Corrosion Behavior And Prediction Model Of Oil And Gas Gathering Pipelines Under Multi-Factors

The oil and gas gathering pipelines are known as the lifeline of oil-gas production and operation,as the connecting line between the units of ground system in oil and gas field development.As most of China’s oil and gas fields enter the middle and late stages of development,the integrated water content rises.Besides,the working environment turns into a complex mixed system under the three-phase flow state of oil-gas-water,while the transport medium often contains acidic components such as H₂S,CO₂and inorganic ions such as Cl^-,Ca²⁺,Mg²⁺,etc.Therefore,steel pipelines are prone to corrosion,perforation and leakage.In general,oil and gas are flammable and explosive if an accident occurs,it may cause huge human and economic losses as well as environmental pollution,and even cause certain negative social impacts.Then,the in-depth study of the corrosion behaviour of oil and gas gathering pipelines and the accurate prediction of their corrosion rates are particularly necessary.In this work,a systematic analysis of the causes and influencing factors of corrosion of typical oil and gas gathering pipelines is employed to clarify the significant factors affecting corrosion.Meanwhile,the response surface methodology is introduced to optimize the corrosion test scheme,and the multivariate high-order regression equations for predicting corrosion rate was established.In addition,the traditional BP artificial neural network is optimized using genetic algorithms to obtain a numerical model for corrosion rate prediction.The specific research carried out and the main conclusions obtained are as follows.（1）By carrying out leakage failure analysis on two typical oil and gas gathering single well pipeline and converging pipeline,the influencing factors of pipeline failure were comprehensively studied based on physio-chemical tests such as metallographic organization and chemical composition,as well as corrosion characterization based on scanning electron microscopy,energy spectroscopy and X-ray diffraction.The results show that the failure of oil and gas gathering pipelines is internal corrosion,which including uniform corrosion and local pitting occurring on the pipe wall.The main corrosion influencing factors come from the pipeline material,water,inorganic ions(Cl^-,Ca²⁺,Mg²⁺),acidic gases（CO₂,H₂S）and operating conditions（temperature,pressure）.（2）The corrosion behaviour of three carbon steels,20#,20G and L360NB steel,commonly used in oil and gas gathering pipelines,was studied by different methods,receptively.There are electrochemical corrosion tests and high temperature and high pressure dynamic reactor corrosion tests,simulating typical oil and gas gathering conditions in the oilfield.Under room temperature and pressure conditions,the open circuit potential,electrochemical impedance spectra and polarisation curves of above steels showed that the trend of corrosion of L360NB steel was relatively small,and the trend of corrosion of 20G and 20#steel was close.Under the simulated high temperature and pressure conditions and corrosive media conditions,however,the three carbon steel materials had obvious uniform corrosion and pitting corrosion,but the difference in corrosion degree was small.Overall,the three pipelines in the same corrosive environment,the corrosion trend is relatively similar.（3）Using the ANOVA method,the high temperature and high pressure dynamic reactor corrosion tests were designed for 20#carbon steel under the interaction of six main factors such as different CO₂partial pressure,H₂S partial pressure,Cl^-content,Ca²⁺+Mg²⁺content,temperature and flow rate to study the correlation between the multi-factor and multi-level effects on the corrosion of the pipelines.The results show that CO₂partial pressure,H₂S partial pressure,Cl^-content and temperature have a very significant effect on the occurrence of uniform corrosion and pitting corrosion of 20#material,while the content of Ca²⁺+Mg²⁺and flow rate do not have a significant effect on them.The significance of uniform corrosion is ranked as:H₂S partial pressure>CO₂partial pressure>temperature>Cl^-content>Ca²⁺+Mg²⁺content>flow rate;The significance of pitting corrosion is ranked as:Cl^-content>CO₂partial pressure>H₂S partial pressure>temperature>Ca²⁺+Mg²⁺content>flow rate.（4）The response surface methodology was used to design a response surface corrosion test scheme under four significant influencing factors,such as CO₂partial pressure,H₂S partial pressure,Cl^-content and temperature,and to construct a theoretical model for predicting the rate of uniform corrosion and pitting corrosion under the interaction of various corrosion influencing factors.The obtained values of R²of the two models were 0.856 and 0.889respectively,which indicated that the observed values and response values of the prediction model had obvious correlation,and the accuracy and stability of the model was good.In addition,the model also identifies a pattern of variation in the combination of different factors:for uniform corrosion,the most significant interactions are between CO₂partial pressure&temperature and CO₂partial pressure&H₂S partial pressure;for pitting corrosion,none of the interactions are significant.（5）With the introduction of traditional BP artificial neural network methods,the rate of uniform corrosion and pitting corrosion prediction models were obtained by training,testing and validating 25 sets of sample data,through setting and selecting key parameters such as network structure,number of nodes in the hidden layer,transfer function,training method,target error and learning rate,etc.The R²of the two models were 0.508 and 0.162,respectively,which indicating that the predicted values had large deviations from the true values.（6）The optimized BP neural network prediction models for the rate of uniform corrosion and pitting corrosion were obtained by integrating genetic algorithms to improve the weights and thresholds of the optimized BP neural network,through the creation of the network structure,initialization of the population,determination of the fitness function,implementation of genetic operations and determination of the operating parameters.Meanwhile,the R²of the two models were 0.999 and 0.991.Based on above results,it further demonstrated that the accuracy of uniform corrosion rate and pitting rate prediction models were significantly improved by the genetic algorithms.In addition,it is more accurate than those of the traditional BP neural network and response surface method.