Research On Real-time Risk Assessment Method For Shale Gas Fracturing Operation And Key Equipment Based On Bayesian Network

In the process of large-scale fracturing operation of shale-gas horizontal wells,downhole incidents,failure of fracturing pump and high-pressure manifold system are main reasons resulting in a long unplanned shutdown of fracturing operations.Down-hole incidents are caused by the coupling effect of causative factors.The existing risk-based warning methods can quantify the uncertainties among causative factors,but neglect the state information of down-hole conditions carried by incident characterization parameters,which could give rise to delayed warnings.The traditional reliability-based remaining life prediction methods neglect the influence of the current system state on future degradation,resulting in the lack of dynamics of prediction,which is difficult to be applied to the fracturing pumps under real working conditions.For a high-pressure manifold system,the degradation of component and internal pressure(a kind of risk influence parameter)are the main causes of system failure.Existing studies have evaluated system risk only from the perspective of the degradation of components,without considering the instantaneous risk increment caused by risk influence parameters,and can not be applied to real-time risk analysis of high-pressure manifold system under variable conditions.This paper has carried out an in-depth research around the above issues,the main work is as follows:(1)To reveal the evolution behavior of incident characterization parameters,a method based on System Dynamics is proposed to study the causation mechanism of down-hole incidents,considering the dynamic characteristics of down-hole incidents under the coupling effect of causative factors.Taking the bridge plug fracturing process in casing pipe of horizontal well as the research object,the system dynamics models of sand plugging near the wellbore zone,sand plugging in formation and the channeling in formation are established respectively.The simulation results show that: a)when a sand plugging occurs near the wellbore zone,the casing pressure shows a rapid upward trend in a short time;b)when a sand plugging in formation occurs,the casing pressure shows a wave-like upward trend;and c)when the channeling in formation occurs,the casing pressure shows a significant downward trend.(2)In view of the phenomenon that ignoring risk characterization parameters can lead to the delay of warning for down-hole incidents,a safety early-warning method by integrating the risk characterization parameters for down-hole incidents is proposed.The risk characterization parameters are integrated into the safety early warning model based on Bayesian network,and its trend characteristics are extracted as the real-time risk information.Then,membership function is introduced to quantify its state probability distribution,and the probability risk of down-hole incidents is corrected in real time.Finally,the warning information is issued according to the warning rules.The proposed method is tested and verified using real-world down-hole sand pluggings,and compared with the traditional warning model without the fusion of risk characterization parameters.For five sand pluggings in formation,the proposed method can advance the warning time by an average of 2.7 min(+0.2 min);for five sand plugging near the wellbore zone,the presented method can achieve accurate early warning for all incidents,while the traditional warning model for only one sand plugging.(3)In view of the fact that traditional reliability-based remaining life prediction method neglects the influence of the current system state on the future degradation,resulting in the lack of dynamic of prediction process,a dynamic residual life prediction method based on dynamic object-oriented Bayesian network for fracturing pump system(DOOBN)is proposed.The method combines the reasoning model for abnormal state of components(RMASC)with degradation trend prediction model(DTPM)to realize the dynamic prediction of the residual life of components(i.e.,the number of remaining fracturing segments)and the development trend of system function index.Taking "the wear of crankshaft" and "the scaling of gas filter" as examples,when the crankshaft is in wear state at 158 h,its residual life is 139 h(+10 h),that is to say,it can be used for another 28(+2)stages.When the gas filter is in scaling state at 365 h,the probability of blockage is higher than that of scaling after 268 h(+5 h).However,the traditional method can only pridict the degradation of components under their natural state,and can not incorporate the change of system state into the prediction process of future degradation.(4)Considering that risk influence parameters can lead to the instantaneous risk increment,a real-time risk assessment method for high-pressure manifold system is proposed,which integrates instantaneous risk increment.The conditional Gauss distribution function is used to quantify the uncertainty between the risk influence parameters and the instantaneous risk increment.The inherent risk caused by degradation of components is effectively fused with the instantaneous risk increment in the hybrid Bayesian network.The simulation results show that the risk influence parameters increase the failure risk of manifolds in varying degrees.When the manifolds are in service for 100 hours and the internal pressure is 80 MPa,the overall risk is three orders higher than the inherent risk.The practical application shows that when the internal pressure is greater than 30 MPa,the system risk fused with the instantaneous risk increment is 1?3 orders of magnitude higher than the inherent risk,indicating the proposed method can be applied to real-time risk assessment of the high-pressure manifold system under actual working conditions.