Abstract:With Due to the massive and heterogeneous nature of power measurement data generated during power grid scheduling, achieving the integration of multi-source heterogeneous data and the construction of a big data warehouse for power grid scheduling error prevention poses a significant challenge in this process. This paper proposes a method for constructing a big data warehouse for power grid scheduling error prevention based on the fusion of multi-source heterogeneous data. Firstly, the Bidirectional Encoder Representations from Transformers (BERT) and the Visual Geometry Group Network with 19 layers (VGG-19) models are used to extract rich semantic and visual features from text and image data, respectively. Then, a joint Kalman filter algorithm is applied for feature fusion of the multi-source heterogeneous data. Secondly, an architecture for the power grid scheduling error prevention big data warehouse is built to enable efficient management, rapid retrieval, and intelligent error prevention of power grid big data. Following that, intelligent retrieval for power grid error prevention is realized based on the big data warehouse and the Classification and Regression Tree (CART) algorithm. Ultimately, the efficacy of proposed method is confirmed through simulated experiments. The simulation results demonstrate that the proposed method has achieved efficient management and rapid retrieval of power grid scheduling data, effectively enhancing the level of intelligent error prevention in power grid scheduling.