The electrical power industry has experienced an unprecedented pace of digital transformation as a prevailing economic trend in recent years. This shift towards digitalization has resulted in an increasing interest in the collection of real-time equipment condition data, which provides opportunities for implementing sensor-driven condition-based repair. As a result, there is a growing need for the development of generator maintenance scheduling to consider probabilistic equipment behavior, which requires significant computational efforts. To address this issue, the research proposes the use of a meta-heuristic league championship method (LCM) for generator maintenance scheduling, considering random generation profiles based on generation adequacy criteria. The experimental part of the study compares this approach and its modifications to widely used meta-heuristics, such as differential evolution and particle swarm methods. The identification and demonstration of optimal method settings for the generation maintenance scheduling problem are presented. Subsequently, it is illustrated that employing random league scheduling expedience can reduce the variance of objective function values in resulting plans by over three times, with values of 0.632 MWh and 0.205 MWh for conventional and proposed techniques respectively. In addition, three approaches are compared to assess generation adequacy corresponding to different schedules. The study emphasizes the efficacy of employing the LCM approach in scheduling generator maintenance. Specifically, it showcases that among all the methods examined, the LCM approach exhibits the lowest variance in objective function values, with values of 38.81 and 39.90 MWh for LCM and its closest rival, the modified particle swarm method (MPSM), respectively.