Data-driven integrated method for frequency security assessment and control in power systems

• A data-driven integrated research framework for frequency security assessment and emergency control is proposed. • Data anomaly detection is implemented using GANomaly to enhance data security. • A frequency security assessment model is constructed by integrating Transformer and LSTM. • Establish a frequency emergency control model based on the DDPG algorithm. With the increasing penetration of renewable energy sources and the extensive deployment of power electronic devices, the frequency security of power systems is facing escalating challenges. When system disturbances cause frequency instability, it becomes essential to promptly assess the system’s frequency security status and implement corresponding emergency control measures. Therefore, this study proposes a data-driven integrated framework for frequency security assessment and emergency control that incorporates data anomaly detection. Specifically, the GANomaly model is employed to detect and filter abnormal data, thereby mitigating the impact of noise, cyberattacks, and other disruptions during the assessment process. Subsequently, the Transformer-based deep learning model is applied to evaluate the frequency warning level and predict the maximum frequency deviation. In addition, a deep reinforcement learning controller based on the deep deterministic policy gradient is designed to connect the assessment stage with the emergency control decision process. Finally, the proposed approach is validated using the IEEE 39-bus system. The experimental results demonstrate that the method can reliably provide frequency security warnings and predict maximum frequency deviations even under data disturbances, while simultaneously generating effective control strategies to enhance frequency stability.