Research On Rock Mechanics Parameters And In-situ Stress Intelligent Prediction Based On Physical Information Constraints

Shale oil reservoir is a tight reservoir with low porosity and permeability,which requires large-scale fracturing to realize commercial development.Accurate acquisition of rock mechanical parameters and ground stress is an important prerequisite for fracturing design and productivity evaluation.At present,a variety of in-situ stress calculation models based on various rock mechanical parameters have been proved to be very effective methods,but the reliability requirements of various calculation models are high,so it is very important to accurately obtain rock mechanical parameters of shale oil reservoir for describing the in-situ stress field.Firstly,the characteristics of various rocks in shale reservoir are studied experimentally.The rock mechanics parameters of shale reservoir were obtained by uniaxial compression experiment.The transverse and longitudinal wave data and rock mineral composition data of shale reservoir were obtained by laboratory rock acoustic wave experiment and X-ray diffraction experiment.In this paper,Pearson correlation analysis method is used to analyze the correlation between well logging data,rock mineral component data and rock mechanics parameters.The well logging data and rock mineral component data which are highly correlated with rock mechanics parameters are selected as input characteristics,and rock mechanics parameters are taken as prediction target values.In this paper,a neural network model based on physical information constraints is established.The model is driven by both physics and data,and can accurately predict rock mechanics parameters with only 100 groups of data.In order to verify the excellent performance of the model,artificial neural network,random forest and XGBoost algorithms are used in this paper for comparison.The results show that the accuracy of the neural network constrained by physical information is higher than85% in predicting rock mechanics parameters with a small amount of data,and the performance is far better than other models.A neural network constrained by physical information is used to predict four rock mechanical parameters,namely Young’s modulus,Poisson’s ratio,pore pressure and density,at different well depths in well A,Cangdong Sag,Bohai Bay.After four kinds of rock mechanics parameters are predicted,the vertical stress,maximum horizontal principal stress and minimum horizontal principal stress of the target reservoir in the study area are calculated based on the rock mechanics parameters and the existing in-situ stress calculation model.The results show that the accuracy of Huang’s model in calculating the maximum and minimum horizontal principal stresses is 87% and 86% respectively,which is the highest among the existing models.Huang’s model is the most suitable for calculating the in-situ stresses of the target reservoir in the study area.The optimized Huang’s model is applied to the in-situ stress calculation of well A in Cangdong Sag,Bohai Bay.The neural network model with physical information constraints proposed in this paper not only has high accuracy in predicting rock mechanics parameters,but also has certain physical significance,which can provide theoretical guidance for field engineering design.