Analysis And Recognition Study On The Abnormal Pressure Fluctuations Occurred In The Crude Oil Pipeline

With the gradually formation of our national oil and gas pipeline network,safe operation of pipeline in service has gradually become the focus of attention for the oil and gas industry.Abnormal pressure fluctuations appears on the interface of SCADA system usually characterizes condition changes in the pipeline transportation.If these changes fails to be judged timely or effectively,adverse effect may be brought on the production plan of the downstream refineries and oil depots,even the pipeline accident emergency.Currently,researches on the abnormal pressure of oil and gas pipeline at home and abroad mainly focus on specific areas such as the pipeline leakage,little work has been done from a systematic point of view to establish a system of the pipe stress analysis and recognition.This is the idea source and the innovation of this thesis.On the basis of the status quo and research advancement of fuzzy mathematics,signal analysis,machine learning and other disciplines from cross frontier research,the thesis goes from two aspects.For one thing,the mechanism,deep inducement,path and induction and decision-making of abnormal fluctuations are explored.The induced fault tree analysis(FTA)model and decision-making tree(DT)model are all established,which provides a reference for the analysis of fluctuation events as well as the optimization of control measures.For another,the application of wavelet transform(WT)on signal processing of the pressure fluctuation is studied.Two mutually complementary are proposed according to the signal complexity.Using knowledge based approach to identify simple pressure signals.The theoretical basis of the traditional pressure-flow method is analyzed.Typical working conditions are classified and summarized based on different mechanisms.Based on the typical station field,realization strategy of intelligent identification is also put forward.Using artificial intelligence to identify complex pressure signals.Characteristic waveforms are achieved based on the results of wavelet singularity analysis.Twenty typical features of signal extraction in time domain,frequency domain and energy domain are analyzed using principal component analysis(PCA).Intelligent judgement could be realized by two plans: constructing the optimal feature vector or establishing feature library.Finally,the conception of industrial application and its conceptual model are put forward.The research shows that internal factors such as oil compressibility,pipeline elasticity,kinetic energy and inertia of oil flow enables pipeline to have the characteristics and dynamics to generate abnormal pressure fluctuations.On the basis of three important events,they could be reduced to a certain extend from five aspects.Signal de-noising and singular analysis based on wavelet analysis yields good results.Typical conditions can be summarized as three categories,namely the dynamic type,blockage type and leakage type.Transportation conditions of typical station can be identified automatically.The feature vectors constructed by the retained principal elements proves to be the optimal feature vectors while the feature library model constructed by the first principal components proves to be the optimal feature library model.Based on the above two approach,the neural networks created accordingly could achieve accurate identification of complex pressure abnormal fluctuation signals.The research content of this paper can provide some reference for engineering application.