To address the issues that inertia support capacity is insufficient in high-proportion renewable energy power systems and it is difficult for existing energy storage systems to adopt the virtual synchronous generator fixed inertia damping control strategies to adapt to the multi-stage characteristics of frequency dynamic response

this paper proposes a multi-stage dynamic frequency control strategy for energy storage based on inertia-damping coordination and optimizes the relevant parameters. First

a four-stage dynamic model encompassing inertia support

primary frequency regulation

and steady-state recovery is established on the basis of the gradient characteristics of the frequency response curve

revealing the differentiated requirements for inertia and damping parameters across different frequency regulation stages. Second

a mechanism for smooth transition between inertia and damping is designed by integrating a hyperbolic tangent buffer function and a sigmoid buffer function. This mechanism enables coordinated switching between inertia and damping through the buffer functions

thereby avoiding secondary disturbances caused by sudden parameter changes. Given that deep learning theory can automatically extract features and optimize parameter values under complex operating conditions

the deep Q-network algorithm is employed to optimize frequency modulation control parameters

thereby further enhancing the system’s frequency response performance. Finally

simulation results show that

compared to the traditional neural network-based inertia-damping coordination tuning strategy

the proposed method reduces the system frequency change rate by 32.58%

reduces steady-state frequency deviation by 15.50%

and shortens recovery time by 20.06% and a 8.08% reduction in overshoot. This validates the effectiveness of multi-stage dynamic frequency control in enhancing the transient stability of high-penetration renewable energy grids

and the introduction of the deep Q-network algorithm significantly improves parameter optimization performance. The proposed scheme provides a theoretical basis and technical support for the large-scale participation of energy storage in system frequency regulation.