应用BP神经网络预测煤质参数及含气量

APPLYING BACK-PROPAGATION ARTIFICIAL NEURAL NETWORKS TO PREDICT COAL QUALITY PARAMETERS AND COALBED GAS CONTENT

煤层气储层物理结构以及煤层气的存储、运移等方面不同于常规天然气，评价煤质参数的测井等效体积模型难以较好地描述煤层这种复杂的物理结构．提出利用ＢＰ神经网络预测煤质参数及煤层气含量的模型和算法，预测的煤质参数以及煤层含气量与煤样分析结果比较表明，预测与煤样分析参数之间的平均绝对误差和相对误差都较小。

The reservoir physical structure, storage and migration of coalked gas are different from those of conventional gas. The usual method of the geophysical well logging to determine coal quality parameters is an equivalent volume model, which can not describe suitablely the complex physical structure. Put forward in this paper is the model using back-propagation artificial neural networks to predict coal quality parameters and coalbed gas content. Compared with the quality parameters and coalbed gas content predicted by BP artificial neural networks and those analysed by coal sample testing, the average absolute error will be less than 1.5 and the average relative error less than 10%, whereby the method is proved to satisfy predicted precision. Accordingly, applying BP neural networks to predict the coal quality information and the coalbed gas content is effective.