There is an issue of low utility rate of solar thermal power plants, high waste rate of wind power, and high carbon emission of traditional gas-fired units and a limitation of "heat-dependent power" operation, etc. To tackle this, there are the following actions to be taken: introducing oxygen-enriched combustion capture to retrofit conventional units, configuring photothermal power plants with heat recovery to realize thermal electrolysis coupling, coupling high-temperature solid oxide electrolysis cells and other energy conversion equipment to build an electricity-thermal- hydrogen low-carbon energy system and its capacity optimization allocation method. First, considering the uncertainty of wind power output and light intensity as well as the temporal correlation with electric load, a multi-run scenario extraction model based on two-stage temporal clustering is established. Second, using probability-based multi-run scenarios, this paper measures the risk caused by uncertainty through conditional value at risk (CVaR) theory, and constructs a low-carbon energy system capacity optimization allocation model with the objective of minimizing the total cost. Finally, the simulation is validated by an arithmetic example, and the results show that the system can reduce annual carbon emission and wind abandonment rate and improve the utilization rate of CSP plants when meeting the load demand. It provides a quantitative basis for decision makers with different risk preferences when facing the problem of system capacity optimization.