Integrated energy systems play a crucial role in the energy transition, but the stochastic fluctuations of wind speed and solar irradiance significantly affect capacity planning and investment decisions. Existing methods struggle to balance adaptability and lifecycle cost optimization. To address this, a bi-level optimization model for capacity allocation is proposed, aiming to minimize the total lifecycle cost while coordinating optimization of capacity allocation and operation scheduling. In the upper-level optimization, a genetic algorithm is used for equipment capacity allocation. Typical wind and solar scenarios are generated using adaptive kernel density estimation combined with an autoregressive model, while Wasserstein distance ensures representativeness and computational efficiency through scenario reduction. The lower-level optimization employs mixed-integer linear programming for operation scheduling, balancing economic efficiency and robustness. The scheduling results are then fed back to the upper level, guiding iterative updates of capacity allocation and forming a bi-level interactive optimization loop. Simulation results show that, compared to traditional optimization methods, the proposed model reduces total lifecycle costs while improving wind and solar utilization and system reliability. This provides both theoretical support and practical reference for the optimal planning of integrated energy systems.