Hybrid cuckoo search optimization algorithms applied to complex wellbore trajectories aided by dynamic, chaos-enhanced, fat-tailed distribution sampling and metaheuristic profiling

Abstract

Hybrid cuckoo search optimization (hCSO) algorithms are described and developed in comparison with the standard cuckoo search algorithm (Yang & Deb, 2009). The hCSO involves potentially eight metaheuristic components that complement each other in their search contributions and operate as a “tool box” of modules such that six of them can be easily switched on or off. The metaheuristics are coordinated to progress through five distinct steps that constitute hCSO: (1) initialization; (2) exchange; (3) modification; (4) replacement; (5) metaheuristic labelling, ranking and carry forward. Key amendments introduced to hCSO involve replacing Levy flight solution space sampling with stochastic random sampling of simpler fat-tailed distributions with dynamic sampling windows that move through the distribution as iterations of the algorithm advance. The randomly-extracted samples are further adjusted with scaled-chaotic sequences to provide more flexibility and control over the granularity of the sampling of the solution space. In addition other metaheuristics are added to the standard CSO that improve the balance of the algorithm between local and global searching. Three of the metaheuristics include chaotic adjustments to dynamic stochastic sampling of search metrics distributions (fat-tailed and other, highly non-linear, stepped ranges). Several configurations of the metaheuristics available in the hCSO algorithm are applied to a well-reported complex wellbore trajectory optimization problem. Their performance is compared with the aid of metaheuristic profiling and statistical analysis of the minimum total measured depth (TMD) found in multiple sequential runs. Several configurations of the hCSO are shown to work efficiently in locating the global optimum, avoiding being trapped by the many local optima within the solution space. They do so requiring less computational time than six other evolutionary algorithms evaluated with the same number of iterations and population of generated solutions and similarly developed in Excel VBA code to facilitate metaheuristic profiling.