Early Gas Kick Detection Based On Pattern Recognition And Inversion Of Formation Information After A Kick

The Gas kick will not only delay drilling process,but also further develop into catastrophic incidents such as blowout.Blowout can cause significant damage to the natural environment,casualties,economic losses and extremely bad social impact.Therefore,the accurate detection of kick is crucial to ensure safety of the drilling operation.And accurate determination of formation information is highly important in the design of kill fluid parameters after a gas kick occurs during drilling.Due to the influence of noisy data and complicated stratigraphy,it is usually difficult for the traditional threshold method to balance between reducing the rate of missing report and reducing the rate of false alarm.In this paper,we focused on the analysis of the time series of monitoring parameters rather than searching for suitable thresholds,process identification of single parameter and the mechanism of multiparameter cooperative diagnosis are considered for kick detection.Based on the theory of data mining and machine learning,a novel and reliable method for kick detection is proposed.In our method,firstly,we proposed a trend and fluctuation based Symbolic Aggregate Approximation(TF?SAX)method.The TF?SAX method can perform a good assessment of the shape and fluctuation characteristics of monitoring curves and can be used to achieve kick detection based on single parameter's time series.Based on this,a multiparameter collaborative diagnosis model which is suitable for kick detection is established based on the support vector machine(SVM).This model takes the result of TF?SAX method as training parameter.The genetic algorithm is used to optimize its parameters to get the best classification results.For the traditional well shut-in method,there exists a difficulty in determining the formation pressure and permeability simultaneously.Furthermore,the standpipe pressure must be stabilized in order to accurately estimate the formation pressure,for which an enormous amount of time is consumed.In this paper,a new method is presented to achieve the simultaneous inversion of the formation pressure and permeability before the well shut-in,which uses a gas kick response parameter,annular multiphase flow model coupled with the reservoir seepage equation,curve segmentation,and genetic algorithm.Owing to the small expansion rate during the early stage of the gas kick,the gas production rate at the bottom of the well can be approximately calculated according to the derivation of the pit gain curve.Therefore,the formation permeability can be calculated from the pit gain curve in the early stage of the gas kick using the reservoir seepage equation.A genetic algorithm can then be used to calculate the formation pressure based on the entire pit gain curve and the transient wellreservoir model.During drilling,gas will enter the wellbore due to different mechanisms because of the complex formation characteristics.The identification of gas kick types and the determination of formation information are of great significance to well kill procedure.Based on the mechanism of gas kick,the coupling model of formation and annulus multiphase flow is established,and the ground observation characteristics of negative pressure gas kick(permeable formation and fractured formation),displacement gas kick,cuttings gas kick and diffusion gas kick are obtained.The identification method of gas kick type based on surface monitoring parameters is obtained,and different suggestive killing methods for different gas kick types are given.Case analysis shows that the proposed kick detection method can lead to accurate diagnosis,which can provide theoretical guideline for kick detection on drilling site.The accuracy of the formation Inversion method was verified based on a sample analysis.In the sample analysis,the average calculation error of the permeability was 9.45% and that of the formation pressure was 0.54%.