Research on drilling rate prediction model based on fusion feature selection algorithm

Abstract:

ROP prediction is an important part of drilling optimization， machine learning algorithms are currently an important means to achieve accurate ROP prediction， and correct feature selection is the key way to ensure machine learning accuracy. Based on the actual drilling data of a well in the South China Sea， this research uses a fusion feature selection method to select 10 drilling characteristic parameters， including well diameter， outlet temperature， inlet density，outlet density， K value， plastic viscosity， filtration loss， overburden pressure， pore pressure， and nozzle equivalent diameter. The optimized parameters are taken as model inputs， and the integrated Gradient Boosting Decision Tree （GBDT） algorithm is introduced to establish a ROP prediction model. The established model is compared with the conventional machine learning algorithm model， and the test results show that the accuracy of the proposed fusion feature selection algorithm model is 2% higher than that of the full feature model， and the average accuracy is 14.5% higher than that of the commonly used machine learning model. The research provides an effective solution for the accurate and rapid optimization of drilling parameters， and have guiding significance and practical application value for improving the drilling rate.