基于优化的BP神经网络地层可钻性预测模型

A Model for Predicting Formation Drillability Based on Optimized BP Neural Network

摘要:

提出了一种粒子群算法（PSO）优化的BP网络模型预测地层可钻性的新方法。利用粒子群算法优化BP网络模型的参数，避免了BP网络陷入局部极小值的缺点，提高了模型的预测速度和精度。结合钻探实例，利用测井资料和地层可钻性级别的关系建立了可钻性级别实时预测模型,并将该模型与传统的BP网络进行对比，结果表明，该模型优于BP网络，具有较高的精度和较快的收敛速度，有一定的适用性。

Abstract:

A new method for predicting formation drillablity was proposed according to the theory of BP networks based on PSO. The use of PSO optimizing the parameters of BP networks is to improve the convergence speed and precision of BP neural networks. Combining with the examples of drilling and based on the relationship of log information formation drillability grade, a real-time formation drillability grade model was established. The results show that the model is superior to BP network with higher accuracy and faster convergence rate and it is an effective way to predict formation drillablity.