Comparing optimization schemes for solving case studies with multiple heat exchangers using high-order pinch point temperature difference methods

Abstract

Heat exchangers (HEs) are often modeled using pinch point temperature difference (ΔT_pinch) methods when optimizing systems with HEs. However, even small inaccuracies in model predictions of HEs will introduce numerical noise that can cause optimization algorithms to fail. A recent study of single HEs suggests that highorder interpolation methods can compute ΔT_pinch much faster than conventional methods. However, the performance of such methods when optimizing HE systems have not previously been tested. Heat pumps with 2 and 3 HEs, with and without an ejector are optimized using different schemes. Results from these case studies show that non-linear constrained gradient-based optimization algorithms are more than 5 times faster than particle swarm (PS), and that the conventional genetic algorithm (GA) should not be used. However, the main conclusion is that the case study optimizations are solved 5–10 times faster if ΔT_pinch is calculated using hybrid high and low-order interpolation methods.