Design of an interconnected power CPS anti-packet-loss oscillation suppression controller based on HyLNN-IE-PID

With the development of interconnected multi-machine power cyber-physical systems (CPS), network packet loss has emerged as a critical issue affecting the control precision of the system. In this paper, the issue of network packet loss in a dual-machine interconnected power CPS is addressed by designing a hybrid liquid neural network (LNN)-based improved Euler (IE) neural adaptive PID controller (HyLNN-IE-PID). The proposed controller integrates the dynamic model of liquid neural networks, the online learning capabilities of conventional neural networks, and numerical solutions based on IE, enabling the conventional PID controller to model the time-series data effectively. This approach eliminates the reliance on historical data, batch updates of new data, and model retraining, achieving rapid response and adaptive adjustment to complex environments such as network packet loss and faults. Considering a dual-machine interconnected power CPS, the performance of the proposed HyLNN-IE-PID controller that uses dual-machine speed feedback is compared with a recurrent neural network-based PID controller (RNN-PID) and a spiking neural network-based PID controller (SNN-PID) with similar feedback structures under varying packet loss conditions. In addition, the proposed controlled is compared with the conventional PID and backpropagation PID controllers, which use single-machine speed feedback. The results demonstrate significant reduction in the oscillations of generator terminal voltage, line voltage, power angle difference between the two machines, and motor speed. This research provides theoretical support and practical reference for the future integration of smart grids with artificial intelligence technologies, such as LNNs.