基于因子分析与BP神经网络的煤体瓦斯渗透率预测

Forecast of Coal Body Gas Permeability Based on Factor Analysis and BP Neural Net

为提高煤体瓦斯渗透率预测准确性,使用因子分析法对BP神经网络模型进行优化、改进,提出一种改进的BP神经网络预测模型。根据煤体瓦斯渗透率相关主要影响因素实例数据,使用因子分析法对4个煤体瓦斯渗透率影响因素原始数据进行降维数据处理,优化得到2个公共因子;以2个公共因子代替原有4个煤体瓦斯渗透率影响因素作为BP神经网络模型输入层参数,建立改进的BP神经网络煤体瓦斯渗透率预测模型,进行实例数据检验改进BP模型预测效果。最终验证结果:20组训练样本预测值与实际值的相对平均误差为0. 63%,证明训练完成的改进BP神经网络模型具有良好的拟合效果;改进BP模型预测样本平均相对误差为3. 16%,传统BP模型预测样本平均相对误差为6. 37%,证明改进BP模型预测精确度优于传统BP模型。

In order to improve forecast accuracy of coal body gas permeability, so an improved BP neural net forecast model was put forward by using factor analysis improved the BP neural net model. According to practical data that related to some main influence factors of gas permeability, and then data dimensionality reduction of original data was done by factor analysis , which influenced by four factors of gas permeability, two optimization common factors were obtained, and the two optimization factors were input as input layers parameters of BP neural net, instead of the above four factors, so improved BP neural net was built, and practical data was verified by the improved model. The ultimately results showed that average error between predicted value of 20 groups training samples and actual value was 0. 63%, the improved appeared good fitting results, and average relative error for improved and traditional model were 3. 16% and 6. 37%, respectively, certified that improved model was better than traditional one.