Abstract: The maintenance space of hydropower units is highly restricted by numerous complex factors related to generation, transmission and utilization. How to quantify the influence of these factors on maintenance arrangements constitutes the key to enhancing the rationality of dispatching plans. To this end, a method for analyzing the influencing factors of hydropower maintenance based on interpretable machine learning is proposed. With the long series of maintenance space of cascade hydropower stations as the basis, the spatial-temporal variation characteristics are refined. The random forest algorithm is employed to establish a nonlinear relationship model between multiple core influencing factors such as water inflow, peak regulation, holidays, ecological dispatch, power exchange, line maintenance, and power supply guarantee and the maintenance space. The cooperative game theory is adopted to establish the analysis method of the influencing factors of cascade hydropower maintenance, quantifying the interactive influence of different factors on hydropower maintenance and determining the dominant factors affecting the maintenance space of each hydropower station. The results indicate that: 1) The random forest model achieves the best fit between multiple influencing factors and maintenance space, with an R² of 0.94, a Mean Squared Logarithmic Error (MSLE) of 44.72, and a Symmetric Mean Absolute Percentage Error (SMAPE) of 0.02; 2) Water inflow, peak regulation, and holidays are the most critical factors affecting maintenance space, with relative importance weights of 52.89%, 20.95%, and 10.86%, respectively; 3) The influencing factors of maintenance for different hydropower stations within a cascade vary significantly depending on the tasks they undertake. The method proposed in this paper is conducive to improving the rationality and refinement level of the power generation and maintenance arrangements for large-scale cascade hydropower stations.