Intelligent Monitoring And Diagnosis Of Underground Working Conditions Based On Mechanism And Data Driven

With the continuous progress of China’s petroleum industry,the focus of petroleum exploration is toward more complex and changeable strata.There are more and more ambiguous and uncertain parameters in the downhole drilling environment,which greatly increases the failure rate during the drilling process.Real-time and accurate measurement of bottom hole flow,accurate and rapid identification of abnormal conditions and even early warning of drilling accidents are still challenging tasks.In view of the inaccurate measurement of many downhole parameters and states,a new adaptive observer is constructed based on the dynamic model of pressure-controlled drilling and the adaptive observer theory.The uncertain parameters of the dynamic model are treated as the unknown parameters.The bottom-hole pressure,flow rate,the backpressure of the standpipe pressure is set to the unknown states of the observer.The unknown parameters and the unknown states are estimated jointly.The simulation results show that the bottom hole pressure,standpipe pressure backpressure data observed can better track the changes of actual pressure data.Aiming at the prediction problem of kick,a prediction model of kick was build based on Independent Component Analysis(ICA).The difference between the actual outlet flow data and the bottom hole flow data observed by the adaptive observer is made.Then the difference is combined with the actual standpipe pressure and back pressure data that were used as the input data of ICA.The independent component of the input data is extracted by ICA,and some statistical values are calculated by the independent component.Some thresholds of those statistical values are constructed and compared to each other to diagnose the drilling conditions.Comparing with other methods,the method proposed in this paper can predict kick condition 5.8 minutes earlier than the conventional method that based on mud pool level.In order to solve the problem that the drilling data are complex and the drilling condition cannot be accurately diagnosed,The Deep Belief Network(DBN)combined with ICA is used in this paper.The independent components of the actual data are used as the input data of DBN and the network parameters are optimized by particle swarm optimization(PSO),then an intelligent drilling condition diagnosis method which combine DBN with ICA is constructed to improve the drilling condition diagnoseresults..Experiments show that the method can diagnose the actual drilling condition quickly and accurately.