Operational flexibility-oriented selection of working fluid for organic Rankine cycles via Bayesian optimization

Abstract

Working fluid selection is a crucial part of organic Rankine cycle (ORC) designs. Traditional selection methods primarily focus on optimizing performance under specific nominal operating conditions, often neglecting potential efficiency losses and feasibility issues that may arise under off-design conditions due to fluctuations in the heat source and sink. This research introduces a novel method for optimizing working fluid selection to achieve robust and efficient operation in the face of environmental variations. Specifically, operational flexibility is analyzed based on the ORC operational model to capture performance deviations from nominal conditions, and is quantified by evaluating the size of the feasible operational region within the uncertain parameter space. Working fluid selection is optimized simultaneously with the cycle configurations, resulting in a computationally challenging mixed-integer nonlinear programming (MINLP) problem, which is addressed through Bayesian optimization. A case study on geothermal brine heat recovery with a recuperative ORC compares flexibility-oriented and conventional working fluid selections, demonstrating a 102% increase in operational flexibility at the cost of an 11.5% efficiency loss. This research underscores the significant impact of working fluid selection on operational flexibility and demonstrates the effectiveness of Bayesian optimization in solving complex MINLP problems for integrated molecule-level and process-level designs.