Prediction Of Hydrate Formation Based On Data-driven

In the development of oil and gas field,hydrate blockage may lead to well production reduction,pipeline pressure difference increasing,and even serious pipeline damage,shutdown and workover.Therefore,hydrate prediction of pipelines has become an important technical issue in oil and gas development.The mechanism model or empirical model is mainly adopted to predict hydrate.But it cannot adapt to various complex components.At present,with the development and application of data-driven in various industries,the technology in the oil industry is still in the research stage.The data-driven methods are proposed to predict hydrate formation in this paper.Firstly,the mechanism model of hydrate formation is investigated.The necessity of improving the model is analyzed through calculation.Different data-driven methods for predicting hydrate formation are analyzed.Through analysis,it is found that the cumbersome measured data won't be conducive to the convergence of the data-driven model.Therefore,dimensionality reduction of parameters in the model is carried out.Then the coefficients in mechanism model are determined.At last,4 groups of experimental data of hydrate formation are collected.Mechanism model,empirical method,data-driven methods are used to calculate the hydrate formation pressure,and the results were compared with the outcome calculated by commercial software Pipesim and experimental results.Finally,the slurry flow mechanism model is used to predict the hydrate flow.The evaluation show data-driven methods in this paper are obviously better than those of the traditional models,and they are in good agreement with the experimental data.Especially,the data-driven method based on the mechanism model has better predictions,the minimum error can be as low as 0.00825 MPa.