Adaptive small-family population-guided swarm intelligence optimization algorithm

Ethylene, a cornerstone of the petrochemical industry, constitutes over 75% of petrochemical products and plays a vital role in the national economy. The ethylene cracking furnace, as the core production unit, directly determines the scale, output, and product quality of ethylene manufacturing. To optimize its operation, this paper proposes a novel adaptive small-family population-guided swarm intelligence optimization algorithm (ASPSIOA). The ASPSIOA operates in two distinct stages, including an early exploration stage and a late exploitation stage. Initially, the population is partitioned into small-family subgroups. During exploration, individuals are adaptively moved toward or away from their subgroup average position to identify promising regions in the search space. Subsequently, during exploitation, individuals are guided toward the global best solution, focusing the search on these identified promising areas. The performance of the ASPSIOA is rigorously evaluated on thirty benchmark functions and three semi-realistic engineering problems. Comparative results against six established algorithms demonstrate that the ASPSIOA achieves competitive performance with faster convergence speed and superior solution accuracy. Furthermore, the ASPSIOA is successfully applied to optimize an industrial ethylene cracking furnace by determining a periodic outlet temperature regulation strategy, resulting in a measurable increase in ethylene yield.