Abstract: By integrating the randomness and stable tendency of cloud droplets, this paper proposes an improved particle swarm algorithm(ICPSO) so as to overcome the premature convergence and easily plunging into the local optimization of the PSO algorithm. And then, the improved algorithm is utilized to optimize the parameters of the fuzzy neural network, which is further applied to the modeling of methanol conversion. The experiment results show that the proposed model has higher precision and better generalization ability, and can realize real time monitoring of the methanol conversion.