钢铁企业高炉煤气受入量的组合预测模型

Combined prediction model of blast furnace gas generation in steel industry

针对钢铁企业中高炉煤气(BFG)受入量难以有效预测的问题,提出了一种基于数据滤波的组合预测模型。首先,采用经验模态分解(EMD)法将原始训练数据分解为相互独立的固有模态函数,根据各模态函数自相关函数的特点滤去噪声分量,采用滤波后的重构序列作为训练样本;然后,采用组合的支持向量机(SVM)模型对受入量进行预测,并利用遗传算法(GA)对支持向量机的参数进行优化;最后,利用现场实际数据验证该模型的预测精度,并与传统预测方法相比较,三组预测的平均绝对百分误差分别为3.22%、4.43%和5.23%。结果表明该方法对高炉煤气受入量的预测具有较高精度,为煤气管网的平衡调度提供了决策支持。

Concerning the prediction problem on Blast Furnace Gas（ BFG） genegration in steel industry, a combined prediction model based on data-filtering was proposed in this paper. Firstly, an empirical mode decomposition approach was employed to decompose the original training data into a group of independent intrinsic mode functions,the noise components were de-noised by the characteristics of the autocorrelation functions, and the training samples were reconstructed using the filtered sequence. Then a combined Support Vector Machine（ SVM） model was proposed to predict the BFG generation,and genetic algorithm was used to optimize the parameters of SVM. Finally the prediction accuracy of the model was validated using the actual data,which is compared with the traditional predicting methods. The mean absolute percentage error of the three groups were 3. 22%, 4. 43% and 5. 23%. The results show that the proposed method keeps high prediction precision on the prediction of BFG generation,which provides decision support for the gas resources scheduling.