Emergency Load-Shedding Strategy for Power System Frequency Stability Based on Disturbance Location Identification

With the evolution of modern power systems, the proportion of renewable energy generation in the grid continues to grow. At the same time, grid operation modes have become increasingly complex and dynamic, leading to heightened uncertainty in disturbance faults. Moreover, power electronic equipment exhibits relatively low-level immunity to disturbances. The issue of frequency stability in power systems is becoming increasingly severe. These factors make the pre-programmed control strategies based on strategy tables, which are widely used as the second line of defense for frequency stability in power systems, prone to mismatches. When a power disturbance occurs, it is crucial to adopt an appropriate emergency load-shedding strategy based on the characteristics of unbalanced power distribution and the network’s frequency profile. In this paper, for a simplified multi-zone equivalent system, the coupling relationship between different load-shedding locations and the system’s frequency response after a disturbance is analyzed. This analysis integrates the power distribution characteristics after the disturbance, a system frequency response (SFR) model, and the frequency distribution law in the network. It is demonstrated that under identical load-shedding amounts and action times, implementing load shedding closer in electrical distance to the disturbance location is more beneficial for stabilizing system frequency. A convolutional neural network (CNN) is employed to localize system faults, and combined with research on the emergency load-shedding amounts based on SFR model parameter identification, a rapid disturbance location-based emergency load-shedding strategy is proposed. This strategy enables prompt and accurate load-shedding actions to enhance the security and stability of the power system. Finally, the effectiveness of the proposed approach is validated using the CEPRI-LF standard arithmetic system.