Abstract: With high-proportion wind and solar integration into the new power system, the conventional static annual average electric carbon factor suffers from insufficient temporal and spatial resolution, and existing factors only account for direct generation emissions, leaving the carbon cost of thermal peak-regulation standby unallocated and hindering refined low-carbon dispatch. This paper proposes a source–load carbon responsibility allocation method based on real-time unit load rates and builds a multi-temporal-spatial accounting model for the time- and area-specific electric carbon responsibility factor (C_r), with case studies on actual grid data and high-renewable penetration scenarios. Results show that C_r exhibits significant temporal and spatial heterogeneity, with its intraday fluctuation being four times that of the conventional factor C. It avoids misleading load reduction in early-morning hours and conforms to practical grid operation rules. Guided by C_r, demand-response regulation raises the renewable-energy consumption rate from 92.2% to 99.3% under 70% renewable-energy-penetration scenario. Nationwide large-scale application can achieve an annual carbon reduction of 600 million tons, delivering an effective low-carbon regulation signal for refined source-load coordinated decarbonization of new power systems.