Iranian Institute of Research and Development in Chemical Industries (IRDCI)-ACECRNashrieh Shimi va Mohandesi Shimi Iran1022-776841220220823Placement Optimization of Production and Injection Wells Using Parallel Genetic Algorithm in a Case StudyPlacement Optimization of Production and Injection Wells Using Parallel Genetic Algorithm in a Case Study377387251404FAMahboobehBaghbanSchool of Chemical Engineering, Iran University of Science and Technology (IUST), Tehran, I.R. IRANMehdiAssarehSchool of Chemical Engineering, Iran University of Science and Technology (IUST), Tehran, I.R. IRANMohammad TaghiSadeghiSchool of Chemical Engineering, Iran University of Science and Technology (IUST), Tehran, I.R. IRANJournal Article20181217<em>Due to the complexities of oil reservoir simulation models, there is a need to find an optimization method that can have good accuracy and speed while reducing computational costs. In this regard, in this study, a parallel genetic algorithm has been used to accelerate the optimizer in order to prevent the increase of the computational load of the well placement problem and the prolongation of the execution time. Instead of one population of chromosomes (i.e., the same optimization variables), this algorithm works with several populations, which exchange chromosomes with each other. The proposed mathematical model is a single-objective problem, which is to maximize the present value of the project in the location of wells. The price of oil and the cost of separating water and gas are considered in this economic function. In evaluating the objective function, the CPU of a single computer, or multiple networked computers, each in parallel, are responsible for calculating the chromosomesâ€™ fitness of parallel populations. After calculating the target function or suitability of the chromosomes of all populations (in which each population has common genetic function), migration between populations takes place. A case study has been used to validate the implementation of this method. In the proposed model, the location of vertical wells in an oil tank has been verified. As a result of the optimization, the wells are located at five points, four injection wells at the corners, and one production well at the center of the model, and in addition to reducing optimization time, the number of simulator performances in parallel processing mode is significantly reduced. In this research, the performance of genetic and parallel genetic algorithms for location optimization on the same reservoir model is shown. The execution time of optimization between genetics and parallel genetics is 7100 seconds and 1800 seconds, respectively. As it is known, the use of PGA has quadrupled the execution time in the studied oil reservoir.</em><em>Due to the complexities of oil reservoir simulation models, there is a need to find an optimization method that can have good accuracy and speed while reducing computational costs. In this regard, in this study, a parallel genetic algorithm has been used to accelerate the optimizer in order to prevent the increase of the computational load of the well placement problem and the prolongation of the execution time. Instead of one population of chromosomes (i.e., the same optimization variables), this algorithm works with several populations, which exchange chromosomes with each other. The proposed mathematical model is a single-objective problem, which is to maximize the present value of the project in the location of wells. The price of oil and the cost of separating water and gas are considered in this economic function. In evaluating the objective function, the CPU of a single computer, or multiple networked computers, each in parallel, are responsible for calculating the chromosomesâ€™ fitness of parallel populations. After calculating the target function or suitability of the chromosomes of all populations (in which each population has common genetic function), migration between populations takes place. A case study has been used to validate the implementation of this method. In the proposed model, the location of vertical wells in an oil tank has been verified. As a result of the optimization, the wells are located at five points, four injection wells at the corners, and one production well at the center of the model, and in addition to reducing optimization time, the number of simulator performances in parallel processing mode is significantly reduced. In this research, the performance of genetic and parallel genetic algorithms for location optimization on the same reservoir model is shown. The execution time of optimization between genetics and parallel genetics is 7100 seconds and 1800 seconds, respectively. As it is known, the use of PGA has quadrupled the execution time in the studied oil reservoir.</em>https://www.nsmsi.ir/article_251404_a73cfbacc4f6d5532920d1cfa49f94c6.pdf