Rate of penetration prediction using hybrid Gaussian process regression model with ARD structure in geological drilling process

Abstract

In complex geological drilling process, efficiency improvement is typically tackled by predicting and optimizing the rate of penetration (ROP). Combining the process characteristics and data distribution characteristics, a hybrid intelligent prediction model is proposed to predict ROP accurately. Firstly, data denoising is performed using multivariate wavelet denoising with principal component analysis while considering the correlation among multiple drilling variables. Then, using mutual information analysis, the input features of ROP prediction model are determined. Next, a hybrid Gaussian process regression model with automatic relevance determination structure is construsted for ROP prediction, with the model hyperparameters optimized using the hybrid bat algorithm. Finally, experiment verification is carried out using actual drilling data from a drilling site in Xiangyang. The results indicate that the proposed method has high prediction accuracy and outperforms other widely used intelligent methods. As a result, this method provides an effective solution to predict ROP in actual drilling process.