摘要
研究表明:多元回归和复合指数回归预测含气量误差较大,尤其针对低含气量煤层;KIM方程预测结果为饱和状态下含气量,预测效果同样较差;BP神经网络基于非线性方法进行预测,预测效果最佳。总体而言,由于含气量与各测井参数存在复杂的非线性关系,现代数学方法必将成为煤储层含气量预测的主要方法。
The study results show that the prediction error of regression analysis and composite exponential regression was large, especially for low gas-bearing coal seam, which is relate to the sensitivity of logging parameter choose. The prediction result of the KIM equation is saturated state, and BP neural network model is higher precision and less errors. In general, there is a complex non-linear relationship between gas content and logging parameters, and modern mathematics method will be the main method to predict the gas content.
出处
《煤炭技术》
CAS
2018年第1期86-89,共4页
Coal Technology
基金
国家科技重大专项资助项目(2016ZX05044)
关键词
临兴区块
测井
含气量预测
Linxing block
well logging
prediction of gas content
