Hybrid Model and Data-Driven Emergency Load Shedding Optimization for Frequency Security in Receiving-End Power Grids

Aiming at the receiving-end power grid frequency security after HVDC blocking events, this paper proposes a hybrid model and data-driven optimization method of emergency load shedding (ELS). Firstly, a decision-making model of ELS considering multiple dynamic security constraints, e.g., frequency and rotor angle, is constructed. Then, the particle swarm optimization (PSO) algorithm is improved by integrating with symmetric opposite learning and Tent chaotic mapping to obtain high-quality ELS schemes. Finally, to boost the optimization efficiency, a data-driven dynamic security assessment model based on the hybrid neural network is constructed and introduced into the solution process of PSO. To ensure the feasibility of the final ELS scheme, the model-driven time-domain simulation method is adopted for validation. The effectiveness of the proposed ELS optimization method is verified on a receiving-end power grid with multi-infeed HVDC lines. It can obtain a high-quality and feasible ELS scheme within 0.7 s.