Emergency Load-Shedding Decision for Frequency Stability of New Energy Power System Based on Constrained Markov Decision Process

Renewable energy systems dominated by powered electronic devices generally exhibit weak disturbance tolerance and limited grid-support capability. Following the blocking of a flexible DC transmission system, emergency load shedding in renewable-rich grid regions may induce overvoltage or undervoltage at the point of common coupling, forcing renewable energy units into a voltage ride-through state. This, in turn, reduces their active power output and threatens the frequency stability of the power system. To address this issue, this paper proposes an emergency load-shedding decision model based on a constrained Markov decision process (CMDP). First, an emergency frequency control model for AC–DC hybrid power systems is established within the Markov decision process framework, thereby formulating power system frequency stability control as a Markov decision problem. Second, Lagrange multipliers are introduced into the CMDP framework to transform the constrained optimization problem with security constraints into an unconstrained objective optimization problem. Finally, the Proximal Policy Optimization (PPO) algorithm is adopted to accelerate the training process and improve the decision accuracy of the intelligent agent. The simulation results, based on the modified IEEE 39-bus system, demonstrate that, compared with the traditional contingency strategy and the conventional Markov decision algorithm, the proposed load-shedding strategy can satisfy system frequency stability requirements, effectively avoid voltage violations at renewable energy grid-connection points, and minimize the total load shedding amount.