Improved Marine Predator Algorithm with Multi-Stage Mutation and Its Application

Aiming at the disadvantages of poor robustness, imbalance between exploration and exploitation and rapid individual assimilation in the later stage of iteration, an improved marine predators algorithm with multistage-mutation is proposed. Monte Carlo mutation is performed on all individuals in the initialization stage of the algorithm to traversal improves individual quality to limit the randomness of operation results in order to enhance the robustness of the algorithm; The golden sine mutation strategy is introduced to update the secondary position of Levy flight group and Brownian motion group, re-screen the high-quality individuals who build the parallel architecture to check and balance the exploitation performance and exploration performance and strengthen the adaptive ability of the algorithm; The non-uniform Gaussian mutation strategy is adopted to realize individual mutation in the later stage of iteration, control the degree of individual assimilation, and effectively reduce the probability of the algorithm falling into local optimization. The benchmark function is used to test the performance of the algorithm. The experimental results show that the optimization performance of the improved marine predator algorithm with multi-stage-mutation is significantly improved compared with the original algorithm. The improved algorithm is used to optimize the parameters of least squares support vector machine and applied to transformer fault diagnosis experiment, and its engineering practicability is verified by the improvement of the fault diagnosis accuracy.