Estimation of flow rates of individual phases in an oil-gas-water multiphase flow system using neural network approach and pressure signal analysis

Up until now, different methods, including; flow pressure signal, ultrasonic, gamma-ray and combination of them with the neural network approach have been proposed for multiphase flow measurement. More sophisticated techniques such as ultrasonic waves and electricity, as well as high-cost procedures such as gamma waves gradually, can be replaced by simple methods. In this research, only flow parameters such as temperature, viscosity, pressure signals, standard deviation and coefficients of kurtosis and skewness are used as inputs of an artificial neural network to determine the three phase flow rates. The model is validated by the field data which were obtained from separators of two oil fields and 6 wells over ten-month with 8 h interval (totally 5400 sets of data). A linear relation can be observed between the actual data and the predictions which were obtained from separators and neural network approach, respectively. Furthermore, it is shown that using feed forward neural network with Levenberg–Marquardt algorithm which has two hidden layers is sufficient to determine the flow rates. Also, it is tried to see the effect of flow regimes on the results of neural network approach by determining kurtosis and skewness coefficients for different flow regimes in a horizontal pipeline.