Distributed Composite Learning Dynamic Event-Triggered Control for Nonlinear Multi-Agent Power Systems

In this paper, a distributed composite learning dynamic event-triggered (ET) control protocol is presented for nonlinear multi-agent power systems (NMAPSs) in the presence of switching topologies, uncertain nonlinearities, external disturbances and limited network resources. A predictor-based continuous emotional self-structuring neural network (NN) is proposed to approximate unknown nonlinearities of NMAPSs. Flexible structure and emotion-based approaches not only can balance the contradiction between computational burden and control performance but also keep a fast response property of NN approximate. The predictor can improve the approximation accuracy and interpretability of NN approximate by introducing a prediction error to update NN's weights. Next, a dynamic ET mechanism is presented, which introduces a dynamic self-regulation variable to prolong the ET interval. On this basis, the designed control protocol is sent to actuator only at the ET instant to further reduce the network burden. Then, a distributed composite learning control protocol is developed for NMAPSs by utilizing the Lyapunov stability theorem. It can guarantee that all signals in the closed-loop system are bounded under a class of switching topologies with the average dwell time, and the Zeno phenomenon is avoided ultimately. Finally, simulation results are provided to demonstrate the effectiveness of the proposed protocol.