Research On Intelligent Optimization Method Of Drilling Parameters Based On Machine Learning

Reasonable optimization of drilling parameters and improvement of ROP are important means to achieve cost reduction and efficiency improvement.This thesis takes big data as the driving force,machine learning algorithm as the core,and intelligent optimization algorithm as the strategy,and conducts research on the intelligent optimization method of drilling parameters based on machine learning,which is expected to provide technical support for drilling parameter optimization and drilling speed.Based on data processing methods such as interpolation,this article optimizes the collected 32 various types of parameters,a total of more than 280,000 lines.Use statistical methods to evaluate the correlation coefficient matrix between various parameters and mechanical parameters,revealing the relationship between various parameters and ROP The internal connection at the data level verifies the feasibility of dataset training.Based on various machine learning algorithms,the ROP prediction models based on support vector regression,random forest regression,BP neural network,and recurrent neural network were established respectively,and the model accuracy of various machine learning models under different hyperparameters was evaluated.Based on 6 types of performance indicators such as explained variance,the LSTM recurrent neural network model is selected as the optimal prediction model.Based on the enumeration method and intelligent optimization algorithm,combined with the optimal prediction model,an intelligent optimization method of drilling parameters with a single parameter,dual parameter,and multi-parameter is established,and the optimization effect and time feasibility of the optimization method is verified by taking the target well as an example.The research results show that after the optimization of the 1000m-3000 m well section of the target well,the average ROP was increased by 19.31m/h,with an increased ratio of 30.15%,and the ROP improvement effect was obvious.The intelligent optimization algorithm can significantly reduce the time overhead growth caused by the increase of the optimized parameters.On the test platform,all three optimization methods can give optimization results within 60 s,and the real-time performance is good.