Given the inherent complexity of human behavior, situational dependencies, and fragmented data, accurately assessing the risks arising from human operational errors in oil and gas pipelines remains a formidable challenge. To address these limitations, this study proposed a three-dimensional fuzzy dynamic risk assessment model with dynamic cloud weights (DW-TDFE). First, based on seven human operational errors, a risk assessment index system was established. Second, historical data (including accident frequency and casualty records) were utilized to determine initial indicator weights using the analytic hierarchy process (AHP). To simulate real-world disturbances, Gaussian noise was introduced. A combination of a time decay factor and a hyperbolic tangent function was employed to smooth fluctuations, yielding dynamic weights. Subsequently, data-driven dynamic membership functions were designed. Based on this, a data augmentation method was introduced to capture local trends in accident frequency and severity. These processes produced a three-dimensional fuzzy matrix, representing the temporal evolution of risk levels across assessment indicators. Finally, cloud models were incorporated to manage data uncertainty using the improved TOPSIS algorithm. Time decay and standard deviation corrections were introduced to optimize dynamic weight management. The DW-TDFE model achieves superior accuracy and reliability relative to conventional approaches. This study offers theoretical and practical references for the dynamic risk assessment of oil and gas pipelines.