Flexible Adaptive Marine Predator Algorithm and Its Application in Fault Detection for Wind Turbines

With the increasing dimensions of operational data for wind turbines, traditional expert empirical methods, such as fuzzy logic systems and fault tree analysis, struggle to provide accurate fault diagnosis amid high-dimensional complex data. Meta-heuristic algorithms excel in analyzing such data, relying on their excellent global optimization capabilities. Although the marine predators algorithm (MPA) is one of the most widely applied meta-heuristics, it has encountered challenges in the fault diagnosis of wind turbines, i.e. low convergence rates and proneness to falling into local optima, which adversely affect the real-time capability and accuracy of diagnosis. To solve these issues, this paper proposes a flexible adaptive marine predator algorithm (FAMPA) by introducing an adaptive parameter mechanism to enhance the original MPA. FAMPA allows for flexibly adjusting population location changes to optimize the balance between global exploration and local exploitation, thus significantly improving both convergence rates and optimization accuracy. In experiments involving nine benchmark functions, FAMPA outperformed traditional MPA in convergence rate, optimization accuracy, and robustness. To facilitate the application of FAMPA in the field of wind turbine fault diagnosis, two new evaluation indexes were established as objective functions, which were then verified using three datasets of faults from actual applications. Verification results showed the ability of FAMPA in effectively extracting key features from high-dimensional data, reducing data dimensions, and improving the accuracy and real-time performance of fault diagnosis models based on machine learning. Through theoretical analysis and empirical research, this study verified the applicability and advantages of FAMPA in the fault diagnosis of wind turbines, providing an efficient and reliable method in this field.