Measurement And Research Of Moisture Content Of Crude Oil Wellhead

In the process of oil exploration，to determine the oil production of oil wells andunderstand the formation of oil and gas content and the change of geological structure，for thewell produced fluid volume ratio of each component in the continuous measurement and thechange of geological structure，in order to optimize production parameters and enhanced oilrecovery. Therefore, crude oil moisture content online accurate measurement has greatsignificance.Microwave method measuring moisture content of crude oil is suitable for the onlinemeasuring of medium and high water cut stage, and this method is sensing the advantages ofsimple structure, convenient installation. In this paper, in order to improve the accuracy ofinstrument we can use on-site measuring water content of crude oil wellhead, to postprocessing of the data when considering the influence of temperature on measurementaccuracy and considering temperature compensation.The core content of this research is mainly divided into two parts: one part of usingSPSS software for field measurement data of multivariate nonlinear regression analysis, so asto build current signal and the function relation between the temperature and moisture contentof crude oil, then build the mathematical model and test the correlation coefficient, furtheranalysis of the influence law of various factors, finally combining with the maple softwarecan calculate the water content of crude oil under any current signal and temperature, thencomparing theoretical calculation value and real value, calculate the average relative error is3.428%, indicate the forecasting model is more reasonable; the other part is through theneural network toolbox in matlab programming,using the BP neural network model forpredicting water cut of crude oil which has been established, and according to theconvergence speed of BP neural network algorithm, easy to fall into local minimum point andother shortcomings, adopt double polarity compression function replace sigmoid function asthe activation function, training of convergence with the LMBP algorithm instead oftraditional gradient descent method to improve the model, ultimately improving themeasuring accuracy of the water cut measuring instrument.