This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper IPTC 20254, “Numerical Simulation of Gas Lift Optimization Using Genetic Algorithm for a Middle East Oil Field: Feasibility Study,” by Mustafa AlJuboori and Mofazzal Hossain, SPE, Curtin University, and Omar Al-Fatlawi, SPE, University of Baghdad, et al. The paper has not been peer reviewed. Copyright 2021 International Petroleum Technology Conference. Reproduced by permission. Optimal allocation of gas-injection rates in large fields through a gas lift network is a challenging task. Traditional gas lift optimization programs may prove inefficient or incapable of modeling gas lift optimization in extremely large networks. The key objective of the complete paper is to assess the feasibility of using the genetic algorithm (GA) technique to optimize the allocation of continuous gas lift injection rate in a network of a Middle Eastern oil field with 43 gas lift injected wells through numerical modeling and simulation studies. GA GA is an optimization technique that solves constrained and unconstrained optimization problems through a natural-selection process based on the concept of evolutionary biology, including the fundamental processes of selection, crossover, and mutation. Instead of considering a single point or solution, a population of solutions is designed. The algorithm modifies the population of individual solutions repetitively, selects individuals randomly from the current population as parents, and uses these parents to produce the children for the next generation. The population evolves over successive generations toward an optimal solution. The solution process of GA follows different principles as compared with conventional approaches. These fundamental differences are as follows: - GA uses objective function. - Instead of using a single point, it uses a population of design points or variables that avoid a local optimum solution. - Strings of binary numbers represent the design variables, which are similar to chromosomes in genetics, so discrete and integer problems are naturally applicable to GA. GA avoids falling into local optimums by sustaining multiple points to multiple-point approaches of each generation. In such a method, the probability of surviving the best, fittest solution is high as compared with a low-fitness solution. As a result, the GA mostly will result in a global optimal solution. GA techniques consists of five basic phases: - Initial population - Fitness function - Selection - Crossover - Mutation The process begins with an initial population comprising a set of individuals (i.e., solutions) within a search space. An individual is characterized by a range of combinations of the design variables (or parameters) termed “genes,” which are joined into a string to form a chromosome (or solution). For each solution, the objective function is assessed with regard to its level of fitness. Each solution is paired with another to form a mating pair, used for reproduction of subsequent populations. The mating pairs produce new solutions as offspring, which retain values of the design variables generated by crossover of the parents’ values. In addition, random mutation is introduced to some of the new solution’s values. The process of GA technique is random, even though its optimization permits one solution to establish the level of control.
Read full abstract