Cluster analysis based on attractor particle swarm optimization with boundary zoomed for working conditions classification of power plant pulverizing system

This paper proposes a cluster analysis method based on Attractor Particle Swarm Optimization with Boundary Zoomed (APSO-BZ) for working conditions classification of power plant pulverizing system. The proposed method could be used on the field data directly and the obtained clusters represent the different working conditions of the power plant pulverizing system. For APSO-BZ, the particle position is updated based on the attractor which equals the random modified value of the own optimal or the global optimal. The boundary zoomed strategy is presented for letting a particle flying outside of the search space be relocated based on the positions of the particle and the attractor. Moreover, the sum of the symmetrical compactness of each cluster is adopted as the fitness function for APSO-BZ. Three real-life datasets from UCI Machine Learning Repository and a field dataset of a real power plant pulverizing system are adopted to evaluate the effectiveness of the proposed method. The experiments results verify that the proposed method has higher clustering capability and avoids the premature convergence under a certain extent. Moreover, the proposed method would implement the working conditions classification more correctly.

► We propose a cluster analysis method based on attractor particle swarm optimization with boundary zoomed. ► Our method could effectively realize the working conditions classification for power plant pulverizing system. ► The attractor mechanism and the boundary zoomed strategy avoid the premature convergence and improve the global exploration ability. ► The symmetrical compactness for fitness function ensures the accuracy of results. ► The experiments results verify the effectiveness of our method and it could be applied successfully.