Determination of optimal well trajectories constitutes one of the problems of field development planning. This problem must be solved when a well pad pattern is being selected; all existing engineering constraints must be taken into consideration; at the same time the cumulative length of all trajectories must be minimal to reduce capital costs for drilling.
Well trajectories are classified according to the number of wellbore intervals. A well section having a steady dogleg severity is taken as an interval. A number of parameters define every well section: length, dogleg severity, inclination angle change. The values of parameters can be changed within the limits of the engineering constraints. A minimum distance between two wells is 1.5 % of the trajectory length.
A genetic algorithm based on the mechanisms of natural selection and inheritance has been chosen to solve the problem. The algorithm defines values for unknown parameters of the trajectories which must meet the engineering constraints. The purpose of this algorithm is to minimize the trajectory length. The algorithm is searching from the initial population of solutions – trajectory options, their values of parameters being filled in randomly. A pair of solutions is selected from the population in each iteration. Their parameters are combined and set to a new solution. If trajectories created from a new solution are shorter than those created earlier, then the new solution is placed into the population. The performance of the search is provided for by including expert actions on selecting the values of unknown parameters into the mechanism of new solution generating. The algorithm generates new solutions until the desired result is reached.
The algorithm described in this work has been implemented in "Trajectory design" software, and it is included into Rosneft corporate software product. The software calculates optimal well trajectories fast, and automatically prevents possible wellbore collisions.
"Trajectory design" software has been tested using Rosneft oil fields models. Calculated trajectories proved to be shorter than real ones by around 13-15%. This proves high performance of the software for well trajectory computations.