钢铁企业蒸汽管网压力预测

Steam Piping Network Pressure Prediction in Iron and Steel Enterprise

为解决钢铁企业中蒸汽管网压力变化无规律很难对蒸汽系统进行实时有效调度的问题,提出一种基于小波变换-贝叶斯神经网络预测方法。首先利用小波变换对原始数据进行降噪处理,以降低数据中的误差干扰;然后利用贝叶斯正则化算法结合BP神经网络,在训练优化过程中降低网络结构的复杂性,避免网络过拟合,提高了网络的泛化能力同时改善了预测效果。实验结果表明:基于小波变换-贝叶斯神经网络预测方法的钢铁生产蒸汽管网压力的预测精度高、性能好,具有良好的实用性,可提高企业蒸汽管网的运行管理水平,为蒸汽的合理调度提供了科学的理论依据。

Piping system's irregular pressure variation results in ineffective scheduling of the steam system and the waste of steam resources in iron and steel enterprises. A wavelet transform-Bayesian neural networkbased prediction method was proposed,in which,having wavelet transform adopted to de-noise original data so as to decrease error interference of the data; then having Bayesian regularization algorithm used to improve BP neural network so that the network structure complexity in process training optimization can be decreased to avoid occurrence of 'overfitting'and to improve generalization ability. The experimental results show that the wavelet transform-Bayesian neural network-based prediction method has high precision,good performance and practicability in predicting steam piping system's pressure,improving the level of operation management and providing scientific theoretical basis for reasonable scheduling of the steam.