Prediction of ROP Method Based on Online Machine Learning and Multi-source Data Preprocessing Technology

Abstract

The drilling operation is a large scale, technically complex and systematic project, which has always taken the top position of investment in oil exploration and development. Speed and efficiency are the primary goals of drilling engineering. Using information technology to optimize drilling parameters can reduce the complexity of accidents, improve the time efficiency of drilling, and significantly shorten the drilling construction cycle and save exploration and development costs. Forecasting drilling rate of penetration (ROP) is an essential component of drilling optimization. This paper introduces a machine learning-based permeability prediction method and its application effects. In response to the drilling rate of penetration (ROP) problem, a drilling rate of penetration (ROP) model based on an integrated learning algorithm is designed and implemented by mining the historical data collected from a specific block. Meanwhile, this approach is compared with traditional machine learning algorithms such as SVM, LR and KNN. The experimental results indicate that the algorithm has better accuracy and applicability than other methods, which can provide a scientific and reliable reference for improving drilling rate of penetration (ROP) and technical support for realizing intelligent drilling.