With high-penetration wind-solar integration into the new power system

conventional static annual-average electricity carbon emission factor suffer from insufficient spatiotemporal resolution. Existing factors only account for direct generation-side carbon emissions

leaving the carbon emissions from thermal units' peak-shaving and reserve services unallocated

which hinders refined low-carbon dispatch. This paper proposes a source–load carbon responsibility allocation method based on real-time unit load rates and constructs a multi-spatiotemporal-scale accounting model for the time-of-use and zonal electricity carbon responsibility factor (Cr). Case studies are conducted using actual grid operating data and high-renewable-penetration simulation scenarios. Results show that Cr exhibits significant spatiotemporal heterogeneity

with its intraday fluctuation amplitude four times that of the conventional carbon emission factor C. It avoids misleading load-reduction commands during early-morning hours and conforms to practical grid-operation requirements. Guided by Cr

demand-response regulation lifts the renewable-energy accommodation rate from 92.2% to 99.3% under a 70% renewable-energy-penetration scenario. Nationwide large-scale annual application can achieve an annual carbon emission reduction of 600 million tons

delivering effective low-carbon regulation signals for refined source-load coordinated decarbonization of new power systems.