The stability of the surrounding rock is the prerequisite for the safe and stable operation of hydraulic tunnel projects. Taking the hydraulic tunnel of a water resources allocation project as the research object, aiming to improve the prediction accuracy of surrounding rock stability grade of long-distance water conveyance tunnels, a multi-factor hydraulic tunnel surrounding rock stability prediction index system was constructed, and the stability grades of the surrounding rock of different tunnel sections were categorized. Then, G1-improved RS theory was adopted to weight the impact indicators subjectively and objectively, and game theory was used to calculate the combined weights. Finally, based on the unascertained measurement theory (UMT), an improved G1-RS-UMT model was proposed to predict the stability grade of the surrounding rock. The results show that among the nine typical sections of the hydraulic tunnel in the B4 section of a water resources allocation project, the surrounding rock of GS19+590-GS20+050 is in an unstable state, and the other eight sections within the mileage range of GS21+340-GS28+061 are all in partially stable state. According to the comparative analysis, the prediction result of the improved G1-RS-UMT model is more accurate than that of the TOPSIS model, and the degree of consistency between the predicted result and the actual situation has increased from 66.67% to 88.89%. This research provides a new approach for the prediction of surrounding rock stability of hydraulic tunnels with more accurate and effective prediction results.