Computational working fluid design under property uncertainties: Application to organic rankine cycle

We present a methodology to integrate fluid property uncertainty (e.g. uncertainty in GC property prediction models) into a computer-aided molecular design (CAMD) framework to identify novel working fluids for thermodynamic cycles. The study is highlight through an organic Rankine cycle (ORC) system for heat recovery from marine diesel engine. The CAMD problem is formulated as an optimization problem that identifies the best chemical structure giving the highest power output of the ORC. The uncertainties of the GC factors are obtained using covariance-based uncertainty analysis. Monte Carlo sampling within the respective uncertainties of the GC factors provides different sets of GC factors. These sets of GC factors are used separately as constraints to the optimization problem, which is solved as a scenario-based optimization. The best performing fluids are further evaluated in the cycle model and their respective property uncertainty values are propagated through the ORC using Monte Carlo based error propagation. In this way a list of working fluid candidates is provided including the new power output and its corresponding uncertainty with 95% confidence. The working fluid candidates can be assessed and selected including their respective property uncertainty.