Research On The Optimization Method Of Fracturing Operation Parameters In The Case Of Small-size Sample Based On Data Mining Method

The hydraulic fracturing is significant for low permeability reservoirs.Reasonable fracturing design is a powerful guarantee for the effect of fracturing.Commercial software is often used to finish a fracturing design based on the physical properties,lithology and rock mechanics properties of the reservoir,which is lack of learning and reference from operation experience of other reservoir and limited to the research progress of the fracture extension models and the difficulty of solving them.At the same time,the data mining technology has developed rapidly in recent years.Many scholars have applied it to the field of oil and gas field development and made some achievements.It shows that the data mining technology has a great research prospect in the fracturing design and optimization.In this paper,a low porosity and low permeability glutenite reservoir in Western China is selected as the research object.There are few fracturing wells which are all exploration wells in target area.Data mining technology is utilized to research on the optimization method of fracturing operation parameters in the case of small-size sample.The achievements are as follows.(1)There is a high risk of fracture height overextension in fracturing and there are obvious speed sensitivity,hydrochloric acid sensitivity,alkali sensitivity and stress sensitivity in target reservoir.The selection of fracturing fluid and proppant conforms to the actual reservoir conditions.The empiricism of the selection of fracturing parameters and the large difference among different well layers affect the fracturing efficiency and fracturing effect.Therefore,it is necessary to optimize the fracturing operation parameters.(2)Taking advantage of the geological data and fracturing operation data of the target area and considering the comprehensiveness,independence and generalization of the factors affecting the productivity after fracturing,the related data of the 53 well layers are collected.Microsoft Access is chosen to establish a fracturing database which supports the storage and renewal of the data after comparing the characteristics of different database system and it is suitable for the research area.(3)A set of preprocessing method of raw data is established for target area.On the one hand,the formation coefficient method is used to split the production capacity of the commingled production wells to realize the full utilization of the data of each production layer.On the other hand,choose an appropriate data dimensionless method by comparison to avoid the non-convergence problem brought by the singular sample points effectively.(4)Select the grey correlation analysis method to analyze the weight of 24 factors affecting the productivity after comparing different weight analysis method.And the order of the weight reduction is:sand volume per meter,the ratio of the preflush,the effective thickness of the reservoir,the Young's modulus,the thickness of the large middle conglomerate,the thickness of the fine sandstone,the brittleness index,the thickness of the small middle conglomerate,the difference between horizontal maximum principal stress and horizontal minimum principal stress,the thickness of the coarse-grained sandstone,permeability,neutron porosity,fracturing displacement,oil saturation,horizontal minimum principal stress,pore structure index,vertical stress,acoustic moveout,Poisson's ratio,mud content,deep lateral logging resistivity,average sand ratio,porosity and density.(5)Based on the support vector machine and Bias neural network in the data mining method,the prediction model for productivity is established under the circumstance of small-size database of the research area.It is found that the prediction accuracy of two models can basically meet the requirements,but the performance of the productivity prediction model based on support vector machine is relatively better.(6)On the basis of the support-vector-machine productivity prediction model and the similar reservoir determination method based on the result of grey correlation analysis,the optimization model of the fracturing operation parameters is established.And it shows good performance when it is applied to a chosen well layer in the target area.(7)Based on the various studies in the paper,a fracturing operation parameters optimization software is developed to through the GUI programming platform in MATLAB to integrate the research achievements.The function of the software includes inquiry of fracturing parameters,prediction of productivity,analysis of influence factors of productivity,optimization of fracturing operation parameters and so on.The software is supposed to provide reference for fracturing design.