This paper presents the Multi-Domain Data Fusion Method for Power Business Data, centered on a Data Middle Platform, designed to address the complexities in modern power systems. This approach integrates diverse data sources using a probabilistic graphical model-based framework, enhancing the accuracy and efficiency of power system analysis and management. The framework combines physical models with machine learning algorithms, ensuring robust and adaptable system analysis. The Gaussian Belief Propagation model is employed for effective state estimation in challenging environments. The study demonstrates the method's potential to improve operational efficiency, reliability, and sustainability of power systems, highlighting its significance in the context of advanced data management and the evolving energy sector.