Multi-objective optimization of cascade storage system in hydrogen refuelling station for minimum cooling energy and maximum state of charge.

Compared with single-stage hydrogen storage refuelling, cascade storage refuelling has more advantages and significantly reduces cooling energy consumption. In the cascade system, the parameters of cascade storage tanks are critical, especially the initial pressure and volume. This article analyzes the thermodynamic processes in a cascade hydrogen refuelling station (HRS) and establishes the simulation model in Matlab/Simulink platform. The state of charge (SOC) of the onboard storage tank and the cooling energy consumption of the refuelling system are obtained from different initial pressures and volumes of the cascade storage tanks by using the simulation model. These data are introduced into the artificial neural networks in Matlab to generate a relationship between the decision variables and objective functions. The decision variables are optimized to minimize the cooling energy consumption and maximize the SOC through the genetic algorithm and Pareto optimization. So that optimal initial pressure and volume of the cascade storage tanks are determined. The research shows that when the ambient temperature is 293.15 K, and the SOC is 0.98–0.99, using the optimal initial pressure and volume of the cascade storage tanks can reduce the cooling energy consumption by up to 11.43%, compared with the baseline situation. Among the factors affecting cooling energy consumption and SOC, initial pressure is more sensitive than volume, so optimizing initial pressure, especially for the high-pressure cascade storage tank, seems more meaningful than volume. This research is instructive for the construction of the cascade HRS. • A thermodynamic model for cascade hydrogen refuelling is developed and well validated. • Effect of initial pressure and cascade tank volume on SOC and cooling energy is studied. • Artificial neural network (ANN) is trained by orthogonal test data from physical model. • The ANN model is used to determine objective functions and constraints of optimization. • Genetic algorithm and Pareto algorithm are combined for multi-objective optimization. [ABSTRACT FROM AUTHOR]