基于双精英进化樽海鞘群算法优化ELM的焦炭价格预测

焦炭是焦化企业生产的重要工业原料之一,准确地预测其未来价格趋势对焦化企业制定排产计划具有重要意义。极限学习机(ELM)泛化能力强,计算速度快,适合作为焦炭价格预测的模型,但ELM的预测性能受模型关键参数影响较大,故需对其参数进行优化。基于此,文中提出了基于双精英进化樽海鞘群算法的ELM焦炭价格预测方法。首先,采用Logistic混沌映射、改进的收敛因子、自适应惯性权重和双精英进化机制来改进樽海鞘群算法,提出了双精英进化樽海鞘群算法(MDSSA),提高算法的搜索能力;其次,运用MDSSA优化ELM的连接权值与阈值,找到ELM的最优参数组合,构建MDSSA-ELM焦炭价格预测模型;最后,在8个基准测试函数上测试MDSSA的收敛性能,在实际焦炭价格数据集上对MDSSA-ELM模型的预测性能进行实验,实验结果表明,MDSSA-ELM相比其他方法预测能力更优,MDSSA相比其他群智能算法搜索能力更强,为焦化企业实现焦炭智慧排产提供了有效的预测工具。

Forecasting the Coke Price Based on the Kalman Filtering Algorithm

Research on coke price forecasting is of theoretical and practical signiifcance. Here, the Kalman ifltering algorithm was used to analyze the price of coke. As the only state variable, the historical coke price is sorted out to build the state space model. The algorithm makes use of innovation composed of the difference between observed and predicted values, and alows us to obtain the optimal estimated value of the coke price via continuous updating and iteration of innovation. Our results show that this algorithm is effective in the ifeld of coke price tracking and forecasting.