Research on attack detection and state estimation for distribution networks based on sparse recovery and unknown input observer

The integration of distributed generation in distribution networks introduces asymmetry and localized coupling characteristics, which render the system vulnerable to control input disturbance attacks. These attacks can affect the system state via the control loop and are difficult to observe directly, thereby exhibiting significant concealment. In this context, a closed-loop state-space model considering control disturbances was formulated in this paper to characterize the dynamic propagation mechanism of disturbance signals and quantify their impact on observed variables and state estimation errors. To achieve both disturbance detection and state estimation, a two-stage method integrating sparse recovery and an Unknown Input Observer (UIO) was proposed. A joint error model was constructed using a sliding window, and sparse optimization was applied to estimate the disturbances. Based on the estimated disturbances, an UIO was introduced to reconstruct the state under disturbance isolation. Furthermore, to address the scenario of non-sparse disturbances, a robust observer was designed using Linear Matrix Inequality(LMI) constraints to ensure bounded state estimation errors. Simulation results based on a typical distribution system demonstrated that the proposed method outperforms conventional filtering approaches in identifying control input disturbances and correcting system states, exhibiting higher estimation accuracy and enhanced dynamic adaptability.