To address the capacity configuration issue of energy storage assisting thermal power units in AGC load variation, an analysis was first conducted on the AGC load variation data of a 660 MW-rated thermal power unit over one-week periods in spring, summer, autumn, and winter, examining the characteristics of the magnitude, direction, and interval time of AGC load variation commands; a capacity configuration calculation method and an auxiliary load variation control strategy were proposed, considering both independent or short-interval continuous step changes of AGC load variation commands and the self-recovery characteristics of the energy storage system’s SOC (state of charge). Finally, using actual load variation data from a typical day, PSO (particle swarm optimization) algorithm was applied, revealing that with a variation rate of 2.45%Pe (rated power), configuring a power-type flywheel energy storage system with a power of 20.082 MW and a capacity of 2.75 MW·h could achieve a maximum return rate of 0.92. As the frequency regulation price and frequency regulation mileage increase, the unit’s return rate continues to rise, exceeding 1 when the frequency regulation price reaches 11 CNY/MW or the frequency regulation mileage reaches 2 800 MW. The capacity accounting method proposed in this paper provides a useful reference for capacity configuration of other types of units.