<p>Accurately predicting wind speed is a critical challenge for the wind power industry, as it is essential to ensuring grid stability and optimizing wind resource allocation. Reliable wind speed prediction hinges on selecting appropriate prediction models, yet this process is hindered by two key issues: the inherent high randomness and indirectness of wind speed data, and the limitations of traditional model evaluation methods, which are typically costly and lack interpretability. To resolve these problems, this paper proposes a lightweight model selection scheme based on the improved fuzzy comprehensive evaluation (IFCE) method. By constructing a data-driven target evaluation system, the proposed scheme effectively alleviates the strong subjectivity of the original fuzzy comprehensive evaluation method. Furthermore, integrating Variational Mode Decomposition and the Multi-objective Salp Swarm Algorithm, we develop a wind speed prediction system that features comprehensive multi-index evaluation capability and high interpretability. To fully verify the system performance and the universality of the IFCE method, experiments are conducted across multiple scenarios, including point prediction, interval prediction, and multi-variable prediction. Results demonstrate that the proposed system achieves excellent performance in all test scenarios, confirming that the IFCE method, as a model selection tool, exhibits strong adaptability and stability. This study provides a novel and effective solution to the model evaluation problem in wind speed prediction.</p>