基于改进PSO算法和集成神经网络的裂解炉在线优化

Online Optimization of Cracking Furnace Based on Advanced PSO Algorithm and Neural Network Ensembles

针对传统粒子群算法(PSO)寻优时易陷入局部最优、后期全局搜索能力下降等不足,提出了基于载波的粒子群算法(CWPSO)。通过粒子基于载波的搜索和载波扩展精确寻优,较好地克服了上述缺点,且寻优时间明显减少。同时,针对工业裂解炉在线优化要求,采用了权值动态集成的集成神经网络(NNE)对双烯收率进行建模预测,并结合CWPSO算法进行了在线滚动优化。仿真结果表明,该方法对裂解炉的优化效果明显,双烯平均收率有了明显提高。

The traditional Particle Swarm Optimization （PSO） algorithm is easily trapped in the local optimum and converges slowly. Due to the shortcomings above, a novel PSO algorithm based on the carrier-wave （CWPSO） is presented in the paper, which searches through the carrier-wave and takes a precise search by means of carrier-wave extending. As a result, it overcomes the above shortcomings better, and has a shorter searching time as well. In addition, towards the online optimal requirements of the industrial cracking furnace, a neural network ensembled with dynamic weights is applied in the predictive modeling of C2 H4 and C3 H6 yield rates, then the online rolling optimization is carried out. The simulating result shows that the optimal method has sound effects for the cracking furnace, and there is a palpable improvement of C2 H4 and C3 H6 yield rates.