Optimizing laying hen diet using particle swarm optimization with two swarms

The highest cost production of the poultry industry is the feed that given to the poultry on daily basis. Unfortunately, manual formulation of poultry diet becomes difficult task when several nutritional requirements with fluctuating price are accounted. Several evolutionary approaches have been employed to solve this complex problem such as particle swarm optimization (PSO). However, in order to prevent premature convergence, PSO highly depends on the diversity of particles that influenced by acceleration component. This study presents a strategy to improve diversity in PSO using two swarms with migration and learning phase (PSO-2S). Numerical experimental results show that swarm size of 20 for each swarm, total iteration of migration phase of 42,000, and total iteration of learning phase of 40,000 are the good choice parameter of PSO-2S. While comparison experimental results show that PSO-2S can provide good solutions with the lowest cost and standard deviation than genetic algorithm, canonical PSO, and another migration strategy in multi-swarm PSO.