Enhancing Digital Twin Accuracy for Energy Efficiency and Decarbonization

Abstract

Abstract The development of Production Optimization Models, exemplified by Mixed Integer Non-Linear Programming Models, aims to optimize process parameters within a network of Gas-Oil Separation Plants, ultimately resulting in minimal energy consumption while achieving production targets. The efficacy of these models hinges upon the accuracy of digital twins for compressors and pumps. This paper presents the methodology employed to ensure that the model reflects actual equipment performance. Additionally, the paper outlines a plan for introducing Artificial Intelligence (AI) into the process model to further validate and optimize production. A process model was established for each major energy user of significance by employing the design performance curves. The model's outcomes, which included pressures, temperatures, and power consumption, were subsequently compared to actual process data retrieved from the data historian and adjusted accordingly. This was particularly crucial for equipment that exhibited considerable deviation from the original design. In some facilities, flow, pressure, or temperature data were inconsistent, prompting the calibration of the model to align it with the process data of the highest level of confidence. Additionally, process data was utilized to estimate the facility's backpressure in the absence of a hydraulic model. Changes made to the system were validated by examining the accuracy of the system curves, which were accomplished by matching predicted outcomes with actual results. The implementation of precisely calibrated digital twins within the process model significantly improved the Production Optimization Models’ ability to forecast the ideal distribution of production. The model's output corresponded closely with the actual energy consumption of the facility, resulting in considerable savings in energy consumption and costs, as well as a reduction in CO2 emissions. As equipment performance is known to evolve over time due to factors such as shutdowns, maintenance, and frequency of changeover, an Artificial Intelligence-based system is being developed to validate the process model and optimize it over time. Various process and equipment parameters are known to affect performance, and these will be taken into account during system development. The process of calibrating process models through the utilization of simulation software to generate a digital twin is a critical stage in guaranteeing the accuracy of the production optimization model. This paper presents a novel approach that incorporates any observed deviations in the digital twin's performance. Furthermore, the paper introduces the concept of using Artificial Intelligence to validate the performance of rotating equipment.