چكيده به لاتين
Measuring the flow rate of the produced fluids from each well is one of the most important parameters for reservoir optimization and management. Conventional methods for measuring the flow rate of produced fluids from wells such as allocation, orifice and choke calculations are not accurate and have uncertainties. These methods also do not measure the flow rates in real-time, and the calculated flow rates using them are reported daily. On the other hand, installing a multiphase flow meter for each well is very costly. This thesis presents a method that can measure the flow rate using temperature, pressure and sound data with a multilayer perceptron neural network. Since the sound signals generated by the fluid stream are not available in operational conditions, experimental setups have been used to collect the required data. These experimental setups are designed to simulate the conditions of fluid production from the reservoir. The results of the developed neural network models show that these models can measure the flow rate of water flowing through the pipe with an accuracy of about 99.15%, flow rate of diesel flowing through the pipe with a certainty 99.09% and flow rate of air flowing through the pipe with an accuracy 99.67% compared to the flow meters used in experimental setups. On the other hand, these models have good response speed, which enables them to be used in real-time to measure the flow rate. Further studies show that the presence of orifice or venturi in the pipe enables neural network models to measure the flow rate more accurately and faster. Overall, the results show that the method presented in this study for measuring flow rate is a fast, inexpensive, accurate, flexible and robust method.
