Gap to Potential Identification through An Online Process Digital Twin

Abstract

Abstract An application has been deployed to identify the gap to potential and provide actionable intelligence to the plant operators on a real time hourly basis to maximize the value-added products and minimize energy consumption in a natural gas processing plant. An online digital twin based on thermodynamic models coordinating the underlying multisite Advanced Process Control (APC) applications, has been implemented enabling multisite gap to potential identification. Model Predictive Control has been for decades the de facto technology adopted in the oil and gas sector to optimize the process operations and improve yield. This falls short of identifying the unconstrained regions of optimization. A fit for purpose online thermodynamic model was developed for the entire NGL value chain. This is then mapped with the underlying APC limits as bounds. A data reconciliation algorithm coupled with a stead state detection algorithm creates first the exact replica or digital twin of the process plant. An optimization algorithm then identifies the true gap to potential and provides the advisory targets. The application solution is scheduled to run every hour mapping the real constraints of the plant in real time. This ensures that the optimized targets provided are realistic in nature and is within the safe operating limits of the plant. The online digital twin resulted in the following: ➢ A true representation of the plant with the mass and energy "balanced" across the plants on an hourly basis representing the true operation of the plant. This has resulted in the identification of suspected process measurements across the plant. ➢ This digital twin has enabled monitoring of the true representation of the different efficiency parameters like the polytropic efficiency of compressors, heat exchanger fouling etc., which can further be used as an input to Machine learning algorithms for the calculation of remaining useful life. ➢ Dynamic calculation of the energy consumption and setting up of the dynamic baseline as against static baseline for energy KPIs ➢ The optimization algorithm further provides advisory for identifying the optimized targets, which on implementation will enhance the yield with reduced specific energy consumption. ➢ Identification of the hardware constraints of the plant with estimated dollar value impact for unit change in that constraint enabling debottlenecking studies. This is a first of its kind implementation, in a geographically separated multisite operations. In an online environment, this digital twin can provide the optimized advisory targets, resulting in increased revenue and an attractive ROI. In an offline environment the same solution can be used for what if analysis and debottlenecking studies. Buoyed with the success of this project, there is a plan to scale up, integrating the upstream and downstream operation.