For optimizing structure parameters of a heat pipe heat exchanger

there is no criterion for selecting either the optimization algorithm among the widely used meta-heuristic algorithms

or its common controlling parameters including population size and maximum iteration number in open literature. The performance of three representative meta-heuristic algorithms

i.e.

particle swarm optimization algorithm (PSO)

non-dominated sorting genetic algorithm (NSGA-Ⅱ)

and teaching and learning optimization algorithm (TLBO)

are compared for the optimization of the heat pipe air preheater in a 58 MW circulating fluidized bed boiler. The air preheater is designed by the conventional logarithmic mean temperature difference method to minimize the total amount of metal material. The effects of the population size and maximum iterations

as well as the random number

on the optimization performance for the three algorithms are investigated in this work. The results show that optimal values for the population size and maximum iterations exist for all three algorithms

and the optimal results of these algorithms are different for each execution

showing the local minima performance. The optimal solutions of these algorithms converge after rounding according to the practical values

and the running time of TLBO is the shortest. Therefore

TLBO is more suitable for the structure parameter optimization of a heat pipe heat exchanger.