Abstract: Distributed energy systems offer various advantages, including clean, low-carbon and efficient energy utilization, making them effective in mitigating energy crises and greenhouse gas effects. However, the fluctuating and intermittent nature of most renewable energy sources poses challenges for precise carbon accounting. In order to enhance the spatiotemporal accuracy of existing carbon accounting methods and facilitate systematic carbon emission management, the "carbon perspective" approach is adopted. Initially, The impact of renewable energy variability on the carbon emission intensity of power generators, introducing dynamic carbon emission factors to characterize varying emission intensities across different time intervals. Subsequently, these dynamic factors are integrated into an optimization model, forming a two-stage real-time optimization model for distributed energy systems. Finally, a three-stage carbon management model encompassing carbon forecasting, optimization and accounting, and carbon trading is proposed. Results indicate that higher variability in unit load intensity corresponds to increased accuracy when using dynamic carbon emission factors compared to static ones. The two-stage real-time optimization algorithm during operational scheduling significantly reduces carbon emissions. Implementation of the three-stage carbon management model leads to a substantial decrease in economic and environmental costs, facilitating fine-grained carbon emission management at the micro level and achieving a bottom-up approach to carbon management.