrictions were detected using comparisons of real-time values and transient modelling results. These will be discussed. Different real cases have been studied. Combining digital RT modelled and real-time measured data in combination with predictive diagnostic messages will improve the decision making and result in less non-productive time and more optimal drilling operations.

An Advanced Drilling Simulation and Engineering Center has been established in Tianjin in China. This center has been set up with simulation systems for the whole life cycle of well construction; from planning to training to operation / Real Time and to Post-analysis. A comprehensive integrated downhole and topside training simulator enables training the crews on a true representation of the real well to be drilled. This behaves like a "Drilling Flight Simulator", and expected problems downhole can be trained on. Furthermore, the center incorporates an advanced planning system with dynamic models for drilling, cementing, displacements and other well related hydraulics as well as multi-phase flows during well control. Also a state-of-the-art drill string and torque & drag model is included. A Real-Time simulation and decision support module provides support during drilling. The drilling simulation is driven by real time data from the ongoing operation itself, and includes a 3D visualization of the downhole drilling in a "virtual well". Automatic look-ahead simulations of ECD and temperatures was performed with the calibrated models on the fly as support for decisions. A unique feature of the Simulation Applications is that the modelling basis and simulation models are the same for planning, for training, for Real Time and for Post-analysis. These are advanced transient state-of-the-art models which have been verified versus numerous well types (HPHT, DW, MPD etc). In preparation, an advanced training session was performed using a dynamic downhole training simulator linked to a topside rig simulator, for training of the drilling teams. Prior to operation data transfer was established so that during the operation simulations was performed in real time using the ongoing operational parameters as input. This paper will first present the Advanced Drilling Simulation and Engineering Center, then elaborate on the use of this in the planning phase, the training phase and the operational phase in well cases in the South China Sea. Experiences from this will be presented, and learnings for future operations will be discussed. Applications of unique technologies contributed to safer and more cost-effective drilling.

The objective of this paper is to demonstrate how advancedrealtime monitoring (ARM) utilizing advanced hydraulic and mechanical modelling of the drilling process provided early detection of anomalies by giving diagnostic messages during drilling operations. These achievements can minimize non-productive time and invisible lost time and maximize the benefits and value of operations; if they are utilized to its full potential by operations. Some well cases are used to illustrate the methodology and its results. Among problems diagnosed are losses, stuck pipe during drilling and casing running, downhole equipment leakage and improper hole cleaning. In some cases, action was taken based on the diagnostics; and the operational conditions were modified to mitigate the situation. In other cases, the warnings were not taken seriously, the situation worsened until the problem was irreversible and a stuck situation occurred. In one well presented in the current study a stuck pipe situation happened during drilling 8½" section which led to a downtime of more than 20 days. By utilizing the ARM, it couldhave been possible to detect some early signs of the stuck conditions in the wellbore and avoid it. Another stuck situation in awell during 14-inch Casing running, led to downtime of more than 10 days which involved breaking out the casing above the stuck point and performing P&A. The ARM provided early signs of stuck casing that was about to occur, and these signs first started appearing about 15 hours before the pipe got completely stuck. This paper will present the Advanced realtime Monitoring ARM System and the modelling behind this. Also, the plans for further implementation and integration of this in the work processes will be discussed, before results from the first year of utilization will be presented with examples.

One of the most common issues faced in the Drilling industry is a Stuck Pipe situation. Stuck Pipes lead to huge losses in cost, energy and productive time. The objective of this paper is to predict stuck pipe incidents prior to their occurrence by harnessing the power of machine learning and deep learning models. Stuck incidents are some of the most difficult and challenging situations. The proposed method uses a two-step model and available historical data from prior drilling operations to predict the occurrence of such an incident well in advance so that it can be avoided. The first step of the model performs time series predictions using a Recurrent Neural Network (RNN) with Walk Forward Validation. The second part of the model uses a Random Forest Classifier to classify the predictions from the previous step and determine if a stuck situation is likely. The classification model is pre-trained using historical drilling operational data collected from old wells. It is possible to determine how far back the model should look at the data and how far ahead it should predict. For the purpose of this paper, the model looks back 5000 timesteps and predicted 3000 timesteps ahead. One timestep is 2 seconds in this case. This model would be able to reduce nonproductive time, and help make drilling operations cleaner, faster and greener. The biggest benefit of such a model is that it can learn on the go and would not require manual intervention. Also, the model can upgrade and modify itself to changes in the drilling operations.