Power system state estimation method based on denoising diffusion probability model data generation and multi-source data fusion

State estimation is fundamental for power system situational awareness but faces challenges of data scarcity, heterogeneous data fusion difficulties, and limited accuracy. This study proposes a novel state estimation method integrating Denoising Diffusion Probabilistic Models (DDPM) with multi-source data fusion. Firstly, we constructed a DDPM data generation model with embedded physical constraints, thereby effectively addressing the issue of data scarcity. Secondly, a weighted fusion strategy based on credibility was designed to integrate heterogeneous data sources. Finally, a least squares estimation algorithm based on dynamic decoupling and adaptive fusion was proposed, which improved the performance of state estimation. The IEEE 57-node system test shows that this method achieves a better balance between data quality generation and training cost. The state estimation error after integrating multi-source data is lower, and the convergence speed is 33% faster. In the presence of noise and data missing, the errors are reduced by 51% and 63% respectively.