Improved Simulation Efficiency for Optimising Inflow Control Device Completions Using Suitably Upscaled Models and Parallel Global Search Optimisation Techniques

Inflow Control Device (ICD) completions can improve well performance by adjusting the inflow profile along the well and reducing the influx of unwanted fluids. The ultimate aim of using ICD completions is to provide maximum oil recovery and/or Net Present Value (NPV) over the life of the field. Proactive ICD optimisation studies use complex reservoir models and high-dimensional nonlinear objective functions to find the optimum ICD configurations over the life of the field. These complex models are generated from fine scale detailed geological models to accurately capture fluid flow behaviour in the reservoir. Although these high-resolution geological models can provide better performance predictions, their simulation runtimes can be computationally expensive and time consuming for performing proactive ICD optimisation studies that often require thousands of simulation runs. We propose a new workflow where we use upscaled and locally refined models coupled with parallelised global search optimisation techniques to improve the simulation efficiency when performing ICD optimisation and decision-making studies. Our approach preserves the flow behaviour in the reservoir and maintains the interaction between the reservoir and the well in the near wellbore region. Moreover, when coupled with parallel optimisation techniques, the simulation time is significantly reduced. We present an in-house code that couples global search optimisation algorithms (Genetic Algorithm and Surrogate Algorithm) with a commercial reservoir simulator to drive the ICD configurations. We evaluate the NPV as the objective function to determine the optimum ICD configurations. We apply and benchmark our approach to two different reservoir models to compare and analyse its efficiency and the optimisation results. Our analysis shows that our proposed approach reduces the run time by more than 80% when using the upscaled models and the parallel optimisation techniques. These results were based on a standard dual-core parallel desktop configuration. Additional results also showed further reduction in run time is possible when employing more processors. Additionally, when testing different ICD completion strategies (ICDs in producers only, ICDs in injectors only, and ICDs in both producers and injectors), the NPV can be increased by 9.6% for the optimised ICD completions. The novelty of our work is rooted in the much-improved simulation efficiency and better performance predictions that supports ICD optimisation and decision-making studies during field development planning to maximize profit and minimize risk over the life of the field.

Validation of Performance-Based Specifications for Surface Asphalt Mixtures in Virginia

The Virginia Department of Transportation (VDOT), like many owner agencies, is interested in ways to facilitate the increased durability of asphalt mixes in an effort to make its roadway network more sustainable, longer lasting, and more economical. The balanced mix design (BMD) method proposes to address this through the incorporation of performance criteria into mix design and acceptance. VDOT has committed to the implementation of the BMD method in an effort to improve asphalt mix performance. The purpose of this study was to continue advancing efforts toward implementation of BMD through the evaluation of 13 asphalt mixes using performance-indicating laboratory tests, validation of the initial performance tests selected for BMD use, and validation of the initial test threshold criteria. Based on the results, the asphalt pavement analyzer (APA) rut test, indirect tensile cracking test (IDT-CT), and Cantabro test were found suitable for continued use in BMD. The current threshold criteria for all three tests were found reasonable based on additional mix testing. The study recommends that APA rut test and IDT-CT results should be compared and correlated to fundamental rutting and cracking tests, respectively, as well as to performance predictions obtained from mechanistic-empirical pavement design simulations, and to field performance for full assurance that test threshold values are appropriate. It was further recommended to evaluate the Cantabro, IDT-CT, and APA rut tests to determine acceptable variability and establish precision statements.

Implementation of machine learning techniques for the quasi real-time blind and electric lighting optimization in a controlled experimental facility

Machine Learning techniques have been recently investigated as an alternative to the use of physical simulations, aiming to improve the response time of daylight and electric lighting performance-predictions. In this study, daylight and electric lighting predictor models are derived from daylighting RADIANCE simulations, aiming to provide visual comfort to office room occupants, with a reduced use of electric lighting. The aim is to integrate an intelligent control scheme, that, implemented on a small embedded 32-bit computer (Raspberry Pi), interfaces a KNX system for a quasi-real-time optimization of the building parameters. The present research constitutes a step towards the broader goal of achieving a unified approach, in which the daylight and electric lighting predictor models would be integrated in a Model Predictive Control. A verification of the ML performance is carried-out by comparing the model predictions to data obtained in monitoring sessions in autumn, winter and spring 2020-2021, resulting in an average MAPE of 19.3%.