Journal of Materials Science and Engineering A 2 (9) (2012) 661-667 A Hybrid Genetic Algorithm to Interpolate and Predict Average Reservoir Pressure Saber Elmabrouk 1, 2 and Ezeddin Shirif 1 1. Faculty of Engineering, University of Regina, Regina, SK, Canada S4S 0A2 2. University of Tripoli, Tripoli, Libya Received: April 18, 2012 / Accepted: May 05, 2012 / Published: September 10, 2012. Abstract: Insight into average reservoir pressure and its change in time, plays a critical role in reservoir development and management, economic evaluation, obtaining a hydrocarbon in place, computing rock and fluid characteristics, reservoir material balance calculations, pressure maintenance projects, surface and subsurface facility designs and predicting reservoir behavior. In particular, almost every well intervention requires insight with respect to average reservoir pressure. Traditionally, average reservoir pressure is obtained via a pressure build-up test which measures long-term built-up pressure when the producing wells are shut in. It is worth noting that the average reservoir pressure measurements should be updated periodically because reservoir pressure changes as fluids (oil, gas and water) are released from the reservoir. However, a significant economic impact is caused by shutting in the producing wells during the entire build-up test. The purpose of the work is to introduce a neural network model as an alternative tool to interpolate and predict average reservoir pressure without closing the producing wells. Key words: Average reservoir pressure, genetic algorithm, neural network model, modeling. 1. Introduction Average reservoir pressure is obtained via a pressure build-up test which measures long-term built-up pressure when the producing wells are shut in. In high permeability reservoirs, this may not be a significant issue, but in medium to low permeability reservoirs, the shut-in period during the entire build-up test may last several weeks before a reliable reservoir pressure can be estimated. This loss of production opportunity, as well as the cost of monitoring the shut-in pressure, is often unacceptable. It is of great practical value if the average reservoir pressure can be obtained from the historical production and reservoir pressure data without having to shut-in the well. Shutting in the wells during the entire pressure build-up test means a loss of income. The purpose of the work, however, is to Corresponding author: Shirif Ezeddin, associate professor, research fields: reservoir simulation, well testing, gas engineering, enhanced oil recovery, in heavy and light oil reservoirs and multiphase flow in porous media secondary recovery. E-mail: ezeddin.shirif@uregina.ca. introduce a neural network with backpropagation (BP-NN) model trained with genetic algorithm (GA) as a new alternative tool to predict and interpolate average reservoir pressure without shutting in the producing wells. This technique is suitable for constant and variable flow rates. A GA was employed, in this study, for the evolution of connection weights in the BP-NN model in order to improve upon model performance and support the simultaneous optimization of the connection weights. Some researchers searched the connection weights of NN using the GA instead of local search algorithms, including a gradient descent algorithm. They suggested that global search techniques, including the GA, might prevent NN from falling into a local optimum [1]. Thus, a combination of BP-NN and the GA based optimization, termed the “BP-NN-GA Model” was proposed to map relationships which controls reservoir oil, gas and water production performance in order to interpolate, predict and estimate the current average DAVID PUBLISHING D