This paper proposes an improved white whale optimization algorithm (AGABWO) that integrates adaptive Gaussian cloud and alpha distribution mutation strategy to address the problems of low search efficiency and susceptibility to local optima in traditional algorithms for trajectory planning of power inspection drones, as well as the insufficient global search capability of the standard white whale optimization algorithm (BWO). Firstly, the PWLCM chaotic mapping quasi adversarial learning initialization strategy is adopted to generate highly diverse initial trajectories, enhancing the algorithm's ability to cover the solution space in complex environments; Secondly, the Gaussian cloud model is introduced to optimize the global search stage of BWO, and the probability distribution characteristics are used to enhance the algorithm's global exploration accuracy for multi-modal trajectory solutions; Further design an adaptive alpha distribution mutation mechanism to dynamically adjust the disturbance range of individual populations, effectively balance global and local search capabilities, and avoid local extremum traps. By constructing a multi-dimensional complex power inspection scenario model, comparative experiments show that compared with standard BWO, particle swarm optimization algorithm, and improved sand cat swarm algorithm, AGABWO improves the average path length, trajectory smoothness, and convergence stability by 18.3%, 24.7%, and 32.5%, respectively, and can still maintain robustness in high threat source density scenarios. The research results provide theoretical support and algorithm tools for efficient unmanned aerial vehicle inspection tasks in complex power grid environments.