Yang Yang

1. Microgrids (MGs) have emerged as an effective solution for improving energy efficiency and sustainability through the integration of renewable energy sources and distributed generation (Chai et ...

Accurate power load forecasting is crucial for smart grid security and efficiency, yet traditional centralized methods fall short in addressing data privacy and communication demands. Federated Learning (FL) offers a promising solution but faces dual challenges: the Non-Independent and Identically Distributed (Non-IID) nature of load data limits model generalization and invites security threats like model poisoning attacks. To address these challenges, this paper proposes a Federated Meta-Learning (FedML) framework integrated with Entropy-Regularized Gradient-Weighted (ER-Grad) aggregation. The framework employs FedML to construct high-precision personalized forecasting models for clients while maintaining training efficiency, thereby significantly enhancing predictive performance. Additionally, the ER-Grad aggregation algorithm is introduced to dynamically calibrate client weights, effectively suppressing abnormal or malicious updates and strengthening the system’s robustness against model poisoning attacks. Experimental results demonstrate that the proposed method substantially improves overall and peak forecasting accuracy, ensures comprehensive robustness against malicious attacks and communication dropouts, and achieves training with high efficiency.

This paper investigates the consensus control problem of multi-agent systems under intermittent communication and unmodeled external disturbances. The main contribution is to overcome the limitation of current synchronous intermittent framework and propose a novel asynchronous intermittent communication methodology on multi-agent systems. In this intermittent methodology, the state space is divided into three distinct regions by introducing both safety and intermittent boundaries, which enables effective monitoring of agent error dynamics.Furthermore, an asynchronous intermittent communication protocol is designed, where the activation and rest intervals are adjusted based on the real-time error states of the agents.By utilizing the distributed extended observer to observe the relative output information and unmodeled disturbances, the novel asynchronous intermittent consensus protocol with disturbance rejection is designed to realize the overall consensus of the multi-agent systems. The proposed spatial-segmentation-dependent intermittent communication methodology can adjust communication and non-communication time of each agent asynchronously according to the communication requirements, under which the multi-agent systems can tolerate more non-communication time and reduce the communication frequency. Finally, numerical simulations are performed to verify our results.

It is a critical issue for achieving privacy-preserving consensus for nonlinear multi-agent systems (MASs) with triggered conditions and consensus performance. A game-based event-triggered privacy-preserving consensus control scheme is proposed with a non-uniform decomposition. For a nonlinear MAS with known dynamics, a non-uniform-decomposed privacy-preserving consensus scheme is constructed, and this scheme is further improved for a nonlinear MAS with unknown dynamics via neural network identification methods. With the non-uniform decomposition technique, each agent is decomposed into interactive substrates and internal substates respectively, and two different positive weight factors are introduced into the interaction between the two substates. Eavesdroppers are not able to reconstruct all information from the non-uniform-decomposed interactive signals within communication networks. In the framework of non-uniform decomposition, a privacypreserving game-based event-triggered mechanism (PP-GETM) is also designed by partial interactive information, where an agent’ performance and its triggered error are regarded as the two players in a zero-sum game in the decomposed substate. Then, with the Lyapunov function, it is proven that, with the non-uniform decomposed consensus control schemes, all the signals in the closed-loop MAS are guaranteed to be bounded with the private-preserving objective. Also, Zeno behaviors are excluded with the proposed PP-GETM. Finally, two numerical examples are given to verify the effectiveness of the proposed scheme.

In this article, a boundary-dependent asynchronous intermittent methodology is proposed to address the optimal cooperative tracking control issue of cyber-physical systems. Different from most works on intermittent control, this intermittent methodology allows each agent to arrange operation time and non-operation time based on their actual control requirements. Within the proposed asynchronous intermittent framework, the real vector space is partitioned into three distinct sub-regions via rest and security boundaries to monitor the error states of individual agents. Then, a new asynchronous intermittent mode is introduced to dynamically allocate operation and non-operation time based on these real-time error states. By developing a resilient fault management framework, the actuator outage behaviors are avoided. Based on this, dynamic fault-tolerant control strategy of cyber-physical systems under asynchronous intermittent scheme is designed, under which the optimal tracking cooperative control can be realized. In contrast to conventional synchronous intermittent frameworks, the proposed boundary-dependent asynchronous intermittent methodology dynamically tailors the operation and non-operation time of individual agents according to specific control requirements. Therefore, it enhances the tolerance of cyber-physical systems for non-operation time and reduces communication frequency. Finally, numerical simulations are provided to verify the effectiveness of the proposed theoretical findings.

Multi-agent systems (MASs) in open networks face dual security threats: Byzantine attacks that steer malicious consensus and eavesdroppers that steal private information. Existing resilient consensus protocol isolates Byzantine attacks by relying on (2f +1)-robust networks, which imposes stringent topological constraints and fails to provide privacy preservation simultaneously. To address this issue, an improved resilient consensus protocol with f (IRCP-f) is proposed, via absolute values of relative states, to defend against f-local Byzantine attacks. This protocol only requires that an undirected and connected graph is (f + 1)-robust instead of (2f + 1)-robust. The properties of the dynamic network processed by the IRCP-f are analyzed, and the graph conditions for achieving average consensus are consequently satisfied. A fully distributed differentially private event-triggered average consensus (DPETAC) control scheme is then developed. With the DPETAC control scheme, convergence analysis, Zeno behavior analysis, accuracy analysis and privacy analysis are presented for the MAS. Finally, a numerical simulation illustrates the feasibility and effectiveness of the proposed privacy-preserving average consensus control scheme.