基于粗集理论的气测录井数据归一化处理

Gas Logging Data Normalization Processing Based on Rough Set Theory

在现代录井工程中,地质情况的不同、钻井工艺的优差等诸多原因都会影响气测录井资料的录取。即使在同一地区、同一层位进行钻井勘探,测量的气测资料结果也会存在很大差异。可想而知如果在不同的地区、不同的层位进行勘探会导致测量气测资料变得更加困难与复杂。因此,快速、有效的规范气测资料及参数选择是现代录井工艺与油气层识别技术中至关重要的步骤。针对RBF神经网络算法具有收敛速度慢且不稳定等缺点,无法有效处理气测录井资料,文中提出了一种基于粗集理论的归一化处理方法,利用粗集理论对气测样本数据归一化处理后提高了RBF神经网络的训练速度。为了验证方法的可行性,以辽河油田的气测录井数据为背景进行仿真计算,实验结果表明此方法有效地提高了RBF神经网络处理气测录井数据速度。

In the modem mud logging engineering, different geological conditions, the good and poor of drilling process, and other reasons will have effects on gas logging data. Drilling exploration, even in the same area, the same horizon, makes the measurement results of gas logging data different highly. It can imagine if in different layers of different area, exploration will result in measuring gas data becomes more difficult and complex. Therefore, fast, effective and standardized gas logging data and parameter selection is essential to modern mud logging technology and oil and gas reservoir recognition technology. In view of the disadvantage of slow convergence and instability of RBF neural networks algorithm, unable to handle the gas effective logging data, present a normalization method based on rough set theo- ry, using rough set theory on gas logging normalized sample data to improve RBF neural network training speed. In order to verify the feasibility of the method, taking the Liaohe oil field gas logging data calculation as the background of the simulation,the experimental re- sults show that this method effectively improves gas logging data speed RBF neural network dealt with.