基于灰色预测和BP的集气管压力集成预测方法

Integration prediction of gas collector pressure based on gray forecasting and BP neural network

针对焦炉煤气集气过程是一个高度复杂的工业生产过程,难以获得焦炉集气管压力的精确数学模型的问题,提出一种基于灰色预测和BP神经网络的集气管压力集成预测模型。该模型首先利用灰色预测和BP神经网络分别对焦炉集气管压力进行预测,然后采用熵值法确定各预测子模型的加权系数,将两个子模型进行加权集成,可以获得较为准确的焦炉集气管压力值。将其预测性能与单一的灰色模型和BP神经网络模型进行比较,运行结果表明:集成预测模型的预测效果和性能优于单一的灰色模型和BP神经网络预测模型,能够获得较高的预测精度。

As the gas collecting process of coke ovens is a highly complex industrial process,for which accurate mathematical model is difficult to obtain,an integrated modeling method is proposed,which incorporates gray forecasting and BP neural network.Firstly,gray prediction and BP neural network are used to predict the gas collector pressures respectively,and the weighted coefficients of the two predicted sub-models are determined with entropy method;then through combining the two sub-models with weighted integration,the more accurate value of the gas collector pressure can be obtained.Finally,the performance of the integration prediction is compared with those of the single gray model or BP neural network model,running result shows that the result and performance of the integration prediction model are better than those of single gray model or BP neural network model,and the proposed method can obtain higher prediction accuracy.