MPSO-Based Operational Conditions Optimization in Chemical Process: A Case Study

摘要

A multi-swarm PSO (MPSO) was proposed, with which the whole swarm is divided into by K-means clustering algorithm randomly to accelerate searching process of global optimum. The big swarm clustering will obey the standard PSO principle to search the global optimal result, which the number of particle is more than a threshold. The small swarm clustering will search randomly inner neighborhood of the global optimal value, and then the outlier particle does not care about the optimal result but flies freely according to themselves velocities and positions. The proposed algorithm enhances its global searching space, and enriches particles' diversity in order to let particles jump out local optimization points. Testing and comparing results with standard PSO and linearly decreasing weight PSO using several benchmark functions show the proposed algorithm is better than other algorithms. Furthermore, the MPSO algorithm is used to optimize the operational conditions in a chemical process case for an ethylene cracking furnace.