Abstract: A novel online identification and detection method based on watermarking and particle swarm optimization (PSO) algorithm is proposed to address the issue of false data injection (FDI) attack detection in the load frequency control (LFC) system under parameter uncertainty. Firstly, a state estimation model for the LFC system is established based on Kalman filtering. To assist the χ² detector in identifying FDI attacks, a transmission strategy of adding a pseudo-random watermark matrix during information transmission is proposed. Prior to transmission, the transmitting terminal multiplies the measurement to be transmitted by a pseudo-random matrix. Subsequently, the receiving terminal decrypts the received data to retrieve the genuine information. Compared with traditional methods, this strategy neither increases computational complexity nor sacrifices estimation performance in the absence of attacks. When an FDI attack occurs, the system’s measurement signals are maliciously tampered with by attackers. During the decryption process at the receiver end, a unique coupling relationship emerges between the pre-embedded watermark matrix and the injected attack signal. This coupling feature is highly distinguishable by the χ² detector, enabling rapid and accurate identification of FDI attacks and thus ensuring the efficient and stable operation of the LFC system. Furthermore, to tackle the parameter uncertainty inherent in LFC systems (a factor that may degrade the accuracy of state estimation and attack detection), the PSO algorithm is introduced for online parameter identification. This algorithm optimizes the key parameters of the LFC system in real time, significantly improving the matching degree between the established state estimation model and the actual operating system. Finally, by fusing the watermarking-based attack detection strategy and the PSO-based parameter identification technique, a complete online FDI attack identification and detection method for LFC systems is formed. The effectiveness and feasibility of the proposed method are verified through simulations on a two-area LFC system.