Towards digital twin modeling of a waste heat recovery organic rankine cycle system

Abstract

Digitalization and climate change mitigation are key drivers of energy sector research aiming to enhance energy production and management. Industrial energy consumption accounts for nearly one-third of greenhouse gas (GHG) emissions, with 50% of generated energy dissipated as waste heat. Achieving carbon neutrality demands optimizing conventional methods, integrating renewable energy sources, and recovering waste heat. Waste Heat Recovery (WHR) is a practical and cost-effective approach to improving energy efficiency, and the Organic Rankine Cycle (ORC) is a leading technology for converting waste heat to electricity. Digital tools like Digital Twins can increase energy efficiency by enabling cost-effective, flexible, and reliable power plant operation.

This dissertation focuses on developing a digital twin prototype for a micro-scale ORC facility that uses octamethylsiloxane (MDM) as the working fluid to recover high-temperature waste heat. The study begins by characterizing the ORC experimental facility to analyze the cycle and turbogenerator behavior under varying operating conditions. Using an electrothermal analogy, a novel reduced-order transient model of the heat exchangers was developed to model the thermal inertia effects in the heat exchangers and validated against experimental data.

MDM’s sub-atmospheric condensing pressures led to air infiltration, resulting in noncondensable gases (NCGs) on the low-pressure side. The dissertation evaluates the impact of NCGs on condensate tank measurements using an analytical model to quantify measurement uncertainties for pure fluid assumption. It analyzes ways to improve the cycle modeling accuracy to account for the effect of NCGs during operation. Finally, the digital twin prototype was realized using a hybrid approach, integrating transient experimental data (physical twin) with a physics-based dynamic model of ORC components and the analytical model (virtual twin). Furthermore, the dissertation discusses the digital twin’s prediction, predictive maintenance, and control potential through different use cases.