基于主成分的BP神经网络煤层气产能预测方法

提出基于主成分的BP神经网络煤层气产能预测方法具有良好的仿真预测功能，预测结果相对误差平均1．67％，相对误差最大3．33％，最小0．22％，预测精度较高；为采用量化的排采参数预测煤层气产能提供了一种新方法。BP神经网络煤层气产能预测方法引入主成分分析，可以有效提高网络运行效率。对于产出关系不确定的煤层气井，可利用该方法获得排采参数到产气量的非线性映射关系，实现产量预测。

A Method Combined Principal Component Analysis and BP Artificial Neural Network for Coalbed Methane (CBM) Wells to Predict Productivity

A method combined principal component analysis and BP artificial neural network for coalbed methane (CBM) wells to predict productivity has been proposed in the paper, which has better simulation and prediction functions. The average relative error between the factual data and predicted data is 1.67%, the max 3.33% and the min 0.22%, showing high precision as well as presenting a means to predict productivity for CBM using production parameters. The approach of BP artificial neural network for CBM wells to predict productivity brings in principal component analysis so that the operating efficiency can be increased validly. As for certain CBM wells without obvious production rules, the method could obtain nonlinear mapping relation, realizing production prediction.