The primary reflector of the Leighton Chajnantor Telescope is usually affected by various environmental loads and panel misalignments, both of which can significantly degrade its radiation pattern performances. To address these effects, this paper focuses on optimizing actuator adjustments and proposes a Mixed-integer Nonlinear Programming model that takes key radiation pattern performance indicators as the optimization objective. Recognizing the limitations of traditional optimization methods, which are prone to falling into local optima due to high-dimensional and complex nonlinear constraints, and struggle to optimize both actuator positions and adjustment values simultaneously, we introduce an enhanced genetic algorithm combined with a heuristic actuator selection rule, which efficiently solves the beam performance optimization problem by regulating part of the actuators. Notably, the approach demonstrates a significant reduction in the magnitude of the pointing deviation, achieving approximately 80% of decrease when utilizing 20 actuators. Furthermore, with the application of 40 actuators, the method successfully reduces the rms surface error induced by gravitational deformation to below 12 μm across all zenith angles. The proposed method is applied to typical deformation scenarios caused by environmental loads and panel misalignments, resulting in actuator adjustment schemes for varying numbers of actuators. A systematic evaluation of each scheme’s impact on beam performances is provided, and further analysis reveals the relationship between actuator position distribution, adjustment values, and surface error. Additionally, the influence of surface error distributions on the selection of key actuators is systematically revealed.