#TITLE_ALTERNATIVE#

• Main Author: HILMI WAHYUDI (NIM : 12208091), ARIF
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/25765
• Institution: Institut Teknologi Bandung

More and more mature field that the reservoir pressure is not enough to drain the fluid to the surface makes people think of the way to be able to get the reservoir fluid. One way to do is artificial lift. The one way of the artificial lift that people doing is ESP or Electric Submersible Pump. ESP is pump which is submerged into the fluid in the well that are not able to flow naturally from a low reservoir pressure. Many factor affects the performance of this ESP will include sand problem, gas present on fluid in the well, and electrical problem. The prediction to lowering of setting depth ESP performed using artificial neural network to determine the factor were most responsible for the performance of ESP. Of course we want the long ESP run time with the rate that have economics value. If the ESP placed on the fluid column in the well is too deep, then the TDH (Total Dynamic Head) will be large and the result that ability of pump is not enough to deliver fluid. If the pump placed too shallow in the fluid column then effect of GOR will make ESP performance is reduces because of the of gas decreases the effectiveness of performance of the ESP. By used the artificial neural network then will be able to get the result of time to lowering setting depth of ESP based on factors that affecting performance of reservoir that change the IPR curve. From the result of running simulator the optimum training of ANN with composition of training set, testing set, and validating set is 8:1:1 and the hidden layer is 8.