Aiming at the deficiencies of the traditional covariance cross-fusion algorithm in terms of computational complexity and fusion accuracy, this paper proposes an improved multi-sensor data fusion and optimization algorithm for enhancing the accuracy and speed of the grid flooding intelligent warning system. The method combines the Drosophila optimization algorithm to optimize the fusion coefficients, the dynamic weight allocation method based on information entropy and the attention mechanism to enhance the ability of capturing the key data features to improve the robustness and adaptability of the system. The efficient fusion of data and decision layers is achieved by designing a two-stage fusion method combined with Kalman filtering, adaptive weighted averaging and ELM neural network with ARO optimization. The results show that the proposed algorithm outperforms traditional methods in terms of root mean square error, fusion time, and warning accuracy, especially in high-noise environments and large-scale sensor networks. The study provides a reliable technical support for the grid flooding intelligent warning system, which has important theoretical value and practical application prospects.
