Data-Driven Sweet-Spot Prediction Andfracturing Optimization Based On Integratedreservoir Model

In recent years,tight oil and gas have accounted for an increasing proportion in the development of oil and gas fields.Due to the low porosity and permeability of the original formation,tight oil and gas are difficult to obtain industrial oil flow in conventional production.Fracturing of tight oil and gas reservoirs is an important means to increase production,and it is widely used in major oilfields at home and abroad.However,there are also certain problems in fracturing stimulation.The post-fracture productivity of different wells in the same reservoir varies.Whether this difference in productivity is caused by differences in geological attributes(porosity,permeability,saturation,etc.)or by factors of fracturing fracturing,there is currently no better way to quantify and evaluate.If geological factors are the main factors of post-fracturing productivity,which combination of geological parameters corresponds to better postfracturing productivity,this quantification process can effectively guide the selection of drilling and fracturing well locations,that is,the prediction of geological sweet spots.If engineering factors are the main factors affecting the productivity after fracturing,whether the fracturing parameters can be optimized,which parameters can be optimized,and how to find the optimal parameter combination are the urgent needs of oilfield tight oil development.In view of the above problems,this study combines data analysis and various machine learning methods,based on the actual data of a tight oil reservoir in Xinjiang Oilfield,and proposes a set of main controlling factors for the post-fracture productivity,geological sweet spot prediction and fracturing of tight oil reservoirs.Optimization of cracking fracturing parameters.In the analysis and research on the main control factors of post-fracture productivity,this thesis extracts the geological attribute parameters(porosity,permeability,rock mechanics parameters,etc.)of each fracturing section of a single well by establishing an integrated geology-engineering model,as well as fracturing fracturing parameters.The method of calculating the importance of characteristic parameters by regression tree evaluates the geological parameters and which parameters in the fracturing fracturing parameters are the main control factors that control the postfracture productivity.On this basis,this study compared the main controlling factor analysis results of various parameter sensitivity analysis methods random forest,gradient boosting,and distance-based general sensitivity analysis methods and the applicability of their methods.The study found that both geological factors and fracturing fracturing parameters have a certain influence on the post-fracture productivity,but the influence of geological factors is slightly larger than that of fracturing factors.Among the fracturing factors,the total amount of fracturing fluid is the main influencing factor.In the prediction of geological sweet spots,this study uses a supervised classification algorithm to establish a classification model of geological parameters and high-yield and non-high-yield well sections,and then applies the classification model to the geological model to obtain the geological sweet spot prediction model in the study area,which is the following One-step drilling provides a sweet spot.For the fracturing parameters,this study uses particle swarm and genetic algorithm to optimize the fracturing fracturing parameters.In the optimization process,the regression model used in the main control factor analysis was used as a proxy model to replace the fracturing forward model.Through research,it is found that both particle swarm optimization and genetic algorithm can optimize fracturing fracturing parameters,and particle swarm optimization takes less time.In the optimization effect of fracturing parameters,the gradient boosting algorithm is better than the stochastic deep forest algorithm.The optimization results recommend total fluid volume,sand addition volume,and displacement fluid volume as the optimization parameters.