Aiming at the problem that drilling data is disturbed by environment noise, which affects the accuracy of stratum identification based on expert experience. Due to the impact of drilling environment and limited resources of drilling experts, experienced experts have been unable to visit the site in person to guide the site in the form of one-to-one. Firstly, the paper introduces remote wireless acquisition system for drilling formation intelligent identification. Secondly, because the drilling data is disturbed by environmental noise, the quality control algorithm of drilling sequence data is applied to eliminate the noise interference of original data. Finally, conjugate gradient optimization BP (CG-BP) neural network model is established to achieve accurate identification of formations. The results show that the error of CG-BP neural network model is 67.1% less than that of BP neural network model in formation identification, and the convergence speed is fast in data training. The prediction accuracy is high in data prediction and the accuracy of formation identification is greatly improved. It has the value of field popularization and application.

针对钻探数据受环境噪声干扰的问题，这会影响基于专家经验的地层识别精度。由于钻探环境的影响以及钻探专家的资源有限，经验丰富的专家无法亲自访问该站点以一对一的形式指导该站点。首先，介绍了用于钻井岩层智能识别的远程无线采集系统。其次，由于钻井数据受到环境噪声的干扰，因此采用钻井序列数据的质量控制算法消除了原始数据的噪声干扰。最后，建立共轭梯度优化BP（CG-BP）神经网络模型，以实现对地层的准确识别。结果表明，CG-BP神经网络模型在地层识别中的误差比BP神经网络模型的误差小67.1％，并且在数据训练中收敛速度快。数据预测的预测精度高，地层识别的精度大大提高。具有现场推广应用价值。