Research on Photovoltaic Inverter Network Security Enhancement Method Based on System Identification Model

Photovoltaic (PV) inverters play a critical role in PV power generation systems by converting direct current (DC) into alternating current (AC). Their dynamic characteristics are influenced by power electronic switching actions, control strategies, load variations, and external disturbances, increasing the complexity of system identification and control. Additionally, the development of smart grids has exposed PV inverters to cybersecurity threats such as false data injection (FDI) attacks and replay attacks, which may lead to abnormal system operation or even grid instability. To enhance the security of PV inverters, a method based on system identification and watermarking signals was proposed. First, a mathematical model of the PV inverter was established, and system identification was performed using an autoregressive moving average model with an out-of-band excitation input to obtain an accurate description of the inverter's input-output relationship. Subsequently, the least squares (LS) method was employed to estimate system parameters, improving identification accuracy. Based on this, a watermarking signal was introduced into the modulation index, and a statistical variance test was utilized to detect abnormal data patterns, enabling the identification of FDI and replay attacks. Experimental platform validation demonstrated that the proposed method could detect attacks within 8 milliseconds, as specified by the IEEE 1547 standard, thereby enhancing system security and robustness. Comparative analysis indicated that the proposed method accurately described the dynamic characteristics of PV inverters, while the randomness of the watermarking signal improved the reliability of attack detection. Experimental data further confirmed the applicability of this method under different attack scenarios, providing both theoretical and experimental support for the cybersecurity protection of PV inverters.