Metaheuristic profiling to assess performance of hybrid evolutionary optimization algorithms applied to complex wellbore trajectories

• Metaheuristic profiling aids development and performance of hybrid evolutionary optimization algorithms.
• A metaheuristics “toolbox” enables effective development of hybrid evolutionary algorithms is supported.
• Genetic, particle swarm, bee colony, ant colony, harmony search, cuckoo search and bat flight algorithms are evaluated.
• Fat-tailed distributions sampled chaotically provide flexible metaheuristics for balancing exploration and exploitation.
• VBA-driven Excel models lend themselves to metaheuristic profiling and performance analysis of hybrid algorithms.

Metaheuristic profiling is proposed as an effective technique with which to evaluate the relative contributions of the metaheuristic components of hybrid evolutionary optimization algorithms in progressing searches of feasible solution spaces to locate global optimum values of their objective functions. Although many useful evolutionary algorithms have been successfully proposed and tested to solve a wide range of complex mathematical optimization problems, when applied to real-world optimization tasks their performance can often be improved by hybridization with other metaheuristics. A case is made here that in developing optimization algorithms for specific practical applications it is better to treat the available evolutionary algorithms as part of a “toolbox” of metaheuristic components that can be configured in various hybridized combinations. The technique of metaheuristic profiling is evaluated as means of identifying the relative contributions of individual metaheuristic components in contributing to the discovery of optimum solutions over multiple iterations of hybrid algorithms. The metaheuristic profiling technique of a toolbox of metaheuristic components is evaluated in terms of applying seven hybrid evolutionary algorithms to optimize a previously studied complex well-bore trajectory optimization problem. The seven hybrid evolutionary algorithms developed with multiple metaheuristics are built upon standard: genetic; particle swarm; bee colony; ant colony; harmony search, cuckoo search and bat flight algorithms. Pseudocode for each of the hybrid algorithms studied are provided in an appendix. These codes identify the metaheuristics included and the sequence in which they are applied in the hybrid algorithms. All seven hybrid algorithms are coded in VBA based in Microsoft Excel with the assistance of the metaheuristic profiling technique, to provide reliably reproducible solutions to well-bore trajectory design optimization. Analysis of metaheuristic performance also confirms the benefits of fat-tailed distributions, sampled chaotically, in a novel way, to drive certain metaheuristics.

Constructing hybrid evolutionary algorithms by selecting appropriate metaheuristics from a “toolbox” of alternative metaheuristics to add to standard evolutionary algorithms is an effective technique to improve and/or refine the performance of the hybrid evolutionary optimization algorithms. The technique of metaheuristic profiling (MHP) is a useful approach to establish the relative contributions of specific metaheuristics to the overall performance of hybrid evolutionary optimization algorithms. The Metaheuristic profile (upper figure) is for a 500-iteration execution of the hybrid cuckoo search algorithm consisting of eight distinct metaheuristics (labelled 1 to 8 on the vertical axis) and applied in sequence 1 to 8 in the algorithm. Each iteration of the algorithm derives 150 new solutions and ranks them with the best solution for the objective function assigned rank#1. Each solution is coded according to the metaheuristic that generated it (labelled 1 to 8 on the vertical scale of the upper figure). The MHP illustrates the Rank#1 to #10 solutions generated by each component metaheuristic of the algorithm in each iteration.Figure optionsDownload high-quality image (267 K)Download as PowerPoint slide