Analysis of single-channel and canopy-to-canopy cold plate cooling systems under cyclic operation

• Single-channel and canopy-to-canopy cold plates are tested under cyclic operation. • A cooling strategy based on controlled pumping activation is developed. • The cycle time is evaluated experimentally and numerically for all configurations. • A correlation relating cycle time to the conduction/convection ratio is proposed. • An additional cold plate is designed and tested to validate the correlation. The performance of PEM fuel cells strongly depends on the operating temperature, being negatively impacted when the equipment works outside of a prescribed temperature range. In this work, a novel liquid-based cooling approach is proposed to keep the average temperature of the heated plate within a temperature range with minimum energy consumption. The strategy consists in switching the coolant pumping on when the average temperature of the cold plate reaches the upper limit of the temperature range, and turning it off once the average temperature coincides with the lower limit. Four canopy-to-canopy and five single-channel configurations are experimentally and numerically tested for three different values of the pumping power. The maximum deviation of the cycle time (plate heating followed by cooling) between the experimental measurements and the numerical simulations is 8 % for all configurations, confirming the validity of the numerical approach. Among the tested designs, canopy-to-canopy configurations with four or more branches exhibit the shortest cycle time under all studied conditions, in both experiments and simulations. The cycle time was found to describe a linear relation with the ratio between conduction in the cold plate and convection in the channel, i.e., distance to the channel and coolant mass flow rate. An additional single-channel configuration was designed and built specifically to test the validity of the linear correlations derived, estimating the cycle time with theoretical, numerical and experimental calculations for the coolant mass flow rate. The deviation between these predictions and the experimentally measured cycle time is below 5 % for all cases.