Short-term high-concurrency thunderstorm activities pose a notable threat to the safe and stable operation of transmission lines. Timely and accurate thunderstorm forecasts are of great significance to guide the management and protection of transmission lines against lightning strikes. This paper selects three typical geomorphic regions in Hubei Province, combined with meteorological and topographic elements, and establishes a three-hour thunderstorm forecast model based on the random forest model. The results indicate that: (1) the random forest-based model can effectively predict the 3 h thunderstorm activity in all three geomorphic regions, and the threat score (TS) is above 88%. (2) The introduction of terrain factors helps to improve the accuracy (PC) and TS of thunderstorm forecasts. Besides, the higher the spatial resolution of the terrain data is, the higher the accuracy of the forecast. However, the impact of terrain factors on forecasts is far smaller than that of meteorological factors. (3) There are certain differences in the accuracy of thunderstorm forecasts between the different geomorphic regions. The forecast results are the best in gentle slope areas, followed by flat slope areas, while the forecast results are the worst in steep slope areas. Dynamic adjustment of the influencing factors of thunderstorms according to the different geomorphological characteristics is an effective means to realize accurate prediction of the regional thunderstorm potential.